Datasets:

Hendrik-a

/

arxiv

like 0

even though approximated , strong coupling non - perturbative qcd amplitudes remain very difficult to obtain . in this article , in eikonal and quenched approximations , physical insights are presented that rely on the newly - discovered property of _ effective locality_.

1207.5017

as one derives the _ effective locality _ property of the fermionic qcd amplitudes @xcite , a bizarre factor of @xmath0 comes about , in an unexpected place , where @xmath1 is the impact parameter of two scattering quarks . inspection shows that this curious factor is not related to any systematic approximation scheme , such as eikonal or quenching . rather , it shows up as an unavoidable feature of any treatment , exact or approximate , and is likely to suggest the neglect of transverse fluctuations of quarks confined inside hadrons . in this respect , quite interesting perspectives offer themselves @xcite . both quenched and eikonal approximations will be used throughout . both seem quite in line with the context of a highly energetic scattering process : the former because one may think that the vacuum back - reaction has not enough time to influence the process , and the latter because it is constructed in a lorentz covariant way from the old bloch - nordsieck approximation @xcite , and is appropriate for high - energy scattering at momentum transfers substantially less than the scattering energies . for such processes , the fourier momenta @xmath2 of the interacting @xmath3- fields are typically of the same order - of - magnitude as the momentum transfer , and the conventional eikonal approximation is built around the idea that @xmath4 ( for `` hard scattering '' , there are eikonal variants of this `` soft '' scattering setting @xcite ) . in this article , our aim is twofold . in the first place , it is made clear that this @xmath0 can not be avoided , and a rigorous mathematical proof is proposed to support this claim . this is the matter of section [ sec:2 ] . then , in section [ sec:3 ] , as a second step , in eikonal and quenched approximations , it is shown how the powerful _ random matrix _ calculus can be used to obtain the generic , closed form expression of qcd fermionic amplitudes in the strong coupling regime . eventually , in section [ sec:4 ] , the forms so obtained will be used to revisit and improve upon a previous estimate of the quark / anti - quark binding potential and related aspects @xcite . a discussion of the results is presented in section [ sec:5 ] , and a short appendix concludes this article . using standard functional notations and manipulations @xcite , and a small re - arrangement that guarantees rigorous gauge invariance for all scattering quark amplitudes @xcite , the qcd generating functional can be written as @xmath5 & = & \mathcal{n } \ e^{\frac{i}{2 } \ , \int{j \mathbf{d}_{c}^{(\zeta ) } j } } \int{\rm{d}[\chi]\ e^{\frac{i}{4 } \int{\chi^{2 } } } \ e^{\mathfrak{d}_{a } } \ e^{\frac{i}{2 } \ , \int{\chi \mathbf{f } } + \frac{i}{2 } \int{a { \mathbf{d}_{c}^{(\zeta)}}^{-1 } a } } } \nonumber \\ & & \times \left . e^ { i \int{\bar{\eta } \mathbf{g}_{c}[a ] \eta } } \ e^{\mathbf{l}[a ] } \right|_{a = \int{\mathbf{d}_{c}^{(\zeta ) } j}},\end{aligned}\ ] ] where @xmath6 is a normalization constant , and the @xmath7 , @xmath8 and @xmath9 stand for bosonic and fermionic sources , respectively . note that the causal gluonic propagator @xmath10 may differ by a factor of @xmath11 from the more customarily used feynman propagator . the ` linkage operator ' @xmath12 involves the covariant gluonic propagator with gauge parameter @xmath13 , @xmath14.\ ] ] the @xmath15-field appearing in @xmath16 $ ] has long been used to _ the exponential of the original @xmath17-dependence that enters the original qcd lagrangian density @xcite : @xmath18\ , e^{{\frac{i}{4}}\ , \int\chi^2 + { \frac{i}{2}}\ , \int \chi^{\mu\ , \nu}_{a}\ , \mathbf{f}_{\mu\ , \nu}^{a } } } .\ ] ] proceeding in this way , the @xmath19-gauge field dependence are made gaussian instead of cubic and quartic , allowing for straight forward linkage operations . in eq . ( [ eq : s2 - 1 ] ) , the original fermionic fields have been integrated out : this has given rise to the term involving the fermionic propagator @xmath20 $ ] in the gauge field background configuration @xmath21 , as well as to the logarithm of the fermionic determinant @xmath22 $ ] ( up to a constant absorbed into the overall normalization @xmath6 ) . that is , @xmath23^{-1 } | y\rangle\,\ ] ] and @xmath24 = { { \mathrm{tr}\ , } } { \ln[1 - i\ , g\ , ( \gamma\ , a\ , t)\ , \mathbf{s}_c ] } , \quad \mathbf{s}_c = \mathbf{g}_c[0].\ ] ] in order to be able to display the property of effective locality , one must devise convenient enough representations for the functionals @xmath20 $ ] and @xmath22 $ ] . schwinger - fradkin representations can be used ; for example , a ` mixed ' ( configuration and momentum space ) expression of @xmath20 $ ] is given by @xmath25 |y \rangle } & = & e^{-i p \cdot y } \ i \int_{0}^{\infty}{\mathrm{d}s \ , e^{-is m^{2 } } } \ e^{- \frac{1}{2 } { { \mathrm{tr}\ , } } { \ln{\left ( 2h \right ) } } } \nonumber\\ \nonumber & & \times \int{\mathrm{d}[u ] } { \left\ { m - i \gamma \cdot \left[p - g a(y - u(s ) ) \right ] \right\ } } \ e^{{i\over 4 } \int_{0}^{s}{\mathrm{d}s ' \ , [ u'(s')]^{2 } } } \ e^{i p \cdot u(s)}\nonumber\\ & & \times \left ( e^{g \int_{0}^{s}{\mathrm{d}s ' \sigma \cdot \mathbf{f}(y - u(s ' ) ) } } \ e^{-ig \int_{0}^{s}{\mathrm{d}s ' \ , u'(s ' ) \cdot \mathbf{a}(y - u(s ' ) ) } } \right)_{+}\end{aligned}\ ] ] and similarly for @xmath22 $ ] , @xcite . in the representation above , one has @xmath26 and @xmath27 , the customary commutator @xmath28 $ ] controls the spin - related contributions . this expression ends up with a subscript @xmath29 to mean that time ordering with respect to schwinger proper - time is in order , because of the @xmath30-lie - algebra valuation of the gauge potentials @xmath31 , and field - strength tensor @xmath32 . the fradkin field variables satisfy @xmath33 and a few words of explanation may be in order . from fradkin s original representation , originally in terms of @xmath34 and here denoted by @xmath35 , this quantity has the interpretation of a particle s @xmath36-velocity or , depending on the mass - scale used , of a @xmath36-momentum . but while undergoing fluctuations at every @xmath37-value , that particle needs not be on its mass shell . it is in the context of an eikonal approximation that @xmath35 can be replaced by its mass shell @xmath36-momentum ( that is , @xmath38 ) , because it is then assumed that the particle s @xmath36-momentum is much bigger than any momentum transfer ( @xmath39 ) , and that the mass shell hypothesis is accordingly a reliable one . note that this approximation should be relevant to the non - perturbative , semi - classical level of description considered here , whereas ` harder ' excitations should concern the next - to - leading order terms of a @xmath40-expansion , that is , the description of the perturbative phase of qcd @xcite . eventually , it must be stressed that none of the features of effective locality is dependent upon this eikonal approximation ; rather , the latter is used to provide a simple enough way to derive an effective quark - binding potential out of the relevant eikonal amplitude , taken as a function of the quark s _ impact parameter _ , _ i.e. _ , the inter - quark separation @xcite . the full derivation of the effective locality property will not be repeated here since its complete non - approximate proof can be found in ref . rather , focus will be put on specific and important technical / physical aspects of effective locality . on ( [ eq : s2 - 2 ] ) , however , and within the quenched approximation , it is immediate to realize that in order to be able to perform the linkage operations in an exact way , then , @xmath41 $ ] factors have to be pulled downstairs by functional differentiations with respect to the fermionic sources @xmath42 and @xmath43 . this is why the property of effective locality is not discernable from the generating functional itself , but after the sums over all gluon exchanges have been performed , on the full set of @xmath44-point fermionic green s functions . though equivalent , and thus not a crucial difference , a form of the effective action is under investigation , on which effective locality could be read off directly . this point would be important when this formulation of qcd amplitudes generates a possible _ most dual _ qcd formulation , as was convincingly proposed in ref . @xcite in the pure yang - mills case , or in relation to the recent results of ref . @xcite . in a first attempt to the effective locality property of qcd @xcite , and within the same approximate context as the one being studied in the present article , one finds in the effective action an exponential factor of @xmath45^{-1}\right|^{\mu\nu}_{ab } \ q^{b}_{\nu}(x)}\ ] ] representing , since all gluon interactions have been summed upon , the action of an effective local interaction term between two colored currents . the interacting currents are @xmath46,\ ] ] where @xmath47 , @xmath48 , stand for their associated quark content . these expressions describe a @xmath49-body scattering process and is related to a @xmath36-point fermionic green s function . in eikonal approximation and in a non - perturbative strong coupling regime , @xmath50 , the leading part of eq . ( [ eq : s2 - 5 ] ) reads accordingly @xcite @xmath51 where the @xmath52 and @xmath53-variables are used so as to extract the @xmath54-field dependence out of the ordered exponential of eq . ( [ eq : s2 - 2 ] ) @xcite . the expression above is obtained in the standard eikonal and quenched approximations , which , for the former , also includes the neglect of spin - related contributions . however , it is important to emphasize that the full non - approximate expression @xcite would manifest exactly the same technical intricacy as the one under consideration , and this is why the point can be made using the simplified example of eq . ( [ eq : s2 - 6 ] ) . the technically important aspect of ( [ eq : s2 - 6 ] ) is the following : what is to be thought of @xmath55 that is , how should one interpret such a factor as @xmath56 ? at face value , for any given pair of values @xmath570 , { s}]\times ] 0 , \bar{s}]$ ] , and any pair of arbitrary functions @xmath58 , each belonging to some infinite dimensional functional space , the probability of coincidence of @xmath59 with @xmath60 is likely to be infinitesimally small , if not zero ( the reason why the points @xmath61 are excluded will appear clear below , in eq . ( [ eq : mbym-2 ] ) ) . besides , in the @xmath49-body center of mass system ( @xmath62 ) , a somewhat heuristic manipulation of the @xmath63- and @xmath64- delta constraints would suggest that @xmath65 , and write @xcite @xmath66 where the dots are to be evaluated shortly . it is assumed that the fradkin fields @xmath67 are @xmath680,s ] , ] 0 , \bar{s}]\rightarrow { \mathbb{r}}^4\right)$ ] . then , out of @xmath69 , and in view of eq . ( [ eq : s2 - 8 ] ) , this leads to a remaining constraint of @xmath70 where @xmath71 is the impact parameter , or transverse distance between the two scattering quarks . given the very place where it comes about , as a multiplicative factor in an exponential , this @xmath72 is awkward . in ref . @xcite , it has been suggested that this factor is a remnant of the implicit existence of asymptotic quark states , an assumption which , beyond the stage of perturbation theory , can not be maintained in qcd , neither theoretically nor experimentally @xcite . this point of view leads to a change of @xmath72 into a modified , smeared and normalized impact parameter distribution of form @xcite @xmath73 where @xmath74 is a typical mass term used to calibrate the transverse momenta distribution of quarks inside a given bound state . in the transverse plane , bound quarks are endowed with transverse momenta taken as independent random variables , exponentially suppressed above a given mass parameter @xmath74 . that is , describing the non - perturbative regime of qcd , in order to avoid the nonsensical @xmath72 , it is necessary to introduce a mass scale where there was nt any before , in agreement with the considerations of ref . @xcite , for example . first fits , with respect to a model pion @xmath75 and a model nucleon @xmath76 @xcite , indicate a value of @xmath74 close to the pion mass , and a value of @xmath77 , on the order of @xmath78 , small enough as it should if one is willing to preserve baryonic linear regge behaviours through a(n almost ) linear confining potential @xcite . eventually , it is worth remarking that taking @xmath79 negative - valued in eq . ( [ eq : s2 - 10 ] ) , then @xmath80 is ( the _ characteristic function _ of ) a _ lvy - flights _ probability distribution ( lvy flight distributions are stable probability distributions ) . they include and generalize gaussian distributions for independent random variables , and comply with a generalized _ central limit theorem _ of statistical physics @xcite ) . that is , starting from quark propagation as ordinarily conceived , _ la _ @xmath81 , in the non - perturbative bound context , one could be lead to think of ( transverse ) quark propagation in terms of _ lvy - flights _ that are _ markovian _ diffusive processes ; a picture which , in view of color confinement , may look pretty sensible indeed . whatever its interpretation , though , the latter will be ruined if eq . ( [ eq : s2 - 8 ] ) does not provide a reliable enough evaluation of eq . ( [ eq : s2 - 7 ] ) . in order to deal with that issue , one may skip to the most achieved realization of a functional space , that is to the wiener functional space @xcite . note that even though that space may be not the only one to be conceived ( and/or taken into account , according to ref . @xcite ) , there is no restriction at all in working out a proof within that space , because fradkin vectorial fields are real valued . then , the following theorem can be proven : _ * theorem * _ : for all pairs @xmath820 , { s}]\times ] 0 , \bar{s}]$ ] , @xmath83 @xmath84 with the wiener measure @xmath85 on @xmath86 , whereas @xmath87 is taken as the wiener measure on the product of spaces @xmath88 , endowed with the topology product . this is made possible thanks to the independence of the random variables @xmath89 and @xmath90 that are here taken to represent either of the 2 possibilities @xmath91 and @xmath92 . this theorem proves that the right hand side of eq . ( [ eq : s2 - 8 ] ) is necessarily proportional to @xmath93 . dimensional and symmetry arguments ( the right hand side must be symmetric both in the exchange of indices @xmath94 , and under the combined exchange @xmath95 and @xmath96 , whereas the normalization is fixed by comparison to the eikonal approximation @xcite ) can be used to complete the right hand side of eq . ( [ eq : s2 - 8 ] ) . one finds @xmath97 and because of the @xmath62-relation @xmath98 , which holds true at eikonal approximation , eq . ( [ eq : s2 - 11 ] ) gets reduced to eq . ( [ eq : s2 - 8 ] ) without the dots . the proof is as follows : let @xmath21 be the set @xmath99 . one has @xmath100 , where @xmath101 because of the obvious inclusion , @xmath102 , @xmath103 , one can write @xmath104 now , @xmath105 is given by @xmath106 since one has the obvious inequality of @xmath107 one gets immediately @xmath108 as a mere consequence of the _ dominated convergence theorem _ , allowing one to take the @xmath109-limit under the integral , whereas the second equation , eq . ( [ eq : measure_normal ] ) , follows from normalization . the unavoidable presence of this surprising factor of @xmath72 is thus demonstrated . then , in order to get a non - trivial ( almost linear ) inter - quark binding potential at moderate distance , no much larger than @xmath110 , ( to be probed , larger distance requires the inclusion of quark loops @xcite , whereas eq . ( [ eq : s2 - 12 ] ) below is obtained with the full quenched approximation ) , @xmath111 a smearing of the initial @xmath72 in the form prescribed by eq . ( [ eq : s2 - 10 ] ) is mandatory @xcite , with that small deformation parameter @xmath79 . fascinating relations ensue thereof , between lvy flight modes of propagation for confined quarks ( at @xmath112 ) , chiral symmetry breaking , confinement and an unexpected non - commutative geometrical aspect of the scattering transverse planes ( _ de moyal _ planes @xcite ) . these relations shall be discussed elsewhere . as a result of effective locality , one finds a 2-body scattering amplitude proportional to @xmath113 \ , \det[gf\cdot\chi(w)]^{-{1\over 2}}\ e^{{i\over 4}\chi^2(w)+ig\varphi(b)\ \omega_1^a[f \cdot \chi(w)]^{-1}\vert^{ab}_{30}\ \omega^b_{2 } } } , \end{aligned}\ ] ] where @xmath114 is the shorthand for @xmath115 , and the dot , ` @xmath116 ' , denotes an euclidean scalar product in a relevant representation space ; for example , the product @xmath117^{bc}=\sum_af^{abc}\chi^a$ ] . the @xmath118s are the @xmath114 hermitian traceless generators of the @xmath30-lie - algebra , taken in the fundamental @xmath119-dimensional representation of @xmath30 . in the adjoint representation to be used shortly , the @xmath118s are given by the structure constants of the @xmath30-lie - algebra : @xmath120 . the form of eq . ( [ eq:1 ] ) is non - trivial and is justified in ref . the full 2-body amplitude requires several prefactors that are not written in eq . ( [ eq:1 ] ) since they have no relevance to the consideration to follow . in the second line of eq . ( [ eq:1 ] ) , the last exponential argument is nothing else than eq . ( [ eq : s2 - 6 ] ) evaluated with the help of identity eq . ( [ eq : s2 - 11 ] ) . in eikonal approximation , one gets in effect @xmath121 \ , [ f \cdot \chi(w)]^{-1 } \vert_{ab}^{\mu\nu}\ \omega_1^a(0)\omega^b_{2}(0)\right\}},\end{aligned}\ ] ] where @xmath122 and @xmath123 , and where the original @xmath0 has been turned into the distribution @xmath80 of eq . ( [ eq : s2 - 10 ] ) . in the @xmath62 of the two scattering quarks of momenta @xmath124 and @xmath125 , eq . ( [ eq:2 ] ) reduces to the last argument in the exponential at the end of the expression in eq . ( [ eq:1 ] ) . it is worth stressing how the peculiar form of eq . ( [ eq:1 ] ) originates from the property of effective locality . because of eq . ( [ eq : s2 - 7 ] ) , one has @xmath126 with @xmath127 as the only spacetime point where the interaction takes place , and the contributions attached to all the other points are eliminated . the net result is that there is no spacetime integration in the argument of the exponential at the end of eq . ( [ eq:1 ] ) ; and accordingly , no spacetime integration either for the first term . this can be understood in a straight forward way by recalling the standard procedure of a generating functional construction @xcite , where the spacetime manifold is broken up into an infinite series of infinitesimal cells . since the interaction , of the contact type , takes place at a unique point , @xmath128 , the contributions of all of the other infinitesimal cells , centered at different points , just contribute a product of infinite multiplicative factors of @xmath129 , thanks to the normalization of each cell . the final result for the amplitude will therefore depend on a length scale that is not required by the interaction term itself , since it is dimensionless , but rather by the action term @xmath130}$ ] , which , because of effective locality , is now to be necessarily understood as @xmath131}$ ] . a non - trivial result will depend upon forming a connection between the infinitesimal mathematical @xmath132 and a physically meaningful @xmath133 related to the scattering process under consideration @xcite . for such a 2-body process , the standard laws of quantum mechanics lead to a cell of time extension of order @xmath134 , while that corresponding to a high - energy @xmath62 longitudinal coordinate should be @xmath135 ; eventually , transverse coordinates should extend over a distance of order @xmath136 each , _ c.f . _ ( [ eq : s2 - 10 ] ) , and the overall interaction cell volume is therefore on the order of @xmath137 . reciprocally , the dimensionless interaction term is now re - written as @xmath138^{-1}(w)\ \omega_2\longrightarrow ig\ [ \varphi(b ) \delta_{\mathrm{ph}}^{2 } ] \ \omega_1 \ [ f \cdot \bar{\chi}(w)]^{-1}\ \omega_2,\ ] ] where the field variables @xmath139 are dimensionless , as is the new combination @xmath140 where @xmath141 . hereafter , the bar over @xmath15-fields will be omitted ; and instead of eq . ( [ eq : s2 - 10 ] ) , the right hand side of eq . ( [ newphi ] ) will be understood as @xmath80 , which is now dimensionless . as it stands , however , the expression eq . ( [ eq:2 ] ) confronts us with a puzzling issue , that is inverting the quantity @xmath117 $ ] . taken as some relevant matrix , in effect , it is immediate to realize that zero belongs to the spectrum of @xmath117 $ ] , so that there seems to be no way out of this dead end . however , two most favorable circumstances allow one not only to cope with that issue , but also to reach the level of actual , sensible calculations . the first one of these two circumstances is as follows . in order to : 1 . invert , at least formally ( for the time being ) the product @xmath117 $ ] , 2 . and take into account the contributions of all of the @xmath142- @xmath143-fields entering @xmath117 $ ] at the spacetime dimension of @xmath144 ( contrast to ref . @xcite , where , for simplicity , only @xmath114 components @xmath145 were considered ) , 3 . as well as the full symmetry of the product @xmath146 $ ] under the combined exchanges , @xmath147 and @xmath148 @xcite , it is appropriate to think of @xmath149 as a tensorial product @xmath150 of operators acting on the product space @xmath151 . the first space , @xmath152 , is the _ vectorial _ minkowski space ( as opposed to the _ affine _ minkowski space , the space of event - points , to which the former space is tangent ) , and the second , @xmath153 , a @xmath30-adjoint representation space with dimension @xmath154 . their product is a @xmath155-dimensional vectorial space that can be endowed with the product basis @xmath156 constructed out of the bases of @xmath152 and @xmath153 spaces , on which act the 2-form @xmath157 and the endomorphism @xmath118 , respectively . in this product space , at real - valued halpern - fields , and with @xmath158 , the operator @xmath159 is represented by a @xmath160 matrix equal to @xmath161 times a real symmetric and traceless matrix @xmath162 , @xmath163 the operator @xmath164 acts upon the @xmath165-component vectors @xmath166 , @xmath167 of the product space , and in the basis @xmath156 , these vectors have components as @xmath168 where , for short , the dependence on @xmath169 , @xmath48 , are hereafter omitted , and where the superscript @xmath170 stands for ` transpose ' . note that unlike the @xmath171s , the @xmath172s are not dimensionless objects since , as quoted before , @xmath173 and @xmath174 . note also that , out of respective scalar products of the 2 spaces , a scalar product on the product space @xmath151 can be defined in a canonical way , and inherits the pseudo - euclidean character of the minkowskian @xmath152 space ( this is exploited in the calculations of section [ sec:4 ] ) . then , there exists an orthogonal matrix @xmath175 that takes @xmath162 to its diagonal form @xmath176 where @xmath177 is the transpose of @xmath175 , and likewise , @xmath178 to its diagonal form also @xmath179 the spectrum of @xmath162 may contain the eigenvalue @xmath180 , so that , for the time being , eq . ( [ eq:6 ] ) can only be given a formal meaning . re - writing eq . ( [ eq:2 ] ) as @xmath181_{ij}\ v_2^j\right),\ ] ] and redefining the @xmath172-fields , @xmath182 one gets for eq . ( [ eq:7 ] ) the expression @xmath183 at eikonal approximation , fixed @xmath184 , one has @xmath185 ; and likewise , a trivial jacobian for the transformation eq . ( [ eq:8 ] ) . by using the reformulation eq . ( [ eq:4 ] ) , the integration with measure @xmath186,\ ] ] which appears in eq . ( [ eq:1 ] ) , gets translated into an integration over the algebra of real symmetric @xmath187 traceless matrices with measure @xcite @xmath188 the @xmath165 eigenvalues of @xmath162 are real , and @xmath189 is a(n unspecified ) function of the @xmath190 extra parameters which complete the parametrization of a given real symmetric matrix @xmath162 , and fully determine the orthogonal matrix @xmath191 . mathematically , this passage from an infinite dimensional functional space of integration , to a finite dimensional matrix space is non - trivial ; it is a direct consequence of the effective locality and can be proven by relying on the _ measure image theorem _ @xcite . the dependences on the extra parameters @xmath192 of the orthogonal matrices @xmath193 will play a role in the calculations of the next section [ sec:4 ] . the second line of eq . ( [ eq:1 ] ) can accordingly be re - written as @xmath194\ , [ \prod_{1\leq i < j\leq n } |\xi_i-\xi_j|\ ] \ e^{-{i\over 8n_c}\ , \xi_i^2}\ \ { \frac{1}{\sqrt{\xi_i}}}\,{e^{-i{g\frac{\varphi(b)}{2ep}}\ { \frac{{v'_1}^i{v'_2}^i}{\xi_i}}}},\end{aligned}\ ] ] where , for the first exponential term , the relation @xmath195 has been used . keeping in line with ref . @xcite , the normalization constant @xmath196 may be so defined as to normalize eq . ( [ eq:11 ] ) to unity when the whole set of functions @xmath197 is replaced by the set of constant unit functions , @xmath198 ( by the way , this turns out to be the normalization used in eq . ( 40 ) of ref . @xcite , and the superscript @xmath199 refers to a @xmath36-point function ) . in order to allow for a concise and closed form expression , one can take advantage of the method used in ref . @xcite ; a simple re - scaling of the eigenvalues is necessary , that is , @xmath200 , taking eq . ( [ eq:11 ] ) to the form relevant to the _ gaussian orthogonal ensemble _ @xcite , @xmath201 in which appears a _ vandermonde determinant _ of @xmath202 with @xmath203 , the absolute value of a sum of @xmath204 monomials in @xmath205 of overall degree @xmath190 . the expression @xmath206 is normalized so that its integration over the full spectrum of eigenvalues @xmath205 , @xmath207 yields unity , and one gets @xcite @xmath208 allowing one to identify the normalization constant of eq . ( [ eq:12 ] ) as @xmath209 things may be expressed in terms of harmonic oscillator wave functions , @xmath210 where @xmath211 is the hermite polynomial of order @xmath2 . their normalization reads as @xmath212 however , contrary to the standard expression of ref . @xcite , eq . ( [ eq:15 ] ) displays a factor of @xmath161 in its exponential term , which , in ref . ( [ eq : normal ] ) , translates into a @xmath213-dependence of the harmonic oscillator wave functions , rather than just @xmath205 . relying on cauchy s theorem , though , it is proven in the appendix that the integration formulae of eqs . ( [ eq:15 ] ) and ( [ eq : normal ] ) defined without @xmath214 , can be analytically continued from real valued @xmath215 to the complex range of @xmath216 . in particular , using this analytical continuation it is possible to check that the harmonic oscillator wave function normalization just undergoes a simple re - scaling of @xmath217 then , proceeding along the standard steps of ref . @xcite , one can check that eq . ( [ eq:12 ] ) can be simply written as @xmath218 the absolute values in eqs . ( [ eq:13 ] ) and ( [ eq:20 ] ) are bothering . at a formal level at least , eq . ( [ eq:20 ] ) can be expressed in terms of normalized harmonic oscillator wave functions under the form @xmath219 , \\ \nonumber & & \quad \quad \quad k , l = 1, .. ,n,\end{aligned}\ ] ] where @xmath220 is the distribution _ sign of x _ , and @xmath221 are those of eq . ( [ eq:11 ] ) with @xmath222 where the rescaling from eqs . ( [ eq:11 ] ) to ( [ eq:12 ] ) has been taken into account . it is remarkable that exactly the same equations as eqs.([eq:11 ] ) , ( [ eq:20 ] ) and ( [ eq:22 ] ) express the closed form of any @xmath223-point fermionic amplitudes ( not yet integrated over the extra auxiliary variables @xmath224 and @xmath171 of eq . ( [ eq:1 ] ) ; see next section [ sec:4 ] ) . in the general case of a fermionic @xmath223-point function , in effect , all of the previous steps can be followed through , and the only change turns out to be the redefinition of the constant @xmath225 . within obvious notations ( @xmath226 , and @xmath227 , in each of the @xmath49 by @xmath49-@xmath228 ) , @xmath229 where , again , the restriction @xmath230 in the summation discards the self - energy effect that will be dealt with in another article . in ref . @xcite , though , a glance at this issue is obtained : amazingly , the same ultraviolet renormalization structure as in the usual qcd perturbative regime comes about , and it is most interesting to note that this point is in line with the explanation proposed in ref . @xcite , based on an analysis of the ads / qcd correspondence . to summarize , at quenched and eikonal approximations , and up to renormalization , a @xmath223-point fermionic amplitude can concisely be written under a form proportional to @xmath231 unfortunately , the _ method of integration over alternate variables _ of ref . @xcite , aimed at circumventing the absolute values of eqs . ( [ eq:20 ] ) and ( [ eq:26 ] ) , is able to yield elegant closed form expressions such as ( [ eq:22 ] ) , but is of no real help to calculational purposes . the general and closed form expressions given above may be most appropriate to numerical simulations . in this section though , for the sake of illustration and a further speculation on non - perturbative qcd , it is enough to get back to an expression close to eq . ( [ eq:20 ] ) , @xmath233 where , anticipating somewhat on a result that does not depend on them , the prescriptions of absolute values have been given up . likewise , in this section , calculations are restricted to the @xmath36-point fermionic function , that is @xmath234 , and the superscript @xmath199 will hereafter be omitted for short . the vandermonde determinant being a polynomial in eigenvalues , one has to cope with generic integrations of type @xmath235 where @xmath236 . defining @xmath237 for @xmath238 and @xmath239 , one has that eq . ( [ eq:28 ] ) evaluates to @xcite @xmath240 where @xmath241 is a _ meijer s special function _ , and the second of the two most favorable circumstances alluded to , after eq.([newphi ] ) , is the following : whereas eq . ( [ eq:29 ] ) holds true at values of @xmath242 and @xmath1 such as specified above , it turns out that @xmath241 is analytic in @xmath243 , and that the parameters appearing after @xmath243 can also be continued to complex values @xcite . on the basis of these analyticity properties , it is therefore possible to give the generic integrations of eq . ( [ eq:27 ] ) the meaning of @xmath244\over 2{\sqrt{\pi}}}g^{30}_{03}\left({i(b^{(i)})^2\over 8 } \biggr| { 2q_i+1\over 4 } , { 1\over 2},0\right),\end{aligned}\ ] ] where the equality , demonstrated in the appendix , is a consequence of _ cauchy s theorem _ , and where the constants @xmath245 , @xmath246 , are defined in eq . ( [ eq:25 ] ) . now , each of the @xmath245 is on the order of @xmath247 , and the small @xmath247 limit can therefore be investigated on the basis of the small @xmath248 expansion , @xmath249 , @xmath250,\end{aligned}\ ] ] where the prime on top of the product indicates that the pole value @xmath251 is omitted , and likewise the asterisk in the @xmath252 generalized hypergeometric series , indicates also that the value @xmath251 is omitted : @xmath253.\end{aligned}\ ] ] for all @xmath254 , one has , at @xmath249 , @xmath255=1+{\cal{o}}(z).\end{aligned}\ ] ] for the situation considered in eq . ( [ eq:29 ] ) , the parameter sets @xmath256 always read as @xmath257 , so that in the small @xmath248-limit , the meijer s special function ( [ eq:29 ] ) yields , for example , @xmath258^{1\over 2 } \right . \\ & & \left . + { \gamma(-{(2q_i+1)/ 4})\gamma({-(2q_i-3)/ 4})\over \gamma((2q_i+5)/4)}[i{(b^{(i)})^2\over 8}]^{2q_i+1\over 4 } \right\ } \ , \left(1+{\cal{o}}((b^{(i)})^2)\right).\end{aligned}\ ] ] one can check that the first term in the right hand side is identical to @xmath259 as it should . this result is interesting in view of the crude approximations we had been relying on in ref . @xcite and @xcite . in particular , the term of @xmath260 of eq . ( [ eq:28 ] ) was expanded at leading linear order in @xmath261 . one can now see that , were it not for the terms at @xmath262 , this expansion would yield the correct leading order estimate of eq . ( [ eq:12 ] ) . one has now to deal with the remaining integrations , as they show up in eqs . ( [ eq:1 ] ) and ( [ eq:26 ] ) , that is , firstly with the subsequent integrations of ( dropping the primes of the @xmath263 ) @xmath264 ^ 2\over 128n_c\ , e^2p^2 } \biggr| { 2q_i+1\over 4 } , { 1\over 2},0\right).\end{aligned}\ ] ] eq . ( [ eq:34 ] ) is obtained by promoting the original @xmath114-component vectors @xmath265 to @xmath165-component ones , defining @xmath266 , for @xmath48 . then , taking eq . ( [ eq:5 ] ) into account , the first line of eq . ( [ eq:26 ] ) can be written as @xmath267 where @xmath175 is the orthogonal matrix of eq . ( [ o ] ) ; eventually , in eq . ( [ eq:34 ] ) , the _ prime_-superscripts are dropped , and likewise , the little hats . rather than integrating a meijer s function expansion that does nt terminate , one can take advantage of _ formula _ 20.5.(8 ) of ref . @xcite , so as to obtain for eq . ( [ eq:34 ] ) the exact result of @xmath268 where the @xmath62-relation @xmath269 is used . then , in view of eq . ( [ eq:1 ] ) , two extra integrations remain to be carried out : defining @xmath270 the @xmath165-vector of matrices @xmath271 , where as in eqs . ( [ eq:1 ] ) and ( [ eq:26 ] ) , @xmath272 stands for the @xmath114-vector of generators @xmath118 , @xmath273 in the adjoint representation , one can write for ( [ eq:26 ] ) without absolute values @xmath274 where @xmath175 is the orthogonal matrix introduced in eq . ( [ o ] ) . for the sake of further integrations , eq . ( [ eq:36 ] ) can be most conveniently re - written as @xmath275 where use has been made of the inversion formula 5.3.1(9 ) of ref . @xcite , @xmath276 by symmetry , eq . ( [ eq:37 ] ) reads indeed as @xmath277\over \alpha_1^i}\int_0^{+\infty } { \rm{d}}\alpha_2^i\ { \sin[\alpha_2^i({\cal{ot}})_i]\over \alpha_2^i } \nonumber\\ & & \quad \times \ , g^{23}_{34}\left ( \left.{8n_c\over -i}\left ( { \alpha_1^i\alpha_2^i \over g\varphi(b ) } \right)^2{e^2p^2\over m^4 } \right| \begin{array}{cccc } \frac{3 - 2 q_i}{4 } , & \frac{1}{2 } , & 1 , & \\ \frac{1}{2 } , & \frac{1}{2 } , & 1 , & 1 \end{array } \right).\end{aligned}\ ] ] considering the product of the @xmath165- integrations , and the trace to be taken over internal color degrees , the first line of eq . ( [ eq:39 ] ) leads to the expression @xmath278^{k'_i+k_i+1}\,,\ ] ] which results from the expansion of the 2 _ sine_-functions of eq . ( [ eq:39 ] ) , and where the normalization constant in eq . ( [ eq:10 ] ) reads @xmath279 by construction , the @xmath280 are such that , @xmath281 , @xmath282 , such that @xmath283 . note that were it not for this average over the extra parameters @xmath284 introduced in eq . ( [ eq:10 ] ) , which does not factor out because of the @xmath284-dependence of the orthogonal matrices @xmath193 , then , instead of eq . ( [ p ] ) , one got an identically null result under the form of @xmath285 where these equalities come from the fact that in the @xmath165-plet of matrices @xmath270 , each of the @xmath114- @xmath118 appears @xmath144 times by construction , from the idempotency of the @xmath286 , and to the fact that they commute , @xmath287=0 $ ] . a complete , exact result for eq . ( [ p ] ) is not straight forward and will be presented elsewhere @xcite . it can be checked that eq . ( [ p ] ) comes out non - vanishing and , in agreement with lattice accurate calculations @xcite , not exactly proportional to the first casimir operator eigenvalue @xmath288 , over any given gauge group representation , @xmath289 . for the time being , we will here content ourselves with the leading contribution to eq . ( [ p ] ) , so that eqs . ( [ eq:1 ] ) and ( [ eq:26 ] ) ( without absolute values ) read eventually @xmath290 where @xmath291 is the standard quadratic casimir operator eigenvalue on the gauge group fundamental representation . recalling the shorthand @xmath292 , and getting back to eq . ( [ 56 ] ) , now integrated over the @xmath293 , one obtains therefore a @xmath36-point fermionic amplitude proportional to @xmath294\int_{0}^{+\infty } { \rm{d}}\alpha_1^j\ { \sin{{}}\alpha_1^j\over \alpha_1^j } \nonumber\\ & & \quad \times \int_{0}^{+\infty } { \rm{d}}\alpha_2^j\ { \sin{{}}\alpha_2^j\over \alpha_2^j } \ , g^{23}_{34}\left ( \left . { 8n_c\over -i}\left ( { \alpha_1^j\alpha_2^j \over g\varphi(b ) } \right)^2{e^2p^2\over m^4 } \right| \begin{array}{cccc } \frac{3 - 2 q_j}{4 } , & \frac{1}{2 } , & 1 , & \\ \frac{1}{2 } , & \frac{1}{2 } , & 1 , & 1 \end{array } \right).\end{aligned}\ ] ] in the second line , the sum runs over the monomials composing the vandermonde determinant , and the @xmath295 is here to mean that each monomial appears with a plus or minus sign . already at @xmath296 , one gets @xmath297 monomials of overall degree @xmath298 in the the eigenvalues @xmath205 . but at @xmath299 one gets @xmath300 monomials of overall degree @xmath301 that require several pages to be written down . one need not specify the number of monomials entailed in the vandermonde determinant at @xmath302 , relevant to the case @xmath303 , not to speak of @xmath304 , for which @xmath305 . before proceeding to the full integration of eq . ( [ 75 ] ) , it is worth observing that the last line , @xmath306 was the essential part of our first and crudely approximated estimate in ref . @xcite , and the result was @xmath307 based on eq . ( [ eq:43 ] ) , we had found for @xmath308 that at large values of @xmath247 , corresponding to small impact parameters , @xmath309 , obviously damped in comparison to the large impact parameter limit of @xmath310 , where @xmath311 . the interpretation proposed in ref . @xcite and based on eq . ( [ eq:43 ] ) , preserves its essential features within the present paper s more accurate calculation where one gets a behavior of @xmath312 at large impact parameter , and a behavior of @xmath313 at small impact parameter . focusing on a given integration , that of index @xmath314 for example , one gets for the integrals of eq . ( [ 75 ] ) , @xmath315}}{x } } \int_{0}^{\infty } { \rm{d}}y\ \frac{\sin{\left[y\sqrt{{\frac{g\varphi{\sqrt{-i}}}{\sqrt{8n_c } } } \frac{m^2}{ep}}\right]}}{y } \ , g^{23}_{34}\left ( x^2 y ^2 \biggr| \begin{array}{c } a_r(j ) \\ b_s \end{array } \right)\end{aligned}\ ] ] with @xmath316 and @xmath317 the sequences of numbers such as displayed in eq . ( [ 75 ] ) . these two sequences of numbers are such that the reduction formula 5.3.1.(7 ) of ref . @xcite , can be used to simplify somewhat the meijer s special function entering eq . ( [ eq:44 ] ) and write @xmath318 this allows one to re - write eq . ( [ eq:44 ] ) as @xmath319}\int_{0}^{\infty } { \rm{d}}y\ { \sin[y\sqrt{{{g\varphi\sqrt{-i}\over{\sqrt{8n_c}}}}{m^2\over ep}}]}\,g^{22}_{23 } \left ( x^2 y^2 \biggr\vert \begin{array}{ccc } 0 , & \frac{1 - 2 q_j}{4 } & \\ 0 , & 0 , & \frac{1}{2 } \end{array } \right),\end{aligned}\ ] ] where the following identity , 5.3.1.(8 ) of ref . @xcite , has been used also , @xmath320 in the end , parameters are such , in the _ sine _ function as well as in the meijer special function , the sequences of numbers @xmath321 and @xmath322 , that @xmath323 @xcite , can be used twice so as to yield for eq . ( [ eq:46 ] ) the result @xmath324 where the inversion formula 5.3.1(9 ) of ref . @xcite , that is , eq . ( [ eq:38 ] ) , has been used again in order to express the result in a form suited to a @xmath325 limit expansion of the net amplitude . gathering all factors , the amplitude under consideration , eq . ( [ eq:26 ] ) without absolute values , can eventually be displayed under the form of @xmath326 \\ & & \left({{\sqrt{8in_c}}\over g\varphi(b)}\right)\times\ g^{42}_{36}\left ( \left ( { g\varphi(b)\over { \sqrt{8in_c } } } { m^2\over{\sqrt{{\widehat{s}}({\widehat{s}}-4m^2)}}}\right)^{2 } \biggr| \begin{array}{cccccc } 1 , & 1 , & & \frac{1}{2 } & & \\ 1 , & 1 , & 1 , & \frac{2 q_j + 3}{4 } , & \frac{1}{2 } , & \frac{1}{2 } \end{array } \right),\end{aligned}\ ] ] which relies on the normalization of eq . ( [ eq:11 ] ) ( in particular , this normalization has the effect of eliminating the instanton - like contributions of @xmath327 that appear in ( [ 75 ] ) ) . a notation typical of perturbative drell - yan scattering processes has been introduced , @xmath328 , the quark s @xmath36-momenta , the @xmath329 , being related to their respective hadronic carriers in the usual _ partonic _ way , @xmath330 , with @xmath331 $ ] . beyond the notation itself , this is certainly far from an innocuous fact , as an explicit link shows up between a perturbative partonic content , essentially controlled by @xmath332 and @xmath333 , and a hadronic non - perturbative component controlled by @xmath247 , eq . ( [ newphi ] ) , and this , within one and the same overall amplitude expression : this is surely an important realization of the effective locality analysis @xcite . it is worth noticing that the small @xmath247-limit expansion of the @xmath334-meijer s functions is the more relevant as their argument is not @xmath247 itself , but rather the combination @xmath335 , that is a small number in the eikonal and light quark mass limit . at leading order , corresponding to the value @xmath336 of the @xmath337- meijer s function parameters , the two last terms of eq . ( [ normaleq:49 ] ) combine into a constant , @xmath247-independent contribution , exactly as it should ( see eq . ( [ limit ] ) below ) . worth observing also is the interplay between the @xmath49-body scattering _ probing energy _ @xmath332 , and the non - perturbative physics controlled by the function @xmath247 ( it is to be recalled that the large coupling limit is considered here , @xmath338 ) . the meijer s special functions argument , in effect , reads basically as @xmath339 so that the relative magnitude of @xmath332 with respect to the scale @xmath74 , decides of either the aperture ( at @xmath340 ) , or of the closure ( at @xmath341 ) , of the non - perturbative physics channel that is related to the @xmath74-dependences of eq . ( [ normaleq:49 ] ) . in the latter case in effect , the meijer s special functions behaviour is such that one has @xmath342^{2 } \biggr| \begin{array}{cccccc } 1 , & 1 , & & \frac{1}{2 } & & \\ 1 , & 1 , & 1 , & \frac{2 q_j + 3}{4 } , & \frac{1}{2 } , & \frac{1}{2 } \end{array } \right ) \nonumber \\ & = & { m^2/ { \widehat{s}}\over { \sqrt{1-{4m^2/ { \widehat{s}}}}}}+{\cal{o}}(\sqrt{g\varphi}),\end{aligned}\ ] ] whose leading order term is free of interaction dependences , exactly as it should , while leaving a residual non - perturbative and sub - leading effect on the order of @xmath343 , which , at high enough probing energy @xmath332 , can be quite small . clearly , as is well known from nuclear physics , non - perturbative aspects are totally washed out beyond a certain energy scale ( of @xmath344mev @xcite ) . also , it is interesting to observe that because the argument eq . ( [ argument ] ) can become small in various ways , one has an indication that qcd should manifest _ asymptotic freedom _ beyond the sole criterion of perturbation theory , as it has been advocated recently on the basis of dyson - schwinger equations analyses @xcite . eventually , whenever eq . ( [ argument ] ) is not that small valued a combination , then an amplitude like eq . ( [ normaleq:49 ] ) describes a non - trivial blending of non - perturbative aspects together with perturbative degrees of freedom . these remarks easily extend to the cases of @xmath223-point fermionic green s functions , describing the @xmath85-bodies scattering processes of nuclei collisions , for example . in this situation , the overall energy of @xmath345 gets distributed among all of the possible @xmath346- @xmath49-body scattering sub - processes , and a summation over the sub - energy distributions complying with the constraint of global energy conservation must be performed . that is , the concise form obtained for the fermionic correlation functions can certainly be exploited in various situations , limits , and along various numerical and/or semi - analytical ways . similarly , when the full flavor and mass and structure function richness of qcd is restored , these generic forms of amplitudes could be used and tested at a phenomenological level to describe hadronic collisions at _ intermediate _ energies , for example . this however , will require that the absolute value prescriptions of eq . ( [ eq:26 ] ) be circumvented in a way or other . in the present article one is interested in understanding quark s binding potential such as derived in ref . @xcite , based on the relationship between an eikonal amplitude and the potential able to generate it . as advertised in the main text of the present article , that result , given in eq . ( [ eq : s2 - 12 ] ) , displays the occurrence of an intriguing @xmath79-parameter whose non - zero value is essential for obtaining a non - trivial confining potential . introducing a convenient representation for the heavyside step function @xcite , we have been able to check that the absolute value prescriptions bring no change to the following analysis , and this is why they have been dropped in the current section . as the derivation of eq . ( [ eq : s2 - 12 ] ) relied upon frequently crude approximations , a natural issue is to know wether that @xmath79-parameter is an _ avatar _ of the approximate calculational scheme used in ref . @xcite , or if it is not . it is this question that can be addressed now that , with respect to our first crude estimations @xcite , the calculational accuracy has been substantially improved . following the standard procedure that relates a 2-body eikonal amplitude to a potential acting between them @xcite , it turns out that from eq . ( [ normaleq:49 ] ) , the next - to - leading order of the @xmath325 limit expansion is required . now , as made obvious by meijer s functions small @xmath347 expansions , eqs . ( [ eq:31 ] ) and ( [ eq:33 ] ) , the next - to - leading contributions are obtained from the parameters @xmath348 whenever @xmath349 , and instead of order @xmath347 , as derived previously @xcite , are indeed on the order of @xmath350 . however , as argued in ref . @xcite , it is the logarithm of that next - to - leading order contribution which is associated to the calculation of @xmath351 . in view of eq . ( [ eq : s2 - 10 ] ) or ( [ newphi ] ) though , one has @xmath352 ( see also ref . @xcite , where a form of @xmath353 is proposed ) . therefore , not solely that power of @xmath336 , instead of @xmath129 , but any other possible next - to - leading power of @xmath247 would never bring the slightest change to the conclusion of ref . @xcite : that is , a non - trivial quark confining potential requires a non - vanishing @xmath79-parameter value . in eikonal and quenched approximations at least , this conclusion is now reached on a substantially improved basis , and it is made accessible thanks to the non - perturbative effective locality property of qcd . if effective locality is a genuine property of the non - perturbative fermionc sector of qcd , then it is worth exploring its consequences , that is , it is worth looking at qcd from the point of view of effective locality . it is the task that has been undertaken in the present article , where we have tried to make more rigorous the arguments of two previous analyses @xcite , while putting forth new interesting aspects . in the first place , evaluating qcd fermionic amplitudes , it is argued that an awkward term of @xmath354 , appearing as an overall multiplicative factor in an exponential argument , can not be avoided and is in no way related to any approximation _ artifact_. inspection rather suggests that the occurrence of this factor should be understood as a compelling indication that the ordinary notion of an elementary particle can not be maintained in the non - perturbative realm of qcd . this term of @xmath354 was accordingly corrected , and , in a first attempt , turned into the smoother form of a normalized gaussian distribution of transverse distance of the scattering quarks , _ i.e. _ , their collision s impact parameter . we could note that in an analysis of `` relativistic quark models in the quasi - potential approach '' , a similar form had been proposed @xcite . against all odds , though , the quark confining potential associated to the scattering eikonal amplitude is then found to be identically zero @xcite . subsequently , an elementary _ deformation _ of the random transverse quark gaussian distribution has been proposed in order to prevent such an odd result from occurring @xcite . fortunately enough , at @xmath112 , that deformation ( which , more precisely , does nt bear on the lvy flight distribution itself , but on its _ characteristic function _ ) , falls into a stable and canonical generalization of random variable gaussian distributions that are compatible with a generalized _ central limit theorem _ of statistical physics @xcite . in a second place , we have made use of the most efficient _ random matrix _ calculus so as to obtain the generic structure of fermionic qcd amplitudes , at least at quenched and eikonal approximations . illustrated on the simpler case of a @xmath36-point green s function , the results can be easily extended to any @xmath44-point function , and could therefore open the route to phenomenological applications . an amazing and non - trivial aspect of the resulting amplitude is that within one and the same expression , it is able to encode some hadronic non - perturbative aspect together with the perturbative partonic content of the hadronic matter . it is worth noting that such a link has been advocated recently ( and the way by which this link must come about ) within the ads / qcd approach to non - perturbative qcd @xcite . moreover , the amplitude s generic forms so obtained comply with a conjecture concerning the form quantum field theory s generating functionals must display in full generality @xcite . in effect , while the latter must be expressed in terms of _ fox special functions _ , our amplitudes turn out to be written in terms of _ meijer s special functions _ that are but a particular instance of the former . however , the conclusions are here obtained on the bases of eikonal and quenched approximations to the full strong coupling fermionic scattering amplitudes . for example , this means that so long as quenched ( at least ) is not relaxed , the relevance of that @xmath79-parameter ( describing a lvy - flight _ markovian _ diffusion mode of propagation for confined quarks ) remains somewhat uncertain ; and this , in spite of its fascinating relation to confinement , to a possible transverse non - commutative geometrical aspect of non - perturbative qcd , to confined quarks lvy - flight modes of propagation , to ( spontaneous ) chiral symmetry breaking and another related aspect of non - perturbative qcd not evoked in this article . since , on the other hand , the results of ref . @xcite are not restricted to any sort of approximation , one may hope that , at least at a qualitative level , our present conclusions could be somewhat preserved while relaxing the quenched approximation . still , that remains to be checked . it is true also that the traceless character of the random matrices @xmath162 has not been implemented in our analysis : so far as could be seen , this point does nt seem to modify in a drastic way the general features put forth in this article ; still , it needs to be evaluated in a cogent way . clearly , more work is necessary to explore on firm grounds the consequences of such an intriguing , non - perturbative property as effective locality . in these perspectives , quark propagator s full calculation and self energy effects should bring some new understanding , and shall be the matter of forthcoming analyses . starting from @xmath355 such as given in eq . ( [ eq:28 ] ) and evaluated thanks to ref . @xcite , at @xmath238 and @xmath239 , @xmath356 ( 10,10 ) ( 1,2)(9,2 ) ( 2,1)(2,7 ) ( 2,2)(5,2 ) ( 5,2)(8,2 ) ( 2,2)6022.5 ( 2,2)622.545 ( 6.242641,6.242641)(4.12132,4.12132 ) ( 4.12132,4.12132)(2,2 ) ( 2,2)1.5045 ( 3.5,2.65)@xmath363 ( 1.75,1.75)@xmath63 ( 9,1.75)@xmath364 ( 1.5,7)@xmath365 ( 8,1.75)@xmath366 ( 8.0,4.25 ) @xmath361 now , the analyticity of meijer s special functions in their arguments , as well as the possibility of continuing their parameters to arbitrary complex values @xcite allows one to define the required integral as @xmath367 v. a. matveev , v. i. savrin , a. n. sissakian and a. n. tavkhelidze , theor . math . phys . * 132 * , ( 2002 ) 1119 . h. m. fried , _ basics of functional methods and eikonal models _ , ( editions frontires , gif - sur - yvette cedex , france 1990 ) .

we introduce a multivariate hawkes process that accounts for the dynamics of market prices through the impact of market order arrivals at microstructural level . our model is a point process mainly characterized by 4 kernels associated with respectively the trade arrival self - excitation , the price changes mean reversion the impact of trade arrivals on price variations and the feedback of price changes on trading activity . it allows one to account for both stylized facts of market prices microstructure ( including random time arrival of price moves , discrete price grid , high frequency mean reversion , correlation functions behavior at various time scales ) and the stylized facts of market impact ( mainly the concave - square - root - like / relaxation characteristic shape of the market impact of a meta - order ) . moreover , it allows one to estimate the entire market impact profile from anonymous market data . we show that these kernels can be estimated from the empirical conditional mean intensities . we provide numerical examples , application to real data and comparisons to former approaches .

1301.1135

market impact modeling ( i.e , the influence of market orders on forthcoming prices ) is a longstanding problem in market microstructure literature and is obviously of great interest for practitioners ( see e.g. , @xcite for a recent review ) . even if there are various ways to define and estimate the impact associated with an order or a meta - order , a large number of empirical results have been obtained recently . the theory of market price formation and the relationship between the order flow and price changes has made significant progress during the last decade @xcite . many empirical studies have provided evidence that the price impact has , to many respects , some universal properties and is the main source of price variations . this corroborates the picture of an `` endogenous '' nature of price fluctuations that contrasts with the classical scenario according to which an `` exogenous '' flow of information leads the prices towards a `` true '' fondamental value @xcite . we will not review in details all these results but simply recall the point of view of bouchaud et al @xcite . these authors propose a model of price fluctuations by generalizing kyle s pioneering approach @xcite according to which the price is written ( up to a noise term ) as the result of the impact of all trades . if @xmath0 stands for the trading time , bouchaud et al . model is written as follows @xcite : @xmath1 where @xmath2 is a white noise while @xmath3 with @xmath4 ( resp . @xmath5 ) if a trade occurs at the best ask ( resp . at best bid ) and @xmath6 describes the volume dependence of a single trade impact . the function @xmath7 accounts for the temporal dependence of a market order impact . even if models like represent a real breakthrough in the understanding of price dynamics they have many drawbacks . first , the nature and the status of the noise @xmath2 is not well defined . more importantly , these models involve discrete events ( through the trading or event time ) only defined at a microstructural level though they intend to represent some coarse version of prices ( indeed , the price @xmath8 in the previous equation can take arbitrary continuous values , moreover , calibrating its volatility at any scale involves some additional parameter , ... ) . moreover , these models can not account for real time ( i.e `` physical time '' ) dynamics or real time aggregation properties of price fluctuations . in that respect , they are not that easy to be used in real high frequency applications such as optimized execution . to make it short , though being defined at finest time scales , they can not account for the main microstructure properties of price variations related to their discrete nature : prices live on tick grids and jump at discrete random times . our aim in this paper is mainly to define _ continuous time _ version of the market impact price model discussed previously . for that purpose , point processes @xcite provide a natural framework . let us not that point processes have been involved in many studies in high frequency finance from the famous ` zero - intelligence' order - book models @xcite to models for trade @xcite or book events @xcite irregular arrivals . in a recent series of papers @xcite , we have shown that self - excited point ( hawkes ) processes can be pertinent to model the microstructure of the price and in particular to reproduce the shape of the signature plot and the epps effect . our goal is to extend this framework in order to account for the market impact of market orders . in that respect , the main ideas proposed in refs . @xcite can be reconsidered within the more realistic framework of point processes where correlations and impact are interpreted as cross and self excitations mechanisms acting on the conditional intensities of poisson processes . this allows us to make a step towards the definition of a faithful model of price microstructure that accounts for most recent empirical findings on the liquidity dynamical properties and to uncover new features . the paper is organized as follows . in section [ sec : defmodel ] we show how market order impact can be naturally accounted within the class of multivariate hawkes processes . our main model for price microstructure and market impact is presented and its stability is studied . numerical simulations of the model are presented . in section [ covmatrix ] , the microstructure of price and market order flows are studied through the covariance matrix of the process . our results are illustrated on numerical simulations . an extension of the model that accounts for labeled agents is defined in section [ m11 ] and results on market impact are presented in section [ sec : mi ] including an explanation on how the newly defined framework allows one to estimate the market impact profile from non labeled data . section [ sec : estimation ] shows how kernels defining the dynamics of trade occurrences and price variations can be nonparametrically estimated . all the theoretical results obtained in the previous sections are illustrated in section [ sec : empirical ] when applied on various high frequency future data . it allows one to reveal the different dynamics involved in price movements , market order flows and market impact . intraday seasonalities are shown to be taken care in a particularly simple way . we also discuss , on a semi - qualitative ground , how the market efficiency can be compatible with the observed long - range correlation in supply and demand without any parameter adjustment . conclusion and prospects are reported in section [ conclusion ] while technical computations are provided in appendices . as recalled in the introduction , in order to define a realistic microstructure price model while accounting for the impact of market orders , the framework of multivariate self - excited point processes is well suited . a natural approach is to associate a point process to each set of events one wants to describe . we choose to consider all market order events and all mid - price change events . let us point out that we will not take into account the volumes associated to each market orders . though this can be basically done within the framework of _ marked _ the point processes , it would necessitate cumbersome notations and make the estimation much more difficult . this issue be discussed briefly in section [ conclusion ] and addressed in a forthcoming work . the arrivals of the market orders are represented by a two dimensional counting process @xmath9 representing cumulated number of market orders arrived before time @xmath10 at the best ask ( @xmath11 ) and at the best bid ( @xmath12 ) . each time there is a market order , either @xmath13 or @xmath14 jumps up by 1 . we suppose that the trade process will be referred indifferently to as the market order arrival process or the trade arrival process . ] @xmath15 is a counting process that is fully defined by * @xmath16 the conditional intensity vector of the process @xmath17 at @xmath10 . the price is represented in the same way . let @xmath18 represents a proxy of the price at high - frequency ( e.g. , mid - price ) . as in refs . @xcite we write @xmath19 where @xmath20 ( resp . @xmath21 ) represents the number of upward ( resp . downward ) price jumps at time @xmath10 . @xmath22 ) for any upward ( reps . downward ) jumps ] thus , each time the price goes up ( resp . down ) , @xmath23 ( resp . @xmath24 ) jumps up by 1 . we set @xmath25 as for the trade process , @xmath26 is fully defined by * @xmath27 the conditional intensity vector of the process @xmath26 at @xmath10 . the 4 dimensional counting process is then naturally defined as @xmath28 as well as its associated conditional intensity vector @xmath29 basically , the model consists in considering that the 4-dimensional counting process @xmath30 is a hawkes process @xcite . the structure of a hawkes process allows one to take into account the influence of any component in @xmath30 on any component of @xmath31 . in its general form the model is represented by the following equation will introduce a generalization of this model including labeled trades . ] @xmath32 where @xmath33 is a @xmath34 matrix whose elements are _ causal _ positive functions ( by _ causal _ we mean functions supported by @xmath35 ) . moreover , we used the matrix convolution " notation , @xmath36 where @xmath37 refers to the regular matrix product . @xmath38 accounts for the exogenous intensity of the trades , it has the form @xmath39 since , by symmetry we assume that the exogenous intensity of @xmath13 and @xmath14 are equal while mid - price jumps are only caused by the endogenous dynamics . the matrix @xmath40 or any of its sub - matrices ( or element ) are often referred to as hawkes kernels . each element describes the influence of a component over another component . thus , it is natural to decompose the @xmath41 kernel @xmath33 in four @xmath42 matrices in the following way @xmath43 where * @xmath44 ( influence of @xmath15 on @xmath45 ) : accounts for the trade correlations ( e.g. , splitting , herding , ) . * @xmath46 ( influence of @xmath15 on @xmath47 ) : accounts for the impact of a single trade on the price * @xmath48 ( influence of @xmath49 on @xmath47 ) : accounts for the influence of past changes in price on future changes in price ( due to cancel and limit orders only , since changes in price due to market orders are explicitly taken into account by @xmath46 ) * @xmath50 ( influence of @xmath49 on @xmath45 ) : accounts for _ feedback _ influence of the price moves on the trades . if we account for the obvious symmetries between bid - ask sides for trades and up - down directions for price jumps these matrices are naturally written as : @xmath51 and @xmath52 where all @xmath53 are causal functions and the upperscripts @xmath54 and @xmath55 stand for `` self '' and `` cross '' influences of the poisson rates ( we use the same convention which was initially introduced in @xcite for @xmath56 ) . thus , for instance , on the one hand , @xmath57 accounts for the influence of the past buying ( resp . selling ) market orders on the intensity of the future upward ( reps . downward ) price jumps . on the other hand , @xmath58 quantifies the influence of the past buying ( resp . selling ) market orders on the intensity of the future downward ( resp . upward ) price jumps . .2 cm * remark 1 :* all these @xmath42 matrices commute since they diagonalize in the same basis ( independently of @xmath10 ) . their eigenvalues are the sum ( resp . the difference ) of the self term with the cross term . this property will be used all along the paper . most of the computations will be made after diagonalizing all the matrices . it is important to point out that a buying market order that eats up the whole volume sitting at best ask results in an instantaneous " change in the mid - price . from our model point of view , it would mean that @xmath15 and @xmath49 have simultaneous jumps with a non zero probability . it is clearly not allowed as is by the model . however , from a numerical point of view , this can be simulated by just considering that the jump in the price takes place within a very small time interval ( e.g. , of width 1ms which is the resolution level of our data ) after the market order has arrived . there is no ambiguity in the direction " of the causality : it is a market order that makes the price change and not the other way around . this would result in an impact kernel @xmath57 which is impulsive " , i.e. , localized around 0 , actually close to a dirac distribution @xmath59 . from a practical numerical point of view , choosing @xmath57 to be a dirac distribution is fairly easy . it basically amounts to considering that it is a positive function of given @xmath60 norm @xmath61 and with a support @xmath62 of the order of a few milliseconds . let us point out that it means that , the price increment between the moment of the trade and @xmath62 milliseconds afterwards , follows a poisson law whose parameter is @xmath61 . this actually allows price jumps ( spread over a few milliseconds ) of several ticks ( greater than @xmath63 ) . of course , from a mathematical point of view , this is not that simple . the limit @xmath64 has to be defined properly . this will be rigorously defined and extensively discussed in a future work and is out of the scope of this paper . the `` practical '' approach described above and the fact that we can formally replace , in all the computations , @xmath57 by a @xmath65 is far enough for our purpose . it is clear that we expect to find an impulsive component in @xmath57 when estimating on real data . though , a priori , we do expect also a non singular component that could have a large support ( e.g. , when the marker order eats up only a part of the volume sitting at best ask ) , we will see in estimations that most of the energy of @xmath57 is localized around 0 . moreover , we will find that @xmath58 is close to 0 . all these remarks will lead us to study a particularly interesting case of the previously defined model for which * @xmath66 and @xmath67 . consequently * @xmath68 , where @xmath69 refers to the identity matrix . this model will be referred in the following as the impulsive impact kernel model . before moving on and study the conditions for our model to be well defined , we need to introduce some notations that will be used all along the paper . let us introduce the following notations : if @xmath70 is a function @xmath71 refers to the laplace transform of this function , i.e. , @xmath72 @xmath59 refers to the dirac distribution , consequently @xmath73 moreover , we will use the convenient convention ( which holds in the laplace domain ) : @xmath74 . + the @xmath60 norm of @xmath75 is referred to as : @xmath76 thus @xmath77 we extend these notations to matrix of functions . thus , if @xmath78 is a matrix whose element are functions of @xmath10 , let @xmath79 denote the matrix whose elements are the laplace transform of the elements of @xmath78 . following this notation , we note @xmath80 if @xmath81 is a matrix , @xmath82 refers to the matrix @xmath81 whose each element has been replaced by its conjugate and @xmath83 refers to the hermitian conjugate matrix of @xmath81 . whenever @xmath31 is a stationary process , we will use the notation @xmath84 where @xmath85 let us point out that the fact that the @xmath86 mean intensities are equal is due to the symmetries of the kernels in and . [ not : delta ] we define the _ kernel s imbalance _ : * @xmath87 and @xmath88 * @xmath89 and @xmath90 * @xmath91 and @xmath92 * @xmath93 and @xmath94 let us point out that since the kernels are all positive functions , one has , replacing @xmath95 by @xmath15 , @xmath49 , @xmath61 or @xmath96 : @xmath97 and consequently @xmath98 the process @xmath30 defined in section [ m1 ] is well defined as long as the matrix @xmath33 is locally integrable on @xmath35 . hawkes , in his original papers @xcite , has formalized the necessary and sufficient condition for the previously introduced model to be stable : the matrix made of the @xmath60 norm of the elements of @xmath40 should have eigenvalues whose modulus are strictly smaller than 1 . this condition can be expressed in terms of conditions on the @xmath60 norm of the different kernels : * ( stability condition ) * [ prop : stability ] the hawkes process @xmath30 is stable if and only if the following condition holds : * the eigenvalues of the matrix @xmath99 have a modulus strictly smaller than 1 . in that case , @xmath30 has stationary increments and the process @xmath31 is strictly stationary . moreover ( h ) holds if and only if @xmath100 where * @xmath101 , * @xmath102 and * @xmath103 . moreover implies that @xmath104 where we used notations [ not : delta ] the proof is in section [ proofstab ] . .2 cm let us point out that in the case there is no feedback of the price jumps on the trades , i.e. , @xmath105 ( or @xmath106 ) , then the stability condition is equivalent to @xmath107 and @xmath108 , i.e. , @xmath109 and @xmath110 . .4 cm the mean intensity vector is given by the following proposition . * ( mean intensity ) * [ prop : lambda ] we suppose that ( h ) holds ( i.e. , ) . then @xmath111 this can be written as @xmath112 and @xmath113 where @xmath114 and where @xmath115 is defined by its laplace transform @xmath116 _ proof _ : the proof is basically an adaptation of a proof previously presented in @xcite . let the martingale @xmath117 be defined as @xmath118 using , we get @xmath119 thus @xmath120 consequently , @xmath121 where @xmath122 is defined by @xmath123 taking the expectation , we get . moreover , we have @xmath124 using remark 1 a the end of section [ m1 ] , one can easily check that @xmath125 where @xmath115 is defined by . the equations and are direct consequences of this last equation combined with . .3 cm * in the following we will always consider that ( h ) holds , i.e. , that holds . * in order to perform numerical simulations of hawkes models , various methods have been proposed . we chose to use a thinning algorithm ( as proposed , e.g. , in @xcite ) that consists in generating on @xmath126 $ ] an homogeneous poisson process with an intensity @xmath127 } ( \lambda^{t\pm}_t,\lambda^{n^\pm}_t)$ ] . a thinning procedure is then applied , each jump being accepted or rejected according to the actual value of @xmath128 or @xmath129 . in order to illustrate the 4-dimensional process we chose to display only the price path @xmath130 and the _ cumulated trade process _ path as defined by @xmath131 in figure [ fig : pathexpzoom ] , we show an example of sample paths of both @xmath18 and @xmath132 on a few minutes time interval . all the involved kernels are exponentials . some microstructure stylized facts of the price can be clearly identified directly on the plot : price moves arrive at random times , price moves on a discrete grid and is strongly mean reverting ( see beginning of next section ) . in the large time limit , one can show that these processes converge to correlated brownian motions ( see @xcite or section [ diffusive ] ) . this is illustrated in fig . [ fig : pathexp ] where the paths are represented over a wider time window ( almost 2 hours ) . as discussed in section [ empirical : correlation ] , since we choose @xmath133 and @xmath134 , the small time increments of @xmath132 remain correlated while the price increments correlations almost vanish . this can be observed in fig . [ fig : pathexp ] where the path of @xmath132 appears to be smoother than the path of @xmath26 . , @xmath135 , @xmath136 , @xmath137 and @xmath138 . , height=302 ] over a wider range of time . one does not see the discrete nature of time variations anymore . @xmath132 and @xmath18 appear as correlated processes.,height=302 ] in the next section we study the microstructure properties of both the price and the market order flow . the model we introduced above can be seen as a generalization of the price microstructure model previously introduced in @xcite . indeed , the 2-dimensional model defined in @xcite can somewhat be seen as the `` projection '' on the price components @xmath26 of the model defined in section [ m1 ] . thus , it easy to show that it will account for all the price microstructure stylized facts already accounted for by the model in @xcite . this includes the characteristic decreasing shape of the _ mean signature plot _ @xmath139 ( which explains why estimating the diffusing variance using high frequency quadratic variations leads to a systematic positive bias ) . as explained in @xcite , this effect is due to the high frequency strong mean reversion observed in real prices and can easily be modeled by choosing kernels such that @xmath140 . we refer the reader to @xcite for all the discussions concerning the signature plot , the problem of variance estimation using high frequency data and the link with mean reversion of price time - series . thus , in this section , we shall mainly focus one price correlations and market order correlations and how they relate one to the other . following @xcite , we define the covariance matrix of the normalized process at scale @xmath141 and lag @xmath142 by @xmath143 where we normalized by @xmath141 in order to avoid a trivial scale dependence . let us note that , since the increments of @xmath30 are stationary ( we suppose that is satisfied ) , the previous definition does not depend on @xmath10 . thus , it can be rewritten as @xmath144 in @xcite , it is proven that the the laplace transform of @xmath145 can be expressed as a function of the laplace transform of @xmath146 ( where @xmath147 ) and of @xmath148 : @xmath149 where @xmath150 is the diagonal matrix @xmath151 from this result , one can easily deduce an analytical formula for the price auto - covariance function . * ( price auto - covariance ) * let @xmath152 be the normalized auto - covariance of the price increment : @xmath153 then @xmath154 before proving this proposition , we first need the following lemma , which is a direct consequence of and . [ lem1 ] @xmath155 where @xmath156 @xmath157 @xmath158 @xmath159 _ proof of proposition : _ using the symmetries of all the kernels , and the fact that @xmath160 , we get @xmath161 thus , if we define @xmath162 and @xmath163 such that @xmath164 we have @xmath165 using remark 1 , we get that @xmath166 which proves . .7 cm using similar computations , one derives an analytical formula for the auto - covariance of the cumulated trade process : * ( trade auto - covariance ) * let @xmath167 be the normalized covariance of the increments of the cumulated trade process @xmath132 defined by . @xmath168 then @xmath169 in order to illustrate these results , in fig . [ fig : correxp ] we have plotted both theoretical ( solid lines ) and estimated ( @xmath170 ) correlation functions . the sample we used for the estimation contains around @xmath171 trading events and corresponds to the same kernels as the ones used for fig . [ fig : pathexp ] . one clearly see that empirical estimates closely match theoretical expressions . moreover , one can observe that the trade increment autocorrelation has an amplitude that is an order of magnitude larger than the price increment covariance . this issue will be addressed in section [ empirical : correlation ] . . ( a ) autocorrelation of the increments of @xmath132 . ( b ) autocorrelation of the price increments ( i.e. , the increments of @xmath18 ) . in both cases , for the sake of clarity with have removed the lag @xmath172.,height=302 ] let us point out that we know @xcite that a centered @xmath173-dimensional hawkes process @xmath30 diffuses at large scales ( @xmath174 ) towards a multidimensional gaussian process : @xmath175 where @xmath176 is a standard @xmath173-dimensional brownian motion and @xmath150 is defined by . this is a very general result that can be applied here to obtain the covariance matrix of the diffusive limit of our 4-dimensional hawkes process @xmath30 . actually , if one is just interested in the diffusive variance of @xmath30 , one can easily obtain it directly from the definition of the covariance matrix ( noticing that @xmath177 corresponds to the variance of @xmath178 ) . one gets that the diffusing variance is @xmath179 . it is then straightforward to obtain the diffusive variance of the price @xmath180 : @xmath181 ( let us point out that shows that the denominator is positive ) . the trade arrival process @xmath15 models _ anonymous _ trades sent by _ anonymous agents _ : this corresponds to the typical trade information sent by most exchanges ( i.e. , the trades are _ unlabeled _ , there is no way to know whether two different trades involves the same agent or not ) . however , there are many situations where one has access to some _ labeled _ data corresponding to some specific _ labeled agents_. in this case , it could be very interesting to be able to analyze the impact of these labeled trades within the same framework . it is naturally the case of brokers who have labels for all their clients ( though some clients may have several brokers , consequently a given broker might not be able to identify _ all _ the trades of a given client ) . in general , this is the case of any financial institutions which has access to the historic of all its own trades . in our model , we chose , to consider only the case of a single labeled agent that is sending market orders at some deterministic time . the case where there are several such agents is a straightforward generalization . the trades arrival of this labeled agent is represented by a 2d deterministic function @xmath182 representing the market orders arrival at the best ask ( @xmath183 ) and at the best bid ( @xmath184 ) . again , @xmath185 ( resp . @xmath186 ) jumps upward by 1 as soon as the agent is buying ( resp . selling ) ( we do not take into account the associated volumes ) . * in all the following , we consider that the agent is active only on a finite positive time interval , i.e , @xmath187 * we are ready now to reformulate the previously introduced model ( see section [ m1 ] ) taking into account the labeled agent . in its most general form the model writes @xmath188 where @xmath189 is a @xmath190 matrix . in the same way as we did for @xmath40 , we decompose the kernel @xmath189 using @xmath42 matrices : @xmath191 * @xmath192 ( influence of the labeled trades on @xmath45 ) : accounts for the herding of the anonymous trades with respect to the labeled trades . * @xmath193 ( influence of the labeled trades on @xmath47 ) : accounts for the impact of a single labeled trade on the price . a priori there is no reason not to consider that the impact of a labeled trade is not the same as the impact of a anonymous trade , i.e. , we will take @xmath194 let us point out that the model introduced in section [ m1 ] is a particular case ( when @xmath195 ) of this model . in that sense , this model is a more general model . let us give an analytical expression for the market impact profile . * ( market impact profile ) * [ th : main ] the expectation of the intensity vector is given by @xmath196 where @xmath122 is defined by its laplace transform @xmath123 the market impact profile between time 0 and time @xmath10 of the labeled agent is defined as @xmath197 it can be written @xmath198 where @xmath199 is defined as @xmath200 * proof of * . it is very similar to the proof of . let @xmath117 be the martingale defined as @xmath118 using , we get @xmath201 which gives @xmath202 which is equivalent to @xmath203 taking the expectation , we get . .2 cm * proof of * + using , we get @xmath204 where @xmath205 is the @xmath206 matrix defined by @xmath207 thus @xmath208 where @xmath209 stands for the vector @xmath210 . let us recall that we chose @xmath211 . using remark 1 , all the @xmath212 matrices involved in this equation are commuting since they are symmetric along both diagonals . thus one gets @xmath213 thus the market impact is @xmath214 using remark 1 again , we deduce let us state a trivial corollary of this last proposition that will be of particular interest in order to study the permanent impact in the following section : [ cor1 ] in the case of an impulsive impact kernel ( see section [ impulsive ] ) and @xmath215 , the market impact profile of a single buying market order sent at time @xmath216 ( i.e. , @xmath217 and @xmath218 ) is given by @xmath219 where @xmath199 is defined as @xmath220 and @xmath221 . the kernel shapes and the parameters have been chosen to roughly match the values observed empirically in section [ sec : empirical ] . the two figures only differ by the values of @xmath222 . ( a ) @xmath223 ( b ) @xmath224.,height=188 ] in fig . [ fig : impact ] are displayed 2 examples of market impact profiles computed according to eq . for a meta - order consisting in buying a constant amount of shares during a period @xmath15 ( i.e. @xmath225 if @xmath226 and @xmath227 otherwise ) . the parameters fig . [ fig : impact](b ) correspond to the ones estimated from empirical data as discussed in section [ sec : empirical ] ( notably @xmath228 ) . one can see that fig . [ fig : impact](b ) reproduces fairly well the empirical shape measured using labeled database ( as , e.g. , in @xcite ) : a concave shape of the profile @xmath229 for times @xmath226 ( the so - called `` square - root '' law @xcite ) and a convex relaxation towards the permanent impact when @xmath230 . this shape is qualitatively explaind in section [ empirical : miestimation ] . for illustration purpose , the parameters of fig . [ fig : impact](a ) are the same as the ones of fig . [ fig : impact](b ) except that they have been tweaked in order to get @xmath231 . the fact that the asymptotic value of the market impact is larger is explained by point iii ) of the next section . one important consequence of is that the asymptotic market impact of a single labeled market buying order at time @xmath232 is of the form ( assuming an impulsive market impact kernel ) : @xmath233 thus controlling how permanent is the market impact essentially amounts in controlling how large @xmath234 is where @xmath235 thus @xmath236 let us note that , the stability condition of the process ( see proposition [ prop : stability ] ) implies that @xmath237 and @xmath238 . this last equation allows us to identify three different dynamics that can lead to a decrease of the permanent impact . they all correspond to different ways of introducing mean reversion into the price : * mean reversion due to microstructure . this case corresponds to the case @xmath239 is very negative , i.e. , @xmath240 much larger than @xmath241 . however this effect is limited by the fact that we know that we can not go below @xmath242 due to the stability condition . * feedback cross influence of the price moves on the trades . this case corresponds to the case @xmath243 is very negative , e.g. , @xmath244 large and @xmath245 . this is definitely an effect that is present in real life ( see section [ sec : empirical ] ) : as the price goes up , traders are sending more and more selling buying orders . * auto - correlation of the trades . the previous effect i ) is even stronger if there is a strong correlation in the signs of the market orders ( i.e. , @xmath246 ) . let us point out that fig . [ fig : impact ] shows a clear illustration of this point : the left market impact curve has a stronger permanent market impact than the one of the right , this is due to the fact that it corresponds to a smaller @xmath247 . as we will see in section [ empirical : estimation ] , all these effects are present in real data . and they will all play a part in reducing the asymptotic market impact ( see section [ empirical : miestimation ] ) as already pointed out , most markets do not provide labeled data . the order flows are made of anonymous orders sent by anonymous agents . a priori , in that case , the only way of quantifying the `` impact '' a given market order has on the price is to estimate numerically the response function @xmath248 . the response function is defined as the variation of the price from time 0 to time @xmath10 knowing there was a ( e.g. , buying ) trade at time 0 . thus it can be written @xmath249 and @xmath250 otherwise . it can be written as @xmath251 notice that within the model of bouchaud et al . , this response function can be explicitly related to the `` bare impact '' function @xmath252 and therefore used to estimate its shape @xcite . it is important to understand that this function is fundamentally different from the market impact profile of a single market order . the market impact isolates all the market orders of a single agent ( e.g. , a meta - order ) and quantifies what the impact of these market orders are . in order to compute it , one , a priori , needs to identify all the market orders of a particular agent . this is not the case of the response function which is `` polluted '' by the impact of all the market orders that are in the same meta - order as the market order that is under consideration . a very important consequence of our model ( as it will be explained in section [ empirical : miestimation ] ) is that our model allows estimation of the market impact profile even if no labeled data are available . it will basically consist in first estimating all the kernels ( see section [ sec : estimation ] ) and then using the analytical formula . let us point out that one can easily obtain an analytical formula for the response function using proposition [ th : cexpect ] of section [ annex_cond ] . we prove that if @xmath253 is the @xmath41 matrix defined by @xmath254 with @xmath255 then @xmath256 using lemma [ lem1 ] , if we define @xmath257 and if we define @xmath258 , @xmath259 , then , @xmath260 using remark 1 , we get @xmath261 in the case of an impulsive impact kernel , it gives @xmath262 where @xmath263 is defined by @xmath264 by solving the fredholm equation @xmath265 for an exponential model . estimations are represented by symbols @xmath266 for the `` self '' kernels and @xmath267 for the `` cross '' kernels . exact exponential kernels are represented by the solid lines . the model corresponds to @xmath268 , @xmath269 , @xmath270 , @xmath271 and @xmath272.,height=340 ] in this section we provide a new method to estimate the shape of the kernels involved in the definition of the model . a former non parametric estimation method has been introduced in @xcite . this method relies on the expression and mainly consists in extracting the square root of the autocorrelation matrix . however , in order to do so univocally , one has to suppose that the process is fully symmetric and in particular that @xmath15 and @xmath49 have the same laws . this assumption is clearly unrealistic , therefore the method of @xcite is not suited to estimate the process defined in section [ sec : defmodel ] . for that reason we introduce an alternative method that does not require any symmetry hypothesis . this method relies on the proposition [ th : fredholm ] of annex [ annex_cond ] ( eq . ): @xmath273 where @xmath253 is the matrix of conditional expectations defined in eq . . this above equation is a fredholm equation of 2nd kind . since @xmath253 can be easily estimated from empirical data , the matrix @xmath274 can be obtained as a numerical solution of the fredholm system . we thus implemented a classical nystrom method that amounts to approximate the integrals by a quadrature and solve a linear system @xcite . in log - log scale . the slope is @xmath22.,height=188 ] in order to illustrate the method , we have reported in fig . [ fig : estimexp ] the estimated kernels in the case when all functions @xmath275 have an exponential shape . the sample used has typically @xmath276 trade events and @xmath277 price change events . one can see that for such sample length , numerical estimates and the real kernels are close enough to determine a good fit of the latter . let us note that we have checked that the method is reliable for various examples of kernels like exponential , power - laws or constant over an interval . in order to investigate the efficiency of the method , one can evaluate the estimation error behavior as a function of the number of events . this error can be defined as the supremum of the mean square error associated with each kernel : if @xmath278 stands for the estimated kernel matrix for @xmath279 trading events , then one can consider the following square error : @xmath280 a log - log plot of @xmath281 as a function of @xmath279 is reported in fig . [ fig : estimerror ] where @xmath281 has been estimated , for each @xmath279 , using 500 trials of the model . the measured slope is close to @xmath22 in agreement with the standard behavior of the estimation error : @xmath282 a comprehensive study of the statistical properties of the method is out of the scope of the present paper and will be addressed in a forthcoming work . in this section we apply the main theoretical results we obtained in the previous sections to real data . we consider intraday data associated with the most liquid maturity of eurostoxx ( fsxe ) and euro - bund ( fgbl ) future contracts . the data we used are trades at best bid / ask provided by quanthouse trading solution . each time series covers a period of 800 trading days going from 2009 may to 2012 september . the typical number of trades is around 40.000 per day while the number of mid price changes is 20.000 per day . , height=302 ] ( a ) and @xmath283 ( b ) in the intraday slot [ 9 a.m. , 11 a.m. ] for eurostoxx . this is a `` zoom '' of fig . [ fig : phisxef ] . one sees that @xmath284 slowly decreases and is very large as compared to @xmath283.,height=188 ] ( a ) and @xmath285 ( b ) in the intraday slot [ 9 a.m. , 11 a.m. ] for eurostoxx . this is a `` zoom '' of fig . [ fig : phisxef ] . one sees that @xmath285 is larger than @xmath286.,height=188 ] and @xmath285 . the kernels are represented in double logarithmic representation for eurostoxx ( @xmath170 ) and euro - bund ( solid lines ) estimates in the intraday slot [ 9 a.m. , 11 a.m. ] . ( a ) @xmath284 ( b ) @xmath285 . these plots suggest that the kernels have the same power - law behavior for both eurostoxx and euro - bund.,height=188 ] in this section we apply the non parametric kernel estimation algorithm presented in section [ sec : estimation ] to our data . since intraday statistics are well known to be non stationary , estimations are based on a 2 hours liquid period from 9 a.m. to 11 a.m. ( gmt ) . the results of the kernel estimation of eurostoxx are displayed in figs.[fig : phisxef ] , [ fig : phitsxe ] and [ fig : phinsxe ] . one first sees that @xmath283 is small as compared to @xmath284 . this appears more clearly on the `` zoom '' presented in fig . [ fig : phitsxe ] where we have removed the first point in order to get a finer scale . the fact that @xmath284 is larger than @xmath283 confirms the well known strong correlation observed in trade signs ( see section [ empirical : correlation ] ) . the mid - price dynamics is well known to be mainly mean - reverting @xcite . this is confirmed by fig . [ fig : phinsxe ] where a greater intensity of the kernel @xmath285 as compared to @xmath286 can be observed . both kernels @xmath284 and @xmath285 are found to decrease as a power - law : @xmath287 with an exponent @xmath288 for @xmath284 and @xmath289 for @xmath285 . the cut - off @xmath55 insures the finiteness of the @xmath60 norm of @xmath274 and is found to be smaller than @xmath290 while the values of @xmath291 are typically between @xmath292 and @xmath293 . this power - law behavior is illustrated in fig . [ fig : philoglogsxefgbl ] where we have reported in log - log scale the estimates @xmath284 and @xmath285 for both eurostoxx ( @xmath170 ) and euro - bund ( solid lines ) . let us notice that power - law behavior of hawkes kernels have already been observed in @xcite using a simple 2d hawkes model of price jumps . this important property suggests the existence of some scale invariance properties underlying the order book dynamics . let us point out that fig . [ fig : philoglogsxefgbl ] also suggests that the kernels are the same for eurostoxx and euro - bund . as far as the impact is concerned , one can see in fig . [ fig : phisxef ] that only @xmath294 is significant and well modeled by an impulsive kernel ( see section [ impulsive ] ) . this confirms the fact that a buy ( resp . sell ) market order increases the probability of an upward ( resp . downward ) movement of the mid - price but only within a very small time interval after the trade . the feed - back kernel that accounts for the influence of a price move on the trading intensity is also found to be well localized but only the cross term is non negligible . it seems that an upward ( resp . downward ) move of the mid price triggers an higher trading activity on the bid side ( resp . on the ask side ) . for various 2 hours slices during the day for eurostoxx . in ( a ) and ( b ) 7 estimated kernels are represented in doubly logarithmic scale corresponding to the 7 slots : [ 8 a.m. , 10 a.m. ] , [ 9 a.m. , 11 a.m. ] , [ 11 a.m. , 1 p.m ] , [ 12 a.m. , 2 p.m. ] , [ 1 p.m. , 3 p.m ] , [ 2 p.m. , 4 p.m ] and [ 3 p.m. , 5 p.m. ] . in ( a ) @xmath284 and in ( b ) @xmath285 . one sees that the kernel shapes are remarkably stable power - laws across intraday time periods . in ( c ) we reported the estimation of the constant rate @xmath295 as a function of the intraday time . one recognizes the classical u - shaped behavior.,height=340 ] in order to check the stationarity of the kernels , we performed various estimations on different intraday slices of 2 hours . the results for @xmath284 and @xmath285 are reported in fig . [ fig : stability](a ) and fig . [ fig : stability](b ) in the case of eurostoxx . one can see that the shape of the kernels does not depend on the intraday market activity and are remarkably stable throughout the day . an estimation of the norms of all kernels allows one to estimate the rate @xmath295 through or . in fig . [ fig : stability](c ) , we see that this parameter follows the classical u - shaped intraday curve . it thus appears that , in the model , the exogenous intensity of trades fully accounts for the intraday modulation of market activity . it is therefore natural to consider the following model with seasonal variations that generalizes , in a particularly simple way , the definition : @xmath296 with @xmath297 where @xmath298 is a u - shaped 1-day periodic function . let us point out that this is particularly elegant and simple way of accounting for an a priori pretty complex phenomenon . few years ago , bouchaud et al . @xcite and lillo and farmer @xcite independently brought evidence that the market order flow is a long memory process . by studying the empirical correlation function ( in trading time ) of the signs of trades they have shown , for different markets and assets that : @xmath299 where @xmath0 is the lag in number of trades ( i.e. , using trading time ) . this scaling law remains valid over 2 or 3 decades and the exponent @xmath300 is in the interval @xmath301 $ ] . the aim of impact price models as described in the introductory section ( eq . ) was to solve this `` long - memory '' puzzle . indeed , since trades impact prices , if trades are long - range correlated , in order to maintain efficiency , the price have to respond to market order through a long - memory kernel . this is precisely the meaning of the kernel @xmath302 in . bouchaud et al . @xcite have shown that , provided the specific shape of @xmath302 is adjusted as respect to the behavior , trade correlations impact on prices can be canceled ( see also @xcite for an interpretation of this price model in terms of prediction error or `` surprise '' ) . in this section , we explain how this issue can be addressed within our model . it is important to point out that , as we will see , the apparent long - range correlation of the trade signs and the price efficiency will be a consequence of 3 empirical findings ( cf section [ empirical : estimation ] ) : * @xmath303 is power - law * @xmath304 is close to 1 , i.e. , it almost saturates the stability condition @xmath305 * @xmath306 , i.e. , the price at high frequency is mainly mean - reverting . .5 cm [ [ about - the - long - range - memory - of - the - trade - sign - process ] ] * about the long - range memory of the trade sign process * + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the exact expression of the autocorrelations of the increments of @xmath307 and @xmath308 as a function of the lag @xmath142 can be hardly deduced from and . in order to discuss the shape of theses correlation functions one can however use qualitative arguments based on classical tauberian theorems @xcite . in what follows the argument @xmath309 of laplace transforms is assumed to be real and positive ( @xmath310 ) . let us first remark that , within the range of parameters observed in empirical data ( and notably the relationship @xmath311 ) , one can show that the terms involving @xmath312 both at the numerator and at the denominator in are subdominant . in this case , this equation , reads ( we neglect the @xmath313 factor and drop the @xmath314 upper - script everywhere ) : @xmath315 let us note that , according to this expression , the stability condition @xmath316 implies that @xmath317 , i.e. , the correlation function of the trades @xmath318 is integrable and that , strictly speaking , there is no long - range memory in supply and demand . let us show however that , under the conditions we observe empirically , @xmath318 can reproduce , on a pretty large ( though finite ) range of @xmath142 , a slow - decay with an exponent smaller than 1 , leading various numerical estimations to conclude long - range memory . as observed in previous section , @xmath319 is close to a power - law ( @xmath320 ) : @xmath321 were @xmath322 is the scaling exponent ( empirically we found @xmath323 ) , @xmath55 is a small scale cut - off ( empirically @xmath324 ) , and @xmath291 is a factor such that the norm @xmath325 . on has obviously : @xmath326 if one computes the laplace transform ( with @xmath327 ) of expression , it is easy to show that , in the limit of small @xmath309 ( in practice that means @xmath328 ) @xmath329 thus , according to , one has @xmath330 and consequently @xmath331 and therefore , thanks to tauberian theorems ( limit of small @xmath309 corresponds to limit of large time ) : @xmath332 however , if we suppose that not only @xmath309 is small but that @xmath222 is close enough to 1 such that @xmath333 then becomes @xmath334 or equivalently @xmath335 let us point out that the inequality holds , as long as : @xmath336 using the estimates @xmath323 and @xmath228 . since @xmath328 , this means that it can take 5 decades to see the `` short range '' nature of the trade correlation function . considering that @xmath337 , for an inter event mean time of @xmath338 , the scaling can extend over 3 decades . since @xmath339 , this exponent is precisely in the range of values reported empirically in @xcite and @xcite . .5 cm [ [ about - price - efficiency ] ] * about price efficiency * + + + + + + + + + + + + + + + + + + + + + + + + from an empirical point of view , in agreement with market efficiency hypothesis , it is well known that price variations are almost uncorrelated after few seconds . let us show , that , under the conditions observed in empirical data , @xmath340 decreases very fast around @xmath172 . we provide the same kind of pedestrian arguments as in previous discussion : we neglect the influence of @xmath341 and suppose that we are in the `` impulsive '' case of the impact kernel , i.e. , @xmath342 . let us remark that , since @xmath343 is bounded ( empirically @xmath344 ) , if @xmath61 is small enough ( empirically @xmath345 ) , the same arguments invoked for @xmath346 shows that in the intermediate range : @xmath347 one has @xmath348 . it results that reduces , in this range , to : @xmath349 in order to study the behavior of this function , we use again a power - law expression for @xmath350 : @xmath351 where empirically @xmath352 ( and thus @xmath353 ) and @xmath354 with @xmath355 . it follows that @xmath356 in the range @xmath357 since @xmath353 , @xmath358 , @xmath359 , @xmath360 , we have @xmath361 and the previous inequality always holds . this shows that the price increments correlation functions decreases very fastly around zero without any `` fine tuning '' of the kernel shapes . and @xmath18 in trading time . the process is a numerical simulation of a hawkes process where the kernels have been chosen to fit the empirical observations found in [ empirical : estimation ] . the correlation functions where estimated on a single realization containing around 40.000 market orders . ( a ) @xmath362 ( @xmath170 ) and @xmath363 ( solid line ) as functions of the lag @xmath0 ( expressed in trading time ) . ( b ) @xmath364 in double logarithmic reprensentation ( @xmath170 ) . the solid line represents the power - law fit with expression where @xmath365.,height=188 ] [ [ numerical - illustrations ] ] * numerical illustrations * + + + + + + + + + + + + + + + + + + + + + + + + + all these considerations are illustrated in fig . [ fig : covtnmodel ] , where we have simulated a hawkes process with the parameters close to the ones we found empirically . let us notice that , in order to mimic the experiments performed in @xcite and @xcite , the correlation functions are computed in trading time , i.e. , a discrete time which is incremented by 1 at each jump of @xmath132 . we checked that setting a lag @xmath0 in trading time roughly amounts to consider a lag @xmath142 in physical time such that : @xmath366 . in that respect , the shapes of the correlation functions in trading time and in physical time are very close to each other , up to a scaling factor . in fig . [ fig : covtnmodel](a ) are displayed @xmath362 ( @xmath170 ) and @xmath363 ( solid line ) as functions of the lag @xmath0 ( in trading time ) . on clearly sees that @xmath364 is slowly decaying while @xmath367 almost vanishes after few lags . in fig . [ fig : covtnmodel](b ) @xmath364 is represented in log - log . as expected , it behaves as power - law with the exponent @xmath368 ( @xmath369 in the example we choose ) represented by the solid line . as discussed in section [ sec : mi ] , once one has estimated all the model parameters , it is possible to compute the shape of the market impact of some particular ( meta- ) order ( proposition [ th : main ] with @xmath215 ) . if one uses the parameters reported previously , one gets a shape of the market impact profile associated with the meta - order @xmath370 similar to the one displayed in fig . [ fig : impact](b ) . one can use similar qualitative arguments than in previous section to explain the shape observed in fig . [ fig : impact](b ) . indeed , according to and using the same notations , parameter values ( notably the fact that @xmath222 is close to 1 ) and approximations as discussed previously , one can show that in a wide intermediate range of laplace parameters @xmath309 , we have : @xmath371 if @xmath372 , then @xmath373 and therefore @xmath374 it follows that the market impact profile behavior reads : @xmath375 that corresponds to the behavior observed in fig . [ fig : impact](b ) . notice that a strict `` square - root '' would correspond to @xmath376 while it seems we rather observe @xmath323 empirically . in general , the determination of the market impact profile is a hard task that requires to possess agent labeled ( e.g. broker ) data . the model presented in this paper allows one to recover a market profile using anonymous order flow data as explained in section [ minon ] . from our knowledge , the model we developed in this paper is the first model that accounts for both stylized facts of market prices microstructure ( including random time arrival of price moves , discrete price grid , high frequency mean reversion , correlation functions behavior at various time scales ) and the stylized facts of market impact ( mainly the concave / relaxation characteristic shape of the market impact of a meta - order ) . analytical closed formula can be obtained for most of these stylized facts . not only it allows us to reveal ( through the estimations of the different kernels ) the dynamics involved between trade arrivals and price moves , but it also allows us to estimate the entire market impact profile from anonymous market data . as far as trade and price dynamics are concerned , we have provided evidence of a power - law behavior of the kernels @xmath377 and @xmath378 and that the model is close to instability ( i.e. @xmath379 is smaller but close to 1 ) . this suggests the existence of some self - similarity properties in the order - book dynamics and sharply contrasts with the usual exponential kernels used in former parametric hawkes modeling in finance . the cross - kernels associated with impact of trade on prices ( @xmath380 ) and feed - back ( @xmath381 ) are well localized in time ( i.e. of `` impulsive nature '' ) . thus , upward ( resp . downward ) price moves are mainly impacted by trades on the ask ( resp . bid ) side . in turn , positive ( resp . negative ) mid - price variations imply an increase of the trading intensity on the bid ( resp . ask ) side . besides these important points , qualitative arguments showed that , provided @xmath382 and @xmath383 , the long memory puzzle `` of the order flow raised by bouchaud et al . @xcite can be addressed without any fine tuning of the model parameters : trades naturally appear as long - range correlated over a wide range of lags while price variations are almost uncorrelated . moreover , the same kind of arguments can explain the concave ( ' ' square - root " law ) / relaxation typical market impact shape and an almost vanishing permanent impact . let us first point out that the model , as is , can be used as a stochastic price replayer " using as an input the true market order arrivals in place of the stochastic process @xmath17 . this allows one to replay " the price of a given historical period and , for instance , using it as an input price for any algorithm designed to estimate or manage a risk . in this paper , we have presented the most basic form of the model . it can be seen as a building block for more elaborate models depending on what it is meant to be used for . there are many ways for extensions . let us just go through some of them we have already developed or we are still working on . for instance , it would be important to have a model which not only takes into account the arrival times of the market orders but their volumes too . this is a pretty easy extension since it can be done within the framework of marked hawkes processes for which straightforward extensions of all computations presented in this paper can be obtained . in the simplest form , one could use i.i.d . volumes @xmath384 for the market orders : at any time @xmath10 a market order arrives ( @xmath385 or @xmath386 ) a volume @xmath384 is chosen randomly ( using a given law ) . in its simplest form the new model consists in replacing the projection of on the last two components by @xmath387 where @xmath75 is a function that describes how the volume impacts the price . it basically corresponds to what is generally referred into the literature by the instantaneous impact function " . in order to go further into the understanding of the underlying dynamics of the order - book , a very natural extension of the model , consists in using more dimensions in order to take into account limit / cancel orders . thus , for instance , one way would be to introduce a new point process @xmath388 where @xmath389 ( resp . @xmath390 ) is incremented by 1 whenever a limit order arrives at the best ask ( resp . best bid ) or a cancel order arrives at best bid ( resp . best ask ) . one would then need to introduce the different kernels that account for the influence of @xmath391 on @xmath15 and @xmath49 and the kernels that account for the influence of @xmath15 and @xmath49 on @xmath391 itself . the estimation procedure of the kernels can follow the exact same procedure described in section [ sec : estimation ] . along the same line , i.e. , by adding new dimensions to the 4d - hawkes model presented in this paper , one could quantify the impact of a given exogenous news on the market order flow or directly on the price by estimating the corresponding kernel . or , alternatively , one could consider a multi - agent models ( e.g. , adding 2 dimensions for each agent ) and model / estimate the influence of a given agent on another one or on all the anonymous agents ( as @xmath392 does in ) . these extended framework would open the door to precise estimations and obvious interpretations in order to get better insights into to full order book dynamics . in this section , we give the proof to the proposition [ prop : stability ] for the process to be stable , we need the eigenvalues of the matrix @xmath99 to have a modulus smaller than 1 ( cf hypothesis ( h ) ) . [ lm1 ] * ( eigenvalues of @xmath99 ) * + if @xmath393 is an eigenvalue of @xmath99 , then it satisfies @xmath394 or @xmath395 where @xmath396 , @xmath397 and @xmath398 . ( let us point out that , using notations [ not : delta ] , @xmath399 , @xmath400 and @xmath401 ) . _ proof _ + from we get @xmath402 since all the matrices are bi - symmetric , they all commute and thus @xmath403 moreover , they all diagonalize in the same basis and their eigenvalues are the sum and the difference between the self term and the cross term . thus the eigein values of @xmath404 satisfy either or . + .4 cm we now have to study when the modulus of the roots of these equations are smaller than 1 . [ lm2 ] * ( condition for the roots to have a modulus smaller than 1 ) * + the roots of the equation @xmath405 have a modulus smaller than 1 iff * either @xmath406 * or @xmath407 _ proof _ + the discriminant of this second - order equation is @xmath408 , thus there are two cases * @xmath409 . in this case the roots are complex and conjugated one of the other . thus their modulus is smaller than 1 iff their product is smaller than 1 , i.e. , @xmath410 which gives the result for ( ii ) * @xmath411 . in this case the roots are real . the condition for them to be smaller than 1 is @xmath412 and the condition for them to be greater than -1 is @xmath413 this synthesizes into case ii of lemma .4 cm using these two lemmas , let us study the stability condition corresponding to and . before starting , let us note that , since all the @xmath274 functions are positive functions ( thus all the @xmath414 are real positive ) , it is clear that @xmath415 , since @xmath416 . moreover is equivalent to @xmath417 which is equivalent to @xmath418 which proves .4 cm * proving is a sufficient condition for ( h ) to hold . ( and proving ) * + for that purpose , we suppose that holds . we want to prove that the roots of both and have modulus strictly smaller than 1 . we have just seen this is the case for the roots of , we just need to check that it is also the case for the roots of . for the case ( i ) reads @xmath419 and the case ( ii ) ( since @xmath420 according to ) @xmath421 thus , merging these last two inequations , we get the following condition for the roots of to have a modulus strictly smaller than 1 @xmath422 which is nothing but . thus , in order to complete the proof of proposition [ prop : stability ] , we just need to prove that this last inequation holds ( i.e. , holds ) . let us first notice that , since @xmath423 and @xmath424 , one has @xmath425 , and consequently @xmath426 since @xmath427 the following inequation @xmath428 is a sufficient condition for to hold . moreover since @xmath429 , using , a sufficient condition for to hold is @xmath430 which is equivalent to @xmath431 since , one has @xmath432 inequation is implied by @xmath433 , we just need to prove this last inequation . .4 cm in the case @xmath434 and @xmath435 have the same sign , this inequation is a actually a strict equality , so it obviously holds . let us suppose that @xmath434 and @xmath435 do not have the same sign . without loss of generality , we can suppose that @xmath436 . we distinguish two cases : * either @xmath437 , in which case @xmath438 * or @xmath439 , in which case @xmath440 in this section , we establish general results on @xmath49-dimensional hawkes process @xmath441 and more particularly on the expectations of @xmath442 at time @xmath10 conditioned by the fact that the @xmath443 jumped at time 0 . we mainly establish two results . the fist one ( prop . [ th : cexpect ] ) links these conditional expectations with the auto - covariance function of @xmath444 and , using previous results @xcite , allows us to derive an analytical formula as a function of the kernel . these results are used in the section [ response ] for characterizing the response function . the second one ( prop . [ th : fredholm ] ) proves that these expectations satisfy a fredholm system that will be used in section [ sec : estimation ] for elaborating a general procedure for the kernel estimation . the hawkes process is defined by its kernel @xmath445 and the exogenous intensity @xmath446 through the equation @xmath447 where @xmath31 is the conditional intensity of @xmath448 at time @xmath10 . we consider that the process is stable , i.e. , the eigenvalues of the matrix @xmath99 have modulus smaller than 1 . we set @xmath449 finally for all @xmath10 and @xmath450 such that @xmath451 , we define @xmath452 with @xmath453 where @xmath454 is the conditional expectation of @xmath455 knowing that @xmath456 jumps at @xmath216 , @xmath59 is the dirac distribution and @xmath457 is always 0 except for @xmath458 for which it is equal to 1 . since @xmath459 is singular at @xmath216 we substracted this singular component . in the following @xmath253 will refer to the matrix whose elements are the @xmath460 , i.e. , @xmath461 we are ready to state our first result . [ th : cexpect ] let @xmath462 be the infinitesimal covariance matrix ( without the singular part ) as defined by @xmath463 with @xmath464 then @xmath253 and @xmath462 are linked through the relation @xmath465 using the analytical formula proved in @xcite for @xmath466,one gets @xmath467 where @xmath468 and @xmath122 is defined such that @xmath469 . in the fourier domain , this last equation corresponds to @xmath470 _ proof of the proposition . _ + this is a pretty straightforward computation : @xmath471 which gives @xmath472 using the stationnarity and dividing by @xmath473 , one gets @xmath474 then , using the definition of @xmath475 @xmath476 which gives . from @xcite , we know that @xmath477 this last equation along with leads to and then to . .3 cm the next result corresponds to the following proposition : [ th : fredholm ] using the notations above , the density @xmath253 satisfies the following fredholm system of integral equations for positive @xmath10 @xmath478 moreover , for @xmath479 we have @xmath480 _ proof of the proposition . _ + * proof of . we consider @xmath320 @xmath481 using , @xmath482 using , we get @xmath483 and consequently @xmath484 however , the vector formulation of @xmath485 is nothing but @xmath486 which is 0 according to , thus @xmath487 which gives . + let us note that we could have derived directly the system from . indeed , gives @xmath488 in the time domain , @xmath489 is supported by @xmath490 , thus going back to the time domain and restricting to @xmath491 directly leads to . * proof of . let @xmath479 . @xmath492 since @xmath448 is stable , @xmath493 is stationary , @xmath494 , thus @xmath495 consequently : @xmath496 the authors thank sylvain delattre , marc hoffmann , charles - albert lehalle and mathieu rosenbaum for useful discussions . we gratefully acknowledge financial support of the chair _ financial risks _ of the _ risk foundation _ and of the chair quantvalley / risk foundation : quantitative management initiative . the financial data used in this paper have been provided by the company _ quanthouse europe / asia _ , http://www.quanthouse.com . , _ modelling financial high frequency data using point processes . _ , in t. mikosch , j - p . kreiss , r. a. davis , and t. g. andersen , editors , handbook of financial time series , springer berlin heidelberg , 2009 . height 2pt depth -1.6pt width 23pt , _ spectra of some self - exciting and mutually exciting point processes _ , journal of the royal statistical society . series b ( methodological ) , 33 - 3 ( 1971 ) , pp . 438443 . , _ market making " behaviour in an order book model and its impact on the bid - ask spread _ , in econophysics of order - driven markets , f. abergel , b.k . chakrabarti , a. chakraborti , and m. mitra , eds . , springer verlag , 2011 .

we apply geometric techniques to obtain the necessary and sufficient conditions on the existence and nonlinear stability of self - gravitating riemann ellipsoids having at least two equal axes .

0712.3081

a riemann ellipsoid is a relative equilibrium for a dynamical model of a rotating self - gravitating fluid mass that remains an ellipsoid at all times , and for which the fluid velocity is a linear function of the coordinates . this model , first studied and formulated by dirichlet , is nowadays known as dirichlet s model . the linearity assumption on the allowed motions makes the study of these deformable bodies very attractive since it implies that their dynamics is governed by a system of ordinary differential equations with a finite number of degrees of freedom . these bodies are also known as affine - rigid bodies or pseudo - rigid bodies . dirichlet s model can be viewed as a first order model for the study of the shape of the earth , and the study of the stability of its solutions can be used in planetary stability research . the study of self - gravitating fluid masses has a long history which can be traced back to newton s times with many contributions from dirichlet , maclaurin , jacobi , dedekind , riemann , liapunov , poincar and cartan , just to name a few . we can distinguish two classical approaches to the study of the stability of riemann ellipsoids . one initiated independently in the latter part of the nineteenth century by poincar @xcite and liapunov @xcite who used expansions in spherical and ellipsoidal harmonics to study the stability of maclaurin ellipsoids and jacobi ellipsoids , not only under dirichlet s assumptions but also under more general conditions ( allowing perturbations not preserving the ellipsoidal shape ) . the other approach occurred in the middle of the twentieth century with the works of chandrasekhar and collaborators who developed the so - called virial method by applying it to the study of the linear stability of riemann ellipsoids . these works are collected in the book of chandrasekhar @xcite which constitutes a comprehensive survey on the subject and related problems with many historical facts on this model . in recent times the subject has had the attention of several researchers , in particular in what respects to the application of new formulations and methods ( @xcite ) to study rotating deformable bodies , not only subject to the self - gravitating potential but also for other potentials modelling nuclei ( see for instance @xcite ) , or elastic bodies ( see @xcite and @xcite ) . our aim is to use geometric methods not only to obtain a complete characterization of the conditions for the existence of riemann ellipsoids having configurations with at least two equal axes but also to obtain the complete description of their nonlinear stability . these geometric methods exploit the geometry and symmetries of the problem and its hamiltonian structure . some works using the same philosophical approach to dirichlet s model are available in the literature for studying several aspects of the problem such as in @xcite which obtains riemann s theorem as a consequence of the symmetry alone , or @xcite where some results on the formal stability and bifurcations of maclaurin spheroids are obtained . we do not address the problem of the stability for riemann ellipsoids with three distinct axes for which the self - gravitating potential is an elliptic integral . in @xcite this has been studied employing numerical analysis techniques . we view dirichlet s model as a hamiltonian system where hamilton s function , @xmath0 , has the form kinetic plus potential energy and is defined on the cotangent bundle , @xmath1 , of the set of all @xmath2 matrices of determinant 1 . furthermore , @xmath0 is invariant for the action of @xmath3 on the phase space . the @xmath4 symmetry reflects the existence of two conserved vector quantities ( by noether s theorem ) : the angular momentum and circulation . the @xmath5 symmetry reflects the reciprocity theorem of dedekind : interchanging the angular velocity and vorticity vectors one obtains another ( physically different ) solution for dirichlet s model with the same geometric configuration . it is well known that noether s conserved quantities are organized as the components of the momentum map . in the case of relative equilibria with configurations having at least two equal axes ( symmetric configurations ) the corresponding momentum map value can be singular . recently , one of the authors developed in @xcite a method appropriated for the study of nonlinear stability of this kind of relative equilibria . this construction extends the so - called reduced energy - momentum method of simo , lewis and marsden @xcite to the case of singular relative equilibria and so we will refer to the method @xcite as the singular reduced energy - momentum method . this is the approach we use in this work to study the nonlinear stability of riemann ellipsoids with symmetric configurations . the preliminary sections of the paper are organized as follow : in section [ sec geometric formulation ] we give the geometric formulation of dirichlet s model , in section [ sec singular reduced ] the singular reduced energy - momentum method is briefly reviewed and in section [ sec potential ] we compute the augmented potential energy for symmetric configurations , as a necessary step towards the stability analysis . the main results of the paper are in the following two sections . in section [ sec existence ] , theorem [ theorem existence ] , we give the complete characterization of all the possible riemann ellipsoids with symmetric configurations . we prove that for dirichlet s model the only relative equilibria with configurations having at least two equal axes are : the spherical configuration which has zero angular momentum and circulation , the _ maclaurin spheroids _ which are oblate spheroids rotating around the ( shortest ) symmetry axis and have angular momentum and circulation aligned with it , and the _ transversal spheroids _ , which have prolate spheroidal configurations that rotate around an axis , say @xmath6 , perpendicular to the ( longest ) symmetry axis and have angular momentum and circulation aligned with @xmath6 . we also prove that there are no symmetric relative equilibria for which the angular velocity and vorticity are not aligned with a principal axis of the relative equilibrium configuration . that is , there are not symmetric configurations which are not of type s in chandrasekhar s terminology . in section [ sec stability ] we apply the singular reduced energy - momentum method to the study of the nonlinear stability of the relative equilibria found in the previous section . the main results of this section are theorems [ spherical stability ] , [ maclaurin stability ] and [ transversal stability ] giving respectively necessary and sufficient conditions for the nonlinear stability of the spherical equilibrium , maclaurin spheroids and transversal spheroids . our results on the eccentricity range for the nonlinear @xmath7-stability of the maclaurin spheroids agree with those already obtained by liapunov and poincar ( see remark [ rem maclaurin ] ) . in the works of these authors there is no reference to the transversal spheroids , however their existence is acknowledged in pag . 143 of chandrasekhar s book @xcite . it is not clear for us what are the results obtained by chandrasekhar with respect to the stability of the transversal spheroids , however his use of the virial method only gives linear stability . to the best of our knowledge , our result on the nonlinear stability of the transversal spheroids is new . in conclusion , this work presents , from a purely geometric point of view , a self - contained and complete study of the nonlinear stability of all symmetric relative equilibria for dirichlet s model . at the same time , the richness of the model helps to clarify the applicability of the singular reduced energy - momentum method , providing also a methodology for its application to other models like . * acknowledgements . * we would like to thank to centro de anlise matemtica , geometria e sistemas dinmicos of the ist , lisbon , for the portuguese translation of original riemann s paper @xcite made by c.e . the work of esd has been supported by the fundao para a cincia e a tecnologia through the program poci 2010/feder . let @xmath8 be a riemannian manifold ( the configuration manifold ) , @xmath9 a lie group that acts by isometries on @xmath10 ( the symmetry group ) and @xmath11 a @xmath9-invariant function ( the potential energy ) . with these ingredients we construct a symmetric hamiltonian system on @xmath12 ( which is a manifold equipped with a natural symplectic structure ) as follows : the potential energy @xmath13 can be lifted to @xmath12 with the pullback of the cotangent bundle projection @xmath14 . we denote this lifted function also by @xmath13 . the riemannian metric on @xmath10 induces an inner product on each cotangent fiber @xmath15 , @xmath16 . then the hamiltonian is defined as @xmath17 the @xmath9-action on @xmath10 induces a cotangent - lifted hamiltonian action on @xmath12 with associated equivariant momentum map @xmath18 defined by @xmath19 where @xmath20 is the fundamental vector field on @xmath10 associated to the generator @xmath21 , defined by @xmath22 this momentum map is @xmath23-equivariant in the sense that @xmath24 for every @xmath25 , @xmath26 . the hamiltonian @xmath0 is @xmath9-invariant for this lifted action ( this follows from the invariance of the metric and of @xmath13 ) . therefore , due to noether s theorem , the components of @xmath27 are conserved quantities for the hamiltonian dynamics associated to @xmath0 . the quadruple @xmath28 is called a _ symmetric simple mechanical system_. let @xmath28 be a simple mechanical system with symmetry . a _ relative equilibrium _ is a point in phase space @xmath29 such that its hamiltonian orbit lies inside a group orbit for the cotangent - lifted action . this amounts to the existence of a generator @xmath30 such that the evolution of @xmath31 is given by @xmath32 . the element @xmath21 is called a _ velocity _ for the relative equilibrium and is defined up to addition of elements in @xmath33 , where @xmath34 is the stabilizer of @xmath31 under the cotangent - lifted action . a useful criterion for finding relative equilibria in simple mechanical systems is given by the following theorem : [ rethm]a point @xmath29 of a symmetric simple mechanical system @xmath28 is a relative equilibrium with velocity @xmath30 if and only if the following conditions are verified : 1 . 2 . @xmath36 is a critical point of @xmath37 , where @xmath38 is defined by @xmath39 . moreover , the momentum @xmath40 of the relative equilibrium is given by @xmath41 note that , in virtue of the above theorem , any relative equilibrium is characterized by a configuration - velocity pair @xmath42 satisfying @xmath43 . the map @xmath44 is called the _ locked inertia tensor _ , while the function @xmath45 is called the _ augmented potential_. we indicate for later use that the kernel of @xmath44 is precisely @xmath46 , the lie algebra of @xmath47 , the stabilizer of @xmath36 for the @xmath9-action on @xmath10 . the knowledge of the pair @xmath48 allows us to compute the stabilizer of the corresponding relative equilibrium @xmath49 with the formula ( see @xcite ) : @xmath50 dirichlet s model is a symmetric simple mechanical system for the motion of a homogenous and incompressible fluid mass of density @xmath51 having as reference configuration the unit ball centered at the origin in @xmath52 and subject to the self - gravitating potential . the only allowed configurations for this model are linear embeddings of the reference ball into @xmath52 preserving volume and orientation . the configuration manifold @xmath10 for a self - gravitating fluid mass under dirichlet s conditions is then @xmath53 , the group of all @xmath2 matrices with determinant equal to 1 , which is equivalent to the space of orientation and volume preserving linear automorphisms of @xmath52 . in what follows we review the geometric formulation of dirichlet s model as a symmetric simple mechanical system on @xmath54 , the group of all @xmath2 matrices with positive determinant , with a symmetric holonomic constraint . the singular value decomposition of any linear automorphism of @xmath52 allows to decompose ( non - uniquely ) any matrix @xmath55 as @xmath56 where @xmath57 , and @xmath58 is a diagonal matrix with positive entries called singular values ( the square roots of the eigenvalues of @xmath59 ) . it follows from this decomposition that the reference unit ball is mapped by @xmath60 into a solid ellipsoid of equation @xmath61 , having principal axes half - lengths equal to the entries of @xmath58 . physically the matrix @xmath62 describes the rigid rotation of the body in space relative to an inertial frame and @xmath63 is related to the rigid internal motion of the fluid with respect to a moving frame . then @xmath58 is an orientation preserving dilatation of the original reference body into an ellipsoid with principal axes aligned with the eigenvectors of @xmath58 . the condition on the volume preservation of the total embedding corresponds to impose the holonomic constraint @xmath64 ( or equivalently @xmath65 ) , which in turn amounts to consider our system as defined on @xmath53 . the tangent space at @xmath66 is isomorphic to @xmath67 , the vector space of @xmath2 matrices . we can define a riemannian metric on @xmath68 as : @xmath69 for @xmath70 , and @xmath71 is a constant depending on the density of the reference body and other physical parameters of the system . in the case of interest here , the reference body is a homogeneous unit ball of constant density @xmath51 , and @xmath71 in is @xmath72 the symmetry group @xmath9 of our model is the semi - direct product @xmath3 , where @xmath73 . several actions of @xmath9 of interest in this paper are : 1 . the @xmath9-action on @xmath9 : if @xmath74 then @xmath75 where for the nontrivial element @xmath76 , @xmath77 , and @xmath4 acts on itself by the direct product of left matrix multiplications . 2 . the adjoint representation of @xmath9 : we identify the lie algebra of @xmath78 , the set @xmath79 of skew - symmetric @xmath2 matrices , with @xmath52 via the usual isomorphism @xmath80 : @xmath81 this is an isomorphism of lie algebras , i.e @xmath82)$ ] is isomorphic by to @xmath83 , where @xmath84 $ ] denotes the commutator of matrices and @xmath85 denotes the vector product of vectors in @xmath52 . + the lie algebra of @xmath9 is then @xmath86 and the adjoint action is given by @xmath87 where @xmath88 and @xmath89 is the rotation of @xmath90 by @xmath91 ( and similarly for @xmath92 ) . + using the standard inner product in @xmath52 we also identify @xmath93 with @xmath94 . under this identification it follows easily that the coadjoint representation has the expression @xmath95 3 . the @xmath9-action on @xmath68 : @xmath96 note that the @xmath5 transposition symmetry on @xmath53 ( first noticed by dedekind ) maps a rigidly rotating configuration without internal motion into one that is stationary in space but with rigid fluid internal motions . that is , for a given ellipsoid there is an adjoint one , obtained by transposition . these adjoint type of ellipsoids are called dedekind ellipsoids . more generally , the transposition symmetry interchanges external rotations and internal motions for any solution of dirichlet s model . any @xmath9-invariant function @xmath97 on @xmath68 can be written as @xmath98 where @xmath99 , and @xmath100 and @xmath101 are the three principal invariants of a @xmath2 matrix , given by @xmath102 with @xmath103 ( also valid interchanging @xmath104 with @xmath59 ) . note that @xmath105 are @xmath9-invariant and that @xmath106 is a regular value of @xmath101 . hence we have @xmath107 as a @xmath9-invariant submanifold of @xmath68 . the condition @xmath108 is the holonomic constraint of the model . note also that the restriction of a @xmath9-invariant function @xmath109 to @xmath110 is given by @xmath111 for @xmath112 . therefore any @xmath9-invariant function on @xmath110 can be written as @xmath113 , with @xmath114 . any @xmath9-invariant function on @xmath53 can be then extended invariantly to @xmath68 by declaring it to be independent of @xmath115 . from now on we will drop the tildes and hats from the corresponding functions unless there is risk of confusion . since @xmath9 acts on @xmath68 by isometries with respect to , the induced metric on @xmath53 ( which we will denote by the same symbol ) is also @xmath9-invariant . for later use we recall that tangent vectors to @xmath53 can be seen as tangent vectors to @xmath68 satisfying the linearization of the constraint @xmath108 . in other words , @xmath116 dirichlet s model is governed by a hamiltonian function of the form kinetic plus potential energy on the phase space @xmath117 given by @xmath118 here @xmath119 is the norm of the covector @xmath120 ( seen as a @xmath2 matrix ) relative to the metric on @xmath53 . the potential energy @xmath13 for a self - gravitating body of homogeneous density @xmath51 under dirichlet s assumptions is given by restricting the function @xmath121 where @xmath122 , @xmath9 is the gravitational constant and @xmath123 the quadruple @xmath124 defines a symmetric simple mechanical system on @xmath53 . the infinitesimal generator for the @xmath9-action on @xmath125 ( and on @xmath110 ) corresponding to @xmath126 is : @xmath127 the vectors @xmath90 and @xmath92 are respectively the angular velocity and vorticity of the fluid motion . we denote the momentum value of @xmath120 by @xmath128 . the components @xmath129 and @xmath130 are respectively the angular momentum and circulation of the instantaneous state @xmath120 ( see for instance @xcite or @xcite ) . a riemann ellipsoid ( a.k.a . an ellipsoidal figure of equilibrium ) is a solution of the hamiltonian system defined by dirichlet s model with angular velocity , vorticity and principal axes lengths all constant . in our setting , riemman ellipsoids correspond exactly to relative equilibria of the underlying symmetric simple mechanical system . therefore , a riemann ellipsoid is represented by a triple @xmath131 , where @xmath132 is the configuration matrix and the lie algebra element @xmath133 is the angular velocity - vorticity pair . in the last years , several are the works studying the stability of relative equilibria of hamiltonian systems by exploiting their symmetry and the geometric properties of their phase space ( see for instance @xcite , @xcite , and @xcite for a overview ) . most of these methods can be used to test the stability of relative equilibria lying in singular level sets of the momentum map , for instance @xcite and @xcite under the hypotesis of @xmath7 compact and @xcite , @xcite using topologic properties . this observation is important since , as we will see , the class of symmetric riemann ellipsoids known as maclaurin spheroids corresponds precisely to non - trivial relative equilibria for dirichlet s model having singular momentum values . we refer the reader to @xcite for a comparison of the applicability of several existing methods . the generally adopted notion of stability for relative equilibria of symmetric hamiltonian systems is that of @xmath7-stability , introduced in @xcite and that we now review in the context of symmetric simple mechanical systems . this notion is closely related to the liapunov stability of the induced hamiltonian system on the reduced phase space . [ patrick ] let @xmath28 be a symmetric simple mechanical system and @xmath29 a relative equilibrium with momentum value @xmath134 . we say that @xmath31 is @xmath7 stable if for every @xmath7-invariant neighborhood @xmath135 of the orbit @xmath136 there exists a neighborhood @xmath137 of @xmath31 such that the hamiltonian evolution of @xmath137 lies in @xmath138 for all time . in the early 1990 s ( @xcite ) a tool known as the reduced energy - momentum method has been developed , providing sufficient conditions for the stability of relative equilibria of a simple mechanical system under the hypothesis that its momentum is a regular value of the momentum map . this method is especially well - suited for simple mechanical systems since in incorporates all of their distinguishing characteristics with respect to general hamiltonian systems . recently , based on the characterization @xcite of the so - called symplectic normal space @xmath139 for a cotangent - lifted action , the reduced energy - momentum method was generalized in @xcite to cover also the case of singular momentum values . in this section we outline the implementation of this singular reduced energy - momentum method following @xcite . our setup will be as in definition [ patrick ] and section 2 . in particular we will fix a relative equilibrium @xmath31 with configuration - velocity pair @xmath48 and momentum @xmath140 . we will also assume that the @xmath9-action on @xmath10 is proper and that there exists a @xmath7-invariant complement to @xmath141 in @xmath142 . these last two conditions are always satisfied for any relative equilibrium in dirichlet s model due to the compactness of @xmath9 . we start by stating some key observations : first , by equivariance of @xmath143 and @xmath14 , one has @xmath144 and @xmath145 . in fact , it is not difficult to prove the characterization @xmath146 we remark that the above formula is not valid in general for covectors @xmath31 other than relative equilibria . second , also by equivariance of @xmath147 together with the bifurcation lemma ( see @xcite ) , if @xmath140 is a singular momentum value then @xmath148 , in which case @xmath149 . third , the properness of the @xmath9-action implies that @xmath34 is compact . this , together with allows to define the following @xmath34-invariant splittings : @xmath150 which by duality induce the splittings @xmath151 here @xmath152 and @xmath153 must be chosen in such a way that @xmath154 . note , from the definition of the locked inertia tensor in theorem [ rethm ] , that @xmath155 , so the restriction @xmath156 is a @xmath34-equivariant isomorphism . then the condition on the above splitting is that @xmath152 and @xmath153 must be orthogonal with respect to the inner product on @xmath157 induced by @xmath158 . we will denote generically the linear projections associated to the splittings and by the letter @xmath159 with an appropriate subindex . for instance @xmath160 or @xmath161 . it is a consequence of noether s theorem that @xmath162 and so we will denote by @xmath163 ( the orthogonal velocity of the relative equilibrium ) . in this work we only need a particular version of the singular reduced energy - momentum method . for , consider the following definitions : * let @xmath164 be the linear orthogonal slice for the @xmath9-action on @xmath10 at @xmath36 , i.e. @xmath165 * define the subspace @xmath166 as @xmath167=0\right\}.\ ] ] * define the space of internal variations @xmath168 * for any @xmath169 , the correction term is the bilinear form on @xmath170 defined by @xmath171\ , , \ , { \widehat{\mathbb{i}}_0}^{-1}\left({\mathbb{p}}_{\mathfrak{p}^*\oplus\mathfrak{t}^*}\left[\left({\mathbf{d}}{\mathbb{i}}\cdot v_2\right ) ( \xi)\right]\right ) \right\rangle.\ ] ] * the arnold form @xmath172 is defined by : @xmath173 \right\rangle.\ ] ] the following theorem ( corollary 6.2 of @xcite ) is the synthesis of the singular reduced energy - momentum method . [ energy miguel ] let @xmath29 be a relative equilibrium with configuration - velocity pair @xmath174 and momentum @xmath175 such that @xmath176 . let @xmath177 be its orthogonal velocity . if the arnold form is non - degenerate at @xmath31 and @xmath178 is positive definite , then the relative equilibrium is @xmath7-stable . when the arnold form is non - degenerate it is also shown in @xcite that the symplectic matrix of the symplectic normal space at @xmath31 has a particularly simple block - diagonal expression . we quote this result which will be essential in the proof of the linear unstability of some riemann ellipsoids . [ symplectic mat ] if the arnold form is non - degenerate at the relative equilibrium @xmath31 with configuration - velocity pair @xmath48 and momentum @xmath140 , the symplectic normal space at @xmath31 , is symplectomorphic to @xmath179 equipped with the symplectic matrix @xmath180 \end{array}\ ] ] where @xmath181 and @xmath182 is the one - form defined by @xmath183 , for all @xmath184 . [ orbitre ] the @xmath9-invariance of @xmath13 and @xmath185 imply the following property . if @xmath48 is a relative equilibrium , then the orbit @xmath186 for every @xmath26 consists of relative equilibria with the same stability or unstability properties . [ augmentations ] the reason why in the previous section we looked at dirichlet s model as a simple mechanical system holonomicaly constrained is that the unconstrained space @xmath68 is an open domain of the vector space @xmath67 , and then the implementation of the reduced energy - momentum method is easier than if one is working directly on @xmath53 . in view of the survey of the method , the strategy will be to use the trivial extension of the self - gravitating potential to @xmath68 and consider its augmented potential with respect to the locked inertia tensor corresponding to the original riemannian metric on @xmath68 . then we further augment this augmented potential with the constraint function @xmath101 and lagrange multiplier @xmath187 . denoting by @xmath188 the resulting twice augmented potential , we have : 1 . relative equilibria for dirichlet s model correspond to triples @xmath131 with @xmath189 such that the following two equations hold : @xmath190 2 . the stability test now follows from the following observation . if we call @xmath191 and @xmath192 the spaces of internal variations for @xmath68 and @xmath53 associated to the triple @xmath131 , according to , we notice that @xmath193 therefore , according to the general method , and standard lagrange multiplier theory , to conclude stability it suffices to study the eigenvalues of the bilinear form @xmath194 where @xmath192 is seen as a vector subspace of @xmath191 . from now on we will omit the superindex @xmath110 for the space of internal variations . in this section we compute the augmented potential @xmath195 in the unconstrained configuration space @xmath68 and collect some results for the self - gravitating potential @xmath13 . the potential @xmath196 at a typical configuration is an elliptic integral except for symmetric configurations ( i.e. with at least two equal singular values ) for which it can be integrated by elementary functions . the extension to the unconstrained space of the potential @xmath13 depends on @xmath55 through the two principal invariants @xmath197 and @xmath198 defined by and respectively . in the study of the existence and stability of relative equilibria of simple mechanical systems will be necessary to compute the first and second derivatives of @xmath199 and the results of this section are essential to this end . in the following we will restrict ourselves to diagonal configurations only . there is no loss of generality with this assumption since , according to the singular value decomposition , every matrix @xmath55 belongs to the @xmath9-orbit of some diagonal configuration @xmath200 by some element @xmath201 . hence by remark [ orbitre ] the qualitative properties of a relative equilibrium @xmath202 are the same as those of @xmath203 . let @xmath204 , with @xmath205 be the following family of integrals : @xmath206 and denote by @xmath207 ( @xmath208 ) the partial derivative of @xmath13 with respect to @xmath209 and by @xmath210 the partial derivative of @xmath207 with respect to @xmath211 for @xmath212 . using , elementary calculus computations give : @xmath213 note that the integrals @xmath204 are all positive as well as @xmath214 and @xmath215 . next proposition gives the values of @xmath204 in the case of spheroidal ( two equal axes ) configurations . [ integrals]let @xmath216 be a spheroidal configuration with @xmath217 . * the integral @xmath204 for the oblate spheroid @xmath60 ( @xmath218 ) and eccentricity @xmath219 is given by @xmath220 * the integral @xmath204 for the prolate spheroid @xmath60 ( @xmath221 ) and eccentricity @xmath222 is given by @xmath223 note that for the diagonal configuration @xmath224 the value of @xmath225 in definition is @xmath226^{1/2}.\ ] ] for ( i ) : making the change of variables @xmath227 we get @xmath228 since the eccentricity is @xmath229 then @xmath230 and @xmath231 because @xmath217 . then , from the above expression for @xmath232 one gets @xmath233 making @xmath234 the result follows . for ( ii ) : the change of variables @xmath235 gives @xmath236 the eccentricity @xmath237 of the prolate spheroid is such that @xmath238 . as @xmath217 then @xmath239 and @xmath240 and the result follows for @xmath234 . as stated in the previous section , in order to find critical points of a @xmath9-invariant function defined in @xmath110 we will work with its extension to @xmath68 subject to the constraint @xmath64 . for , since any such function can be written as @xmath241 , in order to compute the critical points we use the augmented function @xmath242 subject to the condition @xmath243 . for the differentiation of @xmath244 consider the pairing between vectors @xmath245 and covectors @xmath246 to be @xmath247 . then using the chain rule we get that critical points must verify the following set of equations : @xmath248\\ \nonumber & = & 0.\\ \nonumber \det(f ) & = & 1\end{aligned}\ ] ] since @xmath249\\ \label{traceformula3}{\mathbf{d}}\det ( f)\cdot\delta f & = & \det ( f)\mathrm{tr}(f^{-1}\delta f ) .\end{aligned}\ ] ] ( see for instance @xcite or @xcite ) . next proposition gives the expression for the locked inertia tensor . [ lockedprop]the locked inertia tensor for the @xmath9-action on @xmath250 , at a configuration @xmath55 , is defined by @xmath251\ ] ] where @xmath252 and @xmath253 for @xmath208 . under the isomorphism , the locked inertia tensor is also equivalent to : @xmath254 where @xmath103 , @xmath59 and @xmath255 ( @xmath256 denotes the identity matrix ) . by the locked inertia tensor definition in proposition [ rethm ] , the expression for the infinitesimal generators of the @xmath9-action on @xmath68 and the definition for the riemannian metric , we have @xmath257.\end{aligned}\ ] ] using the fact that @xmath258 and @xmath259 are skew - symmetric matrices and the cyclic property of the trace of a matrix , it is straightforward to obtain expression . for the expression we need some standard properties of the isomorphism . in particular , @xmath260 where @xmath261 denotes the standard inner product on @xmath52 . let us compute some terms of the expression since the other are done similarly @xmath262&\quad\text{(by \eqref{eqisosecond})}\\ & = \frac{1}{2 } { \operatorname{tr}\,}\left(\widehat{\xi_l}^t \widehat{\mathbf{i}_s\eta_l}\right)&\quad\text{(by \eqref{eqisothird})}\\ & = \xi_l\cdot \mathbf{i}_s\eta_l&\quad\text{(by \eqref{eqisofirst})}\\\end{aligned}\ ] ] @xmath263&\quad\text{(by \eqref{eqisofour})}\\ & = 2\det ( f ) \xi_r\cdot f^{-1 } \eta_l & \quad\text{(by \eqref{eqisofirst})}\\\end{aligned}\ ] ] as a straightforward consequence we can obtain the momentum of a relative equilibrium for dirichlet s model , that is its angular momentum and circulation . [ momcirc ] the momentum for a relative equilibrium with configuration @xmath60 and velocity - vorticity pair @xmath264 is @xmath265 that is , the angular momentum and circulation of a riemann ellipsoid with configuration by @xmath60 , and angular velocity - vorticity pair @xmath266 are given , respectively , by @xmath267 the expression for the twice augmented potential @xmath268 follows now easily from proposition [ lockedprop ] . in this section we classify symmetric relative equilibria for dirichlet s model . we will treat the spherical case ( i.e. a configuration having three equal principal axes ) as a particular case of a symmetric configuration . from the singular value decomposition and the definition of the action of @xmath270 on @xmath125 ( or on @xmath271 ) it follows that the stabilizer of a symmetric configuration @xmath60 is conjugate to the stabilizer of a diagonal configuration . that is , conjugate to @xmath272 or @xmath273 if @xmath60 has 2 or 1 different singular values , respectively ( see @xcite for a derivation of this result ) . actually , if the configurations are diagonal , these are exactly their stabilizers . here , @xmath274 denotes the diagonal embedding of @xmath275 in @xmath4 and @xmath276 is the subgroup of @xmath277 generated by all the rotations @xmath278 around a given axis @xmath279 in @xmath52 and a rotation , @xmath280 , by @xmath281 around an axis @xmath282 perpendicular to @xmath279 . in case of the diagonal configuration @xmath283 , @xmath284 is the rotation matrix by and angle @xmath285 around @xmath286 and @xmath280 can be chosen to be @xmath287 . we introduce the following subgroups : * @xmath288 , generated by elements @xmath289 , with @xmath290 and @xmath291 , * @xmath292 , generated by elements @xmath293 , with @xmath294 and @xmath291 , * @xmath295 , the cyclic group isomorphic to @xmath5 generated by the element @xmath296 . * more generally , if @xmath297 is a subgroup of @xmath4 , we denote also by @xmath297 the subgroup of @xmath298 generated by elements @xmath299 , with @xmath300 . note that since we are going to impose the constraint @xmath301 , we will consider only two kinds of symmetric configurations , especifically : * spherical : @xmath302 , * spheroidal : @xmath303 , with @xmath217 . to find all the possible riemann ellipsoids with symmetric configurations , we will have to solve with @xmath60 of the above forms and different pairs @xmath266 . the possible solutions are summarized in the following theorem . [ theorem existence ] the relative equilibria for dirichlet s model of a self - gravitating fluid mass are : * the * spherical equilibrium * with spherical configuration @xmath304 , velocity - vorticity pair @xmath305 and lagrange multiplier @xmath306 . its corresponding momentum and isotropy groups are @xmath307 * the family of * maclaurin spheroids * which have oblate spheroidal configurations @xmath283 ( with @xmath308 ) and angular velocity and vorticity parallel to the axis of symmetry @xmath309 . in terms of the parameter @xmath310 defined by @xmath311 , this family is characterized by @xmath312 and the following constraint between @xmath310 and the eccentricity @xmath237 : @xmath313 its corresponding momentum and isotropy groups are : @xmath314 @xmath315 * two branches of * transversal spheroids * which have prolate spheroidal configurations @xmath283 ( with @xmath316 ) . we distinguish the two branches of this family with the signs @xmath317 and @xmath318 . these branches are characterized by the lagrange multiplier @xmath319 , the velocity - vorticity pair @xmath320 with @xmath321 ( where @xmath6 is a unit vector perpendicular to @xmath322 ) and the following constraints between @xmath323 and the eccentricity @xmath237 : @xmath324 the corresponding momentum and isotropy groups are : @xmath325 @xmath326 before proving the theorem we remark that formula has already been obtained by maclaurin in 1742 , as it is claimed in page 4 of chandrasekhar s book @xcite . first , using , and , it is easy to see that the general conditions are equivalent to @xmath327\\ \label{recond2 } 1 & = & \det(f)\end{aligned}\ ] ] [ [ spherical - case ] ] spherical case : + + + + + + + + + + + + + + + if @xmath328 , then , are simply @xmath329\right ) \mathbf{i}+t\left[\xi_l\otimes\xi_l+\xi_r\otimes\xi_r-2\xi_l\otimes\xi_r\right]=0\ ] ] taking @xmath330 and the same sorte of notation for @xmath92 , the off - diagonal terms of this expression are independent of @xmath331 and @xmath187 , and equivalent to the following 6 equations : @xmath332 it follows then that @xmath333 . recall from theorem [ rethm ] that the momentum of a relative equilibrium with configuration @xmath36 and velocity @xmath21 is given by @xmath334 . then , from we have @xmath335 . therefore , @xmath336 , @xmath337 , @xmath338 . now , noting that @xmath339 if @xmath333 , and that the velocity of a relative equilibrium is defined only up to addition of elements in @xmath340 , the relative equilibrium @xmath341 is the same as @xmath342 . then the relative equilibrium conditions are satisfied with @xmath306 . [ [ spheroidal - case ] ] spheroidal case : + + + + + + + + + + + + + + + + we now consider @xmath303 with @xmath217 . since in the @xmath343 plane all directions are equivalent , we can assume without loss of generality that @xmath344 . now , conditions , are equivalent to the system @xmath345 note that these equations imply that @xmath346 . indeed , if @xmath347 then implies @xmath348 and so and imply @xmath346 which is a contradiction . we will now proceed systematically considering four main cases : ( i ) @xmath349 , ( ii ) @xmath350 , ( iii ) @xmath351 and @xmath352 , and ( iv ) @xmath353 and @xmath352 . \(i ) if @xmath349 , then , and are the only non - trivial conditions , which are equivalent to @xmath354 as @xmath214 and @xmath215 are positive and @xmath355 it follows from the last equation that there is no solution . \(ii ) if @xmath350 , then from and we have @xmath356 and so and give @xmath357 . let @xmath358 . then the remaining non - trivial equations , , and , give @xmath359 as @xmath214 and @xmath215 are positive , then last equality implies that the spheroidal configuration @xmath60 is oblate , that is @xmath218 . the eccentricity of the spheroid is @xmath360 and @xmath361 . using the relations for the partial derivatives of the self - gravitating potential , @xmath214 and @xmath215 , the expression @xmath362 for the integrals @xmath204 in the oblate case given by and @xmath363 , then the equation is equivalent to @xmath364 one can easily compute the definite integrals @xmath365 and @xmath366 , although we avoid to display their expressions since they are quite lengthy . however the expression @xmath367 is @xmath368 from which follows . from corollary [ momcirc ] it is trivial to obtain that the momentum of this relative equilibrium . hence , using , the appropriate isotropy groups are also straightforward . \(iii ) if @xmath351 and @xmath369 , it follows from and that @xmath370 . then and give @xmath371 we can set @xmath372 and @xmath373 . so , substituting the above value of @xmath187 into and these equations are @xmath374 from we obtain the solutions @xmath375 , from which follows that the spheroids must be prolate @xmath376 . in terms of the eccentricity @xmath377 we have @xmath378 therefore , the equation gives @xmath379 substituting in this expression @xmath380 and @xmath381 , as well as @xmath382 gives . as before , using the expression of the locked inertia tensor and the remaining results follow . \(iv ) in this case we have @xmath369 , @xmath353 and @xmath383 . note that from and one should also have @xmath384 and @xmath385 . so , let @xmath386 and @xmath387 for some reals @xmath388 . then , using also , the equations and reduce to @xmath389 these equations have solutions @xmath390 comparing and using @xmath386 , @xmath387 and we obtain @xmath391 and so @xmath392 must be @xmath393 or @xmath394 , but this is a contradiction with . so there is no solution for the above system . [ transversal adjoint ] note that the functions @xmath395 and @xmath396 appearing in the characterization of transversal spheroids are inverse of each other . therefore the two families of transversal spheroids belong to a single orbit of the symmetry group . indeed , the @xmath5 symmetry interchanges the @xmath317 and @xmath318 families , since @xmath397 , as it follows from their definitions that @xmath398 . theorem [ theorem existence ] is in agreement with riemann s classification of ellipsoidal figures of equilibrium for dirichlet s model of self - gravitating fluid masses . that is , these ellipsoidal figures of equilibrium must lie in one of the following categories : ( a ) the case of a uniform rotation with no internal motion ( or uniform vorticity and no rotation ) ; ( b ) the case when the directions of the angular velocity @xmath90 and vorticity @xmath92 are the same and coincide with a principal axis of the ellipsoid ( also known as ellipsoids of type s ) ; ( c ) the case when the angular velocity and vorticity are not parallel but lie in the same principal plane . in particular , we show that for dirichlet s model it is not possible to obtain relative equilibria with spheroidal configurations belonging to category ( c ) . the existence of transversal spheroids is referred in chandrasekhar s book @xcite ( see for instance page 143 ) , however their study is not present in the classical works of liapunov @xcite and poincar @xcite . in this section we apply the singular version of the reduced energy - momentum method introduced in @xcite in order to deduce the stability of the symmetric relative equilibria obtained in theorem [ theorem existence ] . in order to apply this method it is essential to compute the second derivative of the twice augmented potential @xmath399 . next lemma gives that result . [ second derivative augmented ] if @xmath60 is a critical point of the twice augmented potential @xmath400 for @xmath401 and @xmath402 as in proposition [ lockedprop ] and @xmath403 , then @xmath404 where @xmath405 and @xmath406 then , applying standard properties of the trace , we get @xmath413.\ ] ] the expression for @xmath411 stated follows now easily . finally , using to differenciate the expression @xmath414 we get @xmath415.\end{aligned}\ ] ] we now study the stability of the spherical equilibrium . notice from the outline of the method in section 3 that for this equilibrium whe have that @xmath416 , the correction term and the arnold form are all trivial , as well as the velocity - vorticity pair @xmath266 . as a consequence , @xmath417 , the orthogonal complement to the @xmath9-orbit at the identity in @xmath110 . hence , to conclude stability of the spherical equilibrium we need to study the definiteness of @xmath418 recall that @xmath419 is the space of traceless matrices . also , the infinitesimal action of @xmath142 on @xmath68 at @xmath256 is @xmath420 . then @xmath421 therefore @xmath422 is the space of traceless symmetric matrices . that is , matrices of the form @xmath423 we fix a basis for @xmath422 with respect to which the components of @xmath58 are @xmath424 . by lemma [ second derivative augmented ] , the expression of @xmath425 , with @xmath426 as given in theorem [ theorem existence ] , reduces to : @xmath427 where we have applied the fact that @xmath58 and @xmath428 are traceless symmetric matrices . therefore in this basis we have @xmath429 as @xmath430 is positive , the eigenvalues of this matrix are @xmath431 and @xmath432 with multiplicities @xmath433 and @xmath434 respectively . these are all positive , therefore the spherical equilibrium is @xmath9-stable . we now study the nonlinear stability of maclaurin spheroids in the setup of previous sections ( theorem [ theorem existence ] ) . as it has been stated , a maclaurin spheroid has an oblate configuration which , with no loss of generality , we suppose diagonal . this configuration is uniquely characterized by the eccentricity @xmath435 . in order to apply theorem [ energy miguel ] one needs first to split @xmath436 according to , that is as @xmath437 recall that for the maclaurin spheroid one has @xmath438 and @xmath439 . one can then choose the following ordered orthonormal bases ( with respect to the euclidean product in @xmath440 ) for each of the spaces of the above splitting : if we define @xmath441 , then @xmath442 it is straightforward to check that these subspaces are invariant for @xmath443 . the orthogonal velocity @xmath444 for the maclaurin relative equilibrium is the orthogonal projection of the velocity @xmath21 onto @xmath152 . then , @xmath445 where @xmath310 must satisfy . as already defined , @xmath446 is the restriction of @xmath44 to @xmath157 . the locked inertia matrix for the configuration @xmath303 is , according to , @xmath447 where @xmath448 and @xmath449 . it is now straighforward to obtain the @xmath446 matrix with respect to the basis @xmath450 . that is @xmath451 or , in terms of the eccentricity @xmath452{1-e^2 } } & 0 & 0 & 0 & 0\\ 0 & \frac{2-e^2}{\sqrt[3]{1-e^2 } } & -2\sqrt[6]{1-e^2 } & 0 & 0\\ 0 & -2\sqrt[6]{1-e^2 } & \frac{2-e^2}{\sqrt[3]{1-e^2 } } & 0 & 0 \\ 0 & 0 & 0 & \frac{2-e^2}{\sqrt[3]{1-e^2 } } & -2\sqrt[6]{1-e^2 } \\ 0 & 0 & 0 & -2\sqrt[6]{1-e^2 } & \frac{2-e^2}{\sqrt[3]{1-e^2 } } \end{bmatrix}\ ] ] we can use @xmath158 and @xmath453 to compute the momentum of a maclaurin spheroid . indeed @xmath454 which is of course the same as the value obtained in theorem [ theorem existence ] under the identification @xmath455 induced by the euclidean product in @xmath440 . first , we compute the space @xmath416 . notice the following relations for the adjoint representation of @xmath9 : @xmath459 and @xmath460 also , recall that under our identification @xmath455 we have @xmath461 , for @xmath462 , and where @xmath51 in the right hand side is identified with an element of @xmath142 . therefore @xmath463 for @xmath464 , hence @xmath456 . as @xmath465 then @xmath466 . then , from we obtain @xmath467 the inverse of the matrix is not difficult to compute . here we just state the values of @xmath468 on vectors @xmath469 of the fixed basis : @xmath470 from these expressions it follows easily that @xmath471 finally , the entries of the arnold matrix are given by @xmath472 using the arnold matrix is then given by @xmath473 for ( i ) : as the arnold form is non - degenerate , @xmath7-stability will follow whenever @xmath478 is positive definite . for , recall from that @xmath479 where @xmath422 is the linear slice for the @xmath9-action on @xmath53 at the oblate configuration , @xmath303 . matrices @xmath480 must verify @xmath481,\quad\forall \xi,\eta\in{\mathbb{r}}^3,\quad\text{and}\\ 0 & = & { \operatorname{tr}\,}(f^{-1}a)\end{aligned}\ ] ] because @xmath58 must belong , respectively to the orthogonal complement to the tangent space to the group orbit through @xmath60 and @xmath482 . these two conditions give that @xmath58 must be of the form @xmath483 therefore we can describe @xmath422 as the set of matrices of the form @xmath484 with @xmath485 . let the vector @xmath486 be @xmath487 with respect to the basis @xmath488 . therefore , using , we have @xmath489 the set @xmath490 is precisely the set of matrices @xmath491 for which @xmath492 where @xmath493 denotes the annihilator of @xmath494 . using @xmath495 and @xmath496 and differentiating , the computation of @xmath497 gives , in terms of the eccentricity @xmath237 : @xmath498\\ \langle t_2,\left({\mathbf{d}}{\mathbb{i}}(f ) \cdot \left(\gamma_{\mathrm{sl}(3)}(f)+a\right)\right ) ( \xi_l,\xi_r)^\perp\rangle&= \frac{t\omega}{2 ( 1-e^2)^{1/3 } } \left[2\sqrt{1-e^2 } \gamma_3- ( 2+e^2 ) \gamma_4\right]\\ \langle t_3,\left({\mathbf{d}}{\mathbb{i}}(f ) \cdot \left(\gamma_{\mathrm{sl}(3)}(f)+a\right)\right ) ( \xi_l,\xi_r)^\perp\rangle&= \frac{t\omega}{2 ( 1-e^2)^{1/3 } } \left[-(2+e^2 ) \gamma_1 + 2\sqrt{1-e^2 } \gamma_2\right]\\ \langle t_4,\left({\mathbf{d}}{\mathbb{i}}(f ) \cdot \left(\gamma_{\mathrm{sl}(3)}(f)+a\right)\right ) ( \xi_l,\xi_r)^\perp\rangle&= \frac{t\omega}{2 ( 1-e^2)^{1/3 } } \left[-2\sqrt{1-e^2 } \gamma_1 + ( 2+e^2 ) \gamma_2\right]\end{aligned}\ ] ] it follows from the above expressions that @xmath499 if and only if @xmath500 . this is equivalent to @xmath501 . let us now compute the correction term restricted to @xmath501 . for any @xmath480 one has @xmath502 and so @xmath503 = \left(\frac{4t\omega \sqrt{2}}{(1-e^2)^{1/6 } } a_1,0,0,0,0\right).\ ] ] from the expression of @xmath504 it is straightforward to obtain @xmath505\right)= \sqrt{2 } \omega ( 1-e^2)^{1/6 } b_1 p,\ ] ] where @xmath506 is the entry of @xmath507 playing the same role of @xmath508 in @xmath58 . then , from @xmath509 the plots of @xmath521 and @xmath522 are shown in figure [ plots maclaurin ] . they show that @xmath521 is always positive in @xmath523 while @xmath522 has a root @xmath524 being positive for @xmath475 and negative for @xmath477 . we used the mathematica programing system for the numerical computation of @xmath525 to obtain @xmath526 . therefore maclaurin spheroids are @xmath7-stable for @xmath475 . for ( ii ) : when @xmath527 we study the spectrum of the linearized hamiltonian vector field @xmath528 , where in our case @xmath139 and @xmath529 are as in theorem [ symplectic mat ] also , from proposition 6.2 in @xcite we have @xmath530 let us start by computing the block @xmath531 of the above matrix , where @xmath531 is the inner product on the dual of @xmath422 induced by the riemannian metric on @xmath422 . for , let as before fix the ordered basis @xmath532 on the slice @xmath422 where @xmath533 recalling that @xmath534 then the matrix that represents @xmath535 in the fixed basis is @xmath536 let @xmath537 be the dual basis of @xmath532 under the identification of @xmath538 with @xmath539 using the inner product @xmath540 . in this basis the induced inner product @xmath531 is represented by @xmath541 . recall that @xmath456 for a maclaurin spheroid . then , from theorem [ symplectic mat ] , for @xmath546 we have : @xmath547 using , the matrix @xmath548 is given by @xmath549 since for a maclaurin spheroid @xmath501 it is immediate from the definition of @xmath550 in theorem [ symplectic mat ] that @xmath550 is the zero matrix . start by considering for @xmath554 the corresponding left - invariant extensions @xmath555 . we have @xmath556 for every @xmath55 . recall that @xmath557 . then according to the definition given in theorem [ symplectic mat ] we have @xmath558 also , since @xmath560 are left - invariant vector fields , the identity @xmath561=x_{uv - vu}$ ] holds and we have , from @xmath562)(f)= -\frac{\omega t}{2}{\operatorname{tr}\,}\left(\widehat{\mathbf{e_3}}(f^tf(uv - vu)+f(uv - vu)f^t)\right).\ ] ] in order to compute @xmath551 let @xmath283 and @xmath563 . the unique left - invariant vector fields extending @xmath58 and @xmath428 are @xmath564 and @xmath565 respectively . then , using together with and it is immediate to obtain @xmath566 . therefore , from theorem [ symplectic mat ] the symplectic matrix @xmath567 and its inverse are @xmath568 so the linearized vector field is @xmath569 where in our basis @xmath570 is given in . as @xmath576 and @xmath521 are positive for @xmath577 , then @xmath578 is always imaginary in that range . however , @xmath579 becomes real if @xmath522 becomes negative , which happens exactly at @xmath525 as seen from the previous stability analysis . hence , if @xmath477 the maclaurin spheroid becomes linearly unstable , therefore unstable . this loss of stability corresponds exactly to a collision at 0 of @xmath579 for @xmath580 , which passes from being pure imaginary to be real . in riemann s work @xcite some conclusions were made concerning the stability of maclaurin spheroids by studying the existence of a minimum of a certain function @xmath9 . the existence of this minimum was not done by studying the second variation of @xmath9 . riemann even says in page 188 : `` the direct analysis of the second variation of @xmath9 when the first variation vanishes would be very complicated ; we can however decide if this function has a minimum by the following form : ... '' . he follows with the analysis of the behavior of @xmath9 . his final conclusion on the stability ends the paragraph 9 of his paper and is the following : from this study it follows that the case of a rotation of an oblate ellipsoid , around its shortest axis , case already known to maclaurin , can only be unstable if the relation between the shortest axis with the others is less than 0.303327 ... " . we note that if the relation between the shortest axis of the maclaurin spheroid and the others is @xmath581 this is equivalent to say that the eccentricity @xmath582 . the value @xmath583 obtained by riemann follows from his study on the existence of oblate spheroids in pages 184185 of @xcite , namely as the root of the last displayed equation in page 184 . we remark that this equation is equivalent to the equation @xmath584 where @xmath522 is as in the proof of ( i ) in theorem [ maclaurin stability ] . indeed @xmath585 taking @xmath586 and @xmath587 one has @xmath588 the equation is the same one appearing in riemann s paper . with respect to the computation of the root of this equation riemann just says : `` ... this equation has , for @xmath589 between @xmath590 and @xmath591 , the unique root @xmath592 ... '' . indeed , this is equivalent to say that @xmath593 \2 . concerning the stability of maclaurin spheroids , it is shown in liapunov s work @xcite that under the hypothesis of the preservation of the ellipsoidal shape ( setup we used ) the value @xmath525 of loss of stability is the same as in riemann s work , but if this hypothesis is dropped then the maclaurin spheroid is only stable for @xmath594 with @xmath595 . the point @xmath596 is exactly the point where the family of maclaurin spheroids bifurcates into a branch of ellipsoids with 3 distinct axes lengths ( jacobi ellipsoids ) . we refer the reader to pages 52 and 61 - 63 of @xcite . all the qualitative properties , including stability , of two relative equilibria lying in the same orbit of the symmetry group are the same . therefore , in view of remark [ transversal adjoint ] , in this subsection we analyze the nonlinear stability of the @xmath317 family of transversal spheroids and the main result , theorem [ transversal stability ] , will follow for both families . we will set @xmath597 , @xmath598 and @xmath599 for notational simplicity . also , to keep the notation consistent with the proof of theorem [ theorem existence ] , we will set @xmath372 . in this case the computation of the splitting is simplified due to the fact that @xmath600 and therefore @xmath601 . introducing the vectors @xmath602 , we choose @xmath603 these subspaces are obviously invariant under the action of @xmath604 . with respect to the basis @xmath605 for @xmath157 we have , @xmath606.\ ] ] or in terms of the eccentricity of a prolate spheroid , @xmath607.\ ] ] it follows immediately that @xmath152 and @xmath153 are orthogonal with respect to @xmath158 , so our choice of the splitting @xmath608 is correct . in this basis the orthogonal velocity is @xmath609 with @xmath610 and @xmath97 as in the @xmath317 family in theorem [ theorem existence ] . recall that the momentum of a transversal spheroid is @xmath612 where @xmath613 $ ] and @xmath614 $ ] . from lemma [ transadjoint ] , and recalling that @xmath461 , we have , for @xmath615 : @xmath616 this is zero iff @xmath617 , with @xmath618 and therefore @xmath619 . a basis for @xmath416 is given by @xmath620 . now , proceeding as for the maclaurin spheroid , we compute @xmath621 @xmath622 differentiating we find @xmath630 in order for @xmath631 all the above expressions must vanish , which happens if and only if @xmath632 and @xmath633 , with @xmath634 . therefore we can chose a basis for the space of internal variations @xmath625 such that any element @xmath635 belonging to it has components @xmath636 with the parametrization @xmath637.\ ] ] it is straightforward to obtain @xmath638 then the correction term , for @xmath639 , is given by : @xmath640 for the restriction of the hessian of @xmath641 at @xmath60 we compute first the following second variations putting all the contributions together , and substituting the integral expressions for the derivatives of the potential and the values of @xmath187 and @xmath610 given in theorem [ theorem existence ] , we find that the matrix representing @xmath646 in the fixed basis of @xmath625 is block diagonal , with a @xmath647 symmetric block in the first two components , and the other block given by the coefficient of @xmath648 , say @xmath649 , which has the following expression : @xmath650.\ ] ] it is clear that @xmath649 is positive since @xmath651 is positive in @xmath523 and @xmath214 and @xmath215 are always positive . the matrix @xmath138 has real eigenvalues . the trace and the determinant of @xmath138 are : @xmath654\ ] ] and @xmath655.\end{aligned}\ ] ] these two quantities are positive in the interval @xmath656 . their plots are displayed in figure [ prolateplots ] . therefore , both families of transversal spheroids are stable for all eccentricities .

by means of space - group symmetry arguments , we argue that the electronic pairing in iron - based high temperature superconductors shows a structure which is a linear combination of _ planar _ s - wave and d - wave symmetry channels , both preserving the 3-dimensional irreducible representation of the corresponding crystal point - group . we demonstrate that the s- and d - wave channels are determined by the parity under reflection of the electronic orbitals through the iron planes , and by improper rotations around the iron sites . we provide evidence of these general properties by performing accurate quantum monte carlo ab - initio calculations of the pairing function , for a fese lattice with tetragonal experimental geometry at ambient pressure . we find that this picture survives even in the fese under pressure and at low temperatures , when the tetragonal point - group symmetry is slightly broken . in order to achieve a higher resolution in momentum space we introduce a bcs model that faithfully describes our qmc variational pairing function on the simulated 4x4 fese unit cell . this allows us to provide a k - resolved image of the pairing function , and show that non - isotropic contributions in the bcs gap function are related to the improper s - wave symmetry . our theory can rationalize and explain a series of contradictory experimental findings , such as the observation of twofold symmetry in the fese superconducting phase , the anomalous drop of with co - impurity in lafeasof , the-to--wave gap transition in bafeas under k doping , and the nodes appearing in the lifeas superconducting gap upon p isovalent substitution

1302.4748

the pairing symmetry of the superconducting state in iron - based superconductors ( ibs ) has been one of the most debated subjects since their first discovery , in both theory and experiments.@xcite its determination is particularly challenging in the ibs for their complex electronic structure , with a strong multiband character , and a fermi surface constituted by many sheets , which can vary with doping and chemical composition . moreover , the ibs families are usually compensated metals , and thus both electrons and holes are involved in the paired state . weak - coupling rpa approaches , coupled to multiband bcs theory , yield a pairing function with global s - wave ( @xmath0 ) symmetry , but with electron and holes pockets having opposite sign . this scenario , dubbed @xmath7 , was first proposed by mazin _ _ et al.__@xcite complemented later with its variants , called `` extended @xmath7'',@xcite in the case that accidental nodes appear on a single fermi sheet without breaking the full rotational symmetry . the latest generalizations include also `` weak '' nodal lines which develop as closed loops on the 3-dimensional ( 3d ) fermi surface.@xcite a variety of experiments has been performed to probe the pairing symmetry of ibs , ranging from thermal conductivity and specific heat measurements to angle - resolved photoemission spectroscopy . while there is no doubt on the spin singlet nature of the pairing state as revealed by the knight shift,@xcite the presence of nodes and the total symmetry of the spatial part of the pairing function are controversial . in fact , the experimental outcome seems to lack universality , as it depends crucially on the `` family '' of tested compound , its doping , its isovalent substitutions , and its level of disorder . for instance , the situation is paradoxical for the `` 111 '' family , where the lifeas material shows a fully gapped superconducting order parameter , while the isovalent substitution of arsenic by phosphorus leads to thermodynamic properties compatible with a nodal pairing function.@xcite for the `` 122 '' family , there is a recent claim , supported by independent experimental probes , that the pairing in the bafe@xmath6as@xmath6 undergoes an @xmath4-to-@xmath5 symmetry change by doping with potassium,@xcite the fully substituted kfe@xmath6as@xmath6 being identified as a @xmath5-wave superconductor.@xcite in the `` 1111 '' family , the dependence of the critical temperature @xmath1 upon disorder has been studied in the lafeaso@xmath2f@xmath3 with @xmath8 . it has been found that co - doping induced disorder makes @xmath1 to fall much more slowly than mn - doping.@xcite this is certainly not compatible with a doping independent pairing function with sign changes ( either @xmath7 or @xmath5-wave ) . finally , specific heat measurements performed on the fese revealed a highly anisotropic order parameter in the `` 11 '' family,@xcite with a twofold symmetry directly observed by scanning tunneling microscopy ( stm).@xcite in this paper , we study the ibs gap structure from a different perspective , namely by looking for its _ universal _ features based on symmetry constraints induced by the 3d _ space - group _ transformations of the crystals . we prove that the ibs pairing function is a linear combination of terms having planar s- and d - wave symmetries , both fulfilling the full 3d @xmath0 representation . these properties are then verified in the fese by performing state - of - the - art quantum monte carlo ( qmc ) calculations from first - principles . our theory provides a general framework to account for the contradictory experimental outcomes . the paper is organized as it follows . in sec . [ improper_symmetry ] we present the derivation of the 2d iron lattice model consistent with the symmetries of the 3d point group of the fese . we show that it is necessary to include improper rotations and a gauge transformation for translations in order to define a 2d square lattice model consistent with the 3d structure . we study what are the implications of the symmetry constraints for the pairing structure , and show that a d@xmath9-wave channel is present beside the extended s - wave one . these predictions are verified by accurate quantum monte carlo simulations , presented in sec . [ qmc_section ] , to determine the global symmetry of the pairing function from first - principles . in sec . [ bcs_modeling ] we model the ab - initio pairing function by means of a bcs hamiltonian . this allows us to compute the bcs gap on a dense k - point grid , and study its nodal structure in the k - space . in sec . [ physical_consequences ] we relate our findings to the experimental outcome in fese , and to other experiments in different families of iron pnictides and chalcogenides . the undistorted structure of fese , together with the other parent compounds of the `` 111 '' and `` 1111 '' families , belongs to the @xmath11 space - group , which is nonsymmorphic due to a glide plane parallel to the iron layers . indeed , its related @xmath12 point - group classes contain certain operations with the nonprimitive lattice translation @xmath13 - expressed in crystal units of the 2-fe unit cell - which brings a fe site into the other in the cell . this is due to the se sites alternating above and below the fe layer ( see fig.[fordummies ] ) . fe@xmath6 ) are below ( lower triangles ) or above ( upper triangles ) the iron layer . it is evident that a symmetry operation like rotation of 90 degrees around an iron atom , interchanges lower triangles with upper ones . therefore a further reflection @xmath14 in the iron plane has to be combined with the 90 degree rotation to define a correct symmetry of the hamiltonian . in a low energy model where se and as atomic centers do not appear , their effect is to define a non trivial spatial symmetry in the iron square lattice where orbital and spatial degrees of freedom are strongly coupled in an unusual way . as noted in ref . , also translation symmetry is not defined , but the composition of translation and reflection @xmath14 is again a well defined symmetry of the model . then , the fictitious broken translation symmetry of the low energy model can be easily gauged out ( see text and ref . ) . ] in particular , the @xmath15 subgroup generated by pure rotations by @xmath16 about the principal axis is not compatible with the structure of the solid . on the other hand , the @xmath10 subgroup , defined with the rotation by @xmath16 about the principal axis combined with the reflection @xmath14 through the fe plane ( @xmath17 ) , is a a pure point - group operation allowed by the @xmath11 space - group , as well as the reflection through the plane perpendicular to the fe ( xy ) plane and bisecting the secondary ( x and y ) axes ( the @xmath18 element).@xcite this has important consequences on the symmetries of the subgroup associated with the 2-dimensional ( 2d ) fe square sublattice , and thus on the low - energy features of the system , as for instance , the superconducting gap structure . indeed , the standard @xmath19 group of the 2d square lattice , generated by the elementary @xmath18 and @xmath15 point - group operations , does not fulfill all symmetries of the @xmath11 space - group , whereas the group generated by @xmath18 and @xmath10 does . moreover , proper ( @xmath15 ) and improper ( @xmath10 ) rotations differ from the physical point of view , because a hamiltonian or a pairing function fulfilling the point - group operations generated by @xmath10 and @xmath18 will transform in a very different way with respect to those based on the standard @xmath19 group , as it will be shown in the following . from ab - initio electronic structure dft and dmft calculations of ibs,@xcite it is well established that the bands crossing the fermi level have a strong d - orbital character . indeed , the minimal tight - binding hamiltonian which reproduces successfully the fermi surface contains all the five atomic ( or wannier ) d - orbitals centered on the fe sites,@xcite namely @xmath20 , @xmath21 , @xmath22 , @xmath23 , and @xmath24 , expressed with the crystal axes oriented according to the fe square lattice directions . a given atomic orbital @xmath25 defining the 2d multiorbital model can be represented by @xmath26 , where @xmath27 lives in the 3d space , @xmath28 is the center of the atomic orbital on the square lattice , @xmath29 is the orbital symmetry , and @xmath30 is the electron spin . a spatial point - group operation @xmath31 acts on the orbital creation operator by changing its position @xmath32 , its phase @xmath33 and its orbital label @xmath34 according to the general rule : @xmath35 for @xmath36 and @xmath29 given , the corresponding @xmath37 and @xmath38 are reported in tab . the transformation rules are different , whether the group is generated by @xmath10 or @xmath15 . . action of the symmetry operations @xmath36 on the five iron atomic d - orbitals . the entries represent the pair @xmath39 and @xmath40 as a function of the orbital index @xmath29 ( first column ) and the symmetry operation @xmath31 considered . note the different @xmath37 between proper @xmath15 and improper @xmath10 rotations in the first two orbitals ( odd with respect to @xmath14 ) . [ cols="<,<,<,<",options="header " , ] as it is readily seen in fig . ( [ fordummies ] ) , also the elementary translation symmetry bringing one iron atom to a nearest neighbor one , is not defined on the plane , because , under this transformation , the upper triangles are interchanged with the lower ones . again the combination of the conventional translation @xmath41 ( @xmath42 ) symmetry with the reflection @xmath14 is a well defined symmetry of the crystal @xmath43 ( @xmath44 ) , isomorphic to the conventional translation group of the square lattice . in a low energy model with a given number of orbitals @xmath45 ( @xmath46 ) behaves like the conventional translation group , only if we take into account a non - translation invariant sign in the definition of the localized basis , namely : @xmath47 for the orbitals @xmath48 that are odd with respect to the symmetry operator @xmath14 . here @xmath49 , where @xmath50 is the lattice space of the square lattice . unless otherwise specified , here and thereafter we use the translation invariant gauge defined in eq . [ gaugeout ] . with these conventions , let us take into account the 2d tight - binding hamiltonian @xmath51 which defines the low - energy model for the ibs . translation invariance means that all the coupling can be written in fourier space : @xmath52 where the vectors @xmath53 belong to the unfolded brillouin zone of the 2d lattice with one iron per unit cell , @xmath54 ( @xmath55 ) is the fourier transform of @xmath56 ( @xmath57 ) , the @xmath58 indexes run over the d - orbital labels , the hoppings fulfill the hermitian condition @xmath59 , and @xmath60 holds for reality . without using the phase factor @xmath61 for the odd @xmath14 orbitals , the hopping between even and odd orbitals would have acquired an alternating sign , breaking the translation symmetry , as correctly pointed out in ref . . after the simple gauge transformation given in eq.([gaugeout ] ) this fictitious broken symmetry is easily removed , and the low energy model becomes translational invariant , with a corresponding unfolded bz . in our approach we want to point out that a low energy two - dimensional model , like the free - band model in eq.([h_0 ] ) , has to be not only translation invariant but has to satisfy also all point symmetry transformations generated by @xmath18 and @xmath10 , the compatible subgroup of the full @xmath11 space - group . the latter point group is isomorphic to the standard @xmath19 group of the square lattice provided , in this group , we identify the @xmath16 rotation with the corresponding improper one . if we apply the transformation rules in eq . [ general_rules ] to @xmath62 , and require its invariance , we obtain a set of relations for the hopping matrix : @xmath63 for the eight possible independent point - group operations , defined by @xmath64 , @xmath10 , and all their products . these transformations have already been derived in refs . and or implicitly assumed in refs . and . however , their implications for the pairing properties have so far been overlooked . the pairing operator @xmath65 is assumed to be translationally invariant ( on the square lattice with one iron atom for unit cell and with the sign convention given in eq . [ gaugeout ] for the odd reflection orbitals ) , so that it is convenient to write it in @xmath66-space as @xmath67 singlet pairing , and no broken time - reversal symmetry imply that @xmath68 , and @xmath69 , respectively . in the following , we derive the spatial - orbital properties of the pairing function latexmath:[$f({\bf r},{\bf r}^\prime)=\langle { \bf r } , { \bf r}^\prime| { \hat f } irreducible representation of the point - group generated by @xmath18 and @xmath10 . this assumption will be verified later by ab - initio qmc calculations with a fully optimized jastrow projected bcs wave function . `` improper s - wave '' is the name we assign to the symmetry of the resulting @xmath65 , to distinguish it from the standard or `` proper '' one , i.e. the @xmath0 irreducible representation of the isomorphic point - group where the @xmath10 class has been replaced by @xmath15 . based on the general rules in eq . [ general_rules ] and the entries of tab . [ sgap ] , we can derive the symmetry constraints for the improper s - wave . as the characters of the @xmath0 representation are - by definition - identically one for all classes within the point - group , the symmetry transformations of @xmath71 read as @xmath72 for @xmath64 , @xmath10 , and all their independent products . the above constraints reduce the variational freedom for the signs of @xmath73 . accidental nodes must not break the _ general _ symmetries in eq . [ improper_s_constraints ] . for the improper s - wave , we demonstrate that @xmath74 has a peculiar , symmetry selected , sign structure . to show this , let us divide the iron d - orbitals into two groups , namely the `` even '' ( @xmath75 ) and``odd '' ( @xmath76 ) ones , according to their parity under @xmath14 . the @xmath77 and @xmath78 orbitals are odd , while @xmath79 , @xmath80 , and @xmath81 are even . thus , @xmath74 can be expanded as @xmath82 with its components defined as follows : @xmath83 the above pairing functions are periodic in both @xmath27 and @xmath84 , with periodicity set by the bravais lattice vectors . in the following analysis , without loss of generality , we are going to pin @xmath27 around an iron site@xcite , while we define @xmath85 as the 2d fourier transform of @xmath74 with respect to the planar ( @xmath86 ) components of @xmath87 . according to the transformation rules in eq . [ improper_s_constraints ] and in tab . [ sgap ] , we obtain that @xmath88 and @xmath89 have an s - wave planar symmetry , e.g. @xmath90 , whereas @xmath91 and @xmath92 , which couple unlike - parity components , have planar @xmath93 symmetry , e.g. @xmath94 . we note here that in a standard s - wave superconductor obeying the @xmath19 point - group transformations , also the @xmath91 and @xmath92 components are s - wave . thus , our improper - s wave pairing function has to have different signs just from symmetry considerations . we note also that a two - band model containing only odd orbitals@xcite , can not account for the anomalous @xmath93 symmetry of the pairing function . our predictions presented in sec . [ improper_symmetry ] and derived from symmetry arguments have been verified against accurate ab - initio qmc calculations based on the energy minimization of a jastrow projected antisymmetrized geminal product ( jagp ) wavefunction , performed for the fese on a @xmath96 iron lattice , subject to periodic boundary conditions . the geminal or pairing function describes the spatial correlations in the singlet electron pairs , giving rise eventually to a superconducting state , if the pairing remains stable in the thermodynamic limit . we used the experimental fese geometry , determined at ambient pressure ( 0 gpa ) and temperature@xcite and under a hydrostatic pressure of 4 gpa at low - temperature.@xcite fe and se atoms have been replaced by pseudoatoms , containing only 16 and 6 valence electrons , respectively . in this work , we used variational quantum monte carlo ( qmc ) methods to find the best many - body wave function @xmath97 which minimizes the variational energy @xmath98 where @xmath99 is the electron first principles hamiltonian in hartree units @xmath100 with @xmath101 the number of electrons in the supercell , @xmath102 the number of nuclei , and @xmath103 the position of the i - th nucleus , fixed at the experimental geometry . in the hamiltonian , the core electrons have been replaced by the pseudopotentials @xmath104 . for iron , we used the energy - consistent scalar - relativistic hartree - fock pseudopotentials generated in ref . , while for selenium , we used the smooth scalar - relativistic hartree - fock pseudopotentials published in ref . . the system is unpolarized with @xmath105 , and the supercell with @xmath101 electrons is subject to periodic boundary conditions to reproduce the periodicity of the crystal , with standard ewald summations of the long range coulomb potential by including also the constant nucleus - nucleus interaction . the wave function @xmath97 is written in the jastrow projected agp form @xmath106 \det[\phi({\bf r}_i^\uparrow,{\bf r}_j^\downarrow ) ] , \label{jagp}\ ] ] where @xmath107 , and @xmath108 the many - body electron configuration . the variational form above has been widely used to study strongly correlated hamiltonians on the lattice@xcite , and then generalized to a large variety of ab - initio systems , ranging from molecules@xcite to solids@xcite . in all systems studied so far , it has been proven to be accurate and capable to describe strong charge localization , polarization effects and weak dispersive forces . the functional form in eq . [ jagp ] is very flexible . the jastrow factor @xmath109 reads as @xmath110 where @xmath111 is the electron - nucleus term , while @xmath112 includes electron - electron correlations . we used a spin - independent parametrization for @xmath109 , therefore the spin indexes are dropped out from eq . the one - body part is developed on gaussian orbitals @xmath113 ( with @xmath114 basis set index , and @xmath115 nuclear index ) , and it depends on the i - th nucleus as follows @xmath116 where the gaussian basis set is built of @xmath117 $ ] contracted @xmath113 orbitals for selenium , while we used a @xmath118 $ ] contracted gaussian basis set for iron . both @xmath119 , the gaussian exponents , and the linear coefficients of the contractions are variational parameters to be optimized . note that there are no electron - nucleus cusps , as the pseudopotentials in refs . and are finite at the origin . the electron - electron part @xmath112 of the jastrow factor is defined as @xmath120 where the homogeneous part @xmath121 fulfills the electron - electron cusp conditions for unlike - spin particles , and the gaussian basis sets @xmath113 are the same as the ones in the one - body part . @xmath122 and the symmetric matrix @xmath123 are variational parameters . the basis set parameters ( gaussian exponents and the linear coefficients ) are optimized together with the one - body part . the pairing function @xmath124 in eq . [ jagp ] reads as @xmath125 where the sum is over @xmath126 molecular orbitals @xmath127 , where each one can be occupied by opposite spin electrons . the orbitals @xmath127 are expanded in a gaussian single - particle basis set @xmath128 , centered on the atomic nuclei as in the jastrow case , i.e. @xmath129 where the sum in the above equation runs over both the basis set and nuclear center indexes ( for simplicity of notations we omit here the explicit dependence on the atomic positions ) . the @xmath130 basis set is uncontracted with size of @xmath131 for fe and @xmath132 for se . @xmath133 , @xmath134 , and the exponents of the uncontracted gaussian basis set @xmath128 are variational parameters . if @xmath135 , the expansion in eq . [ pairing_mo ] is equivalent to a single slater determinant , when the antisymmetrization of @xmath124 is performed . on the other hand , if @xmath136 , correlation is introduced through the pairing function , and one can describe the effective attraction between opposite spin electrons by a reduction of the total variational energy . if this energy gain survives in the thermodynamic limit , it implies that the bcs condensate formed by the electron singlets is stable , and the symmetry of the pairing function is the one of the superconducting gap . in order to treat the periodic system in a finite supercell , both the jastrow gaussian basis set @xmath137 and the one @xmath128 of the pairing function @xmath124 are made of periodic gaussian functions , as described in ref . . to generate the initial molecular orbitals , we performed preliminary density functional theory ( dft ) calculations in the local density approximation , within the same basis described in the previous subsection . the starting gaussian exponents of the @xmath128 basis are even - tempered , as they are defined as @xmath138 for @xmath139 with @xmath140 and @xmath141 . the number of gaussians @xmath142 , @xmath143 and @xmath144 are optimized independently for each angular momentum shell in order to minimize the total dft energy . the number of shells , and the number of gaussians per shell are chosen to give a convergence of @xmath145 10 mhartree per iron atom in the dft energy.@xcite the mo generated by dft are then used to build the @xmath146 wave function . at the beginning , the pairing function is such that @xmath147 for @xmath148 and @xmath149 for @xmath150 , i.e. we start from a jastrow - slater wave function , with the slater part exactly given by the dft orbitals and with no jastrow factor ( @xmath151 ) apart from an initial guess of the variational constant @xmath152 in its homogeneous part @xmath153 . the many - body wave function is optimized in the qmc framework by minimizing the variational energy in eq . [ var_energy ] , by using the stochastic reconfiguration ( sr ) method ( based on the first - order derivatives of the energy with respect to the variational parameters ) combined with second - order contributions coming from the hessian matrix ( `` linear method '' ) . indeed , the sr method described in ref . has been improved by using a statistically more accurate evaluation of the hessian matrix elements in the so called `` linear basis '' as described in ref . . before the full optimization of all the variational parameters in our jagp wave function , we adopt the following intermediate steps : \i ) first the jastrow linear coefficients @xmath119 and @xmath123 , together with the parameter @xmath122 of the homogeneous pairwise jastrow factor , are optimized ; \ii ) then , also the linear coefficients of the jastrow basis contractions @xmath137 are optimized ; \iii ) not only the contraction coefficients but also the jastrow gaussian exponents are relaxed , together with all previous parameters , by means of the linear minimization method ; \iv ) finally , all variational parameters ( in both the jastrow and the slater determinant ) are optimized . in this last step the molecular orbitals evolve according to the algorithm described in ref . , that allows one to exploit the derivatives with respect to their atomic ( uncontracted ) components without changing the rank of the full pairing matrix . in the last step , in order to decrease the number of variational parameters involved in the uncontracted basis for the agp , we perform the minimization by using only one contracted orbital for each angular momentum . this allows us to represent in the most efficient way the correction to the dft orbitals . with this strategy we achieve a dramatic reduction of the number of parameters and a systematic improvement of the variational energy , which substantially decreases with respect to step ( iii ) . once the best jastrow correlated single determinant wave function is obtained , we let the wave function develop a singlet pairing correlation , by expanding it beyond the single slater determinant . we add in eq . [ pairing_mo ] the first @xmath154 additional dft mo orbitals above the fermi level , with their corresponding @xmath133 ( with @xmath115 s.t . @xmath155 ) taken from an educated guess based on a bcs - like estimate of the pairing amplitude , and later fully optimized . the optimal @xmath156 is chosen to match the criterion discussed in ref . , namely it is the sum of the majority spin orbitals in the atomic limit ( @xmath157 , where we take the convention that @xmath158 in the dissociation limit ) . the final global optimization is performed with the same method used before in ( iv ) , with all parameters free to move , and the convergence of the @xmath146 is reached if the variation of @xmath133 with @xmath159 is within the statistical noise . note that the hierarchy followed in the wave function optimization is related to the energy scale of its various components , the smallest and the last one being the energy gain ( also called `` condensation energy '' ) due to the paired state with respect to the fermi liquid reference , namely @xmath160 for @xmath161 . the final wave function yields invaluable information about the correlated ground state , like the symmetry of the superconducting gap , in the case that a paired state is stabilized by the variational approach . this method is unbiased , as we do not provide any constraint in the energy minimization , except for the symmetries imposed by the translational invariance of the lattice structure , which reduce the total number of independent variational parameters . in a large variety of systems studied by this wave function , we reached the chemical accuracy ( in the energy differences ) , without resorting to further projective methods like diffusion monte carlo . we would like also to highlight the fact that this variational method is sign problem free . for the largest system computed , the fese @xmath162 lattice ( 16 fe + 16 se ) with 352 electrons , a full qmc energy minimization has been performed to optimize a total amount of about 10000 independent variational parameters . the full many - body wave function is written in a compact form as in eq . [ jagp ] . in all the calculations quantum effects on nuclei are neglected , within the born - oppenheimer approximation . hence superconductivity is assumed to be non conventional , namely not coming from electron - phonon coupling . @xmath163 , defining the wave function in eq . ( [ jagp ] ) , is an extension of @xmath164 in eq . [ even_odd ] , as it is expanded over the full atomic basis set used to represent the mo s . the resulting @xmath124 , restricted to its correlated part @xmath165 ( obtained by summing the mos in eq . [ pairing_mo ] over @xmath115 with @xmath166 ) , is plotted in figs . [ pairing_0gpa ] and [ pairing_4gpa ] for 0 and 4 gpa , respectively . @xmath165 has been projected over its 4 possible components @xmath167 ( figs . [ pairing_0gpa](a ) and [ pairing_4gpa](a ) ) , @xmath168 , @xmath169 , and @xmath170 ( figs . [ pairing_0gpa](b ) and [ pairing_4gpa](b ) ) , based on the parity with respect to @xmath14 @xcite . of the fese high - t@xmath171 superconductor at @xmath172 gpa obtained by qmc , by optimizing the energy of the variational wavefunction without assuming any point symmetry . the contour plots show @xmath173 with @xmath174 set to be the iron lattice site at the center of the supercell , while @xmath27 spans the plane defined by the @xmath175 lattice . red ( yellow ) balls are iron ( selenium ) sites . arbitrary units blue ( red ) intensity indicates negative ( positive ) regions with corresponding magnitude . ] but for a structure corresponding to a pressure of @xmath176 gpa . ] the final picture is consistent with the improper s - wave pairing function predicted by the 2d point - group symmetries and under the assumption of a 3d @xmath0 irreducible representation . our qmc solution is @xmath0 as a consequence of an unbiased energy minimization , and it has not been imposed a priori in the optimization . thanks to the improper @xmath10 generated point - group , the pairing function @xmath124 shows two planar symmetry channels , s - wave for the @xmath167 and @xmath168 , d@xmath9-wave for the @xmath169 and @xmath170 components . we emphasize that @xmath124 does not possess a well defined symmetry under a proper rotation by @xmath16 , just because such a rotation is not a symmetry of the crystal . moreover , fig . [ pairing_4gpa ] shows that this picture holds even when the tetragonal point - group symmetry is slightly broken , and lowered to the orthorhombic point - group , as in the case of applied pressure and/or low temperature . system ( in hartree ) , and pairing gap @xmath177 ( in mev ) at @xmath172 and @xmath176 gpa , obtained from our qmc calculations , compared to the experimental superconducting temperature t@xmath171 in mev . ] to further check the reliability of @xmath146 to describe the superconducting features of fese , we estimated ( via correlated sampling ) the condensation energy @xmath178 , defined as the difference ( per iron atom ) between the expectation value of @xmath99 computed with the fully paired @xmath146 and the one with the jastrow projected single determinant , obtained by summing the mos with @xmath179 in eq . [ pairing_mo ] . those quantities are much more sensitive to finite size errors than the pairing symmetry , set by the short - range behavior of @xmath124 , which is converged in the @xmath175 supercell . the @xmath178 is shown in fig . [ condensation ] , together with the experimental critical temperature @xmath1 . it turns out that @xmath180 , which is a remarkable result given the tiny energy scale ( of the order of 1 mev ) involved in the gap opening.@xcite moreover , it follows quite well the behavior of the experimental @xmath1 , which increases with pressure , in the pressure range considered here@xcite . therefore , if we assume that the condensation energy should be related to the critical temperature , as it is a common feature of high-@xmath1 superconductors , our pairing function describes reasonably well the experiments . remarkably , we are close to make quantitative predictions on unconventional high temperature superconductors by means of a fully ab - initio method . the determinantal part of our jagp variational wave function , for technical reasons , has not a simple interpretation in terms of a mean - field bcs hamiltonian , as it is common practice in lattice model calculations of strongly correlated systems@xcite . in this case , it is well known that very few short range parameters can describe very accurately the gap function @xmath181 of a correlated superconductors . moreover , whenever the variational description is accurate enough , very small supercells are necessary to describe non trivial physical effects and also phase transitions@xcite . for the above reasons , it is therefore extremely important to describe our variational wave function in terms of a mean - field bcs picture , because it helps to reduce the finite size effects . otherwise , a brute force variational approach based on the jagp , would require high resolution in momentum space , and therefore too large supercells , to describe correctly the opening of the superconducting gap over the non trivial and rather involved hole and electron fermi surfaces of pnictides . for example , we will show in the following subsections , that a short range parametrization of an improper s - wave gap function represents in a quantitative way our paired variational state . this is extremely useful to obtain a detailed information in @xmath182space , and finally provides a rather satisfactory description of the realistic material . we start with the five band model defined on the iron plane in real space : @xmath183 we assume no broken time reversal symmetry and therefore all couplings are real , including the bcs ones , and singlet pairing . thus both @xmath184 and @xmath185 have to be considered symmetric real matrices . due to translation invariance the model is more easily written in @xmath182space : @xmath186 it is easy to show that , for each @xmath66 , both matrices @xmath184 and @xmath185 may be complex , but they have to be hermitian . in * k*-space the relations in tab.([sgap ] ) imply parity symmetry constraints on the @xmath187 matrix elements , as we have seen in sec . [ improper_symmetry ] . the same constraints hold for the bcs pairing matrix @xmath181 . in case of a `` generalized '' @xmath0 real pairing ( dubbed here `` improper s - wave '' ) the symmetry relations are the following : @xmath188 notice that , by the no broken time reversal condition ( real space couplings , i.e. @xmath189,@xmath190 ) , the above functions are purely real or purely imaginary depending on the transformation under reflection , even ( @xmath191 ) or odd ( @xmath192 ) , respectively . a generic quasiparticle with spin up can be written as : @xmath193 where @xmath153 and @xmath194 are arbitrary complex @xmath195 matrices , satisfying : @xmath196 in order to compute these matrices it is enough to write the quasiparticle equation : @xmath197 = e_\nu({\bf k } ) \psi^{\dag}_{{\bf k},\nu,\uparrow } \label{qp_equation}\ ] ] which implies to solve a @xmath198 eigenvalue problem for each @xmath66 : @xmath199 notice that eq . [ bcseq ] has pairs of eigenvalues of opposite values as if @xmath200 is an eigenvector with eigenvalue @xmath201 , @xmath202 has eigenvalue @xmath203 . thus we have 5 positive and 5 negative eigenvalues for each * k*. the pairing function @xmath204 defines the ground state @xmath205 of the bcs hamiltonian by : @xmath206 the function @xmath204 can be determined by requiring that the ground state is the vacuum of all the quasiparticles with positive eigenvalues : @xmath207 for all eigenvalues @xmath208 . by using that @xmath209 c^\dag_{-{\bf k},\mu,\downarrow } \right\},\ ] ] where @xmath210 , we finally obtain that the matrix @xmath204 is fully determined via the quasiparticle equation . by solving @xmath211 , it reads : @xmath212 namely , for each @xmath66 , @xmath204 is a generally full @xmath195 matrix . from eq.([nonso ] ) it also follows that @xmath213 is hermitian for each @xmath66 and can therefore be diagonalized with real eigenvalues . this property is the final consequence of the assumed no broken time reversal symmetry , otherwise @xmath204 is only a symmetric matrix , generally complex . we use the bcs model of eq . [ bcs_model_real_space ] as a way to interpolate on an ultra dense k - grid the realistic pairing @xmath214 which enters the jagp wave function , computed on a @xmath175 k - point mesh by the ab - initio variational qmc method as described in sec . [ qmc_section ] . within the bcs framework , we fit both the tight - binding model and the effective bcs coupling @xmath185 from ab - initio calculations . the tight - binding model in eq . [ bcs_model_k_space ] is fitted based on the wannier interpolation of the lda - dft band structure@xcite calculated for the undistorted ambient pressure fese compound . the number of neighbors taken in the model is the same as the one used in ref . by graser _ et al._. on the other hand , we assume a short - range bcs @xmath185 , by taking only the nearest - neighbors couplings , assumption that is normally verified for strongly correlated superconductors . those couplings are set to reproduce the real space image of the ab - initio qmc pairing function @xmath215 plotted in fig . [ pairing_0gpa ] . we define the real - space bcs pairing function as : @xmath216 where @xmath217 is the solution of the bcs problem ( [ bcs_model_k_space ] ) on a given k - grid as in eq . [ f_solution ] , and @xmath218 are localized functions taken as gaussian periodic orbitals with d - symmetry@xcite and gaussian exponent @xmath219 . moreover , according to eq . ( [ gaugeout ] ) , an overall sign @xmath61 is given to the localized wannier orbitals , centered at @xmath220 and with odd reflection symmetry . this essentially implies that @xmath221 , for such odd reflection orbitals , where @xmath222 . in this way the anomalous part of the pairing function @xmath223 that connects even and odd @xmath14 orbitals , breaks the translation symmetry because couples @xmath66 and @xmath224 physical momenta . however , this broken translation symmetry , noticed in refs . and , can be completely gauged out in the low energy 2d model , and is not so crucial therefore ( e.g. absolutely irrelevant for the spectrum ) , as the improper @xmath225wave symmetry we are pointing out in this work . we choose the non - zero couplings @xmath185 in eq . [ bcs_model_real_space ] such that @xmath226 is as close as possible to the qmc @xmath215 in the largest qmc supercell , by minimizing the distance in the @xmath227 functional space . this is obtained by maximizing the normalized scalar product @xmath228 , which in the gaussian basis reads : @xmath229}{\sqrt{{\rm tr } [ s f^{bcs } s f^{bcs}]~ { \rm tr } [ s \tilde{\phi}^{corr } s \tilde{\phi}^{corr } ] } } , \ ] ] where @xmath230 is the projection of @xmath165 on the single gaussian basis set @xmath231 , and @xmath232 is the overlap matrix between the basis set elements . the largest normalized scalar product found is about @xmath233 , with best couplings which turn out to be ( in ev ) : @xmath234 where @xmath235 , @xmath236 , with @xmath237 the distance between two nearest - neighbors iron sites . the last one in eq . [ delta_bcs_couplings ] is the even - odd / odd - even coupling which triggers the planar d - wave component and leads to the improper s - wave symmetry of the global pairing . it is interesting to note that the resulting pairing @xmath238 is purely real if it couples even - even and odd - odd orbitals , while it is purely imaginary when it couples even - odd and odd - even orbitals . this follows from the @xmath187 and @xmath239 components , which are either purely real or purely imaginary according to their transformation properties , as already highlighted in the previous subsection . in the literature , the k - resolved image of the coupling and its symmetry analysis provided so far , are usually done for the @xmath240 pairing function projected and rotated in the band space ( here @xmath241 and @xmath142 refer to band indexes ) , i.e. @xmath242 , with @xmath243 the unitary matrix mapping the orbital into the band space . we stress that the band space is not the ideal basis to study the improper s - wave symmetry , as only the orbitals are eigenstates of the parity under reflection through the iron plane . the hybrid nature of the bands does not allow to separate the planar s- and d - wave components of the total paring operator @xmath65 in eq . [ f ] . in the next section , we will analyze in detail the k - resolved components of @xmath74 in the orbital basis and the superconductivity gap of the associated bcs model . here , we provide the k - resolved image of @xmath240 in the band basis on the fermi surface , in order to understand its sign properties , and compare it with the previous literature . plotted on the unfolded brillouin zone ( one fe unit cell ) . ] in fig . [ improper_s_band_space ] we plot @xmath240 with @xmath66 s on the fermi surface ( in the unfolded brillouin zone ) , which results from the bcs model interpolation on an ultra dense @xmath244 k - point grid . the figure shows that the global symmetry is the a@xmath245 . the only sign change in this picture occurs within the m electron pockets , already pointed out in previous works@xcite . however , once the global symmetry is fixed , the @xmath246 components can have in general a non trivial structure , which is material dependent , due to its multi - band nature . here , we find that for the fese there are no other sign changes except for the ones in the m pocket . it is hard to say if this comes from the improper s - wave symmetry or the sign change is just `` accidental '' . in the band representation , there is not a clear - cut signature of the improper s - wave components . in the next section , we are going to analyze what are the symmetry implications of the improper - s symmetry , by studying the full pairing function @xmath74 in the orbital space and its related bcs gap . the bcs model derived in the previous subsection allows not only to interpolate @xmath238 on a dense k - point grid , but also to evaluate the k - resolved bcs gap @xmath247 , of paramount importance to make a connection with some angle - resolved experiments as for instance the specific heat measurements performed in ref . for the fese that experiment indicates the presence of gap minima along the @xmath248-m ( fe - fe ) bond directions , and conversely the gap maxima along the fe - se - fe directions , being the angular dependence fourfold . @xmath249{plot_bcs_gap.png } & \includegraphics[width=0.5\columnwidth]{plot_bcs_gap_folded.png } \end{array}$ ] fig . [ bcs_gap ] obtained on a @xmath244 k - point grid from the solution of the bcs model with parameters in subsec . [ k - resolved_coupling ] shows that indeed minima are aligned in the @xmath248-m directions , while maxima are clearly peaked along the x - m directions , in general agreement with the specific heat experiment . @xmath249{plot_bcs_gap_noxy.png } & \includegraphics[width=0.5\columnwidth]{plot_bcs_gap_noxy_folded.png } \end{array}$ ] to understand the role of the even - odd / odd - even couplings in setting the gap modulation , we explicitly set to zero @xmath250 and @xmath251 in eq . [ delta_bcs_couplings ] , while keeping unchanged the other bcs parameters . the resulting bcs gap of this modified model is shown in fig . [ bcs_gap_noxy ] . as one can see , the angular dependence of the bcs gap minima ( black regions ) of the latter model is much more isotropic . thus , the @xmath250 and @xmath251 bcs parameters , which couple orbitals of different parity , play a key role to trigger the angular dependence of the bcs gap . we turn now our attention to the fourier analysis of the physical pairing function @xmath252 . we express the function @xmath73 in the orbital basis , and we resolve it in terms of even and odd components , @xmath88 , @xmath91 , @xmath92 , and @xmath89 , as in eq . [ f_components ] . from the their translational properties , one can easily show that they all depend on the vector @xmath87 connecting the electron pair , and they are periodic functions ( with periodicity set by the lattice vectors ) of the pair center @xmath174 . the dependence on @xmath174 gives rise to an inhomogeneous contribution within the unit cell , while the @xmath253 dependence is set by the symmetry transformations under the 2d lattice point - group operations . therefore , their orbital - summed fourier decomposition can be written in general as @xmath254 where @xmath66 leaves in the 2d brillouin zone of the iron lattice plane , and @xmath255 and @xmath256 are reciprocal lattice vectors of the 3d bravais lattice of the crystal . from eq . [ fourier_decomposition ] , their parity - resolved fourier components are analytically given by @xmath257 where @xmath258 ( @xmath259 ) is the parity under reflection through the iron plane of the left ( right ) orbital , and @xmath260 is the fourier component of the @xmath29 orbital evaluated at the @xmath261 reciprocal vector in the gauge representation , corresponding to the @xmath262 component in the `` physical '' representation . eq . [ fourier_components ] can be evaluated on a dense k - point grid thanks to the bcs interpolation @xmath217 of the qmc pairing function . in the following we compute eq . [ fourier_components ] by taking @xmath217 and the fourier components of the periodic gaussian d - orbital with exponent @xmath219 . within the `` improper s - wave '' picture , @xmath263 have the following properties , according to the z - component values of @xmath255 and @xmath256 : * for @xmath264 and @xmath265 , the only non - zero components are of type `` even - even '' , they are purely real , and they have a standard s - wave symmetry , as seen in fig . [ fourier_components_plot](a ) ; * for @xmath266 , or @xmath267 , the improper symmetry shows up in the `` even - odd '' and `` odd - even '' components , which are purely imaginary , and they have a d@xmath9-wave symmetry ( see , for instance , fig . [ fourier_components_plot](b ) ) ; * all the other `` even - even '' and `` odd - odd '' components are purely real and of s - wave symmetry ( see figs . [ fourier_components_plot](c ) and [ fourier_components_plot](d ) ) . @xmath249{plot_fksum_real_even_even_gz0_gz0_bz.png } & \includegraphics[width=0.5\columnwidth]{plot_fksum_imag_even_odd_gz1_gz1_bz.png } \\ \includegraphics[width=0.5\columnwidth]{plot_fksum_real_even_even_gz1_gz1_bz.png } & \includegraphics[width=0.5\columnwidth]{plot_fksum_real_odd_odd_gz1_gz1_bz.png } \end{array}$ ] we explore now the physical consequences of the improper s - wave from a general perspective , by assuming that also the other ibs families have a @xmath0 superconducting gap , as the fese . first of all , the twofold symmetry of the pairing function observed by stm@xcite in stoichiometric fese with impurities and magnetic field can be explained by the andreev resonance probed by the stm and sensitive to the sign change induced by a planar superposition of the s- and d - wave channels , invariant under @xmath268 rotations . note that the d@xmath9 arrangement is in agreement with the different orientations measured by the stm zero bias conductance . in fig . [ total_qmc_pairing ] , by plotting the qmc pairing function at different distances from the iron plane , we show how the @xmath15 is broken as soon as the experimental local probe is lifted from the iron layer . from ab - initio dft and dmft calculations,@xcite the @xmath272 and @xmath273 orbitals are usually pseudogapped . this allows us to disregard them from the summation in eq . thus , @xmath93 is the only remaining orbital with even parity , and it can trigger the formation of the d - wave channel in the pairing function . indeed , for absent or weak @xmath93 spectral weight at the fermi level , the two orbitals left in @xmath74 are both odd under @xmath14 , and therefore they can only contribute to the s - wave channel . in the `` 111 '' family , dft band structure calculations and de haas - van alphen experiments show that the main effect of p isovalent substitution is the formation of the third hole pocket - with dominant @xmath93 character - , absent in the lifeas.@xcite the enhanced @xmath93 spectral weight in the p substituted compound and the subsequent development of an `` even - odd '' d - wave pairing gap can explain why the lifep pairing is nodal , while the lifeas shows a fully gapped behavior . within the improper s - wave picture , the properties of the `` 1111 '' family modified by impurity substitution as in ref . are consistent with a tendency of @xmath74 to develop nodes by going from co to mn doping through the opening of the d - wave channel . indeed , co impurities are electron donors , while mn impurities dope the system with hole , thus enhancing the @xmath93 spectral weight around @xmath248 . the `` 122 '' family does not belong to the @xmath11 space - group . nevertheless , by inspection of its @xmath274 space - group unit cell , one can see that also in that case the @xmath10 improper rotation is the valid symmetry on the iron layer , thus the improper s - wave theory applies also here . the @xmath4-to-@xmath5 wave transition by doping the bafe@xmath6as@xmath6 with k is one of the most striking experimental situations for us.@xcite in the strong overdoped regime , the band structure certainly shows all three pockets at @xmath248 with again an enhanced @xmath93 spectral weight . if the d - wave channel dominates in the improper s - wave gap , then it is plausible that the experimental thermodynamic features are compatible with a pure d - wave behavior . finally , we would like to point out that the importance of the third hole pocket , with strong @xmath93 character , has already been observed by kuroki and co - workers in relation to the @xmath1 dependence on the pnictogen height.@xcite the improper s - wave symmetry explains the key role played by @xmath93 , particularly in connection with the appearance of gap nodes , the twofold anisotropy , and the stabilization of the d - wave channel , by reconciling a series of experiments which look otherwise contradictory . it suggests also that the pnictogen or the chalcogen , rather than being spectators , take an active role in the electronic pairing , by bridging two next - nearest neighbor iron sites in a @xmath93 fashion via the improper s - wave symmetry . from the theoretical point of view , there is a lot of confusion in literature . for instance , to our knowledge , no rpa and no mean - field studies have been done so far with a low energy 2d model explicitly compatible with the improper rotations and gauged translation invariance . as we have shown translation invariance and @xmath19 have to be combined in an appropriate way , by taking particular attention to the even / odd @xmath14 reflection of the orbitals in the model . in our qmc simulations the improper s - wave symmetry is stabilized , which determines a highly entangled nodal structure of the pairing function , and the possibility for mixing between s - wave and d - wave components , an effect that , as we discussed , is crucial to understand most experiments in the iron - based superconductors . we acknowledge useful discussions with s. biermann , t. cren , m. fabrizio , a. savin , and d. j. scalapino . the hpc resources have been provided under the idris / genci 2012096493 and cineca - miur iscra - hp10a2gzhv grants . 99 j. paglione , and r. l. greene , nature phys . * 6 * , 645 ( 2010 ) . in the following , for simplicity of notations , we identify the subgroup symbols @xmath10 or @xmath15 with the elementary operation ( e.g. rotation by @xmath16 ) that defines the cyclic subgroup . g. r. stewart , rev . * 83 * , 1589 ( 2011 ) . p. j. hirschfeld , m. m. korshunov , and i. i. mazin , rep . . phys . * 74 * 124508 ( 2011 ) . i. i. mazin , d. j. singh , m. d. johannes , and m. h. du , phys . * 101 * , 057003 ( 2008 ) . k. kuroki , s. onari , r. arita , h. usui , y. tanaka , h. kontani , and h. aoki , phys . * 101 * , 087004 ( 2008 ) . s. graser , a. f. kemper , t. a. maier , h. p. cheng , p. j. hirschfeld , and d. j. scalapino , phys . b * 81 * , 214503 ( 2010 ) . k. suzuki , h. usui , and k. kuroki , j.phys.soc.japan * 80 * , 013710 ( 2011 ) . i. i. mazin , t. p. devereaux , j. g. analytis , j .- h . chu , i. r. fisher , b. muschler , and r. hackl , phys . b * 82 * , 180502(r ) ( 2010 ) . f. ning , k. ahilan , t. imai , a. s. sefat , r. jin , m. a. mcguire , b. c. sales , and d. mandrus , j. phys . japan * 77 * , 103705 ( 2008 ) . k. hashimoto , s. kasahara , r. katsumata , y. mizukami , m. yamashita , h. ikeda , t. terashima , a. carrington , y. matsuda , t. shibauchi , phys . lett . * 108 * , 047003 ( 2012 ) . reid , a. juneau - fecteau , r. t. gordon , s. ren de cotret , n. doiron - leyraud , x. g. luo , h. shakeripour , j. chang , m. a. tanatar , h. kim , r. prozorov , t. saito , h. fukazawa , y. kohori , k. kihou , c. h. lee , a. iyo , h. eisaki , b. shen , h - h . wen , and l. taillefer , supercond . technol . * 25 * , 084013 ( 2012 ) . j. k. dong , s. y. zhou , t. y. guan , h. zhang , y. f. dai , x. qiu , x. f. wang , y. he , x. h. chen , s. y. li , phys . 087005 ( 2010 ) . m. sato , y. kobayashi , s. c. lee , h. takahashi , e. satomi , and y. miura j. phys . . jap . * 79 * , 014710 ( 2010 ) . b. zeng , g. mu , h. q. luo , t. xiang , i. i. mazin , h. yang , l. shan , c. ren , p. c. dai , h .- h . wen , nature communications * 1 * , 112 ( 2010 ) . song , y .- l . wang , p. cheng , y .- p . jiang , w. li , t. zhang , z. li , k. he , l. wang , j .- f . jia , h .- h . hung , c. wu , x. ma , x. chen , q .- k . xue , science * 332 * , 1410 ( 2011 ) . d. j. singh and m .- h . du , phys . lett . * 100 * , 237003 ( 2008 ) . m. aichhorn , s. biermann , t. miyake , a. georges , and m. imada , phys . rev . b * 82 * , 064504 ( 2010 ) . werner , m. casula , t. miyake , f. aryasetiawan , a. j. millis , and s. biermann , nature physics * 8 * , 331 ( 2012 ) . s. graser , t. a. maier , p. j. hirschfeld , and d. j. scalapino , new j. phys . * 11 * , 025016 ( 2009 ) . k. nakamura , r. arita , and m. imada , j. phys . . jpn . * 77 * , 093711 ( 2008 ) . t. miyake , k. nakamura , r. arita , and m. imada , j. phys . . jpn . * 79 * , 044705 ( 2010 ) . strictly speaking , the @xmath5 orbitals are zero at the iron sites . the amplitude of the pairing function is computed by integrating those orbitals over a planar mesh ( @xmath275 , with @xmath276 bohr ) around the iron center . the sign of the integrand on the mesh selects their angular symmetry . r. s. kumar , y. zhang , s. sinogeikin , y. xiao , s. kumar , p. chow , a. l. cornelius , and c. chen , j. phys . b * 114 * , 12597 ( 2010 ) . s. margadonna , y. takabayashi , y. ohishi , y. mizuguchi , y. takano , t. kagayama , t. nakagawa , m. takata , and k. prassides , phys . b * 80 * , 064506 ( 2009 ) . m. burkatzki , claudia filippi , and m. dolg , j. chem . phys . * 129 * , 164115 ( 2008 ) j.r . trail and r.j . needs , j. chem . phys . * 122 * , 174109 ( 2005 ) ; j.r . trail and r.j . needs , j. chem . phys . * 122 * , 014112 ( 2005 ) , see also www.tcm.phy.cam.ac.uk/mdt26/casino2_pseudopotentials.html . s. sorella , g. martins , f. becca , c. gazza , l. capriotti , a. parola and e. dagotto , phys . lett . * 88 * , 117002 ( 2002 ) . m. capello , f. becca , m. fabrizio , s. sorella , and e. tosatti , phys . lett . * 94 * , 026406 ( 2005 ) . wen - jun hu , federico becca , and sandro sorella phys . b 85 , 081110 ( 2012 ) . m. casula , c. attaccalite , and s. sorella , j. chem . phys . * 121 * 7110 ( 2004 ) . s. sorella , m. casula , and d. rocca , j. chem . phys . * 127 * , 014105 ( 2007 ) . f. sterpone , l. spanu , l. ferraro , s. sorella , l. guidoni , j. chem . theory comput . * 4 * , 1428 ( 2008 ) . leonardo spanu , massimo lugas , federico becca , and sandro sorella , , 024510 ( 2008 ) . federico becca , luca capriotti , alberto parola , and sandro sorella , 060401 ( 2007 ) . the lda - dft calculations have been performed with the quantum - espresso@xcite code . fe and se atoms are described respectively by ultrasoft ( with 16 valence electrons ) and norm - conserving ( with 6 valence electrons ) pseudopotentials generated with a scalar - relativistic calculation . the plane - wave ( pw ) cutoff is 60 ry for the wave - function , and 240 ry for the charge . a @xmath277 electron - momentum grid and a methfessel - paxton smearing of 0.01 ry are used in the electronic integration . wannierization has been performed with the wannier90@xcite program on a @xmath278 electron - momentum mesh , by including the d states on the two inequivalent fe site of the unit cell . the localized orbitals yield wannierized bands in a very good agreement with those computed in the pw basis set in a window of @xmath279 1.5 ev around the fermi level . p. giannozzi _ et al . _ , j. phys . : matter * 21 * , 395502 ( 2009 ) . a. a. mosto , j. r. yates , y .- s . lee , i. souza , d. vanderbilt and n. marzari , comput . * 178 * , 685 ( 2008 ) . sandro sorella , michele casula , leonardo spanu , and andrea dal corso , phys . b * 83 * , 075119 ( 2011 ) . m. marchi , s. azadi , s. sorella , phys . lett . * 107 * , 086807 ( 2011 ) . s. azadi , c. cavazzoni and s. sorella , 125112 ( 2010 ) . c. j. umrigar , j. toulouse , c. filippi , s. sorella , and r. g. hennig , phys . 98 * , 110201 ( 2007 ) . m. marchi , s. azadi , m. casula , s. sorella , j. chem . phys . * 131 * , 154116 ( 2009 ) . the parity projection has been made by averaging the orbitals at opposite positions above and below the iron layer at a given distance @xmath280 from it ( 0.25 bohr ) . the odd orbitals are strictly zero at the iron layer . however , the pairing function is 3d . averaging those orbitals above and below the layer gives a finite contribution . therefore , the even and odd mo orbitals entering the pairing function @xmath124 in eq . [ pairing_mo ] read as : @xmath281 at ambient pressure , the slightly broken low temperature phase is very close to the tetragonal parent structure , and their difference ( around 0.01 between the @xmath50 and @xmath282 axes ) has been neglected in our calculations . at 4 gpa , the @xmath283 anisotropy is instead more than four times stronger , and thus we used the corresponding experimental low temperature orthorombic structure . c. putzke , a. i. coldea , i. guillamn , d. vignolles , a. mccollam , d. leboeuf , m. d. watson , i. i. mazin , s. kasahara , t. terashima , t. shibauchi , y. matsuda , and a. carrington , phys . lett . * 108 * , 047002 ( 2012 ) . k. kuroki , h. usui , s. onari , r. arita , and h. aoki , phys . b * 79 * , 224511 ( 2009 ) . j. hu and n. hao phys . x * 2 * , 021009 ( 2012 ) . lin , t. berlijn , l. wang , c. c. lee , w.g . yin , and w. ku , 257001 ( 2011 ) .

within the framework of the operator product expansion ( ope ) and the renormalization group equation ( rge ) , we show that the temperature and chemical potential dependence of the zeroth moment of a spectral function ( sf ) is completely determined by the one - loop structure in an asymptotically free theory , and in particular in qcd . logarithmic corrections are found to play an essential role in the derivation . this exact result constrains the shape of sf s , and implies striking effects near phase transitions . phenomenological parameterizations of the sf , often used in applications such as the analysis of lattice qcd data or qcd sum rule calculations at finite temperature and baryon density must satisfy these constraints . we also explicitly illustrate in detail the exact sum rule in the gross - neveu model .

hep-ph9412246

hadronic properties at finite temperature and baryon density are of great importance in the phenomenology of heavy ions collisions , star interior and the early universe . moreover , the theoretical expectation of transitions to a chirally symmetric phase and , perhaps , to a quark - gluon plasma phase contributes to the interest in studying the effect of matter and temperature on the quantum chromodynamics ( qcd ) vacuum . our present understanding of qcd at finite temperature ( @xmath0 ) and baryon density ( or chemical potential @xmath1 ) is mainly limited in the euclidean realm , due to the lack of non - perturbative and systematic calculating tools directly in the minkowski space . typical methods , with qcd lagrangian as the starting point , are the ope and lattice simulations . because these two formulations are intrinsically euclidean , only static quantities are conveniently studied . in order to gain dynamical informations , which are more accessible experimentally , the analytic structure implemented through dispersion relations often have to be invoked within the theory of linear response . in principle , dispersion relations allow the determination of the spectral function ( sf ) , which carries all the real - time information , from the corresponding correlator in the euclidean region . in practice , realistic calculations , e.g. ope or lattice simulations , yield only partial information on the correlators , making impossible a direct inversion of the dispersion relation . therefore , the standard approach assumes a phenomenological motivated functional form with several parameters for the sf , and uses the information from the approximate correlator , mediated by the dispersion relation , only to determine the value of parameters by a fit . this approach has been quite successful at zero temperature and density , thanks to the fortuitous situation that we roughly know how to parameterize the sf s in many cases . two important examples are the qcd sum rules pioneered by shifman , vainshtein and zakharov @xcite , and the analysis of lattice qcd data @xcite . so far , standard parameterizations have included poles plus perturbative continuum @xcite . the success of such approaches heavily rests on our good understanding of the qualitative behavior of sf s at zero @xmath2 . we can find other such favorable examples in the low-@xmath0 regime @xcite , where the shape of the sf has the same qualitatively features of the zero @xmath2 case , or even in the high-@xmath0 regime for simple models @xcite , for which the functional form of the sf is known . the qcd sum rules approach has been extended also to systems at finite temperature @xcite . the lack of experimental data , and of reliable nonperturbative calculations has prompted people to use the same kind of parameterizations that have worked so well at zero temperature with , at most , perturbative corrections . we believe that physical results at finite @xmath2 can be strongly biased by this assumption . in fact , recent interpretations of lattice simulation data @xcite appear to indicate the existence of such problems . the purpose of this work is to derive exact sum rules that constrain the variation of sf s with @xmath2 . in addition , we apply these sum rules to the chiral phase transition , and demonstrate that sf s in some channels are drastically modified compared to both their zero @xmath2 and their perturbative shapes . this result confirm our worries about non - trivial effect of finite @xmath0 or baryon density on the shape of the sf . our derivation of these exact sum rules , based on the ope and the rge , has a closer relation with the derivation of sum rules for deep inelastic scatterings than with the qcd sum rule approach of svz @xcite . in fact , we establish relationships between moments of the sf and corresponding condensates as functions of @xmath2 , without assuming any functional form of the sf . in the derivation process , we find that the logarithmic corrections are essential to establish the exact sum rules . in contrast , the qcd logarithmic corrections are only marginally relevant in the finite energy sum rules , and hence are rarely discussed in the literature . to properly take into account the logarithmic corrections , a repeated partial integration method is used to match the relevant asymptotic expansions . since no further assumptions other than the validity of the ope and rge are involved in the derivation , our sum rules are very general and can be applied anywhere in the @xmath2-plane , even near or at the boundary of a phase transition . the paper is organized as follows . in section ii we present the general derivation of exact sum rules in asymptotically free theories . the matching of asymptotic behaviors of the correlator and the dispersion integral , including explicitly their logarithmic corrections , are carefully discussed . in section iii we illustrate each single step of the derivation in a soluble model , the gross - neveu model in the large-@xmath3 limit . in this model we can calculate exactly all the relevant quantities ( spectral functions , wilson coefficients , condensates , anomalous dimensions and correlators in space - like region , etc . ) , and , therefore , give a concrete example of how our method works . the application of our method to the derivation of exact sum rules for the mesonic channels in qcd is presented in section iv . in the same section , we also discuss the phenomenological consequences of the exact sum rules near chiral restoration phase transitions . finally , in section v we summarize our work , draw some conclusions , and discuss possible future directions . we start this section with a short review of the linear response theory , the ope and the rge . next we introduce a convenient subtraction in the dispersion relation for studying the dependence of sf s on @xmath0 and @xmath1 . then , we present in detail a crucial part of our method : how to match the asymptotic ope expansion with a corresponding asymptotic expansion of the sf and its dispersion integral . this approach is necessary for properly taking into account the logarithmic corrections , and studying the convergence properties of the relevant moments of the sf . more naive approaches not only yield , in general , incorrect sum rules , but might also fail to recognize that a given sum rule does not exist in the first place , since the integral of the sf involved is divergent . finally , a comparison of the two asymptotic expansions leads to the desired exact sum rules . we end this section with some general comments on the derivation and meaning of these sum rules . the real - time linear response @xcite to an external source , @xmath4 , coupled to a renormalized current @xmath5 in the form of @xmath6 is given by the retarded correlator : @xmath7\rangle_{t,\mu}\ , , \label{corr}\ ] ] where the average is on the grand canonical ensemble specified by temperature @xmath0 and chemical potential @xmath1 . disregarding possible subtraction terms , which are @xmath2-independent since they are related to short distance properties of the theory , we can write the following dispersion relation for the frequency dependence of the retarded correlation function : @xmath8 for convenience , we discuss only the uniform limit ( @xmath9 ) and , from now on , drop the momentum label . we wish to emphasize that , even if the derivation of the sum rules requires the analytic continuation of eq . ( [ disp ] ) into the euclidean region ( @xmath10 ) , the dispersion integral is , nevertheless , determined by the singularities of @xmath11 in the minkowski region . one should keep clearly in mind the distinction between these real - time singularities with the ones related to screening phenomena . in asymptotically free theories , the correlator @xmath12 is calculable in terms of the ope @xcite asymptotic series in the limit of @xmath13 . for a given renormalization prescription , the asymptotic expansion reads @xmath14_\kappa\rangle_{t,\mu } \ , , \ ] ] where @xmath15 , @xmath16 , the @xmath17_\kappa$ ] s and the @xmath18 s are , respectively , the current quark mass , the coupling constant , the renormalized composite operators , and their corresponding wilson coefficients at the subtraction scale @xmath19 . here and in the following , we use the symbol @xmath20 to denote asymptotic equality . we explicitly singled out the perturbative term @xmath21 , which corresponds to the identity operator and , being independent of @xmath0 and @xmath1 , is calculated at @xmath22 . this term is the only one that is not suppressed by powers of @xmath23 and can depend on an overall subtraction scale , which we identify , for simplicity , with the same subtraction scale @xmath19 used in the renormalization of the operators . it is important to note that the information of the ensemble average is encoded in the matrix elements of composite operators , while the wilson coefficients , including @xmath24 , are independent of @xmath0 and @xmath1 . although the matrix elements @xmath25_\kappa\rangle_{t,\mu}$ ] can not be determined perturbatively , the @xmath26-dependence of the wilson coefficient @xmath18 is controlled by the following renormalization group equation @xcite @xmath27 c_n(q^2,m^2(\kappa),g^2(\kappa),\kappa)=0 \ , , \ ] ] where @xmath28 [ rge ] and @xmath29 , @xmath30 and @xmath31 are the anomalous dimensions for the current @xmath32 , the operator @xmath33 and the current quark mass @xmath34 respectively . for the purpose of illustration , we are only considering operators that do not mix under renormalization , but the mixing will be properly taken into account when necessary . the standard approach @xcite to the renormalization group equation is the introduction of a running coupling @xmath35 and a running mass @xmath36 . in asymptotically free theories @xmath35 vanishes logarithmically at large @xmath26 . it is therefore meaningful to consider , in this limit , a perturbative expansion of the renormalization group equation functions @xmath37 where @xmath38 , @xmath39 , @xmath34 , while @xmath40 , @xmath41 are pure numbers determined by a one - loop calculation . within this perturbative context , eq . ( [ rge ] ) can be solved @xcite : @xmath42 where @xmath43 is the canonical dimension of the composite operator @xmath33 minus the dimension of @xmath44 in units of a mass , and the @xmath45 s are calculable perturbatively . corrections to this result coming from nonleading powers of @xmath46 in the expansions of the renormalization group equation functions , see eq . ( [ rgef ] ) , are analytic in @xmath47 and hence can be absorbed into @xmath48 . therefore , the leading term in the large-@xmath26 limit is totally characterized by the one - loop structure of the theory . to study the temperature and chemical potential dependence of @xmath44 , we only need to consider the difference @xmath49 : @xmath50 where @xmath51 . this subtraction is crucial to remove @xmath21 , which contains terms not suppressed by powers of @xmath23 and is explicitly dependent on the renormalization point . the ope asymptotic expansion of @xmath49 is then @xmath52_\kappa\rangle \ , \label{dope}\ ] ] where @xmath53_\kappa\rangle$ ] denotes the difference of the expectation values of @xmath17_\kappa$ ] in the ensembles specified by @xmath2 and @xmath54 , respectively . since the current quark mass only runs logarithmically in asymptotically free theories , we can safely ignore the corrections to the @xmath48 s due to their dependence on @xmath55 when in deep euclidean limit . notice that the subtraction has made @xmath49 independent of the renormalization point @xmath19 , i.e. @xmath49 satisfies a homogeneous rge . since @xmath45 is perturbative and hence can be expanded in power of @xmath35 ( and denoting the first non - vanishing power as @xmath56 ) , the left - hand side of eq . ( [ ddisp ] ) can be expressed as a double ( in @xmath23 and @xmath35 ) asymptotic expansion of the form : @xmath57_\kappa\rangle}{q^{d_n } } [ g^2(q)]^{\nu+\eta_n}\ , , \label{asyope}\ ] ] where the exponents @xmath58 and the @xmath26-independent coefficients @xmath59 are again known perturbatively . it is important to emphasize here that the leading term in @xmath35 for a given power of @xmath23 is controlled by the relevant anomalous dimensions . we proceed by making an analogous asymptotic expansion of @xmath60 valid for @xmath61 with @xmath62 suitably large but otherwise arbitrary : @xmath63 with @xmath64^{-\xi_n}}{u^{2(n+1 ) } } \sum_{\nu=0}^\infty a_{n}^{(\nu)}(t,\mu ) [ \ln(u^2)]^{-\nu } \ , , \label{rholn}\ ] ] where we have explicitly isolated in @xmath65 all terms that vanish exponentially when @xmath66 , such as the pole contributions to @xmath60 or terms containing the factor @xmath67 . to simplify the notation we have chosen the units such that the running coupling has the form at one - loop level @xmath68 ( or equivalently , energy scales are measured in units of the relevant @xmath69-parameter ) . this ansatz is sufficient to produce an asymptotic series of the form of eq . ( [ asyope ] ) to one - loop level . more generally , one could replace @xmath70 with the full running coupling @xmath71 in the asymptotic sequence and generalize the method we are going to describe ; the general strategy involved in this generalization can be found , for example , in ref . @xcite . since we are presently only interested in one - loop calculations , we can regard @xmath35 and @xmath72 to be proportional . it is easy to recognize that the sum over `` @xmath39 '' is meant to match the sum over mass dimension in eq . ( [ asyope ] ) , while the sum over `` @xmath73 '' will match the sum over the order in perturbation series . the existence of anomalous dimensions in the ope makes it necessary to introduce @xmath74 in the expansion for the spectral function . in the following we ignore the dependence of @xmath74 on `` @xmath39 '' , as we have already ignored the fact that there generally exist more than one operator at a given dimension , and write @xmath75 to avoid a too cumbersome notation . the complete notation will be restored when necessary . before we proceed further , we wish to discuss whether it is possible that additional terms might appear in the expansion of @xmath76 in eq . ( [ rholn ] ) . in general , our asymptotic expansion procedure is powerful enough to exclude this possibility within the framework of the ope . in fact , it allows to verify unambiguously that terms different from the ones already present generate , when substituted in the dispersion integral , terms that are missing in the ope . let us examine two specific examples that might be suspected to exist otherwise . first , dimensional arguments could suggest terms like @xmath77 . our procedure shows that any @xmath78 term in @xmath60 would generate a @xmath79 term in @xmath49 , which does not correspond to any known condensate , and it is therefore excluded . the second example is given by terms such as @xmath80 , which are naturally produced by elementary perturbative calculations of the spectral function : are such terms present in eq . ( [ rholn ] ) ? it is indeed true that such terms appear in the high-@xmath0 expansion of the spectral function at fixed @xmath81 . however , the expansion of the spectral function that is relevant for comparing to the ope is a high-@xmath81 expansion at fixed @xmath0 . in the next section an explicit calculation in the gross - neveu model will illustrate the general fact that , contrary to the high-@xmath0 expansion , the high-@xmath81 expansion does not generate terms like @xmath80 . at this stage , we can already recognize a fundamental , and often overlooked , characteristic of the spectral function . if we insert the term @xmath65 in the dispersion integral , we only obtain pure powers of @xmath23 , since the exponential convergence allows a naive expansion of the factor @xmath82 ( watson s lemma @xcite ) . therefore logarithmic corrections , i.e. powers of @xmath35 , come solely from the @xmath76 term . because we know that the running coupling @xmath35 is always present in the ope series , the term @xmath76 must be present in the subtracted sf , and it obviously dominates the asymptotic regime ( @xmath66 ) . this fact immediately implies that only a finite number of moments of the subtracted sf can possibly be finite , i.e. the naive expansion of the factor @xmath82 is generally wrong , and that logarithmic corrections play a important role . a standard method to tackle the dispersion integral in the large-@xmath26 limit is the mellin transform . however , the use of mellin transform methods @xcite is extremely cumbersome when logarithms appear in the denominator . since inverse logarithms can not be avoided in the spectral function , we need to resort to other means . we carry out the dispersion integral of the @xmath76 term by splitting the integral in eq . ( [ ddisp ] ) into three intervals : @xmath83 , @xmath84 and @xmath85 . the integral over the first interval can be naively expanded in powers of @xmath23 , since there are no convergence problems : @xmath86 in particular , the leading term in @xmath23 is @xmath87 in the second interval , we use the asymptotic form of @xmath88 given in eq . ( [ rholn ] ) and obtain ( for instance by repeatedly integrating by parts ) @xmath89 where @xmath90 , and @xmath91 $ ] is the incomplete gamma function . notice that when @xmath92 the correct result for the term with @xmath93 is @xmath94 , which also corresponds to its limiting value : @xmath95/\alpha = \ln\frac{\ln q^2}{\ln\lambda^2}$ ] . we then substitute in the second incomplete gamma function that appears in eq . ( [ int2 ] ) the following asymptotic expression which is valid for @xmath96 and @xmath97 : @xmath98\equiv\int_z^\infty dx\ , x^{\alpha-1 } e^{-x } \sim\ , z^{\alpha-1 } e^{-z}\sum_{m=0}^\infty { ( -1)^m\gamma(1-\alpha+m)\over z^m\gamma(1-\alpha)}\ , . \label{asygamma}\ ] ] the resulting expression for @xmath99 is @xmath100 \label{dk2asym } \\ & -&{1\over q^2}\sum_{n,\nu=0}^\infty { a_n^{(\nu)}\over q^{2n}}\biggl [ \sum_{l=0\atop l\neq n}^\infty \sum_{m=0}^\infty \frac{(-1)^{l+m } } { ( n - l)^{m+1}[\ln q^2]^{m+1-\alpha } } \frac{\gamma(m+1-\alpha)}{\gamma(1-\alpha)}- \frac{(-1)^n [ \ln q^2]^{\alpha}}{\alpha } \biggr]\ , , \nonumber\end{aligned}\ ] ] where again @xmath101 and @xmath102 should be understood as @xmath103 and @xmath104 when @xmath105 . let us notice that in the first line of eq . ( [ dk2asym ] ) any term with @xmath106 such that the integral @xmath107 is finite can be formally identified as @xmath108= \int_{\lambda^2}^\infty du^2\,u^{2l}\delta\rho_{\text{power}}(u)\ , . \label{formom}\ ] ] in particular , the leading @xmath23 term in eq . ( [ dk2asym ] ) is @xmath109\\ & + & { 1\over q^2}\sum_{\nu=0}^\infty a_0^{(\nu)}\biggl [ \sum_{l=1}^\infty \sum_{m=0}^\infty \frac{(-1)^{l } } { l^{m+1}[\ln q^2]^{m+1-\alpha } } \frac{\gamma(m+1-\alpha)}{\gamma(1-\alpha)}+ \frac { [ \ln q^2]^{\alpha}}{\alpha } \biggr ] \label{int2asym } + { \cal o}\left(\frac{1}{q^4}\right ) \ , . \nonumber\end{aligned}\ ] ] we can similarly expand the integral over the third interval @xmath110^{m+1-\alpha } } { \gamma(m+1-\alpha)\over\gamma(1-\alpha ) } \nonumber \\ & \sim&\mbox{}-{1\over q^2 } \sum_{\nu=0}^\infty a_0^{(\nu ) } \sum_{l=1}^\infty \sum_{m=0}^\infty \frac{(-1)^{l+m } } { l^{m+1 } [ \ln q^2]^{m+1-\alpha } } { \gamma(m+1-\alpha)\over\gamma(1-\alpha ) } + { \cal o}\left(\frac{1}{q^4}\right ) \ , . \label{int3}\end{aligned}\ ] ] in the end we add the leading @xmath23 contributions from eqs . ( [ int1asym ] ) , ( [ int2asym ] ) and ( [ int3 ] ) to the corresponding contribution from the naive expansion of @xmath111 and obtain @xmath112^{1-\xi}}{q^2 } \sum_{\nu=0}^\infty \frac{a_0^{(\nu)}}{[\ln q^2]^{\nu}}\biggl [ \frac{1}{1-\xi-\nu } \nonumber \\ & & \,\,\,\,\ , + 2\sum_{l=1\atop m=0}^\infty\frac{(-1)^l } { [ l\ , \ln q^2]^{2m+2 } } \frac{\gamma(2m+1+\nu+\xi)}{\gamma(\nu+\xi)}\biggr ] + { \cal o}\left(\frac{1}{q^4}\right)\ , , \label{leadqrho}\end{aligned}\ ] ] where @xmath113 is defined by @xmath114 since the integral @xmath115 can be finite only if @xmath116 , we can use eq . ( [ formom ] ) to identify @xmath113 as the zeroth moment of the subtracted spectral function , only when the asymptotic expansion of the subtracted sf in eq . ( [ rholn ] ) yields @xmath116 . otherwise , if @xmath117 , the zeroth moment is infinite , though @xmath113 as defined by eq . ( [ drhobar ] ) still exists . we derive the sum rules by comparing the coefficient of @xmath23 in eq . ( [ leadqrho ] ) @xmath118^{1-\xi } } { \textstyle ( 1-\xi ) } \\[0.6 cm ] \ln(\ln q^2 ) \end{array } + { \cal o}\left({1\over q^2},{1\over[\ln q^2]^\xi}\right ) \quad \begin{array}{l } \mbox{if $ \xi\neq 1 $ } \\[0.6 cm ] \mbox{if $ \xi=1 $ } \end{array}\right . \label{leadingrh}\ ] ] and the corresponding coefficient in eqs . ( [ asyope ] ) @xmath119_\kappa\rangle [ g^2(q)]^{\eta } + { \cal o}\left(\frac{1}{q^2},[g^2(q)]^{\eta+1}\right ) \ , , \ ] ] where we are only considering cases with @xmath120 and @xmath121 . furthermore , we are presently interested in exact sum rules that can be derived with one - loop calculations and , therefore , we only compare leading orders in @xmath72 . then there exist three possibilities , depending on the value of @xmath122 , which can be calculated using the ope . \(1 ) if @xmath123 , then @xmath124 ( or a bigger integer ) and @xmath125_\kappa\rangle \ , . \label{cons}\ ] ] this is the result one would get by naively expanding @xmath126 without worrying about the convergence of the moments and disregarding logarithmic corrections in the ope . \(2 ) if @xmath127 , then @xmath128 and @xmath129 \(3 ) if @xmath130 ( positive powers of @xmath131 in eq . ( [ dope ] ) ) or if the term @xmath132 appears in eq . ( [ dope ] ) , then @xmath133 or @xmath134 , respectively . this implies that @xmath135 we remark that , even in this case when the moment is not finite , the asymptotic expansion is still well defined . it is nice to see that whether the zeroth moment of the subtracted sf exists is reflected directly through the leading power of @xmath35 in the ope series . our main results , eq . ( [ cons ] ) and eq . ( [ nonc ] ) , can be expressed in physical terms as follows . the zeroth moment of a sf for a current @xmath32 whose ope series yields @xmath136 is independent of @xmath0 and @xmath1 , while the same moment for a current with @xmath123 changes with @xmath0 and @xmath1 proportionally to the corresponding change(s ) of the condensate(s ) of the leading operator(s ) . although @xmath137 and @xmath138_\kappa\rangle$ ] can separately depend on @xmath19 , their product must be independent of @xmath19 , since the zeroth moment is independent of @xmath19 . at this point several general comments are appropriate : \(1 ) our derivation relies on the fact that an asymptotically free theory allows a perturbative expansion at short distances , making practical the use of the ope and of the rge . we understand why only short distance physics is involved if we realize that the integral over frequencies reduces eq . ( [ corr ] ) to the ensemble average of the equal - time commutator of the currents . therefore , results such as eq . ( [ cons ] ) and eq . ( [ nonc ] ) are completely determined by the one - loop structure of the theory and the particular current under exam . \(2 ) flavor , or other non - dynamical quantum numbers , does not change the expansions at the one - loop level in an essential way . therefore , one can derive analogous sum rules by using other kind of subtractions , instead of the one we adopted . one such example is given by the exact weinberg sum rules @xmath139 in the chiral limit @xcite . \(3 ) the derivation of sum rules for higher moments of the sf requires the complete cancelation of all the lower dimensional operator terms , not just the leading @xmath35 . in particular , one also needs current quark mass corrections to the wilson coefficients . without appropriate subtractions , higher moments do not even converge @xcite . \(4 ) it is essential to properly take into account the logarithmic corrections when deriving exact sum rules , since the logarithmic corrections not only dictate whether @xmath113 satisfy eq . ( [ cons ] ) or eq . ( [ nonc ] ) but they also control the very existence of @xmath113 @xcite . this procedure is in sharp contrast with the usual qcd sum rule approach , where the convergence issue is by - passed by applying the borel improvement by explicitly introducing a cut - off parameter ( the borel mass ) . \(5 ) we believe that the @xmath2-dependent part of the leading condensate appearing in eq . ( [ cons ] ) does not suffer from the infrared renormalon ambiguity . in fact , only the perturbative term @xmath24 can generate contributions to the leading condensate that are dependent on the prescription used to regularize these renormalons . but @xmath24 is independent of @xmath0 and @xmath1 : any prescription dependence cancels out when we make the subtraction in eq . ( [ ddisp ] ) . on the contrary , unless we generalize eq . ( [ ddisp ] ) and make other subtractions , sum rules that involve non - leading condensates are , in principle , ambiguous . \(6 ) it is well - known that conserved operators are not renormalized and , barring anomalous violations , verify the same `` classical '' identities that can be derived at the tree level . this fact is also verified in the sum rules . in fact when both the currents and the operators are conserved ( @xmath123 ) and we obtain the result of eq . ( [ cons ] ) , i.e. one can use the `` naive asymptotic expansion '' to derive the sum rule . conservation of the current alone is not enough to warrant a `` classical '' identity . in the preceding section we have derived sum rules valid for any asymptotically free theory . in this section we illustrate the procedure in the 1 + 1 dimensional gross - neveu model @xcite in the infinite-@xmath3 limit . on one hand , we can derive the sum rules in eqs . ( [ cons ] ) and ( [ nonc ] ) by explicitly calculating wilson coefficients , @xmath140- and @xmath141-functions in the vector and pseudoscalar channels , following the general procedure discussed in section ii . on the other hand , since this model is soluble , we can obtain the exact spectral function at arbitrary @xmath2 , and then explicitly verify both that the sum rules are satisfied and that the asymptotic expansion of the spectral function has the form given in eq . ( [ rholn ] ) . moreover , we can also calculate the condensates , and therefore explicitly check that the ope really matches the asymptotic expansion of the exact spectral integral . the lagrangian of the gross - neveu model is @xmath142\ , , \ ] ] where @xmath143 is a two - component dirac spinor and has @xmath3-component in the internal space . equivalently , we can write @xmath144 where @xmath145 and @xmath146 are auxiliary fields . the coupling constant @xmath46 is independent of @xmath3 and held fixed in the limit @xmath147 . this model is asymptotically free when @xmath148 , and the chiral symmetry is dynamically broken at @xmath149 and @xmath150 to the leading order in @xmath151 @xcite . in the following we give the exact solution at @xmath152 with finite @xmath2 . although formulas are explicitly considered in the symmetry breaking phase , they are also valid in the symmetric phase provided that the vanishing limit of certain condensates ( such as the dynamical fermion mass ) is properly taken . we wish to remark on the well - known fact that the limit @xmath147 here should not be interpreted as a starting point for an expansion of the model at finite @xmath3 , but rather as different model in itself , which is in fact the model we decided to use for the purpose of illustration . moreover , there exist arguments @xcite suggesting that this model in the limit @xmath147 is actually more relevant to @xmath153 phenomenology than the model with finite @xmath3 . since the procedure to obtain the exact solution in the large-@xmath3 limit is rather standard and some of the intermediate steps can be found in the literature , see for instance refs . @xcite , we only give the definitions and final results . the gap equation and the phase diagram can be derived from the first derivative of the effective potential @xmath154 we introduce a momentum cutoff @xmath69 and add the counterterm @xmath155 , where @xmath19 is the subtraction point . then the renormalized effective potential becomes @xmath156 where we have defined @xmath157 and , in the second equality , we have eliminated the subtraction point by introducing the dynamical fermion mass at @xmath22 , i.e. @xmath158 . the effective potential is then obtained by integrating eq . ( [ effpotp ] ) : @xmath159\right\}\ , . \label{effpot}\ ] ] the gap equation is simply given by @xmath160 the system possesses `` metastable states '' at those values of @xmath2 for which the gap equation has more than one minimum solutions : the solution with lowest @xmath161 selects the true ground state . when @xmath2 s are such that @xmath162 is the only minimum solution to the gap equation , these points in the @xmath2-plane defines a second order phase transition line , which separate the symmetry - broken phase , i.e. @xmath163 , from the symmetric phase , i.e. @xmath162 . taking the limit @xmath164 in eq . ( [ effpotp ] ) , after some suitable rewriting , this critical line obeys the following equation @xmath165 \cosh[(k-\mu)/2 t ] } - \int_m^\infty { dk\over k}\ , f(k;t,\mu)\ , , \label{gap0}\ ] ] whose numerical solutions are plotted in fig . [ fig1 ] in full line . the full line turns into a dotted line at the heavy dot , where another minimum ( dash line ) appears and becomes lower than the minimum given by eq . ( [ gap0 ] ) . since the mass gap is finite on the dash line ( given by the solution of @xmath166 with @xmath167 , where prime denote a derivative with respect to @xmath168 ) , the dash line is a first order phase boundary . the critical temperature at @xmath150 is given by @xmath169 . the `` tricritical point '' ( the heavy dot in fig . [ fig1 ] ) can be found by imposing that @xmath170 : @xmath171 @xcite . when @xmath149 the chemical potential at which the first order phase transition takes place is @xmath172 @xcite . the basic ingredient to build up the correlator is the fermion bubble graph . in the euclidean region the pseudoscalar bubble , i.e. the free bubble graph between two @xmath173 currents , can be conveniently expressed as @xmath174 where @xmath175 and @xmath176 here and in the following @xmath19 is always the renormalization subtraction point , @xmath177 is the dynamically generated fermion mass , and @xmath178 . we shall also use the short - hand notation @xmath179 for instance , the use of this notation makes possible to write the gap equation as @xmath180 in the large-@xmath3 limit , the correlator in pseudoscalar channel is given by the geometric sum of the bubble graph : @xmath181 in the euclidean region , we can then expand this correlator in the @xmath13 limit as @xmath182 where we have introduced the running coupling constant @xmath183 and used the gap equation eq . ( [ gapsh ] ) . we can recognize the first term in eq . ( [ corrq ] ) as the perturbative contribution at @xmath22 , @xmath184 , which corresponds to the term @xmath185 in the generic ope expansion shown in section [ general ] . the other two terms in eq . ( [ corrq ] ) can be interpreted as contributions from the condensates . as we will verify later , in fact , @xmath186\rangle_{t,\mu}$ ] , while @xmath187\rangle_{t,\mu}$ ] is the condensate of the energy - momentum tensor . we shall see that the perturbative contribution and the coefficients of the condensates are indeed the ones obtained by the ope of @xmath188 , and therefore the expansion of the correlator shown in eq . ( [ corrq ] ) is of the form of the ope . note that in defining the condensates we have absorbed factors of @xmath3 in order to simplify the notation . to obtain the spectral function we need the bubble graph in the time - like region @xmath189 : @xmath190 and @xmath191 where @xmath192 stands for the principle value of the integral . the complete spectral function has the form @xmath193 the pole contribution comes from the bound state ( the pion ) , whose mass is the solution to the equation @xmath194 in the region @xmath195 . this equation has in fact solution @xmath196 , and the coupling constant of this massless pion to its constituents is given by @xmath197 the continuum part of the spectral function is related to the bubble @xmath198 in the time - like region through @xmath199 in order to get an idea of the dependence of the spectral function on @xmath2 , we plot @xmath200 for several typical values of @xmath2 , which are indicated in the phase diagram in fig . [ fig1 ] . in fig . [ fig2 ] we show @xmath200 for values of @xmath2 that go from the symmetry - broken phase to the symmetric phase through the second order line , while in fig . [ fig3 ] we show @xmath200 as @xmath2 go through the first order boundary . we immediately notice that the behavior of the spectral function changes drastically near the second order phase boundary . this is a clear reflection of the low energy critical phenomena related to the continuous phase transition . in fact , the pseudoscalar correlator ( in the symmetric phase only , since it is then degenerate with the scalar correlator ) at @xmath201 can be regarded as the chiral susceptibility , and should diverge as @xmath202 ( with @xmath203 in this model ) . at the boundary of the first order phase transition ( fig . [ fig3 ] ) , but still relatively close to the `` tricritical point '' , the same qualitative deformation is present , since a weak first order transition shares certain qualitative features of a second order transition , even if we do not expect any real divergence . it is this kind of dramatic behavior of the spectral function near phase transitions that makes practical parameterizations nearly impossible . the chiral restoration in qcd is expected to be either continuous , or at least a smooth crossover , and the spectral functions in scalar and pseudoscalar channels should show a behavior qualitatively similar to the gross - neveu model , i.e. strong peak right above the origin near the phase transition . we shall comment more on this issue after we derive the sum rules for qcd in section [ qcd ] . it is instructive to expand @xmath200 in the limit @xmath204 . we shall see that this asymptotic expansion is of the form assumed in section [ general ] . furthermore , once substituted in the dispersion integral , it generates the ope series of eq . ( [ corrq ] ) according to the procedure developed in section [ general ] . to the leading order in @xmath205 , we find @xmath206 and @xmath207\rangle_{t,\mu}\bigr ] + { \cal o}\bigl({1\over\omega^4}\bigr)\biggr\ } \ , , \ ] ] which leads to the following asymptotic form for the continuum part of the spectral function @xmath208\rangle_{t,\mu } \nonumber \\ & + & { 4g^2(\omega)\over\omega^2}\bigl({g^2(\omega)\over g^2}\bigr)^2 \langle[\theta_{00}]\rangle_{t,\mu } \bigl\{1+{\cal o}\bigl(g^2(q)\bigr)\bigr\}\ , . \label{rho_asym}\end{aligned}\ ] ] here we only concern with the @xmath35-dependent terms and postpone the pure @xmath23 terms ( related to @xmath111 ) to a later subsection . we recognize again the first term in eq . ( [ rho_asym ] as the asymptotic perturbative spectral function . moreover , the @xmath2-dependent part of eq . ( [ rho_asym ] ) has indeed the general form of eq . ( [ rholn ] ) . in connection with the comments we made after eq . ( [ rholn ] ) , we point out that eq . ( [ rho_asym ] ) has been obtained by expanding the spectral function in the limit @xmath209 at fixed @xmath0 . had we made instead a high-@xmath0 expansion ( @xmath210 ) at fixed @xmath26 , we would have obtained , for instance , @xmath211 from the third term of eq . ( [ repi ] ) . upon substituting eq . ( [ rho_asym ] ) in the dispersion integral , and using the identity @xmath212 where @xmath213 and we have kept only the @xmath35-dependent part , we find the following series to the leading order in @xmath35 , @xmath214\rangle_{t,\mu}\over q^2 } + 4\bigl({g^2(q)\over g^2}\bigr)^2 { \langle[\theta_{00}]\rangle_{t,\mu}\over q^2}+\cdots \label{gndisint}\ ] ] this result is in agreement with the ope series of eq . ( [ corrq ] ) , which has been directly expanded from the correlator in the euclidean region . the fact that no pure @xmath23 term appears in eq . ( [ gndisint ] ) implies that the zeroth moment of the subtracted spectral function vanishes : @xmath215 . the zeroth moment of the spectral function without the subtraction is not convergent , because of the contribution from the perturbative term , even if one can define it through a proper analytic continuation that yields @xmath216 . we have just verified that in the gross - neveu model the exact correlation function can indeed be expanded in an asymptotic series precisely in the form of the ope . now we shall calculate the wilson coefficients , @xmath140-function and appropriate anomalous dimensions for the ope , and verify that they match the coefficients and exponents of this asymptotic series . furthermore , though it is beyond the scope of the ope and rge , we shall also calculate the condensates . in this way , we explicitly verify that the asymptotic expansion of the exact correlation function is identical to the ope . the one - loop @xmath140-function of the 1 + 1 dimensional gross - neveu model has already been calculated in the original paper @xcite : @xmath217 , i.e. according to our notation @xmath218 . the anomalous dimensions for pseudoscalar current @xmath173 and four - quark operator @xmath219 can be calculated using the feynman diagrams shown in fig . [ fig4]a and fig . [ fig4]b , respectively . according to the notation of eq . ( [ rgef ] ) we find @xmath220 , and @xmath221 . the energy - momentum conservation makes the anomalous dimension for @xmath222 vanish , i.e. @xmath223 . the relevant wilson coefficients corresponding to operators @xmath224 and @xmath222 can be calculated , to leading order in @xmath35 , using the feynman diagrams shown in fig . [ fig5]a and fig . [ fig5]b , respectively . the resulting coefficient are @xmath225 ( @xmath226 according to notation in eq . ( [ asyope ] ) ) and @xmath227 ( @xmath228 ) . these results lead to the exponents @xmath229 and @xmath230 , which are precisely what we found in eq . ( [ gndisint ] ) . these explicit one - loop calculations exactly match the solution in eq . ( [ gndisint ] ) . at last a comment on the one - loop calculation : this result is exact to the leading order in @xmath151 and only involves insertion of the quark - bubble chain . since the quark - bubble behaves like @xmath231 , this insertion is equivalent to substituting the coupling constant @xmath46 with the running coupling constant @xmath35 . so one can easily identify those sets of feynman graphs whose sum leads to the solution of the rge in eq . ( [ rge ] ) . the ope itself does not specify how the relevant condensates are calculated , since the ope is only a rge - improved perturbative procedure . the calculation of the condensates can be done only in a non - perturbative context . here , we carry out this calculation using the @xmath151 expansion . the bare quark condensate @xmath232 is given by the dynamical quark tadpole graph @xmath233 using the standard contour integral technique to carry out the matsubara frequency sum , we obtain @xmath234 the renormalized quark condensate is obtained by replacing @xmath235 with the subtraction point @xmath19 in the above equation @xmath236\rangle_{t,\mu } = m\langle\!\langle e_k^0\rangle\!\rangle -{m\over\pi}\ln\bigl({\kappa\over m}\bigr)=-{m\over g^2}\ , .\ ] ] in the last step we have used the gap equation eq . ( [ gapsh ] ) . notice that the renormalization of this condensate is not independent of the one carried out for the effective potential . in fact , the @xmath145-field is just an auxiliary field : @xmath237 . therefore , the gap equation gives us not only the expectation value of the @xmath145-field , but also of @xmath238 : @xmath239\rangle_{t,\mu}$ ] . so the renormalizations of @xmath238 and @xmath46 are related . the four - quark condensate factorizes in the large-@xmath3 limit , and we find @xmath240\rangle_{t,\mu}= \langle[\bar{\psi}\psi]\rangle_{t,\mu}^2=m^2/g^4 $ ] . the bare kinetic energy expectation value is given by @xmath241 after removing the @xmath2-independent volume and quadratic divergences , and introduced a subtraction point for the logarithmic divergence the renormalized kinetic energy becomes @xmath242\rangle_{t,\mu } = \langle\!\langle e_k^2 \rangle\!\rangle -{m^2\over 2\pi}\ln\bigl({\kappa\over m}\bigr)\ , .\ ] ] after subtracting the trace term from the kinetic energy , we finally obtain the expectation value of the traceless energy - momentum tensor @xmath243\rangle_{t,\mu } \equiv \langle[\bar{\psi}i\gamma_0\partial_0\psi -{1\over 2}m\bar{\psi}\psi]\rangle_{t,\mu } = \langle\!\langle e_k^2\rangle\!\rangle -{m^2\over 2}\langle\!\langle e_k^0\rangle\!\rangle\ , .\ ] ] in the symmetric phase ( @xmath244 ) , the expectation value of the energy - momentum tensor has the very simple form at @xmath150 @xmath243\rangle_{t,\mu=0}={\pi\over 6}t^2\ , .\ ] ] one can verify that inserting these condensates in the ope , whose coefficients and exponents we have already calculated , reproduces the asymptotic expansion of the exact correlator in eq . ( [ corrq ] ) . let us first notice that the sum rule @xmath245 converges very slowly . in fact , the subtracted spectral function has the leading behavior @xmath246 where @xmath247\rangle_{t,\mu})\over g^4(\omega)+4}\biggr\}\ , . \label{drho2}\ ] ] since @xmath248 , the leading behavior in eq . ( [ drho2 ] ) is @xmath249^{-1}$ ] , and it is clear that the sum rule only converges logarithmically , i.e. @xmath250 . moreover , it is only thanks to the the logarithmic corrections that the sum rule is finite . had we disregarded this corrections , we would have found @xmath251 , and the zeroth moment of the subtracted spectral function would not exist . since the exact spectral function can not be integrated analytically , we verify the exact sum rule numerically . we take advantage of the fact that the integral of @xmath252 in eq . ( [ drho2 ] ) can be done analytically to improve the numerical convergence of the integral . we rewrite the sum rule as @xmath253 where the integral of @xmath252 can be done analytically and we obtain @xmath254\rangle_{t,\mu})\over ( \ln a)^2+\pi^2/4}\biggr\}+{\cal o}\bigl({1\over a^2}\bigr)\ , .\ ] ] this formula can be easily verified numerically with high accuracy , since the error is now @xmath255 , rather than @xmath256 . it is reminded that @xmath257 in the above equation should be obtained from the complete sf in eq . ( [ rho_full ] ) , including both @xmath88 and @xmath111 . there are several reasons to study the correlator in the vector channel in the gross - neveu model . first , the vector channel gives us an example where the zeroth moment of the subtracted sf does not exist , but the asymptotic procedure can still be carried out . moreover , in connection with the fact that the vector current is conserved , this spectral function is not affected by logarithmic corrections , and we can use it to illustrate the use of the mellin transform method to obtain the asymptotic expansion . in addition , very much like in the qcd case , we shall see that the baryon number susceptibility changes drastically near the chiral restoration transition , even if we do not expect any real singularity associated with critical phenomena in the vector channel . for convenience , we only consider the case of @xmath150 . the feynman graphs that contribute to the vector correlator are depicted in fig . note that if there were no mixing between vector and pseudoscalar channel , which in fact is only present in 1 + 1 dimensions , the sole contribution would come from the first graph . the total contribution from all the graphs in fig . [ fig6 ] is @xmath258 we define the spectral function in vector channel as @xmath259 then we find the following explicit expression for @xmath260 @xmath261 it is clear that leading asymptotic behavior of the spectral function is @xmath262 , which implies that the zeroth moment of the spectral function does not exist ( it is divergent ) . in the section [ general ] we have shown that , corresponding to this divergence , we must find negative powers of @xmath35 ( positive power of @xmath263 ) in the ope series . when there are no explicit logarithms in @xmath264 , it is easier to make the asymptotic expansion by using the mellin transform method , and in particular the convolution property of the mellin transform @xcite , @xmath265m[f;1-z]\ , , \label{convolut}\ ] ] where @xmath266 $ ] is the mellin transform of @xmath267 , defined by @xmath268\equiv\int_0^\infty dt\ , t^{z-1}h(t)\ , .\ ] ] the value of @xmath269 in eq.([convolut ] ) depends on the asymptotic behaviors of @xmath267 at @xmath270 and @xmath271 at @xmath272 respectively ; details can be found , for instance , in chapter 4 of ref . @xcite . now let us illustrate how to use this method to expand the power part of the spectral function ( the part containing exponential dumping factors can be expanded naively ) @xmath273 whose mellin transform is @xmath274= ( 4m^2)^z\sqrt{\pi}\gamma(1-z)/\gamma(3/2-z ) \ , .\ ] ] using the convolution , eq . ( [ convolut ] ) , and the mellin transform of @xmath82 @xmath275=\pi(q^2)^{-z}/\sin\pi(1-z ) \ , , \ ] ] we find ( in this case @xmath269 in eq.([convolut ] ) obeys @xmath276 ) @xmath277_{z = n } \nonumber \\ & = & { 4m^2\over q^2}\ln\bigl({q^2\over m^2}\bigr ) -{8m^4\over q^4}\bigl[1+\ln\bigl({q^2\over m^2}\bigr)\bigr ] + { \cal o}\bigl({1\over q^6}\bigr)\ , . \label{asymmell}\end{aligned}\ ] ] similarly to the pseudoscalar correlator , it is possible to check this result in two ways . we can directly expand the power part ( temperature independent part ) of eq . ( [ vvceucl ] ) , and we can derive the ope ( @xmath278 and @xmath279 for the @xmath23 term ) . as promised , there indeed appears a positive power of @xmath263 in the ope series . at the same time , the singularities of @xmath280 $ ] at @xmath281 positive integer hints the fact that all the non - negative integer moments of @xmath282 do not exist . on the other hand , the asymptotic expansion in eq . ( [ asymmell ] ) is well defined . even though the vector current is conserved , due to the fact that the anomalous dimension of @xmath224 is not zero , the correct result can not be obtained by a naive asymptotic expansion in this case , in accordance with the remark number ( 6 ) at the end of section ii . this same result could have been obtained also by the general method developed in section [ general ] , but it should be clear by now that the use of the mellin transform , when possible , is more straightforward and require less labour . the baryon number susceptibility ( @xmath283 ) is defined as the vector correlator in the limit @xmath284 . from eq . ( [ vvceucl ] ) we find @xmath285 notice that since we have contracted the lorentz indices when defining our vector correlator , @xmath283 is nonpositive definite . it is also possible that our definition of @xmath283 differs from others by an additive constant , which is irrelevant to the temperature dependence we are considering . in the symmetric phase , the limit @xmath286 should be take with some care : the result is @xmath287 , when @xmath288 . the complete temperature dependence of @xmath283 is displayed in fig . as expected , there is no divergence or critical phenomena in this channel . nevertheless , we do see a sudden rising of @xmath283 in correspondence to the relatively rapid drop of the dynamical quark mass , which we show in fig . [ fig7]a for comparison , near the phase transition region . after a lengthy detour to the gross - neveu model let us now back to the main interest of our work , to derive exact sum rules and to explore the corresponding phenomenological consequences in qcd . as we will see , qcd shares many of the qualitative features observed in the gross - neveu model , implying the simple fact that any drastic change induced by a phase transition or rapid crossover would necessarily be reflected through the spectral function in proper channels at low frequencies , independent of the details of the model . let us consider four correlators of mesonic currents in qcd . the correlation between scalar currents and the one between pseudoscalar currents involve non - conserved operators , @xmath289 and @xmath290 , whose anomalous dimensions are @xmath291 . one the other end , correlators between vector currents and axial - vector currents involve conserved quantities , @xmath292 and @xmath293 , whose anomalous dimensions vanish ( @xmath294 ) . in qcd there exist several dimension - four operators , but all of them have non - positive anomalous dimensions , therefore the two correlators between nonconserved currents have @xmath295 and eq . ( [ nonc ] ) applies , _ i.e. _ the zeroth moments of their sf s are independent of @xmath0 and @xmath1 . on the other hand , the two conserved currents have , in correspondence with conserved operators @xmath296 and a generalization of eq . ( [ cons ] ) applies . in this case in fact there are three dimension - four operators with zero anomalous dimension . two of these operators are lorentz scalars : @xmath297 and @xmath298 , while the third is the energy - momentum tensor : @xmath299 the sum rules for the vector and axial - vector currents ( with the lorentz indices contracted ) are : @xmath300\rangle + { \delta\langle[\alpha_s g^2]\rangle\over 2\pi } + 8\delta\langle [ \theta_{00}]\rangle\ , , \label{va}\ ] ] with @xmath301 and @xmath302 for vector and axial - vector cases , respectively . obviously , the general remarks at the end of section ii apply to the sum rules derived above . however , there is a new feature of qcd which may not be shared by all asymptotically free theories . it is known that qcd ope series has explicit instanton induced corrections that can not be related to condensates of some local operators . but these exact sum rules are not explicitly affected by this problem , although the value of the condensates certainly have instanton contributions . the reason is that the instanton singularities in the borel - plane are located on the positive axis starting at @xmath303 , which in turn implies that the explicit contribution of instantons to correlation functions are of higher order than @xmath23 @xcite in the ope series . finally , let us discuss some of the phenomenological consequences of these exact sum rules . the power of these exact sum rules rests on their generality , in the sense that there is no assumptions on the functional form of the sf s and they can be applied anywhere in the @xmath2-plane . first let us consider the pseudoscalar channel . the sum rule @xmath215 implies that , in the broken - chiral - symmetry phase , the change of the pion pole induced by @xmath0 or @xmath1 is exactly compensated by a corresponding change of the continuum part of the sf . next let us consider the scalar correlation function at zero frequency , i.e. the chiral susceptibility ( a measure of the fluctuation of the chiral order parameter ) , @xmath304\rangle_{t,\mu } = \int_0^\infty\!\ ! du^2\ , { \rho(u;t,\mu)\over u^2}\ , , \label{chisus}\ ] ] which diverges when @xmath2 approaches the phase boundary , provided the chiral restoration is a continuous transition . the divergence of the chiral susceptibility near the phase transition can only be produced in eq . ( [ chisus ] ) by singularities very close to the origin ; singularities not near the origin are not compatible with the sum rule @xmath215 . thus , when approaching the phase boundary in the @xmath2 plane , the threshold of the spectral function vanishes ( since there is no massless pole in the chirally symmetric phase ) , and a strong peak develops right above threshold . in the chirally symmetric phase the pseudoscalar and scalar channels are degenerate , therefore we expect the same behavior for the pseudoscalar sf . this strong peak in the pseudoscalar and scalar sf s , which is intimately connected with critical phenomena of a diverging susceptibility and correlation length near the phase transition , can be interpreted as some kind of quasi - particle , thus confirming the qualitative picture , originally proposed in the context of the nambu - jona - lasinio model @xcite , of the appearance of soft modes near the chiral phase transition . we wish to remark that the sum rule has been used only to exclude the logical possibility of the sf developing singularities at finite energies . but the sum rule is not a necessary condition for the appearance of a peak . the presence of a second order phase transition ( infinite correlation length ) is the real physical cause of the peak . nevertheless , one should not assume that if the phase transition is not second order the zero-@xmath0 or perturbative spectral function is a good approximation to the real spectral function . even if the chiral restoration turns out to be a cross - over or weak first order transition ( finite but large correlation ) , as the lattice data seem to indicate @xcite , we still expect the same qualitative features , though less pronounced : a peak develops but it does not actually diverge at the transition . our calculation in the gross - neveu model is a clear illustration of this situation . while fig . ( [ fig2 ] ) shows the divergent peak of the spectral function near the second order boundary , we see in fig . ( [ fig3 ] ) that the spectral function is also strongly peaked near the first order transition as long as we are not too far away from the `` tricritical point '' . we believe that similar results hold in the vector and axial - vector channels , even if our conclusions can not be as strong as in the previous case , because of two main differences . the first is that now @xmath305 . this problem is not very serious , since we only need that @xmath113 is not singular crossing the phase boundary . and this can still be argued by means of the sum rule in eq . ( [ va ] ) and the fact that , for a continuous phase transition , the changes of the thermal energy @xmath306\rangle$ ] , and of both condensates @xmath307\rangle$ ] and @xmath308\rangle$ ] should behave smoothly across the critical line . the most serious difference is that , in these channels , we can not argue on physical grounds that the corresponding susceptibility diverges . nonetheless , there exist lattice simulations @xcite showing that the so - called quark number susceptibility rapidly increases in the transition . thus it is plausible that also in the vector channel the sf has a vanishing threshold , and accumulates strength just above it . a similar argument in the context of the nambu - jona - lasinio model can be found in ref . we used the operator product expansion and the renormalization group equation to derive exact sum rules at finite @xmath0 and @xmath1 valid for asymptotically free theories . our derivation explicitly show that logarithmic corrections can not be neglected . we found that , depending on the theory and on the current under study , the zeroth moment of a spectral function is either independent of @xmath0 and @xmath1 , or its change is related to the corresponding changes of the condensates of operators of lowest dimension . in particular the zeroth moment of the scalar and pseudoscalar mesonic currents in qcd are independent of @xmath0 and @xmath1 . as a consequence any change in the strength of the pion pole must be exactly compensated by a change in the continuum contribution . we also infer that spectral functions in the scalar and pseudoscalar channel should dramatically change near phase transition . we also find that the zeroth moment of the vector and axial - vector mesonic currents in qcd changes with @xmath0 and @xmath1 , and these changes are related to the corresponding changes of the condensates . due to their generality , these exact sum rules strongly constrain the qualitative shape of sf s , in particular , near phase transitions . since it appears that a strong deformation of the spectral function from its counter part at @xmath22 and perturbative cases is the most likely scenario near phase transitions , the information carried by these exact sum rules is particularly welcome . we urge whoever parameterizes a spectral function , e.g. in qcd sum rule type of calculations or to interpret lattice simulation data , to incorporate these exact constraints . we have also illustrated in great detail the derivation of the sum rules , and confirmed their validity , in the gross - neveu model , where we can also calculate the exact spectral function . even if this model calculation is meant mostly as an illustration of many delicate issues of the derivation ( such as convergence , asymptotic expansion , mellin transform , high momentum expansion vs. high temperature expansion , etc . ) , it is nevertheless comforting to find that all our general expectations about the spectral function are explicitly verified in the model .

monte carlo methods were used to calculate heat capacities as functions of temperature for classical atomic clusters of aggregate sizes that were bound by pairwise lennard - jones potentials . the parallel tempering method was used to overcome convergence difficulties due to quasiergodicity in the solid - liquid phase - change regions . all of the clusters studied had pronounced peaks in their heat capacity curves , most of which corresponded to their solid - liquid phase - change regions . the heat capacity peak height and location exhibited two general trends as functions of cluster size : for to 36 , the peak temperature slowly increased , while the peak height slowly decreased , disappearing by ; for , a very small secondary peak at very low temperature emerged and quickly increased in size and temperature as increased , becoming the dominant peak by . superimposed on these general trends were smaller fluctuations in the peak heights that corresponded to `` magic number '' behavior , with local maxima found at and 49 , and the largest peak found at . these magic numbers were a subset of the magic numbers found for other cluster properties , and can be largely understood in terms of the clusters underlying geometries . further insights into the melting behavior of these clusters were obtained from quench studies and by examining rms bond length fluctuations .

physics0107013

the investigation of small atomic and molecular clusters remains an active area of research , both theoretically@xcite and experimentally.@xcite computer simulations of clusters have revealed a wide diversity of fundamental behavior that is highly sensitive to many factors , such as the cluster aggregate size @xmath4 , the potential used to model the intermolecular interactions , and even the simulation method itself . the lennard - jones ( lj ) potential is the most commonly used model potential for computer simulations of cluster properties , and it is especially useful for modeling rare gas clusters . a key aspect of this model when applied to small and medium sized atomic clusters is that the resulting potential surfaces have global minima that are dominated by icosahedral - like structures , with fivefold symmetries that are markedly different than the closest - packed structure found in corresponding bulk systems . a major motivation for studying clusters is the insight that can be obtained in understanding the transition from finite to bulk behavior . because of their high proportion of surface atoms , many clusters have physical properties that typically exhibit a very irregular dependence on their aggregate size , with certain sizes standing out in particular . for example , the mass spectra of clusters typically exhibit especially abundant sizes that often reflect particularly stable structures , specially reactive clusters , or closed electronic shells,@xcite these `` magic number '' sizes have been of great theoretical interest,@xcite and much work has been done in relating magic number sequences to cluster structure in terms of `` soft '' sphere packings , since many magic number sizes in lj clusters such as @xmath8 , 19 , 55 , and 147 correspond to compact structures that are especially stable.@xcite many cluster properties obtained from simulation studies such as binding energies , root mean square ( rms ) bond length fluctuations and heat capacities also show very irregular dependencies on the cluster size that mostly correspond to their usual magic number sequences . for example , the heat capacity peak values per atom for lj@xmath9@xcite and lj@xmath10@xcite are especially large , which can be attributed to the exceptional thermodynamic stability of their lowest - energy isomers relative to their next higher energy configurations . to more easily compare various cluster property size dependencies in this report , i have generalized the concept of magic numbers for arbitrary cluster properties to refer to those cluster sizes having particularly enhanced values relative to their immediate neighbors . in a previous study , i examined these generalized magic number effects in lj clusters ranging in size from 4 to 24 atoms.@xcite in this study , i extend the sequence to 60 atoms , which completes the filling of the second mackay icosahedral shell ( to @xmath11 ) and just begins the filling of the third shell ( to @xmath12 ) . the issue of cluster phase transitions is fundamentally important , and so has been of great interest . because of their small size , clusters do not have the sharp first order transition characteristic of bulk melting , but instead show changes occurring over a range of temperatures , typically a few degrees k.@xcite although the phase - change range generally decreases as the cluster size increases , coalescing to the bulk transition temperature as @xmath4 approaches infinity , both the range and the location show much variation as a function of @xmath4 . cluster solid - liquid phase - change regions can be determined from computer simulations typically by identifying characteristic changes in certain cluster properties . for example , cluster potential energies often change markedly as functions of temperature in the phase - change regions , resulting in distinct peaks in the corresponding heat capacity curves ; rms bond length fluctuations often rise abruptly with increasing temperature at the beginning of the phase - change regions , and the fractions of quenched cluster configurations corresponding to the lowest - energy isomers often decrease sharply with increasing temperature in these regions . defining the phase - change region on the basis of such behavior is somewhat ambiguous , though , since the temperature ranges associated with these different properties do not always coincide , and occasionally do not even overlap . for a given property , the phase - change region temperature range can also depend on the simulation method ; for example , for ar@xmath9 , the rise in the rms bond length fluctuations as calculated from canonical monte carlo simulations occurred about 7 k lower than that calculated from molecular dynamics simulations.@xcite because of these difficulties , it is convenient to loosely define the solid - liquid phase - change region as the temperature region between the lower - temperature , solid - like region ( where the clusters are mostly confined to their lower energy configurations , undergoing isomerizations only rarely at the higher - temperature end of the region ) and the higher - temperature , liquid - like region ( where the clusters exhibit easy isomerization and diffusion ) . cluster behavior in the solid - liquid phase - change regions is particularly complicated , with some cluster sizes showing coexistence behavior , where `` solid - like '' and `` liquid - like '' isomers dynamically coexist,@xcite while other cluster sizes show a relatively smooth progression of isomerizations between locally similar but globally distinct configurations occurring throughout this region.@xcite although clusters having magic number sizes such as 13 and 55 have pronounced heat capacity peaks that correspond closely to the solid - liquid phase - change region determined from other cluster properties , other cluster sizes such as @xmath13 , 16 and 17 do not have heat capacity peaks in their phase - change regions.@xcite since heat capcity peaks have often been used to identify phase - change regions,@xcite characterizing their temperature dependencies for sequences of lj clusters can help elucidate the underlying reasons for such varied behavior . the accurate determination of cluster properties such as the heat capacity is hampered by the poor convergence of standard simulation methods in the phase - change regions . this poor convergence is a consequence of quasiergodicity , or the incomplete sampling of configuration space.@xcite various methods have been developed in recent years to reduce the systematic errors resulting from quasiergodicity , including histogram methods,@xcite jump - walking methods ( j - walking),@xcite smart walking methods ( s - walking),@xcite and parallel tempering methods.@xcite many of these methods are based on the coupling of configurations obtained from ergodic higher - temperature simulations to the quasiergodic lower - temperature simulations . in my previous study of magic number effects in cluster heat capacities , i used j - walking to successfully overcome quasiergodicity . monte carlo j - walking methods couple the usual small - scale metropolis moves made by a lower - temperature random walker , with occasional large - scale jumps that move the random walker out of the confined regions of configuration space that it could not otherwise easily escape within the duration of the simulation . there are two complimentary implementations of j - walking . in tandem j - walking , higher - temperature walkers and lower - temperature walkers are run in tandem , with the higher - temperature walkers occasionally feeding configurations to the lower - temperature walkers . because j - walking obeys detailed balance only approximately , the walkers become correlated whenever a lower - temperature walker accepts a jump , and so it is necessary to run several extra metropolis passes after a jump is accepted to break these correlations , thereby decreasing the overall efficiency . the second method breaks the correlations by running the higher - temperature walkers beforehand , and storing representative samples of the configurations in external files , which the lower - temperature walkers sample periodically in a random fashion . thus , sampling from an external distribution is more efficient , provided that the required distributions are not too large , and this was the implementation i used in the previous study . it was also successful for the first few clusters considered in this study , but for clusters ranging in size from @xmath14 to 37 , which had two heat capacity peaks , the heat capacity curves obtained with different j - walker distributions showed poor reproducibility for the lower - temperature peaks . since the distribution file sizes already totalled more than 2 gb , it seemed prudent to abandon that implementation and switch to the tandem j - walker implementation . however , a recent study by neirotti , _ et al._@xcite of lj@xmath15 , which is notoriously difficult to simulate accurately , showed that the parallel tempering method was remarkably successful in overcoming quasiergodicity and calculating an accurate heat capacity curve . parallel tempering is nearly identical to a parallel version of tandem j - walking , except that instead of copying the higher - temperature j - walker s configuration to the lower - temperature walker s configuration whenever a jump is accepted , the two configurations are exchanged . this has the advantage of maintaining detailed balance , thereby eliminating the need to run extra metropolis passes after accepting an exchange , which makes the method more efficient than tandem j - walking . thus , i decided to use parallel tempering for all of the cluster sizes examined in this study . i begin in section [ sec : method ] with a brief description of the computational methods used in this work and a summary of the calculations undertaken . section [ sec : results ] describes the results obtained for the cluster potential energies , heat capacities , rms bond length fluctuations and quenched configuration distributions , as functions of temperature and cluster size . in section [ sec : discussion ] , i summarize the important conclusions concerning magic number behavior in lj clusters for @xmath16 . the clusters studied were modeled by the pairwise additive lennard - jones potential , @xmath17.\end{aligned}\ ] ] because small clusters can dissociate at higher temperatures and make the cluster definition ambiguous , it is customary to confine the clusters by a constraining potential . for classical monte carlo simulations , a perfectly reflecting spherical constraining potential of radius @xmath18 , centered on the cluster s center of mass , is most convenient , and was used in this work . the classical internal energy and heat capacity were calculated by the usual expressions for an @xmath4-atom cluster ( in reduced units ) , @xmath19 where @xmath20 , @xmath21 , @xmath22 , and @xmath23 . since so many rare gas cluster simulations have been done for ar , i found it convenient to generate the temperature - dependent curves using ar parameters , with @xmath24 and @xmath25 k. relative rms bond length fluctuations have been calculated in many previous cluster simulations.@xcite these were calculated as @xmath26 where the summations are taken over the entire walk . because of the summations being taken over the entire walk , rms bond length fluctuations can not be determined for the j - walking and parallel tempering simulations , whose configuration jumps and exchanges effectively reduce the long metropolis walks to a collection of very many short excursions about the last accepted jump or exchange configuration . thus , quasi - ergodicity remains a problem for rms bond length calculations . caution must also be exercised when interpreting @xmath27 curves , since they are dependent on simulation parameters such as the walk length , as well as the ensemble that is used.@xcite j - walking runs were done for cluster sizes @xmath28 , using sampling from externally stored distributions as described in the preceding study.@xcite distributions for each temperature consisted of several files , totaling @xmath29 configurations , sampled every 100 , 50 , 25 , or 10 passes , depending on the temperature ( 100 passes for higher temperatures having broad distributions , 10 passes for very low temperatures having very narrow distributions ) . the distributions were generated in stages from higher temperatures to lower , beginning with temperatures well within the cluster liquid region ( 50 to 60 k ) . data accumulation was done with walks consisting of @xmath30 total passes , with jumps attempted randomly with a frequency of 10% . since any systematic errors in the higher temperature j - walker distributions can feed into the lower temperature distributions and corrupt them , heat capacity peaks occurring at very low temperatures can be very difficult to calculate accurately . in previous studies of 13-atom binary clusters,@xcite very low temperature heat capacity peaks corresponding to mixing anomalies could be obtained accurately by repeating the full j - walking temperature scan several times , each time with newly generated distributions , and averaging the independent curves . this procedure was followed in this work as well , and at least five j - walking runs were done for cluster sizes 25 to 30 , with the heat capacity curves showing very good reproducibility in each case . for cluster sizes 31 to 38 though , the curves showed very poor reproducibility for the lower - temperature peaks , and so the method was abandoned . despite the j - walking method s inability to reproduce the lower - temperature heat capacity peaks accurately , the method was still useful for obtaining a listing of the important lowest - energy configurations for each cluster size . in each case , representative subsets of the distribution files were quenched using steepest descent methods@xcite and all of the unique isomers found were extracted . while this crude method can not identify all of the configurations for clusters of these sizes , it is well suited for identifying most of the lower energy isomers that dominate the low temperature behavior of clusters . for example , the quenched j - walker distributions for lj@xmath15 contained all of the lowest - energy configurations found by doye _ et al._,@xcite including the global minimum face - centered - cubic ( fcc ) truncated octahedron , which is notoriously difficult to obtain from randomly initialized simulations . these findings imply that j - walking was successful in accessing all of the important regions of configuration space for these clusters , even though it was unable to do so in the statistically representative manner needed to accurately reproduce the heat capacities at very low temperatures . thus , higher temperature j - walker distributions were also generated for cluster sizes 39 to 60 to obtain sets of the important low - energy quenched configurations . metropolis simulations were also run for each cluster . these provided a check of the j - walking and parallel tempering results for those temperature regions where quasiergodicity in the metropolis runs was not a problem , as well as revealing trends in the systematic errors arising from quasiergodicity in the metropolis runs themselves . temperature scans were also generated using ar parameters , with the temperature mesh size set to @xmath31 k , and sometimes to 0.5 k in the phase - change regions . for each temperature , simulations consisted of @xmath32 warmup passes , followed by @xmath33 passes with data accumulation . the scans in each case were started at the lowest temperature from the global minimum configuration obtained from the corresponding j - walker simulations , and were continued well past the cluster solid - liquid phase - change region , with the final configuration for each temperature used as the initial configuration for the subsequent temperature . to gain a better appreciation of the extent of quasiergodicity in the solid - liquid phase - change region , additional metropolis simulations consisting of @xmath34 total passes per temperature point were run for each cluster over a narrower temperature range encompassing the phase - change region . parallel tempering@xcite is very closely related to the tandem and parallel walker implementations of j - walking,@xcite with the essential difference being that instead of the higher - temperature configuration being copied over the lower - temperature configuration whenever a jump is accepted , the two configurations are exchanged . this close similarity between parallel tempering and tandem j - walking makes it very easy to convert a tandem j - walking simulation program to a parallel tempering one . the parallel tempering method s configuration exchange ensures that it obeys detailed balance ( provided that the simulation has been warmed up enough to be in the asymptotic region ) , so there is no need to run the extra metropolis passes to break correlations between the higher - temperature and lower - temperature walkers whenever an exchange move is accepted , which is required for tandem j - walking . in addition to offering greater efficiency , the obeying of detailed balance allows parallel tempering simulations to vary the exchange - attempt frequency in temperature regions where the exchange - acceptance rate is too low . in j - walking simulations , the jump - attempt frequency was typically kept at 10% , since larger values required even more extra metropolis passes to be run whenever a jump was accepted to break the correlations , thereby offsetting the efficiency gains made by increasing the jump - attempt frequency in the first place . as with j - walking , the number of temperatures used in a parallel tempering simulation and their spacing can not be chosen arbitrarily , but must be selected to ensure that the exchanges are accepted sufficiently often . generally , more temperature walkers and smaller spacing are required for the solid - liquid phase - change region , and for very low temperatures . since the heat capacity curves as a functions of temperature typically had peaks in the solid - liquid phase - change regions whose location and height i wanted to determine accurately , the number of temperatures and their spacing was also affected by the need to properly represent the heat capacity curves with a sufficient density of points to allow for good interpolation . this latter requirement resulted in my choosing temperature spacings in the phase - change region that were small enough that the exchange acceptance ratios were typically quite high . for very low temperatures , the classical heat capacity curve is nearly linear , thus requiring fewer points to represent . however , the exchange acceptance ratio decreases rapidly at lower temperatures , requiring more temperature walkers . since the overall computation time is nearly proportional to the number of walkers , i was able to reduce the number of lower - temperature walkers needed by increasing the exchange - attempt frequency @xmath35 at lower temperatures , according to the relation @xmath36 where @xmath37 . this expression increases the lowest - temperature exchange - attempt frequency @xmath38 by a quarter of the range above the baseline highest - temperature exchange - attempt frequency , @xmath39 , which was set to 10% . the parallel tempering simulations done in this study were similar to those done by neirotti _ et al . _ for lj@xmath15.@xcite for each cluster simulation , a number of random walkers at various temperatures were run in parallel , with each temperature walker occasionally attempting an exchange move with the next higher temperature walker , except for the highest temperature walker , which attempted no exchanges directly ( but did exchange configurations according to attempts made by the next - highest temperature walker ) . since i had the lowest - energy isomers for each cluster from the j - walking distribution quenches , for most of the clusters i initialized the temperature walkers by using the lowest - energy isomer for the lowest - temperature initial configuration , the next lowest - energy isomer for the next lowest - temperature initial configuration , and so on . for the other clusters , i just used the lowest - energy isomer to initialize all of the initial configurations . it did not seem to matter which initialization scheme was used , and of course , random configurations could have been used , but at the cost of longer warmup periods . because i initialized the simulations from the lowest - energy isomers , i could quickly run preliminary simulations that gave qualitatively good results , which i could use to fine - tune the number of parallel tempering temperature walkers and their spacing . these preliminary parallel tempering runs consisted of @xmath32 metropolis warmup passes for each temperature walker , followed by @xmath32 parallel tempering warmup passes and then by 10 parallel tempering walks with preliminary data accumulation , with each walk being @xmath32 passes in length . the configurations for each temperature walker were saved at the end of the warmup period , and at the end of each walk . the heat capacity curves for each walk were examined for consistency , and if significant variations were found , the parallel tempering run was repeated ( without the metropolis warmup passes ) from the last saved configurations , so that in effect , all of the preceding parallel tempering runs served as warmup runs for the final run . for some of the clusters , only parallel tempering warmup passes were run , indicating that the preceding metropolis warmup passes were likely not necessary ( although their computational expense was minor ) . after the preliminary runs appeared to be sufficiently warmed up , a longer parallel tempering run of @xmath32 warmup passes followed by @xmath33 total passes of data accumulation per temperature point was run ( 10 walks of @xmath40 passes each ) . again , the heat capacity curves for each walk were checked for consistency ; for a few clusters , the variations in the curves were still too large , and so the run was repeated from the last saved configurations . for lj@xmath15 , an especially long warmup was required , consistent with the findings of neirotti _ et al_.@xcite one important difference between my implementation of parallel tempering and that of neirotti _ et al _ concerns the choice of the constraining radius @xmath18 . ideally , the value of the constraining radius should be large enough that the cluster properties being simulated are not adversely affected by the constraining potential , but not so large that cluster dissociations lead to overly long convergence times . in my previous study of clusters of size 4 to 24 , i used @xmath41 , a value large enough not to affect the heat capacity curves , except at higher temperatures , well past the solid - liquid phase - change regions . thus , for this study , i also set @xmath42 for the metropolis and j - walker simulations . et al _ used a value of @xmath43 for their simulation of lj@xmath15 , since they had difficulties attaining ergodicity with larger constraining radii . for lj@xmath15 , such a small constraining radius was not a problem since the heat capacity peak occurred at a relatively low temperature . however , for most of the cluster sizes included in this study , such a small value would affect the heat capacity peak height and location . after some preliminary investigations , i chose to set @xmath44 for parallel tempering simulations of cluster sizes 25 to 48 , and to @xmath41 for the larger cluster sizes . figure [ fig : ar30_rc ] illustrates the effects of the constraining radius on clusters ar@xmath45 and ar@xmath46 . the heat capacity peak for @xmath47 is slightly truncated , but the peak for @xmath44 is in good agreement with the @xmath42 peak , even though the high - temperature side of the peak is substantially affected . these two examples also show how well parallel tempering was able to overcome the quasiergodicity evident in the metropolis curves . the two parallel tempering curves for ar@xmath46 show very good agreement for the smaller peak , which is completely absent in the metropolis simulations . the converged heat capacity curve i obtained for ar@xmath15 was mostly similar to the curve obtained by neirotti _ _ in their parallel tempering study of lj@xmath15,@xcite with a small shoulder in the low - temperature side of the peak . the reduced peak temperature 0.1653 was also in good agreement with that found by neirotti _ et al . _ ( 0.166 ) , although my peak value was slightly higher ( @xmath48 compared to 5.62 ) and the higher - temperature part of my curve rose substantially higher ; these discrepancies are mostly due to neirotti _ _ having used the smaller constraining radius . as pointed out by neirotti _ _ , the shoulder in the heat capacity curve is in contrast to the small lower - temperature peak found by doyle , wales and miller in their study of lj@xmath15@xcite ( their larger peak location and height were similar to the parallel tempering results , though , at 0.17 and 5.68 , respectively ) . finally , follow - up parallel tempering simulations of @xmath40 passes per temperature were run for each cluster with periodic quenching of the configurations to determine the population distributions of the lower - energy isomers as functions of temperature . these were initialized using the final configurations from their corresponding preceding parallel tempering runs . since i was primarily interested in using the quench results to provide qualitative insight into the heat capacity results , rather than determining the quench curves themselves with high accuracy , these shorter walk lengths were sufficient . much of the magic number behavior of cluster heat capacities can be understood in terms of the magic number behavior of the underlying potential surfaces . for example , for the small clusters i studied earlier,@xcite the largest heat capacity peaks were observed for sizes @xmath8 , 19 and 23 , corresponding to single , double and triple icosahedra , respectively . the same magic numbers were found in some corresponding structural properties as well , such as the binding energy differences , @xmath49 . \label{vmin}\ ] ] the lowest - energy configurations for each of the clusters i examined in this study were obtained from quenches of j - walker distributions . these were consistent with the lowest - energy configurations found in other studies.@xcite from @xmath1 to 30 , the global minima all consisted of a @xmath50 double icosahedral core with atoms added to tetrahedrally bonded face sites ( anti - mackay overlayers ) , while for @xmath51 to 54 ( except for @xmath52 ) , the added atoms occupied sites corresponding to the outer shell of the @xmath11 mackay icosahedron ( mackay overlayers);@xcite @xmath52 did not fit the pattern , it being a truncated octahedron instead.@xcite clusters for @xmath53 to 60 had added atoms that contributed to the anti - mackay overlayer of the 55-atom core , beginning the building up of the next mackay icosahedron ( @xmath54 ) . although the lowest - energy isomers play a significant role in cluster thermodynamics , other low - energy isomers also play important roles , and the size of the gap between the lowest - energy isomer and the next lowest - energy isomer is an especially important predictor of magic number behavior.@xcite figure [ fig : energylevels ] shows the energy levels of the lower - energy configurations for each cluster size , relative to their global minimum . sizes @xmath55 and 58 have relatively large gaps between their global minimum and next higher minimum , compared to their immediate neighbors , and so can be considered a magic number sequence ; the gaps between the lowest - energy and next lowest - energy isomers are also plotted in figure [ fig : energyn ] as functions of @xmath4 . the mackay icosahedron @xmath11 has an especially large gap . there is also much variation in the densities of the low - lying isomers evident in figure [ fig : energylevels ] , with densities being substantially greater for sizes @xmath56 to 37 than for the other sizes , while sizes @xmath57 to 56 have particularly low densities . the binding energy differences for the clusters are also displayed in figure [ fig : energyn ] . except for @xmath52 , the maxima in the binding energy differences are the same as the maxima in the energy differences between the two lowest - energy isomers . also shown in figure [ fig : energyn ] is a plot of another indicator of structural magic number behavior , the second finite difference of the energy , @xmath58 these magic numbers are also in agreement with the magic numbers obtained for the binding energy differences , again , except for @xmath52 . the discrepancy in the magic number sequence for @xmath52 is due to that cluster being unique in this range by having an fcc structure as its global minimum , so that those measures that are relative to the cluster s immediate neighbors are considerably different than those of the other clusters , whose immediate neighbors are structurally similar . figure [ fig : del2e_nt ] shows a generalization of the zero - temperature @xmath59 curve shown in figure [ fig : energyn ] , with the curves extended to non - zero temperatures to form a @xmath60 surface . the curves were obtained from parallel tempering simulations ; the temperature scale is for ar . the surface can be seen to encompass two general regions : a flat region at higher temperatures , corresponding to the clusters liquid - like regions , and a corrugated region at lower temperatures , corresponding to the clusters solid - like regions . the plot shows that many of the clusters are locked in their lowest - energy isomers for extended temperature ranges , especially the magic number sizes @xmath61 , 29 , 39 , 43 , 46 , 49 and 55 the @xmath62 peak values differ very little from their zero - temperature values well into the phase - change region , until their abatement over a relatively narrow temperature range . the plot also shows that the phase - change temperature regions have a strong dependence on the cluster size , with the melting temperature generally decreasing from about 30 k for @xmath61 to about 20 k for @xmath63 , and then generally increasing to a maximum of about 40 k for @xmath64 . figure [ fig : potential ] plots the reduced potential energy per particle for each cluster size over a temperature range spanning the solid - like and liquid - like regions . magic number behavior can be seen for sizes @xmath61 , 29 , 32 , 36 , 39 , 43 , 46 , 49 and 55 by the relatively large differences at lower temperatures between their potential energies and those of their preceding cluster size . the mackay icosahedron @xmath11 is so stable relative to its neighbors that its curve actually crosses several of the other curves . the phase regions are apparent in figure [ fig : potential ] by the transition from the irregular spacing between the curves seen for the lower temperature , solid - like regions , to the regular spacing seen for the higher temperature , liquid - like regions . as was seen in the @xmath60 surface in figure [ fig : del2e_nt ] , the melting temperature generally decreases as the cluster size increases from @xmath1 , reaching a minimum of about 20 k for @xmath4 about 36 , and then increases to a maximum for @xmath11 . curves of the reduced heat capacity per particle for each cluster size are shown in figures [ fig : cv1 ] , [ fig : cv2 ] and [ fig : cv3 ] , while figure [ fig : magic ] shows the corresponding peak height and temperature as functions of the cluster size ; the peak parameters are also listed in table [ tbl : heat capacity peak ] . the peak parameters were obtained by smoothing and interpolating the combined parallel tempering data ( 10 walks of @xmath40 passes per temperature point each ) . the associated parameter uncertainties were estimated by determining the peak height and location for each of the 10 walks separately , and calculating the standard error of the set of these values . as a check , additional runs of @xmath33 total passes per temperature were done for a few of the clusters , and the standard deviations of the resulting set of peak parameters were compared to the corresponding uncertainty estimates . these were found to be in good agreement . the extra check runs also showed the parallel tempering heat capacity results to be very reproducible , unlike the corresponding j - walking results . each of the clusters had at least one peak in its heat capacity curve , and overall , while their was much variation in the curves , systematic trends are evident . cluster sizes @xmath65 and 55 had peak heights that were substantially greater than those of their immediate neighbors , and thus represent the magic numbers for the heat capacity sequence . these cluster sizes were also magic numbers for the @xmath59 sequence and the lowest - energy isomer gaps shown in figure [ fig : energyn ] , but the other structural magic numbers ( @xmath66 and 32 ) did not show magic number behavior in their heat capacity curves . instead , for cluster sizes 25 to 36 , both the heat capacity peak height and the peak temperature showed remarkably smooth variation , with the peak temperature slowly increasing , while the peak height slowly decreased in a nearly linear fashion , until the peak disappeared at @xmath2 . as can be seen in figure [ fig : cv1 ] , this behavior was somewhat affected by the choice of the constraining radius . for the metropolis simulations run at @xmath41 , the heat capacity curves had greater slopes at higher temperatures , resulting in the peak disappearing by @xmath67 . more interesting was the emergence and evolution of a second heat capacity peak , as the first peak gradually disappeared . beginning with @xmath3 , a small , lower - temperature peak can be seen in figure [ fig : cv1 ] to develop and gradually increase in both height and temperature , until it becomes the dominant peak at @xmath63 . this trend can be seen to continue in figures [ fig : cv2 ] and [ fig : cv3 ] , until the peak reaches its greatest height and highest temperature for the magic number size @xmath11 . this behavior suggests partitioning the heat capacity peaks into two sequences , with the first sequence containing the small-@xmath4 peaks that evolve into the large , broad high - temperature peaks as the second mackay shell initially builds up , while the second sequence originates from the smaller , narrower , lower - temperature peaks that evolve into the dominant peaks as the second mackay shell is completed . the possible beginning of a third sequence corresponding to the initial building up of the third mackay shell can be inferred from the curves in figure [ fig : cv3 ] , where ar@xmath68 has a small , lower - temperature peak , and ar@xmath69 and ar@xmath70 have discernable shoulders in the low - temperature sides of their peaks . the emergence of the smaller peak nearly coincides with the transition at @xmath51 from the anti - mackay overlayer in the lowest - energy configuration to the mackay overlayer , and the cluster sizes where two prominent heat capacity peaks are evident correspond to those sizes having the greatest density of low - lying isomers ( figure [ fig : energylevels ] ) . _ found that for clusters in this size range , the high density of isomers was due to the small differences in energy between the various configuration types , anti - mackay , mackay , decahedral , fcc , etc;@xcite similar findings were obtained for clusters bound by a morse potential whose range parameter is similar to the lennard - jones potential.@xcite the two heat capacity sequences make the determination of the cluster melting temperature from heat capacity peak temperatures ambiguous for sizes @xmath51 to 36 . peaks in heat capacity curves have typically been associated with cluster phase changes , particularly solid - liquid changes.@xcite however , peaks can arise from other circumstances as well , such as mixing anomalies@xcite and premelting effects , which can make the assignment of the cluster melting temperatures difficult . calvo and spiegelmann defined the melting temperatures of the sodium clusters they studied whose heat capacity curves had multiple peaks , as the temperature associated with the highest peak.@xcite applying such a definition to the clusters shown in figure [ fig : cv1 ] that have two heat capacity peaks would imply that the reduced melting temperature would jump from @xmath71 for @xmath72 to 0.1427 for @xmath63 . given the structural similarities of these two clusters , such a large difference in the melting temperature would be surprising , and as will be seen later when other cluster properties are examined , the two clusters do in fact have similar melting temperatures . also , the potential energy curves shown in figure [ fig : potential ] and the @xmath60 surface shown in figure [ fig : del2e_nt ] both suggest that the melting temperature for these clusters is about @xmath73 , and thus intermediate to the two heat capacity peak temperatures . there is much evidence of systematic error due to quasiergodicity in the metropolis results , especially for sizes @xmath51 to 38 , where the metropolis heat capacity curves had lower - temperature peaks that were either missing ( @xmath74 to 34 ) or poorly formed ( @xmath72 to 38 ) . for the other cluster sizes , the metropolis heat capacity curves showed very good agreement with the parallel tempering results , except for temperatures near the heat capacity peak , where the metropolis results for @xmath33 total passes per temperature point were substantially lower than the parallel tempering results , and even the metropolis results for @xmath34 total passes were typically slightly lower . the fact that parallel tempering simulations of @xmath33 passes could provide significantly better results than similar metropolis simulations of @xmath34 passes , for not much more computational overhead than metropolis simulations of @xmath33 passes , clearly shows the superiority of parallel tempering over the metropolis method . there have been several previous studies of 55-atom cluster heat capacities done.@xcite the study most directly comparable to the parallel tempering results i obtained for @xmath11 is a j - walking study of ar@xmath10 done by lpez.@xcite he obtained peak parameters of @xmath75 and @xmath76 , compared to my parallel tempering results of 17.45 and 0.294 , respectively . lpez used the same constraining volume radius as i did for @xmath11 , @xmath42 , but his j - walker distributions consisted of only @xmath77 configurations , which are considerably fewer than the @xmath29 configurations i used for my j - walking studies of cluster sizes @xmath1 to 38 . given the poor results i obtained with j - walking using larger distributions for smaller cluster sizes , i suspect that the discrepancy in the heat capacity peak parameters is due to lpez having used insufficiently representative j - walker distributions in his study . the heat capacity curve obtained from a metropolis simulation of @xmath34 total passes per temperature that is shown in figure [ fig : cv3 ] had peak parameters of 17.00 and 0.294 , respectively , which are close to the parallel tempering results . the small discrepancy between these metropolis results and the parallel tempering results is similar to the discrepancies obtained for the other clusters of similar size . labastie and whetten obtained peak parameters of 15.4 and 0.3 for ar@xmath10 using histogram methods , but their smaller peak height was most likely due to their constraining radius being considerably smaller at @xmath78 . doye and wales obtained peak parameters of 17.4 and 0.298 for ar@xmath10 using the superposition method,@xcite which is in very good agreement with my results . the rms bond length fluctuation @xmath79 is another measure that has been used often in cluster simulations to estimate the melting temperature.@xcite according to the lindemann criterion , melting occurs in bulk matter when @xmath79 is about 10% or greater . however , much care must be exercised in interpreting rms bond length fluctuations obtained from cluster simulations . first , as calvo and spiegelmann have pointed out , the lindemann criterion was originally proposed for bulk systems , implying that it does not necessarily extend to finite clusters , and the choice of 10% for the critical value is somewhat arbitrary.@xcite second , @xmath27 is dependent on the simulation walk length , with the sharp rise in the curve generally occurring at lower temperatures as the walk length is increased.@xcite for cluster simulations of very long length , a value of @xmath80 can be obtained at temperatures where the cluster spends most of its time in some solid - like form , with only occasional isomerizations to other solid - like forms , rather than undergoing the ready diffusion typical of liquid - like behavior . because @xmath79 is obtained from averages over entire metropolis walks or molecular dynamics trajectories , it can not be estimated from simulations where configuration scrambling occurs , such as j - walking and parallel tempering , and so quasiergodicity issues further complicate matters . using very long walk - lengths to reduce systematic errors arising from quasiergodicity can result in @xmath27 curves whose transitions occur well before the melting phase - change . despite these caveats , rms bond length fluctuation curves have qualitative validity and can provide useful insights , especially in conjunction with other cluster properties . curves of @xmath27 for each cluster size are shown in figures [ fig : rms1 ] , [ fig : rms2 ] and [ fig : rms3 ] for averaged walk - lengths of @xmath40 and @xmath33 passes per temperature , while figure [ fig : rmsn ] shows the @xmath4-dependence of the temperatures corresponding to the lindemann - like threshold @xmath81 . the threshold temperatures were estimated by smoothing and interpolating the @xmath27 curves , and then searching for the interpolated temperature corresponding to @xmath81 . i chose to set the threshold slightly higher than the usual lindemann criterion of @xmath82 , since some of the @xmath27 curves had too much noise near @xmath82 to obtain a reliable estimate of the temperature corresponding to that threshold , and because the threshold is somewhat arbitrary anyway ; the value @xmath81 was also typically near the inflection point of the smoothed curves transition regions . the @xmath27 curves shown in figures [ fig : rms1 ] to [ fig : rms3 ] are mostly similar , but there are some interesting differences . generally , the steepness of the rise in @xmath79 in the transition region increases with cluster size , with the magic number sizes @xmath83 and 55 having especially sharp rises . evidence of quasiergodicity can be clearly seen in some curves : the @xmath27 curves for @xmath51 each have a noisy shoulder just before the transition region , while the curves for @xmath84 each have a small peak ; the curves for @xmath52 show the cluster to be locked in its global minimum for all of the lower - temperature metropolis simulations of length @xmath33 passes or fewer , until finally escaping at @xmath85 k. evidence of pre - melting in the anti - mackay overlayer of the third mackay shell can be seen in sizes @xmath53 to 60 by the emergence and evolution of a shoulder in the lower - temperature part of the @xmath27 curves . also indicated on the temperature axes in figures [ fig : rms1 ] to [ fig : rms3 ] are the corresponding heat capacity peak temperatures obtained from the parallel tempering simulations ( likewise , figures [ fig : cv1 ] to [ fig : cv3 ] have the corresponding temperatures associated with the lindemann - like threshold of @xmath81 indicated on their axes ) . these clearly show that the heat capacity peaks do not necessarily correspond to a cluster phase change . for cluster sizes @xmath1 to 29 , the heat capacity peak temperatures are well within the liquid - like region , according to the @xmath27 curves , while for cluster sizes @xmath3 to 36 , where two heat capacity peaks are evident , the higher - temperature heat capacity peak is also clearly in the liquid - like region . moreover , these temperatures are high enough that quasiergodicity in these regions of the @xmath27 curves is not an issue . likewise , most of the lower - temperature heat capacity peaks are found to be in the solid - like region of the @xmath27 curves , well before the transition region ( although ergodicity issues regarding the @xmath27 curves in these regions could be involved ) . only for sizes @xmath72 and @xmath5 , are the lower - temperature heat capacity peak temperatures similar to the @xmath27 transition regions . for the remaining cluster sizes shown in figures [ fig : rms2 ] and [ fig : rms3 ] , however , the heat capacity peak temperatures correspond quite closely to the end of the @xmath27 transition regions , where the curves quickly change slope to mark the start of the liquid - like region . magic number behavior in the rms bond length fluctuations is evident in figure [ fig : rmsn ] , especially for the threshold temperature values obtained from the longer metropolis simulations of @xmath33 passes . the threshold temperatures are significantly higher for the magic number sizes @xmath86 and 55 , relative to their immediate neighbors . these are the same magic numbers seen in figure [ fig : energyn ] for @xmath59 and for the energy gaps between the lowest - energy and next lowest - energy isomers , except for @xmath87 . the @xmath79 data also account for the absence of magic number behavior in the heat capacities for sizes @xmath66 and 32 . since the large , broad , higher - temperature heat capacity peaks seen slowly diminishing in figures [ fig : energyn ] and [ fig : cv1 ] are located well past the cluster solid - liquid phase - change regions , whatever magic number effects might exist for @xmath66 and 32 are masked by the systematic evolution of the heat capacity peak itself with increasing cluster size . classical cluster heat capacities are inherently a measure of the width of the potential energy distributions . typically , for temperatures near the cluster melting temperature , the distributions are much wider than those at lower temperatures ( which span mostly lower - energy , solid - like forms ) and those at higher temperatures ( which span mostly higher - energy , liquid - like forms ) , since these distributions span both types of forms . however , the distribution width in the liquid region also typically increases with temperature , since the number of accessible configurations also increases . thus , depending on the energy distribution of the higher - energy isomer potential minima , the inverse dependence of the heat capacity on temperature in eq . ( [ eq : cv ] ) could result in a peak occurring in the liquid - like region of the curve that would have little relation to the cluster s solid - liquid phase change . the threshold temperatures corresponding to @xmath81 in figure [ fig : rmsn ] also imply that the cluster melting temperature generally decreases as the cluster size increases from @xmath1 , reaching a minimum near @xmath2 , then generally increases with cluster size , reaching a maximum for @xmath11 , and decreases thereafter ; magic number effects superimposed on these general trends account for the local variations . these general trends are consistent with those inferred from the @xmath60 surface in figure [ fig : del2e_nt ] and the potential energy curves in figure [ fig : potential ] . casero and soler calculated melting temperatures for lj clusters of size @xmath88 to 34 in the microcannonical ensemble , defining the melting temperature to be the temperature where the free energies of the solid and liquid clusters become equal.@xcite their results indicate a generally increasing trend with cluster size , with reduced melting temperatures for sizes @xmath1 to 34 all above @xmath89 . these results are very similar to the temperatures seen for higher - temperature heat capacity peak temperatures in figure [ fig : magic ] , and thus are contrary to the melting temperatures implied by the rms bond length fluctuation data shown in figure [ fig : rmsn ] . since the melting temperatures deduced from the @xmath79 data are in good agreement with the melting temperatures obtained from the heat capacity peak temperatures for cluster sizes greater than about @xmath72 , and these temperatures are well below @xmath89 for @xmath90 , it is likely that the reduced melting temperatures for sizes @xmath1 to 34 are also well below @xmath89 . the temperature dependence on the distribution of quenched isomers is shown in figure [ fig : quench_magic ] for those ar clusters whose heat capacities showed magic number behavior , and in figures [ fig : quench1 ] to [ fig : quench3 ] for some selected non - magic number ar clusters . each plot shows the percentage of quenches to each of the three lowest - energy isomers , as well as the total percentage of quenches to the remaining higher - energy isomers ; the data for all of these plots were all obtained from parallel tempering simulations . much of the behavior exhibited by these curves can be rationalized in terms of the energy spacings and densities of the lower - energy isomers shown in figure [ fig : energylevels ] . the plots for the magic number sizes are all very similar in form , with quenches in the lower - temperature , solid - like region represented solely by the lowest - energy isomer , and quenches to the other isomers only appearing at temperatures near the phase - change region , and higher . this is a consequence of the large gap between the lowest - energy and next lowest - energy isomers that is typical of magic number clusters . the only significant differences between the different cluster sizes is the temperature where the fraction of quenches to the lowest - energy isomer drops off substantially , which increases with cluster size , and can be seen to be very similar to the heat capacity peak temperature . the plots for some non - magic cluster sizes shown in figure [ fig : quench1 ] for @xmath91 are considerably different . the curves for @xmath1 have a similar form to the magic number curves in figure [ fig : quench_magic ] , but both the decrease in the fraction of quenches to the lowest - energy isomer and the increase in the fraction of quenches to the higher - energy isomers are much more gradual . this gradual transition is consistent with the more gradual transition for the @xmath27 curve , seen in figure [ fig : rms1 ] . also , the transition region occurs well below the heat capacity peak temperature ( but near the transition region indicated by @xmath27 ) . ar@xmath92 has two very closely spaced lowest - energy isomers that are well below its next - higher energy isomers , and most of the lower - temperature quenches are to the second lowest - energy isomer . the temperature where the fraction of quenches to the second lowest - energy isomer exceeds the fraction to the lowest - energy isomer is similar to the peak temperature of the very small lower - temperature heat capacity peak . again , the temperature region where the fraction of quenches to the higher - energy isomers increases substantially corresponds well to the transition region in the @xmath27 curves , and is well below the higher - temperature heat capacity peak temperature . the quench curves for @xmath51 and 32 are similar to the @xmath93 curve , with the fraction of quenches to the lowest - energy isomer dropping rapidly at very low temperatures corresponding closely to the small heat capacity peak temperatures , and the fraction of quenches to higher - energy isomers gradually increasing at temperatures well below the peak temperature of the larger , higher - temperature heat capacity peaks . for ar@xmath94 , the density of lower - energy isomers is so great that quenches to higher - energy isomers dominate at temperatures even below the transition region for the @xmath27 curves shown in figure [ fig : rms1 ] . the lower - temperature shoulders appearing in the @xmath51 @xmath27 curves and the behavior of the corresponding quench curves are indicative of pre - melting and an extended phase - change region . ar@xmath46 is a magic number according to such structural properties as @xmath95 and @xmath59 , but not according to its heat capacity peak parameters . its quench curves are also different than the typical magic - number quench curves seen in figure [ fig : quench_magic ] . the @xmath84 @xmath27 curves show unusual peaks in the solid region , just before their sharp rise . this region just follows the lower - temperature heat capacity peak , and according to the quench results , almost all of the quenches are to the second lowest - energy isomer . this indicates that the peaks in the @xmath27 curves are a result of quasiergodicity ( the metropolis heat capacity curves in figure [ fig : cv1 ] also show small , spurious peaks in this region ) . the quench curves for @xmath96 and 34 indicate easy access to the other low - lying isomers at low temperatures near the small heat capacity peak temperatures . figure [ fig : quench2 ] shows quench results for some non - magic number clusters for @xmath2 to 48 . ar@xmath97 , like ar@xmath94 and ar@xmath98 , has an especially high density of low - lying isomers , and its quench curves also show a substantial fraction of quenches to these isomers at temperatures corresponding to the heat capacity peak temperature . for @xmath52 , the temperature region where the fraction of quenches to the lowest - energy fcc isomer drops off and the fraction of quenches to the lowest - energy icosahedral - like isomer ( the second lowest - energy isomer ) increases corresponds to the region just before the heat - capacity peak where a shoulder can be seen in figure [ fig : cv2 ] , which is consistent with other studies of this cluster.@xcite the quench curves for @xmath99 and 48 are similar to those for @xmath2 , since these clusters all have accessible lower - energy isomers that are only slightly higher than the lowest - energy isomer , but the curves for @xmath100 are similar to the curves for the magic number clusters in figure [ fig : quench_magic ] , indicating that its low - lying isomers are not very accessible . the isomer energy levels for @xmath101 are similar to those for @xmath102 and 48 , and the corresponding quench curves shown in figure [ fig : quench3 ] are likewise similar to those for @xmath102 and 48 . ar@xmath103 and ar@xmath104 have some isomers having closely spaced potential energies only slightly above the lowest - energy isomer , and then very few isomers until @xmath105 . this type of isomer energy distribution is similar to the distributions seen for binary ar - kr clusters , where many mixed isomer configurations had energies similar to the lowest - energy isomer,@xcite and the quench results are similar too , with the low - lying isomer quench curves showing plateaus throughout much of the solid - like temperature region . however , neither ar@xmath103 nor ar@xmath104 exhibited the small , very low temperature heat capacity peaks that were evident in the binary cluster curves . ar@xmath106 has two low - energy isomers corresponding to the two unique ways a lone atom can be positioned on a 55-atom mackay icosahedral core , separated by a large energy gap from the other isomers corresponding to rearrangements of the 55-atom core . the @xmath107 quench curves show substantial fractions of quenches to both isomers throughout much of the solid - like region until the phase - change region , when the higher - energy isomers become accessible . for @xmath108 , there are several lower - energy isomers below @xmath109 that were represented in the quench distributions , but as with @xmath110 and 53 , there was no corresponding lower - temperature heat capacity peak . both the @xmath108 and 59 quench curves show evidence of premelting , consistent with their corresponding @xmath27 curves in figure [ fig : rms3 ] . compared to the irregularity with respect to cluster size @xmath4 that characterized the properties of the small clusters of size @xmath111 that i studied previously,@xcite the properties for the clusters sized @xmath112 showed more regularity and consistency . the lowest - energy isomers for these sizes were all dominated by icosahedral based geometries , except for @xmath52 , whose lowest - energy isomer had fcc geometry ( the lowest - energy icosahedral - like isomer for @xmath52 was only slightly less stable ) . for the icosahedral - like clusters , configurations having anti - mackay overlayers were the lowest - energy isomers for @xmath113 , while configurations having mackay overlayers were the lowest - energy isomers for the remaining sizes up to @xmath11 , after which the filling of the third shell began again with anti - mackay overlayers . correspondingly , the heat capacity peak parameters formed two overlapping sequences as functions of cluster size , with those clusters having anti - mackay overlayer lowest - energy isomers showing a gradual evolution in their heat capacity peaks to higher temperature and smaller size , while those clusters having mackay overlayer lowest - energy isomers had small , lower - temperature peaks that generally shifted to higher temperature and grew in size as @xmath4 increased . for cluster sizes @xmath114 , the heat capacity peak temperatures did not correspond well to the cluster melting temperature ranges inferred from the temperature dependence of other cluster properties , such as the rms bond length fluctuations and the distributions of quenched isomers . for each of the cluster sizes from @xmath63 to 60 , though , the melting temperature ranges obtained from the three measures were in qualitatively good agreement . the general trend in the cluster melting temperatures as a function of the cluster size over the range studied was an initial decrease , reaching a minimum near @xmath63 , then an increase until a maximum at @xmath11 , where the second mackay shell was completed , followed by another decrease as atoms were added to the third mackay shell . superimposed on this general trend were local maxima corresponding to the magic numbers @xmath61 , 29 , 32 , 36 , 39 , 43 , 46 , 49 and 55 . the heat capacity peak heights had a magic number series that was a subset of this sequence , with local maxima occuring at @xmath63 , 39 , 43 , 46 , 49 and 55 ; for the other magic number sizes where the heat capacity peaks did not show magic number behavior ( @xmath61 , 29 , and 32 ) , the heat capacity peaks were located well away from the solid - liquid phase - change regions . the magic number sequence for the cluster melting temperatures coincided very well with the magic number sequences obtained for structural properties such as the binding energy difference @xmath95 , the second finite difference of the energy @xmath59 , and the gap between the lowest - energy and next lowest - energy isomers . the only differences were @xmath115 being a magic number for the @xmath95 sequence instead of @xmath116 , and @xmath87 , which was a magic number for all three structural properties , but was not for either the heat capacity or the rms bond length fluctuation sequences . this similarity in the magic number sequences is consistent with the widely held view that the thermodynamic behavior of clusters in the phase - change region is dominated by their potential energy surfaces . the parallel tempering method was found to be especially useful for overcoming quasiergodicity and accurately calculating the heat capacities of these clusters , particularly for the range @xmath3 to 38 , where the metropolis method was hard pressed to provide even qualitatively correct results . even the j - walking method , which had worked very well for smaller cluster sizes , was unable to determine heat capacities reproducibly for clusters in this range . given also the relatively small computational overhead for the method compared to metropolis monte carlo , parallel tempering currently stands as the method of choice for cluster monte carlo simulations . the support of the natural sciences and engineering research council of canada ( nserc ) for part of this research is gratefully acknowledged . i thank northern digital , inc . for generously providing me the use of their workstations for some of the calculations reported , and the university of waterloo for the use of their facilities . i also thank david l. freeman for helpful discussions . m. r. hoare and p. pal , adv . * 20 , * 161 ( 1971 ) . r. s. berry , t. l. beck , h. l. davis and j. jellinek , in _ advances in chemical physics , _ edited by i. prigogine and s. a. rice ( wiley , new york , 1988 ) , vol . 70b , p. 75 . h. l. davis , j. jellinek , and r. s. berry , 6456 ( 1987 ) . t. l. beck , j. jellinek , and r. s. berry , 545 ( 1987 ) . j. e. adams and r. m. stratt , 1332 ( 1990 ) . j. d. honeycutt and h. c. andersen , j. phys . * 91 , * 4950 , ( 1987 ) . j. a. northby , 6166 ( 1987 ) . d. g. vlachos , l. d. schmidt , and r. aris , 6880 ( 1992 ) ; * 96 , * 6891 ( 1992 ) . k. doye , d. j. wales , and r. s. berry , , 4234 ( 1995 ) . j. p. k. doye , m. a. miller , and d. j. wales , , 8417 ( 1999 ) . d. d. frantz , 3747 ( 1995 ) . r. casero and j. m. soler , 2927 ( 1991 ) . t. ikeshoji , b. hafskjold , y. hashi , and y. kawazoe , 5126 ( 1996 ) . o. echt , k. sattler , and r. recknagel , 1121 ( 1981 ) . e. k. parks , b. j. winter , t. d. klots , and s. j. riley , 1882 ( 1991 ) ; e. k. parks , k. p. kearns , and s. .j riley , 10207 ( 1998 ) . d. prekas , c. lder , and m. velegrakis , 4450 ( 1998 ) . m. sakurai , k. watanabe , k. sumiyama , and k. suzuki , 235 ( 1999 ) . d. d. frantz , d. l. freeman , and j. d. doll , 2769 ( 1990 ) . d. d. frantz , d. l. freeman , and j. d. doll , 5713 ( 1992 ) . g. e. lopez and d. l. freeman , 1428 ( 1993 ) . d. d. frantz , 10030 ( 1996 ) ; 1992 ( 1997 ) . g. e. lpez , 6650 ( 1996 ) . a. matro and d. l. freeman , and r. q. topper , 8690 ( 1996 ) . e. curotto , d. l. freeman , and j. d. doll , 1643 ( 1998 ) . h. xu and b. j. berne , 10299 ( 1999 ) . c. j. geyer and e. a. thompson , j. am . * 90 , * 909 ( 1995 ) . m. r. ghayal and e. curotto , , 4298 ( 2000 ) . neirotti , f. calvo , d. l. freeman and j. d. doll , 10340 ( 2000 ) ; f. calvo , j. p. neirotti , d. l. freeman and j. d. doll , 10350 ( 2000 ) . j. p. k. doye , d. j. wales and m. a. miller , , 8143 ( 1998 ) . j .p . k. doye , m. a. miller and d. j. wales , , 6896 ( 1999 ) . p.125in*2p0.75inp0.75inp.125in*2p0.75inp0.75 in & & & & & + & & & & & & & & & + 25 & & & [email protected] & [email protected] & 43 & & & [email protected] & [email protected] + 26 & & & [email protected] & [email protected] & 44 & & & [email protected] & [email protected] + 27 & [email protected] & [email protected] & [email protected] & [email protected] & 45 & & & [email protected] & [email protected] + 28 & & & [email protected] & [email protected] & 46 & & & [email protected] & [email protected] + 29 & & & [email protected] & [email protected] & 47 & & & [email protected] & [email protected] + 30 & [email protected] & [email protected] & [email protected] & [email protected] & 48 & & & [email protected] & [email protected] + 31 & [email protected] & [email protected] & [email protected] & [email protected] & 49 & & & [email protected] & [email protected] + 32 & [email protected] & [email protected] & [email protected] & [email protected] & 50 & & & [email protected] & [email protected] + 33 & [email protected] & [email protected] & [email protected] & [email protected] & 51 & & & [email protected] & [email protected] + 34 & [email protected] & [email protected] & [email protected] & [email protected] & 52 & & & [email protected] & [email protected] + 35 & [email protected] & [email protected] & [email protected] & [email protected] & 53 & & & [email protected] & [email protected] + 36 & [email protected] & [email protected] & [email protected] & [email protected] & 54 & & & [email protected] & [email protected] + 37 & & & [email protected] & [email protected] & 55 & & & [email protected] & [email protected] + 38 & & & [email protected] & [email protected] & 56 & & & [email protected] & [email protected] + 39 & & & [email protected] & [email protected] & 57 & & & [email protected] & [email protected] + 40 & & & [email protected] & [email protected] & 58 & [email protected] & [email protected] & [email protected] & [email protected] + 41 & & & [email protected] & [email protected] & 59 & & & [email protected] & [email protected] + 42 & & & [email protected] & [email protected] & 60 & & & [email protected] & [email protected] +

ground - based spectroscopic follow - up observations of the pulsating stars observed by the satellite mission are needed for their asteroseismic modelling . we aim to derive the fundamental parameters for a sample of 26 doradus candidate stars observed by the satellite mission to accomplish one of the required preconditions for their asteroseismic modelling and to compare our results with the types of pulsators expected from the existing light curve analysis . we use the spectrum synthesis method to derive the fundamental parameters like , , $ ] , and from newly obtained spectra and compute the spectral energy distribution from literature photometry to get an independent measure of . we find that most of the derived values agree with the values given in the kepler input catalogue . according to their positions in the hr - diagram three stars are expected dor stars , ten stars are expected sct stars , and seven stars are possibly sctstars at the hot border of the instability strip . four stars in our sample are found to be spectroscopic binary candidates and four stars have very low metallicity where two show about solar c abundance . six of the 10 stars located in the sct instability region of the hr - diagram show both sct and dor - type oscillations in their light curves implying that dor - like oscillations are much more common among the sct stars than predicted by theory . moreover , seven stars showing periods in the sct and the sct- dor range in their light curves are located in the hr - diagram left of the blue edge of the theoretical sctinstability strip . the consistency of these findings with recent investigations based on high - quality data implies the need for a revision of the theoretical dor and sct instability strips . [ firstpage ] stars : variables : delta scuti stars : fundamental parameters stars : abundances .

1202.1116

the _ kepler _ satellite was launched in march 2009 with the primary goal to search for transiting exoplanets in the solar neighbourhood . it delivers single band - pass light curves of micromagnitude precision and has found hundreds of planet candidates @xcite . the long , uninterrupted , and high - precision time series photometry taken for a huge number of stars has also led to the discovery of many new pulsating stars and is an ideal basis for an in - depth asteroseismic analysis ( see e.g. * ? ? ? * ) . for this analysis and the subsequent asteroseismic modelling , precise knowledge of the fundamental parameters of the stars is essential . these parameters can not be determined from the single band - pass photometry delivered by _ kepler _ alone , however . hence , ground - based spectroscopic follow - up observations have been undertaken to determine the stellar parameters . in this paper , we are concerned with @xmath1 doradus candidates found from data assembled with the _ kepler _ mission . such pulsators are named after their prototype , gamma doradus , whose multiperiodic variable nature was first reported by @xcite . @xcite discovered a multiperiodic photometric variability with an amplitude of about 0.1 mag and periods of 2.275 and 1.277 d in 9 aurigae and reported on similar behaviour in @xmath1 doradus and hd96008 . following these discoveries , @xcite introduced a new class of variable stars named after the prototype star @xmath1 doradus . dor stars are assumed to pulsate in high - order , low - degree non - radial gravity modes driven by the flux blocking mechanism near the base of their convective zones @xcite . the typical masses of these stars lie in the range of 1.51.8 m@xmath7 @xcite . according to @xcite , @xmath1 dor - type stars can be characterised as follows : ( 1 ) spectral type a7f5 and luminosity class iv , iv - v , or v , ( 2 ) low - amplitude photometric variations with periods between 0.5 and 3 days as well as spectroscopic variability seen as both line - profile and low - amplitude radial velocity ( rv ) variations . 1.9 mm rlcrll & & @xmath8 & & & + 01571152 ' '' '' & bd+363535 & 2 & 9.3 & f0 & may - june + 02166218 & bd+373490 & 3 & 9.5 & f0 & may - june + 03217554 & bd+383415 & 2 & 9.6 & a5 & september + 03453494 & bd+383666 & 3 & 9.6 & a5 & september + 04847411 & hd225314 & 4 & 9.8 & a7 & june - july + 05088308 & hd180099 & 1 & 8.7 & f5 & may + 05164767 & hd175537 & 6 & 7.8 & f0 & april - june + 05446068 & bd+403704 & 7 & 9.7 & & june - july + 05785707 & hd181902 & 3 & 9.0 & a & august + 06289468 & bd+413389 & 5 & 9.4 & a2 & may + 06509175 & bd+413248 & 2 & 10.0 & a2 & august + 06587551 & bd+413207 & 3 & 9.8 & a0 & june - july + 06756386 & hd175939 & 2 & 8.7 & a2 & august + 07748238 & hd181985 & 2 & 9.5 & a & may - july + 08623953 & bd+443134 & 3 & 9.3 & a5 & august + 08738244 & hd176390 & 2 & 8.1 & a3 & august + 08750029 & bd+443113 & 3 & 9.7 & a5 & june + 09413057 & bd+452954 & 5 & 9.6 & a2 & may - june + 09764965 & hd181206 & 2 & 8.8 & a5 & august + 09812351 & hd174019 & 3 & 7.9 & a0 & april - may + 10119517 & tyc3544 - 1245 - 1 & 3 & 9.9 & & august + 10451090 & hd174789 & 3 & 9.2 & a & august + 10616594 & tyc3561 - 971 - 1 & 3 & 9.8 & & august + 10977859 & hd184333 & 2 & 8.8 & a2 & august + 11498538 & hd178874 & 2 & 7.3 & f5 & september + 12353648 & hd234859 & 3 & 9.6 & a2 & august + [ table : observations ] @xmath1 dor stars are located close to the red edge of the classical instability strip in the hr - diagram . the theoretical @xmath1 dorinstability strip overlaps with the red edge of classical instability strip where the @xmath6 sct pulsators are located . while the low - order p - modes of @xmath6 sct stars are characterized by short periods ranging from 18 min to 8 h , typical @xmath1 dor high - order g - modes have periods of the order of a day ( e.g. , * ? ? ? multiperiodicity is found for most of the @xmath1 dor class members from ground - based photometry @xcite and spectroscopy ( e.g. , * ? ? ? * ; * ? ? ? pulsators in the overlapping region of the @xmath6 sctand @xmath1 dor instability strip are expected to show the two pulsation characteristics , i.e. , high - order g - modes probing the core and low - order p- and g - modes probing the outer layers . while the frequency patterns of these two types of oscillations are in principle easy to distinguish in the co - rotating frame of reference , the frequencies start to overlap in an inertial frame of reference , particularly for the fast rotators . moreover , the overall beating patterns are complex and hard to unravel from interrupted ground - based data . some hybrid pulsators were already found previously ( e.g. , * ? ? ? * ; * ? ? ? * ) , but the _ kepler _ data make it clear that hybrid pulsators turn out to be numerous , both for af - type stars @xcite and for b - type stars @xcite . @xcite were the first to report a connection between the @xmath9 bootis ( @xmath9 boo ) type stars and the @xmath1 dor variables . boostars are popi hydrogen burning metal poor ( except of c , n , o , and s ) a - type stars @xcite showing significant underabundances of fe - peak elements ( up to 2 dex compared to the solar composition ) . they belong to the class of non - magnetic , chemically peculiar stars . up to now , only two further reports @xcite on a possible connection between the @xmath9 boo stars and @xmath1 dor - type variability appeared . a recent analysis of a sample of 18 @xmath1 dor stars performed by @xcite revealed no principal difference between the abundances of the analysed stars and the chemical composition of non - pulsating a- and f - type stars . in this paper , we investigate a sample of 26 among the brighter stars in the _ kepler _ field which have been proposed to be candidates for @xmath1 dor variables @xcite . we aim to evaluate fundamental stellar parameters like effective temperature @xmath2 , surface gravity @xmath3 , projected rotational velocity @xmath5 , and microturbulent velocity @xmath10 as well as the chemical composition of the target stars from newly obtained high - resolution spectra . based on the derived parameters , we present a classification of the sample stars according to the expected type of variability . the derived chemical composition in turn allows to check for a possible connection between @xmath1 dor - type variability and @xmath9 boo - type abundance patterns . 1.2 mm lclrllrllll & @xmath11 & @xmath3@xmath12 & & @xmath13 & & & @xmath5(kms@xmath14 ) & @xmath10(kms@xmath14 ) & & + 01571152@xmath15 ' '' '' & 7048 & 3.164 & + 0.05 & 7065@xmath16 & 4.46@xmath17 & 0.18@xmath18 & 90.1@xmath19 & 2.03@xmath20 & f0.5 iii & f1 v + 02166218 & 7153 & 3.345 & 0.11 & 7062@xmath21 & 3.88@xmath22 & 0.37@xmath23 & 99.7@xmath24 & 2.82@xmath25 & f0 iv - iii & f1 iv + 03217554@xmath15 & 7801 & 3.504 & + 0.06 & 7667@xmath26 & 2.78@xmath27 & 1.20(fe ) & 225.5@xmath28 & 4.82@xmath29 & a6.5 iv - iii & a7.5 iii - ii + 03453494 & 7806 & 3.843 & 0.33 & 7737@xmath30 & 3.71@xmath31 & 0.95(fe ) & 210.8@xmath32 & 3.24@xmath33 & a7 iv & a7 iv + 04847411@xmath34 & 6563 & 4.517 & 1.95 & 7466@xmath35 & 3.83@xmath36 & 0.52@xmath37 & 139.9@xmath38 & 2.62@xmath39 & f4 v & a8.5 v - iv + 05088308@xmath15 & 6567 & 4.035 & 0.68 & 6708@xmath40 & 2.67@xmath41 & 0.35@xmath42 & 40.7@xmath43 & 4.01@xmath44 & f4 v - iv & f4 iii - ii + 05164767@xmath34 & & & & 6933@xmath45 & 3.59@xmath46 & 0.19@xmath47 & 163.9@xmath48 & 2.62@xmath49 & & f1.5 iv + 05446068@xmath50 & 5337 & 4.476 & 0.69 & 5763@xmath51 & 3.37@xmath52 & + 0.24@xmath18 & 7.8@xmath53 & 0.0@xmath54 & g9.5 v & f9.5 iv + 05785707 & 8009 & 3.615 & 0.14 & 7965@xmath55 & 3.37@xmath56 & 0.56@xmath57 & 171.3@xmath58 & 2.84@xmath59 & a5.5 iv & a6 iv - iii + 06289468@xmath34 & 8267 & 3.741 & 0.30 & 8107@xmath55 & 3.30@xmath60 & 0.48@xmath61 & 149.7@xmath62 & 2.67@xmath63 & a4.5 iv & a5 iii + 06509175 & 7299 & 3.522 & 0.21 & 7510@xmath35 & 3.20@xmath52 & 0.45@xmath18 & 132.4@xmath64 & 2.81@xmath65 & a9 iv - iii & a8 iii + 06587551 & 8377 & 3.929 & 0.07 & 8826@xmath66 & 3.76@xmath23 & 0.11@xmath67 & 139.8@xmath68 & 2.80@xmath69 & a4 v - iv & a2.5 iv + 06756386 & 7992 & 3.513 & 0.51 & 7891@xmath70 & 3.19@xmath71 & 0.59@xmath18 & 192.8@xmath72 & 2.99@xmath73 & a6 iv - iii & a6 iii + 07748238 & 7228 & 3.470 & 0.22 & 7264@xmath74 & 3.96@xmath75 & 0.37@xmath76 & 120.8@xmath77 & 3.53@xmath78 & a9.5 iv - iii & a9.5 v - iv + 08623953 & 7725 & 3.738 & 0.11 & 7726@xmath35 & 3.43@xmath79 & 0.35@xmath76 & 84.8@xmath80 & 2.76@xmath81 & a7 iv & a7 iv - iii + 08738244 & 8167 & 4.152 & + 0.41 & 8154@xmath82 & 3.24@xmath83 & 0.27@xmath67 & 133.9@xmath84 & 2.70@xmath85 & a5 v - iv & a5 iii + 08750029 & & & & 7341@xmath86 & 3.70@xmath87 & 0.56@xmath18 & 166.1@xmath88 & 2.95@xmath89 & & a9 iv + 09413057@xmath34 & 8465 & 3.868 & 0.06 & 8588@xmath90 & 3.59@xmath91 & 0.56@xmath92 & 171.0@xmath93 & 2.83@xmath94 & a4 v - iv & a3 iv + 09764965 & 7455 & 4.085 & 0.18 & 7478@xmath95 & 3.74@xmath96 & 0.27@xmath60 & 85.1@xmath97 & 3.55@xmath36 & a8.5 v - iv & a8 iv + 09812351 & 7794 & 3.470 & 0.29 & 7833@xmath70 & 3.20@xmath98 & 0.90(fe ) & 55.6@xmath99 & 2.18@xmath100 & a6.5 iv - iii & a6 iii + 10119517@xmath34 & 6225 & 4.375 & 0.45 & 6438@xmath101 & 4.21@xmath102 & 0.24@xmath23 & 77.9@xmath103 & 1.26@xmath104 & f8 v & f5 v + 10451090 & 7577 & 4.134 & 0.05 & 7633@xmath35 & 3.58@xmath105 & + 0.04@xmath60 & 44.2@xmath106 & 3.00@xmath107 & a8 v - iv & a7.5 iv - iii + 10616594@xmath108 & 5161 & 3.762 & 0.75 & 5327@xmath109 & 2.81@xmath110 & + 0.16@xmath111 & 7.24@xmath112 & 0.63@xmath75 & g8.5 v - iv & g3.5 iii + 10977859 & 8052 & 3.935 & + 0.15 & 8195@xmath30 & 3.60@xmath113 & 0.11@xmath113 & 63.0@xmath114 & 3.08@xmath115 & a5.5 v - iv & a5 iv + 11498538@xmath108 & 6287 & 4.036 & 0.29 & 6428@xmath30 & 2.90@xmath116 & 0.15@xmath23 & 39.7@xmath117 & 2.54@xmath118 & f7 v - iv & f5.5 iii + 12353648 & 7414 & 3.473 & 0.38 & 7163@xmath119 & 3.49@xmath120 & 1.05(fe ) & 192.0@xmath121 & 2.32@xmath122 & a8.5 iv - iii & f0 iv - iii + [ table : fundamentalparameters ] 0.9 mm cclllllllllllllllll kic ' '' '' & @xmath4 $ ] & & & & & & & & & & & & & & & & & + & & 3.65 & 3.38 & 4.51 & 4.53 & 5.73 & 4.59 & 5.87 & 8.99 & 7.14 & 6.40 & 6.65 & 9.83 & 9.87 & 8.04 & 7.12 & 5.81 & 9.45 + 01571152 ' '' '' & 0.18@xmath18 & + 0.05 & & 0.10@xmath123 & 0.10 & 0.25@xmath123 & 0.15@xmath123 & + 0.00 & + 0.00 & 0.25@xmath123 & 0.15@xmath123 & 0.15 & 0.20 & + 0.80 & & & 0.20@xmath123 & + 02166218 & 0.37@xmath23 & 0.10 & & 0.15@xmath123 & 0.20 & 0.20@xmath123 & 0.45@xmath123 & 0.10 & 0.35 & 0.40@xmath123 & 0.40@xmath123 & 0.40 & 0.35 & + 0.60 & & & 0.45@xmath123 & + 03217554 & 1.20(fe ) & 1.10 & & 0.35@xmath123 & 0.10 & 1.35 & 1.20@xmath123 & & 1.15 & 1.05 & 1.70 & & & 1.10 & & & 0.80 & + 03453494 & 0.95(fe ) & 0.60 & & 0.10@xmath123 & 0.10 & 1.25 & 0.95@xmath123 & & 0.75 & 0.95 & 1.00 & & & 1.05 & & & 0.70 & + 04847411 & 0.52@xmath37 & 0.20 & & 0.20@xmath123 & 0.10 & 0.70@xmath123 & 0.55@xmath123 & + 0.05 & 0.45 & 0.75@xmath123 & 0.55@xmath123 & 0.40 & & + 0.25 & & & 0.50@xmath123 & + 05088308 & 0.35@xmath42 & 0.30 & & 0.30@xmath123 & 0.50 & 0.25@xmath123 & 0.50@xmath123 & 0.30 & 0.35 & 0.55@xmath123 & 0.45@xmath123 & 0.45 & 0.10 & + 0.40 & & & 0.30@xmath123 & + 05164767 & 0.19@xmath47 & + 0.15 & & + 0.00@xmath123 & + 0.15 & 0.20@xmath123 & 0.35@xmath123 & + 0.50 & 0.05 & 0.65@xmath123 & 0.10@xmath123 & 0.05 & 0.60 & + 0.60 & & & 0.25@xmath123 & + 05446068 & + 0.24@xmath18 & 0.30 & & + 0.10@xmath123 & 0.25@xmath123 & 0.05@xmath123 & + 0.15@xmath123 & + 0.10 & + 0.20@xmath123 & + 0.30@xmath123 & + 0.25@xmath123 & + 0.35@xmath123 & + 0.25@xmath123 & + 0.50 & + 0.55@xmath123 & + 0.35@xmath123 & + 0.10@xmath123 & + 0.10 + 05785707 & 0.56@xmath57 & 0.20 & & 0.25@xmath123 & + 0.15 & 0.50 & 0.50@xmath123 & & 0.50 & 0.85 & 0.70 & & & 0.40 & & & 0.45 & + 06289468 & 0.48@xmath61 & 0.30 & & + 0.05@xmath123 & 0.05 & 0.75 & 0.50@xmath123 & & 0.55 & 0.65 & 0.60 & & & 0.50 & & & 0.55 & + 06509175 & 0.45@xmath18 & 0.15 & & 0.05@xmath123 & + 0.05 & 0.35@xmath123 & 0.55@xmath123 & + 0.10 & 0.45 & 0.65@xmath123 & 0.55@xmath123 & 0.20 & & 0.60 & & & 0.45@xmath123 & + 06587551 & 0.11@xmath67 & + 0.00 & + 0.00 & + 0.20@xmath123 & + 0.30 & 0.20 & 0.10@xmath123 & & 0.05 & 0.35 & 0.10 & & & 0.40 & & & 0.15 & + 06756386 & 0.59@xmath67 & 0.20 & & + 0.00@xmath123 & + 0.20 & 0.50 & 0.65@xmath123 & & 0.75 & 0.70 & 0.65 & & & 1.30 & & & 0.60 & + 07748238 & 0.37@xmath18 & 0.15 & & 0.10@xmath123 & + 0.00 & 0.25@xmath123 & 0.45@xmath123 & + 0.10 & 0.35 & 0.45@xmath123 & 0.45@xmath123 & 0.35 & & + 0.50 & & & 0.40@xmath123 & + 08623953 & 0.35@xmath18 & 0.10 & & 0.10@xmath123 & + 0.00 & 0.25 & 0.40@xmath123 & + 0.10 & 0.35 & 0.55 & 0.45 & 0.30 & & 0.35 & & & 0.30 & + 08738244 & 0.27@xmath67 & 0.25 & & + 0.20@xmath123 & + 0.25 & 0.40 & 0.35@xmath123 & & 0.50 & 0.60 & 0.20 & & & 0.55 & & & 0.35 & + 08750029 & 0.56@xmath18 & 0.20 & & 0.35@xmath123 & + 0.00 & 0.60@xmath123 & 0.60@xmath123 & 0.05 & 0.60 & 0.90@xmath123 & 0.50@xmath123 & 0.70 & & + 0.35 & & & 0.65@xmath123 & + 09413057 & 0.56@xmath92 & 0.45 & & 0.05@xmath123 & + 0.10 & 0.70 & 0.60@xmath123 & & 0.85 & 0.75 & 0.65 & & & 0.30 & & & 0.60 & + 09764965 & 0.27@xmath76 & 0.05 & & 0.10@xmath123 & + 0.00 & 0.15@xmath123 & 0.30@xmath123 & + 0.05 & 0.25 & 0.45@xmath123 & 0.30@xmath123 & 0.30 & & 0.15 & & & 0.30@xmath123 & + 09812351 & 0.90(fe ) & + 0.00 & & 0.50@xmath123 & 0.40 & 0.85 & 0.90@xmath123 & & 0.85 & 1.00 & 0.85 & & & 1.20 & & & 0.65 & + 10119517 & 0.24@xmath23 & + 0.05 & & 0.20@xmath123 & 0.15 & 0.05@xmath123 & 0.20@xmath123 & + 0.10 & 0.35 & 0.40@xmath123 & 0.20@xmath123 & 0.15 & 0.40 & + 0.40 & & & 0.25@xmath123 & + 10451090 & + 0.04@xmath60 & 0.10 & & + 0.10@xmath123 & + 0.00 & 0.25 & + 0.00@xmath123 & + 0.25 & 0.15 & 0.15 & + 0.15 & + 0.00 & & + 0.75 & & & + 0.20 & + 10616594 & + 0.16@xmath111 & 0.30 & & + 0.00@xmath123 & 0.35@xmath123 & + 0.05@xmath123 & + 0.05@xmath123 & + 0.10 & + 0.10@xmath123 & + 0.10@xmath123 & + 0.10@xmath123 & + 0.20@xmath123 & + 0.20@xmath123 & + 0.45 & + 0.30@xmath123 & + 0.05@xmath123 & + 0.00@xmath123 & 0.20 + 10977859 & 0.11@xmath113 & 0.05 & & + 0.20@xmath123 & + 0.10 & 0.05 & 0.10@xmath123 & & 0.15 & 0.35 & 0.10 & & & 0.25 & & & 0.05 & + 11498538 & 0.15@xmath23 & 0.15 & & + 0.20@xmath123 & 0.15 & 0.15@xmath123 & 0.20@xmath123 & 0.15 & 0.20 & 0.30@xmath123 & 0.10@xmath123 & 0.15 & 0.40 & + 0.65 & & & 0.20@xmath123 & + 12353648 & 1.05(fe ) & 0.25 & & 0.55@xmath123 & 0.25 & 0.90@xmath123 & 1.05@xmath123 & & 1.35 & 2.05@xmath123 & 1.25@xmath123 & & & 0.75 & & & 1.05@xmath123 & + [ table : individualabundances ] we base our analysis on high - resolution , high signal - to - noise ratio ( s / n ) spectra taken with the coud - echelle spectrograph attached to the 2-m telescope of the thringer landessternwarte tautenburg , germany . the spectra have a resolution of 32000 and cover the wavelength range from 4720 to 7400 . table [ table : observations ] represents the journal of observations and gives the kepler input catalog ( kic ) number , an alternative designation , the number of obtained spectra , the visual magnitude , the spectral type as is indicated in the simbad database , and the period of observations in 2010 . the number of acquired spectra is different for different stars since we aimed to reach a s / n of about 100 for the mean , averaged spectrum of each object . the data have been reduced using standard eso - midas packages . the data reduction included bias and stray - light subtraction , cosmic rays filtering , flat fielding by a halogen lamp , wavelength calibration by a thar lamp , and normalisation to the local continuum . all spectra were additionally corrected in wavelength for individual instrumental shifts by using a large number of telluric o@xmath124 lines . the cross - correlation technique was used to estimate the rvs from the individual spectra so that the single spectra finally could be shifted and added to build the mean , high s / n ratio averaged spectrum of each star . the rvs computed at this step have also been used to check for possible variations due to binarity , high - amplitude stellar oscillations and rotational modulation . the `` classical '' method of spectrum analysis by means of equivalent width measurements and subsequent fitting of the ionization equilibria of different elements requires the star to rotate slowly , so that a sufficient number of clean , un - blended spectral lines can be identified and measured in the stellar spectrum . in the case of rapidly rotating stars , this method fails due to the high percentage of blended lines . the method of spectrum synthesis , on the other hand , is based on the comparison between observed and theoretical spectra in a certain wavelength range . its advantage is that the effect of line blending can be taken into account when computing the synthetic spectra and thus no restrictions with respect to the rotational velocity occur . because of the large number of fast rotators in our sample , we use the second method comparing the observed spectra with a huge number of synthetic spectra computed on a grid in the stellar parameters . against the much faster approach of solving the so - called inverse problem , i.e. to determine the physical parameter values directly from the observations using some non - linear optimization method , the grid search has the advantage that it will always determine the globally best solution if the grid is dense enough . its principal disadvantage of much longer computing time plays no crucial role in our analysis since the synthetic spectra are calculated using a large library of pre - computed model atmospheres ( table[table : atmospheremodels ] ) and the analysis runs very fast on up to 300 processor cores of a cluster pc . our code gssp ( grid search in stellar parameters ) finds the optimum values of effective temperature @xmath2 , surface gravity @xmath3 , microturbulent velocity @xmath10 , metallicity @xmath4 $ ] , and projected rotational velocity @xmath5 from the minimum in @xmath125 obtained from a comparison of the observed spectrum with the synthetic ones computed from all possible combinations of the before mentioned parameters . the errors of measurement ( 1@xmath126 confidence level ) are calculated from the @xmath125 statistics . a detailed description of the method is given in @xcite ( paper i ) . for the calculation of synthetic spectra , we use the lte - based code synthv @xcite which allows the computation of spectra based on individual elemental abundances . the code uses pre - calculated atmosphere models which have been computed with the most recent , parallelised version of the llmodels program @xcite . both programs make use of the vald database @xcite for a pre - selection of atomic spectral lines . the main limitation of the llmodels code is that the models are well suitable for early and intermediate spectral type stars but not for very hot and cool stars where non - lte effects or absorption in molecular bands may become relevant , respectively . 2.3 mm rccc & @xmath127 & @xmath2 & @xmath2 ( dered ) + 01571152 ' '' '' & 0.03 & 6820@xmath128140 & 6980@xmath128190 + 02166218 & 0.01 & 7050@xmath128140 & 7110@xmath128200 + 03217554 & 0.04 & 7650@xmath128160 & 7910@xmath128250 + 03453494 & 0.03 & 7650@xmath128150 & 7840@xmath128240 + 04847411 & 0.05 & 7290@xmath128150 & 7600@xmath128220 + 05088308 & 0.02 & 6730@xmath128150 & 6840@xmath128190 + 05164767 & 0.06 & 6770@xmath128130 & 7100@xmath128190 + 05446068 & 0.04 & 4950@xmath128110 & 5060@xmath128130 + 05785707 & 0.03 & 7840@xmath128160 & 8060@xmath128260 + 06289468 & 0.03 & 8130@xmath128170 & 8300@xmath128280 + 06509175 & 0.08 & 7080@xmath128140 & 7560@xmath128220 + 06587551 & 0.05 & 8280@xmath128170 & 8760@xmath128350 + 06756386 & 0.01 & 7860@xmath128150 & 7930@xmath128240 + 07748238 & 0.05 & 7150@xmath128150 & 7450@xmath128220 + 08623953 & 0.04 & 7720@xmath128150 & 7990@xmath128250 + 08738244 & 0.01 & 8110@xmath128160 & 8190@xmath128240 + 08750029 & 0.05 & 7250@xmath128140 & 7560@xmath128220 + 09413057 & 0.05 & 8270@xmath128170 & 8720@xmath128340 + 09764965 & 0.01 & 7450@xmath128170 & 7510@xmath128230 + 09812351 & 0.01 & 7750@xmath128150 & 7830@xmath128230 + 10119517 & 0.00 & 6380@xmath128160 & 6380@xmath128160 + 10451090 & 0.01 & 7700@xmath128160 & 7750@xmath128240 + 10616594 & 0.00 & 5240@xmath128140 & 5240@xmath128140 + 10977859 & 0.01 & 8130@xmath128170 & 8200@xmath128260 + 11498538 & 0.03 & 6490@xmath128160 & 6530@xmath128190 + 12353648 & 0.02 & 7350@xmath128180 & 7470@xmath128230 + [ table : sed ] derived spectroscopically ( open circles ) with the kic ( filled boxes ) and photometric values ( open triangles and stars ) . see text for detailed description . ] 0.0 mm lclclc + & @xmath129 $ ] & & @xmath130@xmath2(k ) & & @xmath130@xmath3 + 0.8+0.8 & 0.1 & l 450010000 ' '' '' + 1000022000 + & c 100 ' '' '' + 250 + & c 2.55.0 ' '' '' + 3.05.0 + & 0.1 + & * 41888 * + [ table : atmospheremodels ] the method has been tested on spectra of vega and successfully applied to _ kepler _ @xmath131cep and spb candidate stars ( paper i ) . in paper i , the chemical composition of the stars has been determined by means of an iterative procedure involving ( 1 ) the estimation of @xmath2 , @xmath3 , @xmath5 , @xmath10 , and @xmath4 $ ] , ( 2 ) the determination of the individual abundances , element - by - element , by fixing all parameters derived in the previous step and taking the abundance table corresponding to the derived metallicity as a first guess , and ( 3 ) the re - estimation of @xmath2 , @xmath3 , @xmath5 , and @xmath10 based on the chemical composition evaluated in the second step . in the latest version of the gssp code , we still iterate the individual abundances element - by - element after the first step , but together with @xmath2 , @xmath3 , @xmath5 , and @xmath10 . this allows us to avoid the additional third step . in order to allow the possibility of an incorrect normalisation to be taken into account , and to minimize its influence on the results , we also introduced an additional free parameter that allows the adjustment of the observed continuum relative to the synthetic one during the fitting procedure . table [ table : fundamentalparameters ] summarises the results of spectrum analysis for all 26 stars of our sample . the first four columns of the table represent correspondingly the kic - number of a star , the effective temperature @xmath132 , surface gravity @xmath133 , and metallicity @xmath4^{\rm{k}}$ ] as is indicated in the kic . the five following columns list the stellar parameters derived from our spectra , while the last two columns represent the spectral types as estimated from @xmath2 and @xmath3given in the kic and determined in this work , respectively . in both cases , the spectral types and the luminosity classes have been derived using an interpolation in the tables by @xcite . we achieve a mean accuracy of about @xmath134 for @xmath2 , about @xmath128 0.16 dex for @xmath3 , and about @xmath135 for @xmath5 . table [ table : individualabundances ] lists the elemental abundances derived for each target star . the metallicity given in the second column of the table refers to the initially derived chemical composition and was used as initial guess for the determination of the individual abundances . all abundances are given relative to solar values , i.e. negative / positive values refer to an under-/overabundance of the corresponding element compared to the solar composition . we assume the chemical composition of the sun given by @xcite and these values are listed in the header of table [ table : individualabundances ] below the element designations . for some of the stars we have reached the metallicity limit in our grid of atmosphere models ( kic03217554 , 03453494 , 09812351 , and 12353648 ) . in these cases , we give the derived fe - abundance instead of the metallicity . in all other cases , the derived fe abundance matches the derived metallicity within the measurement error . the abundance errors are estimated to be about @xmath1280.1 dex for the elements showing a sufficient number of strong spectral lines in the considered region and about @xmath1280.2 dex for the elements represented in the spectrum by only few and rather faint spectral lines . the derived @xmath2 and @xmath3 are discussed in sect.[section : kic ] . here , we focus on metallicity and abundance anomalies based on table [ table : individualabundances ] , and on possible binarity of the target stars . _ stars of lower metallicity . _ only three stars in our sample of 26 show metallicities slightly higher than the sun ; all other stars have lower metallicity . fifteen stars have a metallicity of more than 0.3 dex lower than the sun . the four stars of lowest metallicity ( kic03217554 , 03453494 , 09812351 , and 12353648 ) show underabundances of the fe - peak elements and ca of about 1 dex but much less for mg and si . two of them ( kic09812351 and 12353648 ) have c abundances comparable to the solar value , resembling the characteristics of @xmath9 boo stars . _ abundance anomalies . _ for eleven of the analysed stars the ba abundance is found to deviate by more than 0.4 dex from the derived metallicity . in only one of them ba is underabundant , all other are ba enhanced . since the ba abundance was determined from only one resonance line at @xmath9 4934 which is known to be strong and sensitive to non - lte effects , this result must be interpreted with caution . 2.0 mm rcrcc & bjd-245000 & & drv & max . diff + & & ( kms@xmath14 ) & ( kms@xmath14 ) & ( kms@xmath14 ) + 04847411 ' '' '' & 362.552194 & 2.189 & 0.058 & + & 365.476249 & 2.136 & 0.137 & + & 381.512639 & 0.968 & 0.204 & + & 388.449139 & 3.357 & 0.267 & 5.5 + 05164767 & 316.464427 & 1.493 & 0.129 & + & 316.486187 & 0.798 & 0.152 & + & 316.509927 & 1.886 & 0.083 & + & 318.457963 & 0.599 & 0.109 & + & 318.479550 & 0.194 & 0.123 & 3.4 + 06289468 & 338.560657 & 0.248 & 0.285 & + & 345.482244 & 2.466 & 0.512 & + & 345.510347 & 0.107 & 0.368 & + & 345.532246 & 0.507 & 0.420 & + & 345.554214 & 2.314 & 0.394 & 4.8 + 09413057 & 340.573022 & 5.208 & 0.104 & + & 351.507244 & 6.058 & 0.082 & + & 363.430141 & 0.895 & 0.540 & + & 365.404034 & 7.092 & 0.736 & + & 376.484499 & 3.279 & 0.517 & 13 + 10119517 & 428.472508 & 110.03 & 0.066 & + & 429.421495 & 80.42 & 0.058 & + & 430.379879 & 29.61 & 0.059 & 190 + [ table : rvs_indiv ] most of the stars show overabundances of mg and si compared to the derived metallicity , and for eight stars na is found to be significantly enhanced . additionally , ti is found to be depleted in the atmospheres of kic05164767 , 08750029 and 12353648 . for kic03217554 , we did not find consistent results for the iron peak elements , there is a large scatter among the abundances of fe , cr , and ni . _ rv variations . _ five stars in our sample show differences in the measured rvs that are much larger than the errors of measurement . they are listed in table [ table : rvs_indiv ] ( note that the rvs are on a relative scale so that the mean rv is zero for each star ) . from an inspection of the rvs , we suspect for two of these stars ( kic04847411 and 09413057 ) a periodic variation - although the time basis is too short to distinguish between possible periods . kic09413057 is found by @xcite to be a hybrid pulsator . one of the possible periods that fits the observed rv variations of this star is about 1.6 d and could be caused by pulsations . for one of the five stars ( kic10119517 ) we observed an extremely large difference in rv of 190 kms@xmath14 within one day . the effective temperature can also be determined from the spectral energy distribution ( sed ) . for our target stars we constructed seds using literature photometry from 2mass @xcite , tycho @xcite , tass @xcite , usno - b1 @xcite and ucac3 @xcite . @xmath2 was determined by fitting solar - composition @xcite model fluxes to the photometry . the model fluxes were convolved with the photometric filter response functions . a weighted levenberg - marquardt non - linear least - squares fitting procedure was used to find the solution that minimised the difference between the observed and model fluxes . since @xmath136 is poorly constrained due to lack of uv flux measurements , we fixed log@xmath137 = 4.0 for all the fits . this introduces additional errors of about 50 k into the determined effective temperature for the stars showing significant deviations from the assumed value of @xmath3 . since spectral energy distributions can be significantly affected by interstellar reddening , we measured the equivalent widths of the interstellar nad lines if present in our spectra and determined @xmath127 using the relation given by @xcite . table [ table : sed ] lists the results . columns 3 and 4 correspondingly give the temperature values derived without and with the effects of interstellar reddening taken into account . figure [ temperature_comp ] compares the spectroscopically derived @xmath2 with the photometric and kic values ( typical errors of the kic data are @xmath128200 k for @xmath2 and @xmath1280.5 dex for both @xmath3 and metallicity ) . the stars are sorted by the spectroscopic @xmath2 value starting with the coolest object . for most of the targets we find a rather good agreement between the spectroscopically determined temperature ( open circles ) and the value listed in the kic ( filled boxes ) . whereas in most cases the @xmath2 from the uncorrected sed fit ( open triangles ) is slightly lower than the spectroscopic one , the temperatures corrected for the interstellar reddening ( asterisks ) show rather a good agreement . in the following , we discuss stars that show larger deviations from this general tendency in @xmath2 based on fig.[temperature_comp ] or large deviations from the kic values in the other parameters based on table[table : fundamentalparameters ] . _ kic05446068 : _ the spectroscopically derived temperature exceeds the kic value by 400k and the de - reddened photometric one by 700k . our fit with the best synthetic spectrum is rather poor and the spectrum of a second star can clearly be seen in the residuals . we assume the star to be a sb2 star so that none of the derived temperatures may be valid . for labels ) and the @xmath1 dor ( dashed lines ) and @xmath6 sct ( solid lines ) theoretical instability strips in the hr - diagram . filled circles indicate suspected binaries , open circles the stars for which no reliable fit has been obtained . ] ' '' '' & light curve & label + 01571152@xmath15 ' '' '' & & & a + 02166218 & @xmath1 dors or hybrids & @xmath1 dors & b + 05164767 ' '' '' & & & g + 03453494 ' '' '' & & & d + 06509175 & & & k + 07748238 & & [ -1.5ex]hybrids & n + 09764965 & & & s + 04847411@xmath138 ' '' '' & & & e + 08623953 & [ -1.5ex]@xmath6 scts & & o + 08750029 & & & q + 09812351 & & [ -1.5ex]@xmath6 scts & t + 10451090 & & & v + 12353648 & & & z + 06289468 ' '' '' & & & j + 06587551 & & & l + 06756386 & & hybrids & m + 08738244 & possibly @xmath6 scts & & p + 09413057 & & & r + 05785707 ' '' '' & & & i + 10977859 & & [ -1.5ex]@xmath6 scts & x + 05446068@xmath50 ' '' '' & & hybrid & h + 10616594@xmath139 & too cool & @xmath6 sct & w + 10119517 & & not pulsating & u + 03217554@xmath15 ' '' '' & & @xmath6 sct & c + 05088308@xmath15 & too evolved & @xmath1 dor & f + 11498538@xmath108 & & no classification & y + + + [ table : classification ] _ kic06587551 : _ according to the spectroscopic findings , this is the hottest star of our sample , in agreement with the photometrically evaluated @xmath2 corrected for the interstellar reddening . both the kic and the uncorrected photometric values are by about 500 k lower . this is an interesting fact since we already showed in paper i that , for the hotter stars ( @xmath2@xmath1408000k ) , the kic values of @xmath2 are systematically lower than the spectroscopic ones , suspecting that the reason may be that the interstellar reddening was not properly taken into account when deriving the kic temperatures . _ kic04847411 : _ this example shows that the parameters listed in the kic can be unreliable . figure [ compprofiles ] ( top panel ) compares the observed h@xmath141 line profile ( solid , black line ) with the synthetic ones computed from the parameters derived by us ( dashed , red line ) and from those listed in the kic ( dotted , green line ) . obviously , the green spectrum does not match the observations at all . since the metallicity value of @xmath142 listed in the kic is much lower than the limiting value in our grid of atmosphere models , we expect the deviation from the observations to be even larger , in particular for all metal lines . our spectra show rv variations with an amplitude of about 5.5kms@xmath14 and a period of @xmath14310 d. we need more spectra to reveal the nature of this variability . _ kic08738244 : _ this star shows a large discrepancy between the derived values of @xmath3 and @xmath4 $ ] and those listed in the kic . figure [ compprofiles ] ( second panel ) compares the observed spectrum with synthetic spectra in one metal line region and shows that the model based on the kic values gives no reliable fit . the same is the case for the balmer line profiles . _ kic09812351 and 12353648 : _ these are metal poor stars with metallicities below the limit of our grid of atmosphere models . both show overabundances of c , mg , and si compared to the derived fe abundance , while the spectrum of kic12353648 additionally exhibits a strong depletion of ti . the derived metallicities , represented in this case by the fe abundance , are much lower than the values listed in the kic and kic12353648 additionally shows a discrepancy of about 300 k in @xmath2 . the effect of the different parameter sets on the synthetic spectra is illustrated in two lower panels of figure [ compprofiles ] . figure [ hr_diagram ] shows the positions of all stars of our sample in the log(l / l@xmath7)log@xmath2 diagram , together with the @xmath6 sct and @xmath1 dor instability strips . the latter were reconstructed from ( * ? ? ? * figures 2 and 9 ) . the edges of the @xmath6 sct instability region have been computed with a mixing - length parameter of @xmath144=1.8 for the fundamental mode ( solid thin lines ) and for a radial order of @xmath145=4 ( solid thick lines ) . the edges of the @xmath1 dor instability regions computed with @xmath144=2.0 and 1.5 are represented by dashed thin and thick lines , respectively . to place the stars into the diagram , we estimated their luminosities from the spectroscopically derived @xmath2 and @xmath3 by means of an interpolation in the tables by @xcite . the luminosity error bars represent a combination of the errors in @xmath2 and @xmath3 , and so in some cases they appear to be significantly larger than the uncertainties in @xmath2 . beside that , the luminosity errors can still be underestimated due to the uncertainties in the empirical relations . realizing that , we base our classification mainly on the position of the stars in the hr - diagram according to the derived temperatures . table [ table : classification ] classifies the stars according to their type of variability , listing the classifications expected from their location in the hr - diagram ( fig.[hr_diagram ] ) and derived by @xcite from the frequency analysis of the _ kepler _ light curves . there are six `` outliers '' in the log(l / l@xmath7)-log@xmath2 diagram ( fig . [ hr_diagram ] ) . three of them ( labels c , f , and h ) are suspected binaries and two ( labels w and y ) are the stars for which no reliable fit of the observed spectrum could be obtained . for these objects we can not give a certain classification . for the remaining , sixth object , kic10119517 ( label u ) , no pulsations could be found from the light curve analysis . for most stars of our sample , the classification based on the light curve analysis appears to be fully consistent with the position of the objects in the log@xmath2-log(l / l@xmath7 ) diagram . we confirm three @xmath1 dor stars ( labels a , b , and g ) , and 10 @xmath6 sct pulsators lying in the expected region of the hr - diagram . four of them , however , have been classified by @xcite as hybrid pulsators although they do not fall in the overlapping region between the @xmath1 dor and @xmath6 sct stars in our hr - diagram . one star , kic04847411 , was not analyzed by @xcite . its light curve classification listed in table [ table : classification ] is based on our own analysis of the first quarter of _ kepler _ data . there are ten further stars that show @xmath6 sct - like oscillations in their light curves . six of them have been classified by @xcite as hybrid pulsators but do not fall in the overlapping region in the hr - diagram . five stars ( labels i , j , m , p , and x ) are close to the hot border of the @xmath6 sct instability region , two other ones ( labels l and r ) are distinctly hotter than given by this border . four stars are too cool ( labels h , w ) or too evolved ( labels c , f ) to be hybrid , @xmath1 dor or @xmath6 sct pulsators . four stars of our sample are reported by @xcite to be binaries . three of them could be identified as sb2 stars . for the fourth one , more observations are needed to confirm its binarity spectroscopically . fifteen of the analysed targets show metallicities which are lower by more than 0.3 dex than the metallicity of the sun . the four stars of lowest metallicity show underabundances of about 1 dex . two of them , kic09812351 and 12353648 , have a c abundance comparable to the solar value which might be a sign of @xmath9 boo nature . additionally to the c abundance , this type of variable stars is characterised by solar abundances of n , o and s. we did not find any spectral lines of these elements in the considered wavelength range that could be used for an abundance determination , however . we also find that most of the analysed stars are rather fast rotators with projected rotational velocities above 90kms@xmath14 . we determined the fundamental parameters of 26 stars in the _ kepler _ satellite field of view proposed to be candidates for @xmath1 dor - type variables @xcite . the analysis was done by means of the spectrum synthesis method based on the comparison between the observed and synthetic spectra . as an additional test of the derived @xmath2 , we computed seds by using photometry from literature and determined @xmath2 by fitting solar - composition @xcite model fluxes to the photometric data . a comparison of the results from the different methods was made . besides some outliers , where the reasons can be explained , the @xmath2 derived from the spectrum analysis shows a good overall agreement with the values given in the kic . for the hottest star of our sample , the kic value appears to be underestimated . this agrees with our finding in paperi that the @xmath2 given in the kic are in general too low for the hotter stars because the interstellar reddening was not properly taken into account . the @xmath2 following from the sed fitting are systematically lower . this can be explained by the interstellar reddening . our correction for this effect by using the equivalent widths of the interstellar nad lines to derive @xmath146@xmath147@xmath148 , improves the situation although in some cases the resulting @xmath2 is found to be slightly overestimated . the accuracy of the values for @xmath3 and @xmath4 $ ] in the kic is rather poor . an uncertainty of @xmath1280.5dex is stated in the catalogue for both parameters , in some cases we also find larger deviations from our analysis so that the values given in the catalogue are not suited to check for the quality of our findings . the spectroscopically derived fundamental parameters allow us to place the stars in a hr - diagram and to compare their location with the classification made by @xcite based on the oscillation frequencies found in the _ kepler _ light curves . in the result , we observed most of the stars in a relative compact region of the diagram that reaches from the cool edge of the @xmath6 sct instability strip to a region left of its hot border . for all of the six outliers that are too cool or too evolved to fall into the @xmath6 sct instability region we could find an explanation either by binary nature or insufficient convergence of the parameter determination . we find three stars ( labels a , b , and g ) out of the 14 stars that show oscillations in their light curves typical for @xmath1 dor or @xmath1 dor-@xmath6 sct hybrid pulsators that certainly fall into the @xmath1 dorrange of the diagram . ten further stars are found to be located in one of the @xmath6 sct regions of the hr - diagram , six of them show @xmath6 sct and in four @xmath6 sct and @xmath1 dor - typical oscillations co - exist . this shows that oscillations with periods in the @xmath1 dor range are much more common among the @xmath6 sct stars than described by the theoretical @xmath1 dor instability region . this finding is in agreement with @xcite who investigated a sample of 234 _ kepler _ stars and found a significant number of hybrid pulsators , whereas theory predicts the existence of hybrids in only a small overlapping region of the instability strips . we find seven stars close to the left or left of the blue edge of the @xmath6 sct instability strip calculated for fourth radial overtone pulsations . only two of these stars show @xmath6 sct - like oscillations but five of them show oscillations with periods in the @xmath6 sct and in the @xmath1 dor range . similar to our results , @xcite found a significant number of stars showing @xmath6 sct or @xmath6 sct-@xmath1 doroscillations which are hotter than predicted by the theoretical blue edge of the @xmath6 sct instability strip calculated for fourth radial overtone pulsations . we found four stars with very low metallicities in the -1 dex range . two of them have about solar c abundance which could be a sign of @xmath9 boo nature . both stars show @xmath6 sct - like but no @xmath1 dor - typical oscillations . thus we did not find any hint for a relationship between the @xmath9 boo stars and @xmath1 dor - type variability in our sample . the research leading to these results has received funding from the european research council under the european community s seventh framework programme ( fp7/20072013)/erc grant agreement n@xmath149227224 ( prosperity ) . this research has made use of the simbad database , operated at cds , strasbourg , france . aerts , c. , christensen - dalsgaard , j. , & kurtz , d. w. 2010 , asteroseismology , springer , heidelberg balona l. a. , krisciunas k. , & cousins a. w. j. , 1994 , mnras , 270 , 905 balona , l. a. , guzik , j. a. , uytterhoeven , k. , et al . 2011a , mnras , 415 , 3531 balona , l. a. , pigulski , a. , cat , p. d. , et al . 2011b , mnras , 413 , 2403 borucki w. j. , koch d. g. , basri g. , et al . , 2011 , apj , 728 , 117 bruntt h. , de cat p. , & aerts c. , 2008 , a&a , 478 , 487 cousins , a. w. j. 1992 , the observatory , 112 , 53 cuypers j. , aerts c. , de cat p. , et al . , 2009 , a&a , 499 , 967 de cat p. , eyer l. , cuypers j. , et al . , 2006 , a&a , 449 , 281 droege t. f. , richmond m. w. , sallman m. p. , & creager r. p. , 2006 , pasp , 118 , 1666 dupret m .- a . , grigahcne a. , garrido r. , gabriel m. , & scuflaire r. , 2005 , a&a , 435 , 927 gilliland r. l. , brown t. m. , christensen - dalsgaard j. , et al . , 2010 , pasp , 122 , 131 gray , r. o. & kaye a. b. , 1999 , aj , 118 , 2993 grigahcne a. , antoci v. , balona l. , et al . , 2010 , apj , 713 , l192 grevesse n. , asplund m. , & sauval a. j. , 2007 , space sci . , 130 , 105 gustafsson b. , bell r. a. , eriksson k. , & nordlund a. , 1975 , a&a , 42 , 407 guzik , j. a. , kaye , a. b. , bradley , p. a. , cox , a. n. , & neuforge , c. 2000 , apjl , 542 , l57 handler g. , & shobbrook r. r. , 2002 , mnras , 333 , 251 henry g. w. , fekel f. c. , & henry s. m. , 2007 , aj , 133 , 1421 hoeg e. , bssgen g. , bastian u. , et al . , 1997 , a&a , 323 , l57 kaye a. b. , handler g. , krisciunas k. , poretti e. , & zerbi f. m. , 1999 , pasp , 111 , 840 king h. , matthews j. m. , rowe j. f. , et al . , 2007 , arxiv:0706.1804 krisciunas k. , aspin c. , geballe t. r. , et al . , 1993 , mnras , 263 , 781 kupka f. g. , ryabchikova t. a. , piskunov n. e. , stempels h. c. , & weiss , w. w. , 2000 , baltic astronomy , 9 , 590 kurucz r. l. , 1979 , apjs , 40 , 1 kurucz r. , 1993 , atlas9 stellar atmosphere programs and 2 km / s grid . kurucz cd - rom no . 13 . cambridge , mass . : smithsonian astrophysical observatory , 1993 . , 13 , kurucz r. l. , 1995 , astrophysical applications of powerful new databases , 78 , 205 lehmann h. , tkachenko a. , semaan t. , et al . , 2011 , a&a , 526 , a124 marquardt d. , 1963 , siam j. appl . , 11 , 431 mathias p. , le contel j .- , chapellier e. , et al . , 2004 , a&a , 417 , 189 monet d. g. , levine s. e. , canzian b. , et al . , 2003 , aj , 125 , 984 munari u. , & zwitter t. , 1997 , a&a , 318 , 269 paunzen e. , weiss w. w. , heiter u. , et al . , 1997 , a&as , 123 , 93 paunzen e. , 2004 , mnras , 224 , 443 rodrguez e. , & breger m. , 2001 , a&a , 366 , 178 rodrguez e. , surez j. c. , moya a. , et al . , 2007 , communications in asteroseismology , 150 , 131 rowe j. f. , matthews j. m. , cameron c. , et al . , 2006 , communications in asteroseismology , 148 , 34 sadakane k. , 2006 , pasj , 58 , 1023 schmidt - kaler th . , 1982 , in landolt - brnstein , ed . k. schaifers , & h. h. voigt ( springer - verlag ) , 2b shulyak d. , tsymbal v. , ryabchikova t. , sttz c. , & weiss w. w. , 2004 , a&a , 428 , 993 skrutskie m. f. , cutri r. m. , stiening r. , et al . , 2006 , aj , 131 , 1163 tsymbal v. , 1996 , m.a.s.s . , model atmospheres and spectrum synthesis , 108 , 198 uytterhoeven k. , et al . , 2011a , arxiv:1111.1667 uytterhoeven k. , moya a. , grigahcene a. , et al . , 2011b , a&a 543 , 125 zacharias n. , finch c. , girard t. , et al . , 2010 , aj , 139 , 2184

in paper i we studied the theory of gravitational microlensing for a planar distribution of point masses . in this second paper , we extend the analysis to a three - dimensional lens distribution . first we study the lensing properties of three - dimensional lens distributions by considering in detail the critical curves , the caustics , the illumination patterns , and the magnification cross - sections of multiplane configurations with two , three , and four point masses . for point masses that are widely separated in lagrangian space , we find that there are critical curves in total , but that only of these produce prominent caustic - induced features at the high magnification end of . in the case of a low optical depth random distribution of point masses , we show that the multiplane lens equation near a point mass can be reduced to the single plane equation of a point mass perturbed by weak shear . this allows us to calculate the caustic - induced feature in the macroimage magnification distribution as a weighted sum of the semi - analytic feature derived in paper i for a planar lens distribution . the resulting semi - analytic caustic - induced feature is similar to the feature in the planar case , but it does not have any simple scaling properties , and it is shifted to higher magnification . the semi - analytic distribution is compared to the results of previous numerical simulations for optical depth , and they are in better agreement than a similar comparison in the planar case . we explain this by estimating the fraction of caustics of individual lenses that merge with those of their neighbors . for , the fraction is , much less than the for the planar case . = cmmib10 scaled 1200 = cmmib10 = cmbsy10 scaled 1200 = cmbsy10 = = = = = = = `` 710b = ' ' 710c = `` 710d = ' ' 710e = `` 710f = ' ' 7110 = `` 7111 = ' ' 7112 = `` 7113 = ' ' 7114 = `` 7115 = ' ' 7116 = `` 7117 = ' ' 7118 = `` 7119 = ' ' 711a = `` 711b = ' ' 711c = `` 711d = ' ' 711e = `` 711f = ' ' 7120 = `` 7121 = ' ' 7122 = `` 7123 = ' ' 7124 = `` 7125 = ' ' 7126 = `` 7127 = ' ' 7272 = " 7201 # 1#1 # 1

astro-ph9608139

the gravitational lensing effect of an ensemble of point masses has long been a subject of investigation ( see , e.g. , schneider , ehlers , & falco 1992 for a summary of earlier work ) . most of the studies have focused on the situation where the spatial distribution of the point mass lenses is `` compact , '' i.e. , the size of the region within which the lenses are distributed is much smaller than the other distances ( those between the source and the lenses and between the observer and the lenses ) in the problem . for gravitational lensing calculations , a compact distribution of lenses is usually approximated by its projection onto a single lens plane . the important case of low optical depth , single plane microlensing has been investigated in great detail ( kaiser 1992 ; mao 1992 ; rauch 1992 ; kofman 1996 , hereafter paper i ; see also schneider 1992 and references therein ) , and we now understand reasonably well the caustic structure , the illumination pattern , the probability distribution of macroimage magnification , and the flux variability of macroimages in this case . there are , however , situations in which the lensing objects are modeled by an `` extended '' ( or three - dimensional ) distribution of point masses , i.e. , by a point mass distribution that is comparable in scale to the other distances in the problem . a well - known example is the microlensing of stars in the large magellanic cloud and the galactic bulge by foreground stars and possibly massive compact halo objects ( machos ) ( alcock 1993 ; aubourg 1993 ; udalski 1993 ) . the theoretical analysis of this problem is relatively simple because the optical depth to microlensing is very small ( @xmath9 ) and the point mass lenses are essentially acting independently ( paczyski 1986 , 1991 ; griest 1991 ) . ( note , however , that microlensing by physical binary systems is also possible ; mao & paczyski 1991 ; udalski 1994 ; alard , mao , & guibert 1995 . ) a more complicated problem is the possibility that a substantial fraction of the dark matter in the universe is in the form of compact objects ( e.g. , carr 1994 ) . then the universe as a whole has a significant optical depth to microlensing ( press & gunn 1973 ; dalcanton 1994 ) . this latter problem has been studied by several authors using numerical simulations ( refsdal 1970 ; schneider & weiss 1988a , b ; rauch 1991 ) , and it was found that there are qualitative differences between two- and three - dimensional microlensing . for example , for @xmath5 , the three - dimensional macroimage magnification distribution does not show the prominent caustic - induced feature produced by a planar distribution of point masses ( rauch 1992 ; but see 4 below ) . the usual approach to gravitational lensing by a three - dimensional distribution of lenses is the multiple lens - plane approximation ( blandford & narayan 1986 ; kovner 1987 ) . the properties of multiplane gravitational lensing and , in particular , the reasons for many of the differences between two- and three - dimensional microlensing have not been investigated in detail , and they are not yet well understood . besides the statistical microlensing studies mentioned above , there have been few studies of the basic mathematical properties of multiplane gravitational lensing ( levine & petters 1993 ; levine , petters , & wambsganss 1993 ) or the properties of specific multiplane configurations ( kochanek & apostolakis 1988 ; lee & paczyski 1990 ; erdl & schneider 1993 , hereafter es ) . this was perhaps due to the lack of lensing candidates with lenses at different redshifts . however , recent observations of the lens systems 2345@xmath10007 ( pell 1996 ) , 2016@xmath10112 ( lawrence , neugebauer , & matthews 1993 ) , and 0957@xmath10561 ( bernstein , tyson , & kochanek 1993 ; but see dahle , maddox , & lilje 1994 ) suggest that these systems may involve lenses at different redshifts . as we shall see in 3 , to understand some of the differences between two- and three - dimensional microlensing , it is useful to study in detail specific cases of simple three - dimensional lens configurations the study by erdl & schneider ( es ) of the lens system consisting of two point masses at different distances from the observer is a step in this direction . they considered primarily the critical curves and the caustics and were able to obtain a complete classification of their topological properties with respect to the model parameters in this particular case . the main motivation of this study is to give a systematic theory of gravitational microlensing by a random distribution of point masses . in the accompanying paper i we refined the theory of gravitational microlensing for a planar distribution of point masses . in particular , we obtained a practical semi - analytic expression for the macroimage magnification distribution , @xmath4 , for low optical depth @xmath11 , by allowing for the shear perturbations of neighboring lenses . in this paper we study the properties of gravitational microlensing by a three - dimensional distribution of point masses . we are interested in the critical curves , the caustics , the illumination pattern , and especially the macroimage magnification distribution . we shall not consider other interesting aspects such as time delays and the temporal behavior of microlensing events ( note , however , that the latter is given by the magnification along tracks through the illumination pattern ) . in 2 we write down the lens equations . in 3 we study in detail multiplane configurations with two , three , and four point masses to find out what new features are expected in three - dimensional microlensing . we also explain qualitatively why these new features arise . the case of low optical depth three - dimensional microlensing is studied in 4 , using an approach similar to that used in paper i to study the caustic - induced feature in @xmath4 for the planar case . the resulting semi - analytic @xmath4 is compared to the numerical results of rauch ( 1991 ) , and the differences between the two- and three - dimensional results are discussed . our conclusions are summarized in we begin by establishing the relevant equations for gravitational lensing by a three - dimensional lens distribution . as in paper i , we shall only consider gravitational lensing in an einstein - de sitter cosmological background weakly perturbed by the gravitational field of the lenses . our notation is identical to that used in paper i , and the readers are referred to 2 of paper i for the definitions of terms not explicitly defined here . let us consider an isotropic point source at the origin of our coordinates and an observer plane at a comoving distance @xmath12 from the source ( see fig . 1 ) . in the absence of perturbations , the observer plane would be uniformly illuminated and a light ray which leaves the source with angle @xmath13 would pierce the observer plane at @xmath14 . this defines our planar lagrangian coordinates @xmath15 . in the presence of density inhomogeneities , the light ray would suffer deflections and pierce the observer plane at the eulerian coordinates @xmath16 . for a three - dimensional ( or extended ) lens distribution , gravitational lensing is usually studied within the framework of the multiple lens - plane approximation ( blandford & narayan 1986 ; kovner 1987 ) . it is assumed that the lens distribution can be approximated by a series of @xmath17 geometrically thin configurations , i.e. , lens planes , and that the separations between the lens planes are sufficiently large . then the deflection of a light ray by one of the lens planes is not influenced by the others and a light ray is deflected @xmath17 times on the way from the source to the observer ( see fig . 1 ) . after @xmath18 deflections , the location at which the light ray pierces the @xmath19th lens plane , projected onto the observer plane , is @xmath20 where @xmath21 is the deflection by the @xmath22th plane ( which can be expressed as the gradient with respect to @xmath23 of an effective surface gravitational potential on plane @xmath22 ) , @xmath24 is a dimensionless distance parameter defined by @xmath25 and @xmath26 is the comoving distance between planes @xmath22 and @xmath19 . note that @xmath24 can take on values between 0 ( in the limit @xmath27 ) and 1 ( in the limit @xmath28 or @xmath29 ) . if we denote the observer plane as plane @xmath30 , then the lagrangian to eulerian mapping is @xmath31 where @xmath32 unlike the single plane lens mapping , this multiplane lens mapping is not in general a gradient mapping . for the particular case where the lens planes are populated by point masses , @xmath33 where @xmath34 is the cosmological scale factor at the redshift of lens plane @xmath22 and @xmath35 is the location of the @xmath36th lens ( of mass @xmath37 ) on lens plane @xmath22 . note that @xmath35 is the position projected onto the observer plane , i.e. , @xmath38 where @xmath39 is the angular position of the lens from the source . henceforth , we shall assume for simplicity that all the point masses have the _ same _ mass @xmath40 . then @xmath41 where @xmath42 is the ( lagrangian ) einstein radius of a point mass on the @xmath22th lens plane . for a random distribution of point masses , the optical depth to microlensing is @xmath43 , where @xmath44 and @xmath45 are , respectively , the projected ( onto the observer plane ) surface number density and the physical surface mass density of point masses on lens plane @xmath22 . when we consider the situation where a significant fraction of the dark matter in the universe is in the form of cosmologically distributed compact objects , since the mean of the density inhomogeneities should be zero , we should compensate the randomly distributed point masses on each plane by a _ negative _ smooth surface mass density . then a term of the form @xmath46 should be added to the right hand side of equation ( [ eq5 ] ) . as we mentioned in 1 , numerical simulations of gravitational microlensing by a three - dimensional random distribution of point masses have revealed differences between two- and three - dimensional microlensing ( refsdal 1970 ; schneider & weiss 1988a , b ; rauch 1991 ) . to find out and understand qualitatively some of these differences , it is useful to study in detail specific cases of simple lens configurations . in this section we study numerically the properties of multiplane gravitational lensing by a few point masses . to reduce the number of parameters , we restrict our study to cases where the distance between the source and the observer is much smaller than the horizon scale , i.e. , @xmath47 . if there are @xmath48 point masses on lens plane @xmath22 ( we shall denote the total number of point masses as @xmath49 ) , the sum of the squared ( lagrangian ) einstein radii is @xmath50 for the numerical computations , it is convenient to express all positions in units of @xmath51 and to define a dimensionless mass parameter @xmath52 so that @xmath53 . then the multiplane lens equation with identical point masses ( eqs.[[eq1 ] ] and [ [ eq5 ] ] ) can be written as @xmath54 first let us consider in detail configurations with two point masses . without any loss of generality , we can choose a coordinate system such that @xmath55 and @xmath56 . since we assume that @xmath47 , the two remaining parameters , @xmath57 and @xmath58 , are related by @xmath59 . the topology of the critical curves ( and the caustics ) as a function of the parameters @xmath60 has been classified by es ( see also schneider & weiss 1986 for the @xmath61 case ) . in general , for given @xmath57 and @xmath58 , there are five transitions in the topology of the critical curves as @xmath62 increases from @xmath63 ( see fig . 6 of es ) . however , in the limit @xmath64 ( i.e. , in the limit of two point masses on the same plane ) , three of the transitions merge at @xmath65 and the number of transitions is reduced to two . for comparison , we show in figures 27 two sets of models with @xmath61 and @xmath66 . ( we have also computed models with @xmath67 , but the results are qualitatively similar to those with @xmath66 and are not shown here . ) the configurations shown in figures 27 were chosen to sample regions of the parameter space with different critical curve topologies . the parameters @xmath68 and the critical curve topology types ( in the notation of es ) of these configurations are listed in table 1 . ( note that the topology type @xmath69 is not represented ; with @xmath59 , type @xmath69 only occurs at very small @xmath62 . ) in figures 2 and 5 we show the critical curves of the configurations with @xmath61 and @xmath70 respectively . the corresponding caustics and iso - magnification contours ( on the observer plane ) are shown in the left and right panels of figures 3 and 6 . in figures 4 and 7 we plot the `` normalized '' differential cross - section @xmath71 , where @xmath0 is the differential cross - section for magnification @xmath72 and @xmath73 is the differential cross - section of a single isolated point mass lens with einstein radius @xmath51 as defined in equation ( [ eq7 ] ) . as we saw in paper i ( see also 4 below ) , the function @xmath74 is a useful measure of the deviations of the cross - section of a more complex lens configuration from that of the point mass lens . the cross - sections and the iso - magnification contours shown in these ( and all subsequent ) figures were obtained using the ray - shooting method ( see , e.g. , kayser , refsdal , & stabell 1986 ; schneider & weiss 1986 , 1988b ) . since the magnification factor is calculated in this method by collecting rays into a grid of square pixels , it is the magnification for an observing device ( or equivalently a source ) of the size of one pixel . to assess the effects of finite pixel size , we performed two sets of calculations , with the linear pixel sizes in one set a factor @xmath75@xmath76 smaller ( see table 1 ) . the pixel sizes were chosen so that all ( except the very small ) caustics are resolved . we show in figures 3 and 6 the contours from the calculations with the smaller pixels and in figures 4 and 7 the function @xmath74 from both sets of calculations . as expected , the finite pixel size introduces an upper limit to the magnification factor and produces extra `` bumps '' in @xmath74 ( rauch 1992 ) , but it does not have significant effects on the position or the strength of the real caustic - induced features in @xmath74 . the @xmath61 ( and @xmath77 ) models have already been studied in detail by schneider & weiss ( 1986 ) ; hence , we only give a brief description of these models here . for @xmath78 , the region near each of the point masses is perturbed by the weak shear from the other point mass . so we have two critical lines that are slightly flattened ellipses and two caustics that have the familiar `` astroid '' shape ( see 3 and fig . 2 of paper i ) . as @xmath79 , the critical curves and the caustics become asymmetric ( figs . 2_a _ and 3_a _ ) . the corresponding differential cross - section ( fig . 4_a _ ) shows caustic - induced features that are similar to those found in the case of a point mass perturbed by shear ( see fig . 3 of paper i ) . there is , in addition , a bump ( at @xmath80 ) which is associated with the last iso - magnification contour that encloses both of the caustics . as @xmath62 decreases below @xmath76 , the number of critical curves ( and caustics ) changes from two to one to finally three , and the topology of the critical curves ( and the caustics ) changes from @xmath81 to @xmath82 to finally @xmath83 ( see table 1 ) . in all cases , there are caustic - induced features in the differential cross - section at moderate to high @xmath72 and the cross - section at small @xmath72 is reduced . each of these features is associated with either a local minimum or saddle point in the magnification pattern or an iso - magnification contour that touches the caustics at one or more points . in the latter case , the contours of slightly larger @xmath72 are broken into disconnected shorter segments by the caustics and the loss of the area inside the caustics results in a drop in @xmath74 . ( an example that has been discussed in detail in 3 of paper i is the feature associated with the contour osculating the astroid - shaped caustic of the point mass plus weak shear lens . ) when the two point masses are not on the same lens plane , there are new features and transitions in the critical curves and the caustics ( see figs . 5 and 6 ) . at large lagrangian separation @xmath62 ( fig . 5_a _ ) , besides the two critical curves that are similar to the ones seen in the @xmath61 case ( see 4 for an explanation ) , there is an extra critical curve inside the critical curve that encloses point mass @xmath84 ( i.e. , the point mass closer to the source ) . this critical curve is offset from point mass @xmath84 in the direction opposite point mass @xmath76 and both its size and its distance from point mass @xmath84 increase with decreasing @xmath62 . the corresponding caustic is located at a large distance along the positive @xmath85-axis ( fig . this caustic has the ( asymmetric ) astroid shape of the two caustics that are known from the @xmath61 case , and it increases in size and moves toward the other two caustics as @xmath62 decreases . the critical curve ( and caustic ) configuration at large @xmath62 that we have just described is denoted as topology type @xmath72 by es . except for the presence of the new critical curve and caustic , the topological transitions shown in figures 5_a__c _ and 6_a__c _ are similar to those shown in figures 2 and 3 . the remaining transitions do not exist in the single lens plane case . however , as es pointed out , except for the elliptic umbilic transition between @xmath86 and @xmath87 , all transitions are the beak - to - beak type . we can explain the new critical line ( and caustic ) and its qualitative behaviors if we focus on the configurations of topology types @xmath72@xmath86 . for these configurations , it is clear that the new critical line is composed of the set of rays that are strongly deflected by point mass @xmath84 ( on lens plane @xmath84 ) to the vicinity of point mass @xmath76 ( on lens plane @xmath76 ) and are then infinitely magnified by point mass @xmath76 . obviously this process can only occur when the point masses are not on the same lens plane . the new critical line is closer to point mass @xmath84 for larger @xmath62 because a light ray must be more strongly deflected by point mass @xmath84 to reach the vicinity of point mass @xmath76 . in the magnification patterns ( fig . 6 ) , there are local minima and saddle points that are associated with the new caustic ; there are also iso - magnification contours that touch the new caustic at two points ( e.g. , the second outermost contour in fig . these new features in the magnification patterns are expected to produce corresponding features in the cross - sections . however , as we can see in figure 7 , the `` normalized '' differential cross - sections @xmath74 are actually quite similar to those shown in figure 4 . features associated with the new caustic are not noticeable until @xmath88 , and even in cases where the new caustic ( which is transformed at small @xmath62 ) encloses a large area of the observer plane , the new features are typically weak and at @xmath89 . ( they are stronger and at higher @xmath72 for the @xmath67 models . ) if we examine the magnification patterns ( and fig . 8) more carefully , we see that the new caustic is usually located in regions where the background magnification is small . in addition , the magnification throughout most of the region inside the new caustic and , in particular , the minimum magnification inside the new caustic are only slightly higher than the low background magnification . hence the new features in the magnification patterns and the corresponding features in @xmath74 have small @xmath72 values . next , let us consider specifically the feature associated with the iso - magnification contour that touches the new caustic at two points ( e.g. , the second outermost contour in fig . this contour is long and , except for the distortion near the new caustic , nearly circular . contours of slightly higher magnification are broken by the new caustic into a short arc around one of the cusps and a long arc that encloses the other caustic(s ) . this results in a decrease in the cross - section @xmath0 because an area that would otherwise be enclosed by two of these contours ( of magnification @xmath72 and @xmath90 ) is lost to the interior of the caustic . although the area lost to the caustic can be large , it is typically much smaller than the area enclosed by the long ( and the short ) arcs . consequently , the decrease in @xmath91 , which is approximately the fractional decrease in @xmath0 , is small . similarly , the increase in @xmath74 at @xmath92 , where @xmath93 is the minimum magnification inside the new caustic , is small because @xmath93 is typically small and the area enclosed by the contours of magnification @xmath93 and @xmath94 outside the new caustic ( mainly two long arcs ) is much larger than the area enclosed by the contour @xmath94 inside the new caustic . the relatively weak influence of the new caustic on the magnification pattern is demonstrated more clearly in the sample of magnification profiles plotted in figure 8 . although the magnification must be infinite on the caustic , only a very small region near the caustic has @xmath95 . in fact , the numerically obtained @xmath72 , which is finite at a caustic because of the finite pixel size of our calculations , shows a maximum at the new caustic that is quite small . this behavior of the magnification pattern near the new caustic matches the description of the `` thin caustics '' found in the microlensing simulations of schneider & weiss ( 1988b ) . we have therefore found in the simple two - point - mass case a clear demonstration of the occurrence of thin caustics . as we mentioned earlier , the new caustic consists of light rays that are strongly deflected by the first point mass to the vicinity of the second point mass and are then infinitely magnified by the second point mass . in the simulations of schneider & weiss , which have a large number of point masses distributed on several lens planes , a light ray can be strongly deflected on more than one lens plane . it is likely that most of the caustics that are strongly deflected on one or more lens planes ( i.e. , the secondary caustics defined in 3.2 below ) are thin caustics . this identification for the thin caustics is consistent with the conjecture by schneider & weiss that the thin caustics are caused by steep deflection gradients because the regions ( in lagrangian space ) with strong deflections are also regions with steep deflection gradients . now let us consider briefly configurations with three and four point masses . two configurations with @xmath96 point masses and one with @xmath97 are shown in figures 911 . these configurations have one point mass on each lens plane ( i.e. , @xmath98 ) and their parameters are listed in table 2 . the configuration _ a _ with @xmath99 was chosen because it has the critical curve topology of point masses widely separated in lagrangian space while the smaller critical curves are clearly resolved . the other configurations are `` typical '' configurations resulting from randomly distributing the point masses within a cone of constant lagrangian radius . we have ( arbitrarily ) chosen a cone radius such that a light ray passing within the cone would be infinitely magnified if the mass of the lenses were uniformly distributed within the cone . a striking feature of figure 9 ( and fig . 5 ) is the rapid increase in the number of critical curves with @xmath1 . for a configuration with @xmath1 point masses well separated in lagrangian space and with each point mass on its own lens plane ( i.e. , @xmath98 ) , we can calculate the number of critical curves by extending the discussion given in 3.1 for the origin of the new critical curve . light rays from the source can reach the vicinity of the point mass on the @xmath22th lens plane after being strongly deflected by any combination ( or none ) of the previous @xmath100 point masses . for each combination , there is a set of rays that are infinitely magnified by the @xmath22th point mass ( and the weak deflections on the remaining planes ) . therefore , there are @xmath101 critical curves associated with the @xmath22th point mass and the total number of critical curves is @xmath102 . the critical curves shown in figures 5_a _ and 9_a _ are consistent with these numbers . ( the total number of critical curves is smaller for configurations with @xmath103 and approaches @xmath1 in the limit @xmath104 , i.e. , in the single lens plane case . ) hereafter , we shall refer to the caustics that are not strongly deflected as the _ primary _ caustics and those that are strongly deflected on at least one lens plane as the _ secondary _ caustics . although the number of critical curves increases exponentially with @xmath1 , we can see in figure 9 that the size of a critical curve decreases rapidly as the number of strong deflections associated with the critical curve ( or equivalently the number of critical curves within which it is nested ) increases . on the other hand , as in the two - point - mass case , a caustic formed after a larger number of strong deflections is not necessarily smaller . for example , in figure 10_c _ , the caustic located near @xmath105 [ which corresponds to the critical curve located near @xmath106 and the caustic with six cusps [ which corresponds to the largest critical curve ] are comparable in size . in the two - point - mass case , due to the absence of overlapping or self - intersecting caustics , a complete classification of the critical curve topology also provides a complete classification of the caustic topology ( es ) . this is not true when @xmath107 . in particular , there are overlaps of caustics that belong to different critical curves . examples of such overlapping caustics are the caustics located near @xmath108 in figure 10_a _ and those in the lower left corner of figure 10_c_. ( there are no self - intersecting caustics in the configurations shown in fig . 10 , but they may be present in other configurations . ) finally , we discuss the normalized differential cross - sections plotted in figure 11 . for the `` typical '' configurations _ b _ and _ c _ , the qualitative behaviors of @xmath74 agree with what one might expect from the two - point - mass configurations studied in 3.1 . the cross - section at small @xmath72 is smaller than that for a single point mass . there are prominent features at moderate to high @xmath72 that are associated with the primary caustics . features associated with the secondary caustics are typically weak , but some of them are noticeable at @xmath89 . for the configuration _ a _ , the features associated with the primary caustics , i.e. , those at @xmath109 , are similar to the features in figures 4_a _ and 7_a_. this is expected since configuration _ a _ has the critical curve topology of widely separated point masses . however , the features associated with the secondary caustics are at unusually high @xmath72 : the `` plateau '' between @xmath110 and @xmath111 consists of a series of such features . this is due to the fact that the parameters of configuration _ a _ were specially chosen so that the smaller critical curves are clearly resolved . for these parameters , the caustics are unusually close to one another ( see fig . 10_a _ and notice the change in scale between figs . 10_a _ and 10_b _ ) . as a result , the secondary caustics are located in regions of relatively high background magnification and the associated features in @xmath74 are at unusually high @xmath72 . the study of simple lens configurations in the previous section shows that microlensing by a three - dimensional distribution of point masses is , in general , much more complicated than microlensing by lenses confined to a single lens plane . but there are also indications that , in the low optical depth limit , some aspects of the three - dimensional problem can be reduced to the planar one . for example , the multiplane configurations with point masses widely separated in lagrangian space show primary caustics and corresponding features in @xmath74 that are similar to those found in the single plane configurations . in 4.1 we confirm this analytically for the limiting case of a point mass lens perturbed by the weak shear from the other lenses . another useful result from the study in 3 is that the significant features in @xmath74 at moderate to high @xmath72 are associated with the primary caustics . therefore , we can neglect the features associated with the secondary caustics when we analyze the macroimage magnification distribution , @xmath4 , at high @xmath72 . in 4.2 we derive the feature induced by the primary caustics at the high magnification end of @xmath4 for a three - dimensional low optical depth lens distribution . the resulting semi - analytic @xmath4 is compared to previous numerical results in 4.3 , and the differences between the two- and three - dimensional results are discussed in 4.4 . in the low optical depth limit , lensing ( other than low magnification events ) by a three - dimensional distribution of point masses is , as in the planar case , typically dominated by a single point mass . without loss of generality , we shall designate one of the point masses residing on the @xmath112th lens plane as the dominant one and locate this point mass at the origin of the lens screen . on all of the other screens , the light ray is only weakly perturbed , and the perturbations are expanded to the first order in @xmath113 . before the light ray passes the @xmath112th screen with the dominant lens ( i.e. , for @xmath114 ) , the lens mapping ( eqs.[[eq1 ] ] and [ [ eq5 ] ] ) is simplified to @xmath115 where the constant deflection @xmath116 and the shear matrix @xmath117 ( similar to @xmath118 and @xmath119 defined in eq.[13 ] of paper i ) represent the perturbative influence of the lenses on the @xmath22th plane . equation ( [ eq11 ] ) can also be written as @xmath120 where @xmath121 and @xmath122 is the identity matrix . once the light ray has traversed the @xmath112th lens screen , the location at which the light ray intersects subsequent ( i.e. , @xmath123 ) screens is given by @xmath124 where @xmath125 denotes the influence of the dominant lens and @xmath126 note that , in the absence of the dominant lens , @xmath127 is the path followed by a light ray with @xmath128 . on the observer plane [ which is the @xmath129th screen , but we also denote it with the subscript `` o '' ] , the eulerian position is given by @xmath130 if we use the fact that @xmath131 and transform the eulerian variable according to @xmath132 , the above equation can be cast into the form @xmath133 where @xmath134 , @xmath135 although there is only shear on each lens screen , the matrix @xmath136 is in general a combination of shear and convergence ( i.e. , it is symmetric but not traceless ) . thus equation ( 19 ) is formally equivalent to the single plane lens equation for the point mass lens plus shear and smooth surface density ( compare with eq.[15 ] of paper i for the point mass plus shear lens ) . this is a specific example of the `` non - linear telescope '' ( or generalized quadrupole lens ) discussed by kovner ( 1987 ; see also schneider 1992 ) . we are interested in the case where the shear perturbations on all the lens screens are weak . to first order in @xmath137 , the expressions for @xmath138 , @xmath139 , and @xmath136 are @xmath140 and @xmath141 respectively . hence @xmath136 is reduced to a shear matrix : @xmath142 , where @xmath143 is the magnitude of the `` effective '' shear and @xmath144 then equation ( 19 ) reduces to the single plane lens equation for the point mass plus weak shear lens ( eq.[15 ] of paper i ) . in 3.12 of paper i , we studied in detail the properties of single plane lensing by the point mass plus weak shear lens . the results are applicable to the mapping from @xmath145 to @xmath146 ( eq.[19 ] ) . we can then infer the properties of the lens mapping from @xmath15 to @xmath16 ( eq.[[eq18 ] ] ) by taking into account the effects of the transformations @xmath147 and @xmath148 . first , it is easy to see that , to first order in @xmath137 , both transformations are a combination of translation and shear distortion and are one - to - one . second , since the deformation tensor @xmath149 there is a simple correspondence between the caustics ( and the critical curves ) of the mappings ( [ eq18 ] ) and ( 19 ) . in addition , because the determinants @xmath150 and @xmath151 are @xmath152 , the magnification @xmath153 at a point @xmath16 is , to first order in @xmath143 , identical to the `` magnification '' @xmath154 at the corresponding point @xmath155 . similarly , the correction factors for the differential cross - section are @xmath152 . all of the corrections that we have just considered are of the same order as the terms already ignored in the weak shear analysis in paper i. therefore , the effects of the transformations are negligible , and the properties derived in 3.1 - 2 of paper i are also applicable to the lens mapping ( [ eq18 ] ) . briefly , for a point mass lens perturbed by weak shear in three dimensions , the critical line is a slightly flattened ellipse ( of size @xmath156 ) ; the caustic has the shape of an astroid , with the four cusp catastrophes ( at a distance @xmath157 from the center of the astroid ) connected to each other by four fold catastrophes ( see fig . 2 of paper i ) ; the differential cross - section has the scaling property ( see eq.[23 ] of paper i ) @xmath158 ) = { 2 \pi x_{e , l}^2 \over ( a^2 -1)^{3/2 } } \ , \varphi(a s'[\chi_{lo } ] ) , \label{eq23}\ ] ] where the normalized differential cross - section @xmath159 depends on @xmath72 and @xmath143 only through the combination @xmath160 ; and @xmath159 , which is shown in figure 3 of paper i , shows strong caustic - induced features at the critical values @xmath161 and @xmath162 . as we mentioned in 2 , when we consider the situation where a significant fraction of the dark matter in the universe is in the form of cosmologically distributed compact objects , we should compensate the randomly distributed point masses on each lens plane by a negative smooth surface mass density . the derivation in this section can be generalized to this case if we replace @xmath137 in all the equations ( except eq.[[eq12b ] ] ) by @xmath163 . the results remain valid , but the magnification @xmath72 should be interpreted as the magnification with respect to an `` empty beam '' ( i.e. , a light beam that is not affected by the point masses ) . as we discussed earlier , in the case of low optical depth microlensing by a three - dimensional distribution of point masses , features associated with the secondary caustics are negligible at the high magnification end of the macroimage magnification distribution @xmath4 . therefore , as in the two - dimensional case ( see 3.3 of paper i ) , @xmath4 at high @xmath72 can be approximated as a superposition of the cross - sections of the individual point mass lenses , with the shear perturbation due to the other point masses varying from lens to lens . let us assume that the point masses are distributed randomly on @xmath17 lens planes , each with projected ( onto the observer plane ) surface number density @xmath164 . then @xmath4 can be written as @xmath165 where @xmath166 is the cross - section given in equation ( [ eq23 ] ) , @xmath167 is the probability distribution of the effective shear @xmath143 for a dominant lens at @xmath168 , @xmath169 is the probability that the dominant lens is located on the @xmath112th screen , and @xmath170 is the total projected surface number density . since @xmath169 is simply @xmath171 we have @xmath172 where @xmath173 the function @xmath174 has the same structure as @xmath4 of the two - dimensional case ( eq.[29 ] of paper i ) , and it is the contribution to the macroimage magnification distribution by the lenses on the @xmath112th lens screen as the dominant lenses . the effective shear @xmath143 that enters into the cross - section @xmath166 is a weighted superposition of the shear induced by all except the dominant lens ( eq.[[eq20b ] ] ) . in the appendix we show that @xmath175^{3/2 } } , \label{eq28}\ ] ] where @xmath176 is the `` effective '' optical depth at @xmath168 . [ note that the effective optical depth @xmath177 is smaller than the ( usual ) optical depth @xmath178 because the extra factors @xmath179 ( for @xmath180 ) and @xmath181 ( for @xmath182 ) are less than unity . ] substituting equations ( [ eq23 ] ) and ( [ eq28 ] ) into equation ( [ eq27 ] ) , we find that @xmath174 is similar to @xmath4 for a single lens plane ( compare with eq.[31 ] of paper i ) : @xmath183,\ ] ] where @xmath184 is the function introduced in 3.3 of paper i to describe the caustic - induced feature in @xmath4 in the single lens plane case ( see fig . 4 of paper i ; also fig . 12 of this paper ) . thus @xmath174 exhibits the same caustic - induced bump ( with a maximum enhancement of @xmath185 at @xmath186 ) and the same scaling property ( with @xmath187 being a function of the combination @xmath188 ) as @xmath4 of the single lens plane case . there is , however , a significant difference between the two- and three - dimensional cases . to obtain the macroimage magnification distribution in the three - dimensional case , we must sum @xmath174 over all the lens planes ( eq.[[eq26 ] ] ) : @xmath189 = { 2 \tau \over ( a^2 - 1)^{3/2 } } f_n ( a ; \tau ) , \label{eq31}\ ] ] where @xmath190 , \label{eq32}\ ] ] and @xmath191 is the ( usual ) optical depth . although we denote the caustic - induced feature in the @xmath17-plane case as @xmath192 for simplicity , it should be clear from equation ( [ eq32 ] ) that @xmath192 depends on the location and the surface density of the lens planes in a non - trivial manner and does not have any simple scaling properties ( when @xmath193 ) . note also that the differences between @xmath194 and @xmath192 are due to the differences between the optical depths @xmath11 and @xmath177 . to proceed further , we need to assume a model for the distribution of the point mass lenses . as in the numerical simulations reported by several authors ( refsdal 1970 ; schneider & weiss 1988a , b ; rauch 1991 ) , we consider a lens distribution with _ constant _ comoving mass density @xmath195 . we divide the volume between the source and the observer into @xmath17 shells of equal comoving thickness ( @xmath196 ) and place the point masses in each shell on a lens plane at the middle of the shell . in figure 12 we plot the numerically evaluated @xmath192 for @xmath104 , @xmath76 , @xmath197 , @xmath198 , and @xmath199 , and for source redshift @xmath200 and @xmath201 . it is clear that @xmath202 has converged by @xmath203 . this agrees with the conclusion from a study of lensing by smoothly variable mass distribution that a fully three - dimensional mass distribution is well approximated by as few as eight or @xmath199 screens ( lee & paczyski 1990 ) . it is also clear that @xmath204(@xmath205 ) is very insensitive to the source redshift , at least for @xmath206 . in the limit @xmath207 , the summations in the expressions for @xmath11 , @xmath177 , and @xmath192 are replaced by integrals : @xmath208 ; \\ f_\infty ( a ; \tau ) & = & { 4 \pi g \rho_l a_0 ^ 3 \over \tau } \int_0^{\chi_{so } } d\chi { \chi \left(\chi_{so}-\chi\right ) \over a(\chi ) \chi_{so } } f_1 \left[a \tau { \tau ' \over \tau}\right ] .\end{aligned}\ ] ] equations ( 33a b ) can be evaluated analytically , and they are particularly simple in the limit @xmath200 ( i.e. , in the limit @xmath209 ) : @xmath210 .\end{aligned}\ ] ] thus the contribution to @xmath204 is dominated by the lens planes at @xmath211 with @xmath212 . this is consistent with the numerical results shown in figure 12 . the most significant change from @xmath187 to @xmath204 is the shift in the location of the bump from @xmath213 to @xmath214 . otherwise , the peak value ( @xmath215 enhancement ) and the width of @xmath204 are only slightly smaller and wider . several authors ( refsdal 1970 ; schneider & weiss 1988a , b ; rauch 1991 ) have used numerical simulations to calculate the macroimage magnification distribution , @xmath4 , produced by a three - dimensional distribution of point masses . in the most recent study by rauch , distributions with high resolution in @xmath72 were obtained from monte carlo simulations for three cases with @xmath216 , @xmath217 , and @xmath218 . the `` equivalent @xmath219 value '' is related to the average magnification ( with respect to an empty beam ) on the observer plane , @xmath220 , by @xmath221 , i.e. , @xmath219 is the optical depth @xmath11 of a _ two - dimensional _ lens distribution with the same @xmath220 . for a three - dimensional lens distribution , @xmath219 and @xmath11 are not identical in general , but @xmath222 in the limit @xmath223 . we wish to compare the semi - analytic @xmath4 derived in 4.2 to the distributions obtained by rauch . ( we shall focus on his results for an einstein - de sitter cosmological background , but rauch also found that the distributions for the same @xmath219 but different background cosmology are nearly identical . ) since our derivation assumes low optical depth , the best case for comparison should be the @xmath216 case . unfortunately , even with @xmath224 simulations , the monte carlo simulations only provide useful information about @xmath4 for @xmath225 ( see fig . 2 of rauch 1991 ) . this is not sufficient for detecting the caustic - induced feature because significant deviation from @xmath226 is not expected for @xmath227 ( @xmath228 for @xmath229 ) ( see @xmath204 in fig . we shall not show the comparison for this case and simply note that the numerically obtained @xmath4 ( for @xmath225 ) agrees with the low optical depth semi - analytic result , with or without the caustic - induced feature . in figure 13 we show the distributions obtained by rauch for @xmath230 and @xmath218 ( histograms ) . the data are those shown in figures 3 and 4 of rauch ( 1991 ) , but they are plotted in the form @xmath231 and we have increased the bin sizes at large @xmath72 to reduce statistical fluctuations . as rauch ( 1992 ) pointed out , the distributions in the three - dimensional case do not show the prominent bump at @xmath232 that was found in the planar case . however , the distributions are not featureless . this can be seen most clearly in the @xmath233 case ( note , e.g. , the inflection point at @xmath234 ) . in a plot of @xmath231 , the low optical depth distribution without the caustic - induced feature is a horizontal line of amplitude @xmath235 , while the semi - analytic distribution with the caustic - induced feature is @xmath236 ( eq.[[eq31 ] ] ; recall that @xmath204 is very insensitive to the source redshift ) . since @xmath230 and @xmath218 are not much less than unity , two corrections to the low optical depth semi - analytic results are necessary when we compare them to the numerical ones . first , it is no longer accurate to assume that @xmath237 . for these particular models , @xmath230 and @xmath218 correspond to @xmath238 and @xmath239 , respectively . second , the normalization @xmath240 can be significantly different from the low optical depth result @xmath241 . in 5.1 of paper i , we discussed the renormalization of @xmath4 due to the net convergence of light rays and multiple streaming for the single lens plane case . from an analytic result for @xmath4 in the high magnification limit , which is exact even for finite optical depth @xmath11 ( see also schneider 1987 ) , we adopted the normalization @xmath242 ( for @xmath223 , @xmath243 and @xmath244 , and the correction to the normalization @xmath241 is small . ) the analysis that led to equation ( [ eq35 ] ) can not be easily generalized to the three - dimensional case ( it would require the statistical properties of the matrix @xmath245 which depends on the shear matrices @xmath137 _ non - linearly _ ; see eq.[14 ] ) , and an equivalent expression for @xmath240 for a three - dimensional lens distribution is not known . nevertheless , in the range of @xmath11 of interest here ( i.e. , for @xmath246 ) , it is likely that equation ( [ eq35 ] ) is a good approximation for @xmath240 in the three - dimensional case . in equation ( [ eq35 ] ) , the dominant correction to the normalization @xmath241 is the factor @xmath220 . ( the correction due to the factor @xmath247 is second order in @xmath11 and less than @xmath248 for @xmath246 . ) the factor @xmath220 arises from converting a probability distribution in lagrangian space to a distribution in eulerian space ( see 5.1 of paper i ) . the conversion ( and hence the factor @xmath220 ) should remain in any generalization of the analysis to the three - dimensional case . furthermore , as in the planar case ( see 5.1 of paper i ) , we can derive the factor @xmath220 using a different argument : in 4.2 we assumed that the surface density of the astroid - shaped caustics on the observer plane is the same as the projected surface density of point masses @xmath249 ( see eq.[[eq24 ] ] ) ; there is , however , a net convergence of the light rays by the overall lens distribution ; this increases the density of astroids by a factor @xmath220 , and introduces an additional factor @xmath220 in the normalization of @xmath4 in equation ( [ eq31 ] ) . therefore , we shall adopt equation ( [ eq35 ] ) for the normalization @xmath240 in the three - dimensional case . in figure 13 we plot @xmath240 ( dashed lines ) and @xmath236 ( solid lines ) for @xmath230 and @xmath218 . they should be compared to the numerical results at @xmath250 ( see 4.5 for a brief discussion of the deviations at small @xmath72 ) . as we can see , the adopted normalization @xmath240 is in good agreement with the amplitudes of the numerical results . in addition , the numerical results show deviations from the horizontal dashed lines that are similar in shape to the semi - analytic caustic - induced feature @xmath204 ( note , e.g. , that the inflection point in the numerical result for @xmath233 coincides with the location @xmath251 where @xmath204 starts to deviate significantly from unity ) . in the @xmath230 case , there is little , if any , difference between the numerical and semi - analytic results ; the only possibly statistically significant deviation is that the numerical distribution may be higher near @xmath252 . in the @xmath233 case , the difference between the numerical and semi - analytic results is more pronounced , with the numerical distribution slightly higher than the semi - analytic distribution over the range @xmath253 . note , however , that the number of events in the histograms at @xmath254 is small ( see the fluctuations amongst the same bins in the @xmath230 case ) , and the true distribution may not be as flat as the histograms shown . the comparison in 4.3 shows that , in the case of a three - dimensional lens distribution , the function @xmath204 , which was derived in the low optical depth limit , provides a reasonably good fit to the caustic - induced feature in @xmath4 for @xmath11 as large as @xmath239 . this is very different from the two - dimensional case studied in paper i. for a two - dimensional lens distribution , the caustic - induced feature in the numerically obtained @xmath4 is more prominent than the low optical depth semi - analytic feature @xmath187 for @xmath11 as small as @xmath217 . this was shown in figure 6 of paper i where we compared the numerical and semi - analytic distributions for @xmath255 and @xmath218 in a plot of @xmath256 . the numerical distributions show a clear minimum at @xmath257 and are significantly higher than the semi - analytic results in the range @xmath258 . for both the two- and three - dimensional lens distributions , the low optical depth analysis assumes that the point mass lenses are well separated in lagrangian space . then the deflection near each point mass is due to that point mass and the shear perturbation from the other lenses , and an astroid - shaped ( primary ) caustic is associated with each point mass . the low optical depth analysis is not valid if the lens density is high enough that a significant fraction of the lenses have close neighbor(s ) , and their caustics are complex structures produced by the collective effect of two ( or more ) point masses . in 5.3 of paper i , we estimated for the two - dimensional case the fraction of point masses whose caustics are not isolated astroids , @xmath259 , by considering a point mass and its nearest neighbor as a two - point - mass lens : @xmath260 . for @xmath6 , @xmath261 . this explains the relatively strong contribution to @xmath4 by the collective effect of two ( or more ) point masses ( which results in the enhanced caustic - induced feature at moderate @xmath72 ) for @xmath11 as small as @xmath217 . to understand the difference between the two- and three - dimensional cases , we need to estimate @xmath259 for the three - dimensional case . we now extend the method used in paper i to estimate @xmath259 to the three - dimensional case . we begin by considering lensing configurations with two point masses on different lens planes . since our results in 4.2 and the numerical results of rauch ( 1991 ) show that @xmath4 is not sensitive to either the source redshift or the background cosmology , we shall simplify the analysis by assuming that @xmath209 . then the two point mass configurations of interest are identical to those studied in 3.1 and are specified by two parameters : the dimensionless distance @xmath24 between the lens planes and the dimensionless lagrangian separation @xmath62 between the lenses [ the actual lagrangian separation is @xmath262 ; see eq.([eq7 ] ) ] . note that we have changed the notation slightly and the point masses are on lens planes @xmath22 and @xmath19 instead of lens planes @xmath84 and @xmath76 . as we saw in 3.1 , the topology of the caustics ( and the critical curves ) changes with decreasing @xmath62 from type @xmath72 to @xmath69 . the only topology with two astroid - shaped primary caustics is type @xmath72 ; the other topology types are produced by the strong effect of both point masses ( see fig . 6 ) . in figure 14 we plot the separation @xmath263 at which the topology changes from type @xmath72 to type @xmath264 as a function of @xmath24 . ( the data were obtained using the numerical code that generated the caustics and critical curves shown in figs . the separation @xmath263 decreases with increasing @xmath24 . in the limit @xmath265 , the two point masses are on the same plane and we have @xmath266 ( schneider & weiss 1986 ) . in the limit @xmath267 ( i.e. , in the limit @xmath268 or @xmath269 ; see eq.[[eq2 ] ] ) , @xmath270 ; this is due to the focusing of the light rays by the point mass on lens plane @xmath22 and the fact that @xmath271 . let us consider first a simple extension of the method used in paper i to estimate @xmath259 . it is convenient to consider the projected distribution ( i.e. , the distribution in lagrangian space ) of the point mass lenses . if we select randomly one of the point masses , the probability that the point mass is on lens plane @xmath22 is @xmath272 , where @xmath170 is the total projected surface number density of point masses . this point mass has an absolute nearest neighbor in lagrangian space . since the point masses are distributed randomly , the probability that the absolute nearest neighbor is on lens plane @xmath19 and at a lagrangian distance less than @xmath273 is @xmath274 $ ] . if we ignore the deflection due the other lenses and consider the point mass and its nearest neighbor as a two - point - mass lens , the discussion in the previous paragraph tells us that the primary caustic associated with the point mass is not an isolated astroid if @xmath275 . therefore , an estimate of the fraction of point masses whose caustics are not isolated astroids is @xmath276 \right\ } . \label{eq36}\ ] ] in the three - dimensional case , since the ( lagrangian ) einstein radius @xmath156 increases from @xmath63 to @xmath277 as @xmath168 increases ( see eq.[[eq6 ] ] ) , it is possible , e.g. , for the primary caustic ( and critical curve ) of a point mass to merge with that of its second nearest neighbor ( from plane @xmath36 ) , but not with that of its absolute nearest neighbor ( from plane @xmath19 ) , if @xmath278 . thus the function @xmath279 , which takes into account the absolute nearest neighbor only , probably underestimates the fraction of point masses whose caustics are not isolated astroids . we can derive a better estimate by taking into account the nearest neighbors from all of the lens planes . for a randomly selected point mass ( on lens plane @xmath22 with probability @xmath272 ) , the probability that the nearest neighbor from lens plane @xmath19 is at lagrangian distance greater than @xmath273 is @xmath280 . if we ignore the deflection due the other lenses and consider the point mass and its nearest neighbor from lens plane @xmath19 as a two - point - mass lens , the two - point - mass lens produces two astroid - shaped primary caustics if @xmath281 . thus the probability that the primary caustic associated with the point mass is an isolated astroid is @xmath282 $ ] . therefore , a better estimate of the fraction of point masses whose caustics are not isolated astroids is @xmath283 \right\ } . \label{eq37}\ ] ] as with @xmath192 , although @xmath284 and @xmath285 are denoted as functions of @xmath11 , they do depend on the location and the surface density of the lens planes . for @xmath104 , i.e. , for a single lens plane , both @xmath286 and @xmath279 reduce correctly to @xmath287 . in the limit @xmath223 , @xmath288 in figure 15 we show @xmath279 for @xmath104 , @xmath76 , @xmath197 , @xmath198 , and @xmath277 ( dashed lines ) and @xmath289 ( solid line ) , using the constant comoving density model adopted in 4.2 . both @xmath284 and @xmath285 have converged by @xmath290 . the difference between @xmath291 and @xmath292 is small for @xmath293 , but it increases to @xmath294 at @xmath295 . it is clear from figure 15 that @xmath289 is significantly smaller than @xmath296 ( the highest dashed curve in fig . 15 ) . in the limit @xmath223 , @xmath297 while @xmath298 . for @xmath6 , @xmath299 and @xmath261 ; even for @xmath300 , @xmath301 only . thus , in the three - dimensional case , the fraction of point masses whose ( primary ) caustics are not isolated astroids and the contribution to @xmath4 by configurations of two ( or more ) point masses are much smaller . this explains why the low optical depth semi - analytic feature @xmath204 provides a reasonably good fit to the caustic - induced feature in @xmath4 for @xmath11 as large as @xmath239 . at sufficiently large @xmath11 ( @xmath302 ) , @xmath303 and @xmath4 should deviate significantly from the semi - analytic result . finally , we discuss briefly the deviations of @xmath4 from the semi - analytic distribution @xmath304 at low magnification ( see fig . 13 ) . as in the planar case , the superposition of the cross - sections of individual lenses that results in the caustic - induced feature @xmath204 is only valid for the magnifications seen by the observers close to one of the astroid - shaped primary caustics . the semi - analytic distribution also requires modification at low magnification as it diverges as @xmath305 at @xmath306 and is not normalizable . in the planar case , an observer not particularly well aligned with any single lens sees a primary image close to the unperturbed position of the source and a large number of faint secondary images ( one close to each lens ) produced by strongly deflected rays . in paper i , we took into account the correlations between the magnifications of the primary and the secondary images and derived the modification of @xmath4 at low magnification . we have not attempted a similar calculation in the three - dimensional case because the problem is significantly more complicated . first , for an observer that is not particularly close to either a primary or a secondary caustic , there are now @xmath307 secondary images ( where @xmath1 is the number of lenses ) because a light ray can reached the observer after being strongly deflected by one or more lenses . although the secondary images that have been strongly deflected by more than one lens are typically much fainter than those that have been strongly deflected by only one lens , they are far more numerous and we may have to account for them . a more serious problem is the presence of weak caustic - induced features at low magnification that are associated with the secondary caustics ( see 3 ) . unlike the features associated with the primary caustics at high magnification , there are no simple semi - analytic descriptions of the features associated with the secondary caustics . in this paper and paper i , we have sought to develop a systematic theory of gravitational microlensing . we focused on microlensing by a planar distribution of point masses in paper i. in this paper we have studied gravitational microlensing by a three - dimensional distribution of point masses , focusing on the nature of the critical curves , the caustics , the illumination pattern , and especially the macroimage magnification distribution @xmath4 . in the process , we have explained many of the differences between two- and three - dimensional microlensing that have been seen in previous numerical simulations . in 3 we studied the lensing properties of three - dimensional lens distributions by considering in detail multiplane configurations with two , three , and four point masses . the lensing properties of a three - dimensional lens distribution is fundamentally different from that of a planar distribution in that a light ray could have arrived in the vicinity of a lens either having suffered only weak deflections along the way from the source , or after having been strongly deflected one or more times by the intervening lens . the latter situation can not arise in a purely planar distribution of lenses , and it gives rise to secondary caustics and critical lines that have no analogues in the planar case . in addition , in the case of more than two point masses , the caustic topology is no longer simply related to the critical curve topology because caustics corresponding to different critical lines may overlap each other . for @xmath1 point masses that are widely separated in lagrangian space , there are @xmath3 primary critical curves ( i.e. , critical curves that have not suffered any strong deflections ) and @xmath2 critical curves in total . features in the `` normalized '' differential cross - section , @xmath74 , that are associated with the secondary caustics are relatively weak and at low magnifications . significant features occuring at moderate to high magnifications are invariably associated with the primary caustics . in 4 we derived a semi - analytic expression for the caustic - induced feature at the high magnification end of @xmath4 for a low optical depth lens distribution . in the low optical depth limit , we showed that the multiplane lens equation near a point mass is formally equivalent to the single plane equation of a point mass perturbed by shear . consequently , the ( primary ) caustic has the familiar astroid shape , and the differential cross - section exhibits the same scaling behavior found in paper i ( i.e. , @xmath308 depends on magnification @xmath72 and the effective shear @xmath309 through the combination @xmath310 ) . by modeling the illumination pattern as a superposition of the patterns due to individual `` point mass plus weak shear '' lenses , we found that the caustic - induced feature in the macroimage magnification distribution @xmath4 can be written as a weighted sum of the semi - analytic feature derived in paper i for a planar lens distribution ( eq.[[eq32 ] ] ) . the resulting semi - analytic caustic - induced feature is similar to the feature in the planar case , but it is shifted to higher magnification . in the moderate to high @xmath72 regime , our semi - analytic @xmath4 agrees remarkably well with the magnification distribution obtained from numerical simulations by rauch ( 1991 ) , even for optical depth as high as @xmath311 . this is very different from the two - dimensional case studied in paper i. in the two - dimensional case , we have found that the caustic - induced feature in @xmath4 tends to be more pronounced than the low optical depth result , even for optical depth as low as @xmath5 . we argued that this particular difference between two- and three - dimensional microlensing is due to the fact that for a given value of optical depth , the fraction of point masses whose ( primary ) caustics are not simple astroids is significantly higher in the planar case than in the three - dimensional case . finally , we discussed briefly the deviations of the numerical @xmath4 from the semi - analytic distribution at low magnification . we thank k. rauch for providing us with the results of rauch ( 1991 ) . we also thank s. mao and k. rauch for helpful discussions . a.b . and l.k acknowledge the hospitality shown to them at cita during their visits in 1995 . this work was supported in part by nserc ( canada ) , the ciar cosmology program , cita , the institute for astronomy ( l.k . ) , the dudley observatory ( a.b . ) , and a cita national fellowship ( m.h.l . ) . in this appendix we evaluate @xmath167 , the probability distribution of the effective shear @xmath143 at a given @xmath168 , due to a low optical depth three - dimensional distribution of point masses . it is assumed that the point masses are distributed randomly on @xmath17 lens planes , each with projected ( onto the observer plane ) surface number density @xmath164 . specifically , on each lens plane , @xmath312 point masses are distributed randomly within a circle of lagrangian radius @xmath313 . at a given @xmath168 , the effective shear ( in matrix form ) is ( eq.[[eq20b ] ] ) @xmath314 where ( eq.[[eq12b ] ] ) @xmath315 it is convenient to represent the two distinct components of the shear matrix @xmath137 as a vector : @xmath316 where @xmath317 . then the effective shear @xmath136 can be written as an appropriately weighted vector superposition : @xmath318 where we have defined @xmath319 to simplify the notations . since all the @xmath320 are statistically independent and have similar statistical properties , the formal expression for the probability distribution of @xmath321 is @xmath322 where @xmath323 is the two - dimensional delta function . to evaluate equation ( a5 ) , it is convenient to express the delta function and @xmath324 as fourier transforms : @xmath325 @xmath326 where the characteristic function @xmath327 is ( nityananda & ostriker 1984 ) @xmath328 substituting equations ( a6 ) and ( a7 ) into equation ( a5 ) , we can perform the integration with respect to @xmath320 : @xmath329 next , we integrate with respect to @xmath330 , grouping together the @xmath48 integrals associated with the @xmath22th screen : @xmath331 in the limit @xmath332 . hence , @xmath333 where the `` effective '' optical depth @xmath177 is @xmath334 performing the integration with respect to @xmath335 in equation ( a11 ) , we obtain @xmath336 since @xmath337 depends only on the magnitude of the effective shear @xmath143 , it is convenient to consider the probability distribution for the latter : @xmath175^{3/2}}.\ ] ] we have expressed the probability distribution explicitly as a conditional probability that depends on the value of @xmath168 . it should be noted that @xmath167 has the same functional form as the shear distribution due to a planar random distribution of point masses ( eq.[30 ] of paper i ) . cccc 0.0 & 2.4 & @xmath81 & 0.020 , 0.0320.0 & 1.0 & @xmath82 & 0.016 , 0.0240.0 & 0.6 & @xmath83 & 0.016 , 0.0240.3 & 2.0 & @xmath72 & 0.020 , 0.0400.3 & 1.0 & @xmath264 & 0.018 , 0.0360.3 & 0.6 & @xmath86 & 0.016 , 0.0320.3 & 0.5 & @xmath87 & 0.016 , 0.0320.3 & 0.4 & @xmath338 & 0.016 , 0.0320.3 & 0.2 & @xmath338 & 0.020 , 0.040 ccclrc a & 3 & 1 & 0.1 & @xmath339 & 0.010 , 0.024 & & 2 & 0.5 & @xmath340 & & & 3 & 0.9 & @xmath341 & b & 3 & 1 & 0.55032 & @xmath342 & 0.012 , 0.024 & & 2 & 0.79370 & @xmath343 & & & 3 & 0.94104 & @xmath344 & c & 4 & 1 & 0.5 & @xmath345 & 0.012 , 0.024 & & 2 & 0.72112 & @xmath346 & & & 3 & 0.85499 & @xmath347 & & & 4 & 0.95647 & @xmath348 &

we study the gauss - manin connection for the moduli space of an arrangement of complex hyperplanes in the cohomology of a nonresonant complex rank one local system . aomoto and kita determined this gauss - manin connection for arrangements in general position . we use their results and an algorithm constructed in this paper to determine this gauss - manin connection for all arrangements . partially supported by louisiana board of regents grant leqsf(1999 - 2002)-rd - a-01 and by national security agency grant mda904 - 00 - 1 - 0038 ] partially supported by national security agency grant mda904 - 02 - 1 - 0019 ]

math0207114

let @xmath0 be an arrangement of @xmath1 ordered hyperplanes in @xmath2 , and let @xmath3 be a local system of coefficients on @xmath4 , the complement of @xmath5 . the need to calculate the local system cohomology @xmath6 arises in various contexts . for instance , local systems may be used to study the milnor fiber of the non - isolated hypersurface singularity at the origin obtained by coning the arrangement , see @xcite . in mathematical physics , local systems on complements of arrangements arise in the aomoto - gelfand theory of multivariable hypergeometric integrals @xcite and the representation theory of lie algebras and quantum groups . these considerations lead to solutions of the knizhnik - zamolodchikov differential equation from conformal field theory , see @xcite . here a central problem is the determination of the gauss - manin connection on @xmath6 for certain arrangements , and certain local systems . a complex rank one local system on @xmath7 is determined by a collection of weights @xmath8 . associated to @xmath9 , we have a representation @xmath10 , given by @xmath11 for any meridian loop @xmath12 about the hyperplane @xmath13 of @xmath5 , and an associated rank one local system @xmath3 on @xmath7 . several authors have considered gauss - manin connections with various conditions on the weights : aomoto @xcite , schechtman and varchenko @xcite , and kaneko @xcite studied discriminantal arrangements ; kanarek @xcite studied the connection arising when a single hyperplane in the arrangement is allowed to move ; aomoto and kita @xcite determined the gauss - manin connection matrices for general position arrangements ; and terao @xcite computed these connection matrices for a larger class of arrangements . in this paper , we determine gauss - manin connection matrices for _ all _ arrangements and all weights @xmath9 which satisfy the nonresonance condition ( stv ) of @xcite stated in theorem [ thm : vanish ] . given weights @xmath9 , we used stratified morse theory in @xcite to construct a complex @xmath14 which computes @xmath15 . this construction is reviewed in section [ sec : local ] . in section [ sec : moduli ] , we compactify @xmath5 by adding the infinite hyperplane @xmath16 to obtain the projective closure @xmath17 . two arrangements are said to have the same combinatorial type if there is an order - preserving isomorphism between their intersection posets . if @xmath18 is the combinatorial type of @xmath5 , we denote by @xmath19 and @xmath20 the independence and dependence sets of @xmath18 . these consist of all @xmath21 element subsets of @xmath22=\{1,2,\dots , n+1\}$ ] for which the corresponding hyperplanes in @xmath23 are independent , and dependent , respectively . the sets @xmath24 ) and @xmath20 may be used to describe the moduli space of all arrangements of type @xmath18 , defined and investigated by terao in @xcite . let @xmath25 be a smooth , connected component of this moduli space . there is a fiber bundle @xmath26 . the fibers of this bundle , @xmath27 , are complements of arrangements @xmath28 of type @xmath18 , so they are diffeomorphic to @xmath7 since @xmath25 is connected . since the bundle @xmath29 is locally trivial , there is an associated vector bundle @xmath30 , with fiber @xmath31 at @xmath32 for each @xmath33 , @xmath34 . the transition functions of this vector bundle are locally constant . fixing a basepoint @xmath32 , the operation of parallel translation of fibers over curves in @xmath25 in the vector bundle @xmath30 provides a complex representation @xmath35 a key idea in this paper is to extract information about arbitrary arrangements using information about general position arrangements . let @xmath36 denote the combinatorial type of a general position arrangement of @xmath1 hyperplanes in @xmath2 , so @xmath37 . let @xmath38 be its moduli space , @xmath39 , @xmath40 the corresponding arrangement , with complement @xmath41 . in section [ sec : reps ] , we compare the representation with the analogous representation @xmath42 . to this end , we introduce a space @xmath43 which contains @xmath25 and @xmath38 as disjoint subspaces . let @xmath44 and @xmath45 denote the natural inclusions . given @xmath46 , we show that there is a _ compatible _ @xmath47 , a loop for which the homology classes @xmath48 $ ] and @xmath49 $ ] satisfy @xmath50)=(j_{{\mathcal t}})_*([{{\gamma } } ] ) $ ] . for compatible loops , the automorphisms @xmath51 and @xmath52 are related by @xmath53 here @xmath54 is the map in local system cohomology induced by a certain chain map @xmath55 , and @xmath56 denotes equality up to conjugacy . these compatible classes are studied further in section [ sec : compatible ] . the moduli space @xmath38 of general position arrangements is the complement of a divisor @xmath57 in @xmath58 . the components @xmath59 of this divisor are irreducible hypersurfaces indexed by @xmath21 element subsets @xmath60 of @xmath22 $ ] . for each such @xmath60 , let @xmath61 be a meridian loop in @xmath38 about a generic point in @xmath59 . we prove that the homology group @xmath62 is generated by the classes @xmath63 $ ] . note that @xmath61 is a loop in @xmath43 for any combinatorial type @xmath18 since @xmath64 . we say that the type @xmath18 covers @xmath65 if @xmath66 has complex codimension one in the closure @xmath67 . if @xmath46 is a simple loop in @xmath25 about a generic point in @xmath66 , we show that the homology class @xmath48 $ ] satisfies @xmath68 ) = \sum_{j\in \operatorname{dep}({{\mathcal t}}',{{\mathcal t } } ) } m_j \cdot [ \gamma_{\!j}],\ ] ] where @xmath69 , and @xmath70 is the order of vanishing of the restriction of a defining polynomial for @xmath59 to @xmath67 along @xmath66 . if @xmath3 is a nontrivial rank one local system on the complement @xmath71 of a general position arrangement of @xmath1 hyperplanes in @xmath2 , then the cohomology @xmath72 vanishes in all dimensions @xmath33 except possibly @xmath73 . for an arbitrary arrangement @xmath5 , a collection of weights @xmath9 and the corresponding local system @xmath3 are called _ nonresonant _ if the cohomology @xmath74 vanishes in all dimensions @xmath33 except possibly @xmath73 . we review the ( stv)-condition for @xmath18nonresonance @xcite and the @xmath75**nbc * * basis for the nonvanishing group @xcite in section [ sec : weights ] . we show that for @xmath18nonresonant weights @xmath9 , the chain map @xmath76 induces a surjection @xmath77 . the vector bundle @xmath78 , with fiber @xmath79 at @xmath32 , supports a gauss - manin connection corresponding to the representation . we study this connection in section [ sec : gm ] for @xmath18nonresonant weights , where only @xmath73 is relevant . let @xmath46 be represented by @xmath80 . a connection matrix , @xmath81 , for the induced connection on the pullback of the vector bundle @xmath82 along @xmath83 is called a gauss - manin connection matrix . let @xmath84 be a loop whose homology class is given by above , @xmath49=\sum_{j\in \operatorname{dep}({{\mathcal t}}',{{\mathcal t } } ) } m_j \cdot [ \gamma_{\!j}]$ ] . then @xmath85 and @xmath86 are compatible , and it follows from that @xmath87 . since the homomorphism @xmath88 is surjective for @xmath18nonresonant weights , this determines the gauss - manin connection matrix @xmath81 . aomoto and kita @xcite obtained explicit formulas for the gauss - manin connection matrices @xmath89 in the case of general position arrangements , see also @xcite . we combine our results with these formulas to obtain the main result of this paper . it is a combinatorial algorithm for calculating gauss - manin connection matrices for an arbitrary combinatorial type @xmath18 , and @xmath18nonresonant weights , in terms of those for the type @xmath36 of general position arrangements . let @xmath18 be a combinatorial type which covers the type @xmath65 . let @xmath90 a simple loop in @xmath25 about a generic point in @xmath66 . we agree to write @xmath91 in place of @xmath92 in this situation . similarly , we write @xmath93 in place of @xmath94 . fix @xmath18nonresonant ( and hence @xmath36nonresonant ) weights @xmath9 . for @xmath95 and @xmath96 , the corresponding local system cohomology groups of the fibers , @xmath97 and @xmath98 , depend only on the combinatorial types @xmath18 and @xmath36 and vanish for @xmath99 . so we write @xmath100 and @xmath101 . this notation makes it clear that the gauss - manin connection matrix @xmath102 depends only on the combinatorial types involved . the matrix @xmath102 is obtained by the combinatorial formula recorded below . let @xmath18 be a combinatorial type which covers the type @xmath65 . let @xmath9 be a collection of @xmath18nonresonant weights , and let @xmath103 be the matrix of the surjection @xmath104 in the respective @xmath75**nbc * * bases . then a gauss - manin connection matrix @xmath102 is determined by the matrix equation @xmath105 illustrations of the algorithm provided by this result are given in section [ sec : examples ] . in this section , we record a number of results concerning a fixed arrangement . we use notation and results from @xcite . let @xmath0 be an arrangement of @xmath1 ordered hyperplanes in @xmath2 . assume that the first @xmath106 hyperplanes of @xmath5 are linearly independent , so @xmath5 is essential . choose coordinates @xmath107 for @xmath2 , and for each hyperplane @xmath13 of @xmath5 , let @xmath108 be a linear polynomial such that @xmath13 is defined by the vanishing of @xmath109 . then @xmath110 is a defining polynomial for @xmath5 , and the complement @xmath111 is given by @xmath112 . choose coordinates @xmath113 for @xmath114 , and consider the arrangement @xmath115 in @xmath114 defined by @xmath116 . since the first @xmath106 hyperplanes of @xmath5 are linearly independent , @xmath115 is an arrangement of @xmath1 linearly independent hyperplanes in @xmath114 , a boolean arrangement , and the complement @xmath117 is the complex @xmath1-torus . define @xmath118 by @xmath119 then @xmath120 , and we have the following . [ prop : slice ] let @xmath5 be an essential arrangement of @xmath1 hyperplanes in @xmath2 . then the complement @xmath7 of @xmath5 is an affine section of the complex @xmath1-torus @xmath121 . let @xmath122 be the orlik - solomon algebra of @xmath5 , generated by one dimensional classes @xmath123 , @xmath124 . it is the quotient of the exterior algebra generated by these classes by a homogeneous ideal , hence a finite dimensional graded @xmath125-algebra . there is an isomorphism of graded algebras @xmath126 . in particular , @xmath127 where @xmath128 denotes the @xmath33-th betti number of @xmath7 with trivial local coefficients @xmath125 . this construction is realized topologically by the map @xmath129 defined above . the cohomology @xmath130 is an exterior algebra , generated by classes @xmath123 dual to meridian loops @xmath12 about the hyperplanes of @xmath115 ( respectively @xmath5 ) , and one can show that the map @xmath131 induces a surjection @xmath132 in ( de rham ) cohomology . let @xmath133 be a general position arrangement of @xmath1 hyperplanes in @xmath2 , a generic @xmath106-dimensional section of the boolean arrangement @xmath115 in @xmath114 . the orlik - solomon algebra @xmath134 is the rank @xmath106 truncation of the exterior algebra @xmath135 . we have @xmath136 for @xmath137 and @xmath138 for @xmath139 . the complement of the arrangement @xmath5 in @xmath2 has the homotopy type of an @xmath106-dimensional complex , so the cohomology of the complement vanishes in higher dimensions , @xmath140 for @xmath139 . these facts yield the following result . [ prop : gpsurjection ] the map @xmath120 induces a surjection @xmath141 . a collection of weights @xmath8 gives rise to compatible rank one local systems on the complements @xmath121 , @xmath71 , and @xmath7 of the arrangements @xmath115 , @xmath133 , and @xmath5 as follows . associated to @xmath9 , there is a rank one representation @xmath142 given by @xmath143 for any meridian loop @xmath12 about the @xmath144-th hyperplane of @xmath115 , and a corresponding rank one local system @xmath3 on @xmath121 . let @xmath120 and @xmath145 be the maps from proposition [ prop : slice ] realizing @xmath7 and @xmath71 as sections of @xmath121 . via these maps , there are induced local systems on @xmath7 and @xmath71 corresponding to representations of @xmath146 and @xmath147 defined as above . for brevity , we also use @xmath3 to denote these local systems . one can use stratified morse theory to construct a cochain complex @xmath14 , the cohomology of which is naturally isomorphic to the local system cohomology @xmath15 , see @xcite . briefly , let @xmath148 be a flag in @xmath2 which is transverse to the arrangement @xmath5 , and let @xmath149 for each @xmath33 . let @xmath150 , and let @xmath151 be the boundary homomorphism of the triple @xmath152 . [ thm : kdot ] 1 . [ item : kdot1 ] for each @xmath33 , @xmath153 , we have @xmath154 if @xmath155 , and @xmath156 is equal to the @xmath33-th betti number of @xmath7 with trivial local coefficients @xmath125 . [ item : kdot2 ] the system of complex vector spaces and linear maps @xmath157 , @xmath158 is a complex @xmath159 . the cohomology of this complex is naturally isomorphic to @xmath6 , the cohomology of @xmath7 with coefficients in @xmath3 . we now consider families of arrangements with a given combinatorial type . fix a pair @xmath160 with @xmath161 . we consider ( multi-)arrangements of @xmath1 ordered hyperplanes in @xmath162 . we assume that the arrangement @xmath5 contains @xmath163 linearly independent hyperplanes and that these are the first @xmath163 in the linear order . let @xmath164 be the coordinates of @xmath162 . choose linear polynomials @xmath165 which define the hyperplanes of @xmath5 . note that the matrix @xmath166 , @xmath167 is invertible by assumption . we embed the arrangement in projective space and add the hyperplane at infinity as last in the ordering , @xmath16 . the moduli space of all arrangements combinatorially equivalent to @xmath5 may be viewed as the set of matrices @xmath168 whose rows are elements of @xmath169 , and whose @xmath170 minors satisfy certain dependency conditions , see ( * ? ? ? * prop . 9.2.2 ) . given an @xmath171 element subset @xmath172 of @xmath22:=\{1,\dots , n , n+1\}$ ] , let @xmath173 denote the determinant of the submatrix of @xmath174 with rows specified by @xmath172 . corresponding to each such @xmath174 , we have an arrangement @xmath175 , with hyperplanes defined by the first @xmath1 rows of the matrix equation @xmath176 , where @xmath177 , that is combinatorially equivalent to @xmath5 . let @xmath178 be the complement of @xmath175 . let @xmath18 denote the combinatorial type of the arrangement @xmath5 , and let @xmath25 denote a smooth , connected component of the moduli space of all arrangements of type @xmath18 . let @xmath179 and define @xmath180 by @xmath181 . a result of randell @xcite implies that @xmath180 is a bundle , with fiber @xmath182 , the complement of the arrangement defined by @xmath95 . we now construct a number of spaces and bundles related to the moduli space @xmath25 . for any combinatorial type @xmath18 , let @xmath19 denote the set of all @xmath171 element subsets @xmath183 of @xmath22 $ ] for which @xmath184 in type @xmath18 . if @xmath18 is realizable , @xmath19 is the set of all subsets @xmath172 for which @xmath185 is linearly independent in the projective closure of an arrangement @xmath5 of type @xmath18 . similarly , let @xmath20 be the set of all @xmath171 element subsets @xmath60 of @xmath22 $ ] for which @xmath186 in type @xmath18 . given a type @xmath18 , let @xmath187 then the moduli space @xmath25 may be realized as @xmath188 note that if @xmath36 is the type of a general position arrangement , then @xmath37 , so @xmath189 . for any type @xmath18 , the moduli space @xmath38 of all general position arrangements may be realized as @xmath190 if @xmath191 , then @xmath25 and @xmath38 are disjoint subspaces of @xmath43 . let @xmath192 and @xmath45 denote the natural inclusions . let @xmath193 , and define @xmath194 by @xmath195 . then @xmath196 is the moduli space of all multi - arrangements with @xmath197 linearly independent . given @xmath198 , let @xmath199 denote the boolean arrangement of @xmath1 hyperplanes in @xmath200 with defining polynomial @xmath201 given by @xmath202 where @xmath203 are the coordinates for @xmath200 . let @xmath204 be the complement of the arrangement @xmath199 . note that @xmath205 is a complex @xmath1-torus . define @xmath206 the natural map @xmath207 defined by @xmath208 is a fiber bundle projection , with fiber the complex @xmath1-torus @xmath209 . this may be established using the thom isotopy lemma by modifying the argument given by randell @xcite . for each type @xmath18 for which @xmath210 , let @xmath211 denote the natural inclusion . recall that @xmath212 denotes the inclusion of the moduli space of @xmath18 in @xmath43 . by pulling back the above bundle along the inclusion maps @xmath211 and @xmath213 , we obtain bundles over @xmath43 and @xmath25 , with fiber the complex @xmath1-torus . denote these bundles by @xmath214 respectively . the total spaces of these bundles may be realized as @xmath215 the bundle @xmath216 and the moduli space bundle @xmath180 are related as follows . [ prop : bundlemap ] there is a bundle map @xmath217 covering the identity map of @xmath25 . let @xmath95 . corresponding to @xmath174 , we have an arrangement @xmath28 of type @xmath18 in @xmath218 , and a boolean arrangement @xmath199 in @xmath200 . these arrangements have defining polynomials @xmath219 where @xmath220 are coordinates for @xmath218 and @xmath221 are coordinates for @xmath200 . it follows that the map @xmath222 , @xmath223 , defined in restricts to a map @xmath224 of the complement of @xmath28 to the complement of @xmath199 , see proposition [ prop : slice ] . defining @xmath217 by @xmath225 yields the desired bundle map . let @xmath5 be an essential arrangement of @xmath1 hyperplanes in @xmath218 of combinatorial type @xmath18 , and let @xmath25 be a smooth , connected component of the corresponding moduli space . for each @xmath95 , a collection of complex weights @xmath226 determines a rank one local system @xmath227 on the complement @xmath228 of the arrangement @xmath28 , with monodromy @xmath229 about the hyperplane @xmath230 . since the fiber bundle @xmath26 is locally trivial , there is an associated vector bundle @xmath30 , with fiber @xmath31 at @xmath32 for each @xmath33 , @xmath34 . the transition functions of this vector bundle are locally constant . fixing a basepoint @xmath32 , the operation of parallel translation of fibers over curves in @xmath25 in the vector bundle @xmath30 provides a complex representation @xmath231 write @xmath232 to indicate the dependence of this representation on the type @xmath18 . by theorem [ thm : kdot ] , the local system cohomology of @xmath233 may be computed using the morse theoretic complex @xmath234 . the fundmental group of @xmath25 acts by chain automorphisms on this complex , see ( * ? ? ? 3.2 ) , yielding a representation @xmath235 as above , write @xmath236 to indicate the dependence of this representation on the combinatorial type @xmath18 . [ thm : inducedrep ] the representation @xmath237 is induced by the representation @xmath238 . the constructions of the previous section provide additional representations of the fundamental group of @xmath25 . recall that @xmath193 , and assume @xmath239 . then there is a bundle @xmath216 , with fiber @xmath240 , the complement of the boolean arrangement @xmath199 in @xmath200 . recall also that the the weights @xmath9 give rise to a local system on @xmath241 , which we also denote by @xmath242 since it is compatible with the local system on @xmath233 . as above , the fundamental group of @xmath25 acts by chain automorphisms on the cochain complex @xmath243 associated with this arrangement , yielding a representation @xmath244 by proposition [ prop : bundlemap ] , there is a bundle map @xmath245 , which restricts to @xmath246 on fibers . the map @xmath247 induces a chain map @xmath248 , see ( * ? ? ? 2.11 ) , which is compatible with the representations @xmath249 and @xmath250 of and in the following sense . [ prop : acd ] for each @xmath251 , there is a commutative diagram @xmath252 [ rem : gp ] recall that @xmath36 denotes the combinatorial type of a general position arrangement of @xmath1 hyperplanes in @xmath162 . if @xmath39 is a point in the moduli space of all such arrangements , let @xmath253 denote the corresponding arrangement . the above discussion , when applied to the arrangement @xmath253 , yields the following commutative diagram @xmath254 where @xmath47 and @xmath255 is the chain map induced by the bundle map @xmath256 of proposition [ prop : bundlemap ] in this particular case . the complement @xmath257 of the general position arrangement @xmath253 may be realized as a generic @xmath163-dimensional section of the complement @xmath258 of the boolean arrangement @xmath259 , and the map @xmath260 as the inclusion of this generic section . in the notation established in the construction of the morse theoretic complex @xmath261 preceding theorem [ thm : kdot ] , we have @xmath262 . from this construction , it is clear that the complex @xmath263 is the rank @xmath163 truncation of the complex @xmath264 : @xmath265 see @xcite and ( * ? ? ? 2.7 ) for explicit constructions of these complexes . the following is a consequence of the above discussion . [ prop : gpphi ] the representations @xmath266 and @xmath267 satisfy @xmath268 for @xmath269 . now recall the bundle @xmath270 , with fiber @xmath271 , and the inclusion map @xmath272 from section [ sec : moduli ] . as discussed above , an element @xmath90 induces a chain automorphism @xmath273 of the cochain complex @xmath274 . let @xmath275 be a loop representing @xmath85 . then @xmath276 is a loop representing an element , say @xmath277 , and @xmath278 also induces a chain automorphism of the complex @xmath274 , which we denote by @xmath279 . [ lem : same ] the chain automorphisms @xmath273 and @xmath279 of the complex @xmath274 coincide , @xmath280 . let @xmath281 denote the pullback of the bundle @xmath282 along the map @xmath275 , and let @xmath283 be the pullback of @xmath284 along @xmath285 . the chain automorphisms @xmath273 and @xmath279 are induced by the monodromies of the bundles @xmath281 and @xmath283 respectively . checking that these two bundles are identical , we conclude that the two monodromies are the same . the result follows . for any combinatorial type @xmath18 , there is an inclusion @xmath286 of the moduli space @xmath189 of all general position arrangements in the space @xmath43 . more generally , if @xmath287 and @xmath18 are types satisfying @xmath288 , there is an inclusion @xmath289 . [ prop : onto ] if @xmath287 and @xmath18 are combinatorial types with @xmath290 , the map @xmath291 induced by the inclusion @xmath292 is surjective . it suffices to consider the case where @xmath293 for some @xmath294 $ ] with @xmath295 . let @xmath296 be the type with @xmath297 , and let @xmath298 and @xmath299 . then @xmath300 , and @xmath301 , where @xmath302 . since @xmath303 is of complex codimension two in @xmath58 for each @xmath33 , the inclusion of @xmath304 in @xmath58 induces an isomorphism @xmath305 . thus @xmath306 . with these observations , the mayer - vietoris sequence in homology is of the form @xmath307 since the map @xmath308 is given by @xmath309 , the result follows . [ cor : gponto ] for any combinatorial type @xmath18 , the inclusion @xmath310 induces a surjection @xmath311 . the above results provide a means for comparing the representations @xmath312 corresponding to an arbitrary arrangement of @xmath1 hyperplanes in @xmath2 and a general position arrangement of @xmath1 hyperplanes in @xmath2 . [ thm : conj ] let @xmath18 be a combinatorial type with @xmath210 . let @xmath175 be an arrangement of type @xmath18 , corresponding to a point @xmath313 . let @xmath40 be a general position arrangement , corresponding to a point @xmath314 . then for any element @xmath315 , there is an element @xmath316 so that the diagram @xmath317 commutes up to conjugacy . let @xmath275 be a loop representing @xmath46 . then @xmath318 is a loop representing @xmath319 . using proposition [ prop : acd ] and lemma [ lem : same ] , we have the commutative diagram @xmath320 where @xmath279 is the chain automorphism induced by @xmath278 . denote the homology class of @xmath278 in @xmath321 by @xmath322 $ ] . by corollary [ cor : gponto ] , the inclusion @xmath323 induces a surjection @xmath324 . let @xmath325 be a basepoint ( the point @xmath32 is not in @xmath38 if @xmath191 ) . let @xmath316 be an element whose homology class @xmath49 $ ] satisfies @xmath50 ) = [ \xi]$ ] . then the chain automorphisms induced by @xmath326 and @xmath327 are conjugate , so we have the commutative diagram @xmath328 recall that @xmath189 . over @xmath38 and @xmath43 , we have bundles @xmath329 and @xmath330 , each with fiber @xmath331 , the complement of the boolean arrangement @xmath259 . let @xmath332 be a loop representing @xmath47 . it is then readily checked that the pullbacks @xmath333 and @xmath334 of the two bundles above are identical . consequently , the chain automorphisms @xmath335 and @xmath336 of @xmath264 induced by @xmath86 and @xmath337 are equal . this fact , together with , yields the commutative diagram @xmath338 since @xmath339 and @xmath340 for @xmath341 by remark [ rem : gp ] and proposition [ prop : gpphi ] , and @xmath342 for @xmath139 , this completes the proof . given @xmath46 , we say that @xmath316 is _ compatible _ with @xmath85 if their homology classes satisfy @xmath50)=(j_{{\mathcal t}})_*([{{\gamma } } ] ) $ ] . the results established above show that for every @xmath85 there exists a compatible @xmath86 . a key tool in this paper is the relationship between the automorphism @xmath343 for an _ arbitrary _ arrangement and the automorphism @xmath344 for a _ general position _ arrangement provided by theorem [ thm : conj ] for compatible @xmath85 and @xmath86 . this relationship extends to cohomology . by theorem [ thm : inducedrep ] , the representation @xmath345 in cohomology is induced by the representation @xmath346 . let @xmath347 denote the map in cohomology induced by @xmath348 . if @xmath84 is compatible with @xmath46 , then theorem [ thm : conj ] implies that , up to conjugacy , we have @xmath349 for each @xmath33 , @xmath153 . we will pursue the implications of this relationship for gauss - manin connections in section [ sec : gm ] . for the type @xmath36 of general position arrangements , the closure of the moduli space is @xmath350 . recall that points in @xmath58 are given by matrices @xmath174 of the form , and that @xmath351 denotes the determinant of the submatrix of @xmath174 with rows specified by @xmath21 element subsets @xmath352 of @xmath22=\{1,\dots , n+1\}$ ] . the divisor @xmath353 is given by @xmath354 , whose components , @xmath355 , are irreducible hypersurfaces indexed by @xmath60 . choose a basepoint @xmath96 , and for each @xmath21 element subset @xmath60 of @xmath22 $ ] , let @xmath356 be a generic point in @xmath59 . let @xmath61 be a meridian loop based at @xmath357 in @xmath38 about the point @xmath358 . note that @xmath359 and that @xmath61 is a ( possibly null - homotopic ) loop in @xmath43 for any combinatorial type @xmath18 . [ prop : h1yt ] for any combinatorial type @xmath18 , the homology group @xmath321 is generated by the classes @xmath360 \mid j \in \operatorname{ind}({{\mathcal t}})\}$ ] . given @xmath18 , let @xmath361 , so that @xmath362 . denote by @xmath363 the singular set of @xmath364 , and let @xmath365 and @xmath366 . consider the corresponding gysin sequence in homology with integer coefficients , as discussed in @xcite . this sequence is of the form @xmath367 where @xmath368 denotes the inclusion . since the complex codimension of @xmath363 is at least two in @xmath58 , the inclusion of @xmath369 in @xmath58 induces an isomorphism @xmath370 and thus @xmath371 . so for small @xmath372 , the above sequence is of the form @xmath373 the connected components of @xmath366 are in one - to - one correspondence with the irreducible components @xmath59 of @xmath364 . hence , @xmath374 is free abelian , of rank equal to the cardinality of @xmath19 . a basis for @xmath374 is given by the classes @xmath375 $ ] , @xmath376 , where @xmath358 is the generic point in @xmath59 chosen above . thus the classes @xmath377)$ ] generate @xmath378 . finally , let @xmath379 be a tubular neighborhood of @xmath380 , and let @xmath381 denote the projection from the boundary of @xmath379 to @xmath380 . then the map @xmath382 is given by @xmath383)=[p^{-1}({\mathbf{z}})]$ ] . since @xmath384 $ ] clearly coincides with @xmath63 $ ] up to sign for each @xmath60 , this completes the proof . [ cor : h1bg ] the homology group @xmath62 is generated by the classes @xmath63 $ ] , where @xmath60 ranges over all @xmath21 element subsets of @xmath22 $ ] . for a combinatorial type @xmath18 , recall that @xmath20 denotes the set of all @xmath21 element subsets @xmath352 of @xmath22 $ ] for which the determinant @xmath385 vanishes for @xmath95 . if @xmath18 is realizable , @xmath20 is the set of all @xmath60 for which @xmath386 is linearly dependent in the projective closure of an arrangement @xmath5 of type @xmath18 . impose a partial order on combinatorial types as follows : @xmath387 . note that the combinatorial type @xmath36 of general position arrangements is the maximal element with respect to this partial order . write @xmath388 if @xmath389 . in this case we define the relative dependence set @xmath390 . if @xmath388 , we say that @xmath18 _ covers _ @xmath65 and @xmath65 is a _ degeneration _ of @xmath18 if there is no combinatorial type @xmath391 with @xmath392 . [ lem : cover ] the moduli space @xmath66 has complex codimension one in the closure @xmath67 of the moduli space @xmath25 if and only if @xmath18 covers @xmath65 . if @xmath66 is codimension one in @xmath67 , then clearly @xmath18 covers @xmath65 . conversely , if @xmath18 covers @xmath65 , then @xmath393 is defined by the vanishing of @xmath394 for any @xmath395 . recall that @xmath212 denotes the natural inclusion . [ thm : lincomb ] let @xmath18 be a combinatorial type which covers the type @xmath65 . let @xmath396 be a point in @xmath66 , and @xmath315 a simple loop in @xmath25 about @xmath396 . then the homology class @xmath48 $ ] satisfies @xmath397 ) = \sum_{j \in \operatorname{dep}({{\mathcal t}}',{{\mathcal t } } ) } m_j \cdot [ \gamma_{\!j}],\ ] ] where @xmath70 is the order of vanishing of the restriction of @xmath394 to @xmath67 along @xmath66 . by proposition [ prop : h1yt ] , the classes @xmath360 \mid j \in \operatorname{ind}({{\mathcal t}})\}$ ] generate @xmath321 . so the image of @xmath48 \in h_1({{\sf b}}({{\mathcal t}}))$ ] in @xmath321 is of the form @xmath398 ) = \sum m_j \cdot [ \gamma_{\!j}]$ ] , where the sum is over all @xmath399 , and @xmath400 . to determine the coefficients @xmath70 , let @xmath401 be a local defining equation for @xmath66 in @xmath67 near @xmath174 . by hypothesis , the winding number of @xmath402 about @xmath85 is one : @xmath403 for each @xmath399 , let @xmath404 be a disk with @xmath405 , @xmath406 , and @xmath407 . suppose also that @xmath408 defines the divisor @xmath409 locally near @xmath174 , and let @xmath410 denote the inclusion of the closure of @xmath25 in @xmath58 . the coefficient @xmath70 is then given by the winding number of @xmath411 about @xmath85 : @xmath412 if @xmath413 , then @xmath414 . so for such @xmath60 , the restriction of @xmath415 to the disk @xmath404 is never zero . consequently , we have @xmath416 , and thus @xmath417 if @xmath418 . if @xmath395 , then the restriction , @xmath419 , of @xmath415 to @xmath67 defines a hypersurface which contains @xmath66 . so near @xmath174 , we have @xmath420 , where @xmath421 , and @xmath422 is the order of vanishing of @xmath411 along @xmath66 . hence for @xmath395 , the coefficient @xmath423 is as asserted . recall that we say @xmath47 is compatible with @xmath46 if the homology classes @xmath49 $ ] and @xmath48 $ ] satisfy @xmath50)=(j_{{\mathcal t}})_*([{{\gamma } } ] ) $ ] . [ cor : compatible ] if @xmath18 covers @xmath65 and @xmath90 is a simple loop in @xmath25 about @xmath424 , then any loop @xmath47 whose homology class satisfies the condition @xmath49=\sum_j m_j \cdot [ \gamma_{\!j}]$ ] given by is compatible with @xmath85 . if @xmath3 is a nontrivial rank one local system on the complement @xmath71 of a general position arrangement @xmath133 in @xmath2 , then the cohomology @xmath72 vanishes in all dimensions @xmath33 except possibly @xmath73 . see @xcite for a proof of this fact using the complex @xmath425 . for an arbitrary arrangement @xmath5 , a collection of weights @xmath9 and the corresponding local system @xmath3 are called _ nonresonant _ if the cohomology @xmath74 vanishes in all dimensions @xmath33 except possibly @xmath73 . we now recall a combinatorial condition due to schechtman , terao , and varchenko @xcite which insures nonresonance . an edge of an arrangement @xmath5 is a nonempty intersection of hyperplanes in @xmath5 . an edge is called _ dense _ if the subarrangement of hyperplanes containing it is irreducible : the hyperplanes can not be partitioned into nonempty sets so that , after a change of coordinates , hyperplanes in different sets are in different coordinates . this is a combinatorially determined property , see @xcite . for each edge @xmath369 , define @xmath426 . let @xmath427 be the projective closure of @xmath5 , the union of @xmath5 and the hyperplane at infinity in @xmath428 , see @xcite . set @xmath429 . [ thm : vanish ] let @xmath7 be the complement of an essential arrangement @xmath5 in @xmath2 of combinatorial type @xmath18 . if @xmath3 is a rank one local system on @xmath7 whose weights @xmath9 satisfy the condition + ( stv ) @xmath430 for every dense edge @xmath369 of @xmath23 , + then @xmath431 for @xmath432 and latexmath:[$\dim h^{{\ell } } ( { { \sf m}};{{\mathcal l } } ) = characteristic of @xmath7 . thus the ( stv)-condition implies nonresonance . in this paper , we call local systems @xmath3 whose weights @xmath9 satisfy the ( stv)-condition _ @xmath18nonresonant_. [ rem : gpnr ] the hyperplanes are among the dense edges for any arrangement , and are the only dense edges for a general position arrangement . consequently , if @xmath9 is @xmath18nonresonant , then @xmath9 is @xmath36nonresonant . any collection of weights @xmath9 determines an element @xmath434 in the orlik - solomon algebra of @xmath5 . since @xmath435 , multiplication by @xmath436 defines a differential on @xmath122 . the resulting complex @xmath437 may be identified with a subcomplex of the twisted de rham complex of @xmath7 with coefficients in @xmath3 . theorem [ thm : vanish ] above may be established by showing that , for @xmath18nonresonant weights @xmath9 , there is an isomorphism @xmath438 @xcite , and that @xmath439 for @xmath432 and @xmath440 , see @xcite . for @xmath18nonresonant weights , one can exhibit an explicit basis , the @xmath75*nbc * basis of @xcite , for the single nonvanishing cohomology group @xmath441 . recall that the hyperplanes of @xmath442 are ordered . a circuit is an inclusion - minimal dependent set of hyperplanes in @xmath5 , and a broken circuit is a set @xmath443 for which there exists @xmath444 so that @xmath445 is a circuit . a frame is a maximal independent set , and an * nbc * frame is a frame which contains no broken circuit . since @xmath446 is essential , every frame has cardinality @xmath106 . an * nbc * frame @xmath447 is a @xmath75*nbc * frame provided that for each @xmath372 , @xmath448 , there exists @xmath449 such that @xmath450 and @xmath451 is a frame . let @xmath75*nbc*@xmath452 be the set of all @xmath75*nbc * frames of @xmath5 . given @xmath447 in @xmath453 , define @xmath454 by @xmath455 , where @xmath456 and @xmath457 . denote the cohomology class of @xmath458 in @xmath459 by the same symbol . [ thm : bnbc ] let @xmath7 be the complement of an essential arrangement @xmath5 in @xmath2 of combinatorial type @xmath18 . if @xmath3 is a rank one local system on @xmath7 corresponding to @xmath18nonresonant weights @xmath9 , then the set @xmath460 is a basis for the only nontrivial local system cohomology group @xmath461 . [ exm : gpbnbc ] if @xmath133 is a general position arrangement of @xmath1 hyperplanes in @xmath2 , then @xmath462 . for @xmath36nonresonant weights @xmath9 , it follows that the @xmath75*nbc * basis for @xmath463 consists of monomials @xmath464 , where @xmath465 , @xmath466 , and @xmath467 . [ thm : nonresprojection ] for @xmath18nonresonant weights @xmath9 , the map @xmath120 induces a surjection @xmath468 . the map @xmath120 induces a surjection @xmath469 by proposition [ prop : gpsurjection ] this is a chain map @xmath470 , as is readily checked . so there is an induced map @xmath471 in cohomology . by remark [ rem : gpnr ] , if @xmath9 is @xmath18nonresonant , then @xmath9 is @xmath36nonresonant . for such weights , we have isomorphisms @xmath472 and @xmath461 . furthermore , the elements of the @xmath75*nbc * basis for @xmath459 are linear combinations of the monomials @xmath473 , which form the @xmath75*nbc * basis for @xmath474 as noted in example [ exm : gpbnbc ] . this fact , together with the surjectivity of the map @xmath141 completes the proof . this leads to the following consequence of theorem [ thm : conj ] . [ cor : hconj ] let @xmath9 be a collection of @xmath18nonresonant weights . if @xmath46 and @xmath47 are compatible , then the conjugacy class of the automorphism @xmath475 is determined by @xmath476 and the surjection @xmath477 . the vector bundle @xmath78 , with fiber @xmath79 at @xmath32 , supports a gauss - manin connection corresponding to the representation @xmath478 . we now study this connection using results of the previous section . in light of corollary [ cor : hconj ] , for an arrangement @xmath5 of combinatorial type @xmath18 , we focus on @xmath18nonresonant weights and the case @xmath73 . over a manifold , there is a well known equivalence between local systems and complex vector bundles equipped with flat connections , see @xcite . let @xmath479 be such a bundle , with connection @xmath480 . the latter is a @xmath125-linear map @xmath481 , where @xmath482 denotes the complex @xmath483-forms on @xmath369 with values in @xmath484 , which satisfies @xmath485 for a function @xmath402 and @xmath486 . the connection extends to a map @xmath487 for @xmath488 , and is flat if the curvature @xmath489 vanishes . call two connections @xmath480 and @xmath490 on @xmath484 isomorphic if @xmath490 is obtained from @xmath480 by a gauge transformation , @xmath491 for some @xmath492 . the aforementioned equivalence is given by @xmath493 , where @xmath494 is the local system , or locally constant sheaf , of horizontal sections @xmath495 . there is also a well known equivalence between local systems on @xmath369 and finite dimensional complex representations of the fundamental group of @xmath369 . note that isomorphic connections give rise to the same representation . under these equivalences , the local system on @xmath496 induced by the representation @xmath497 corresponds to a flat connection on the vector bundle @xmath78 , the gauss - manin connection . for @xmath18nonresonant weights , the @xmath75*nbc * basis of theorem [ thm : bnbc ] provides a basis for each fiber @xmath498 , independent of @xmath174 . thus the vector bundle @xmath82 is trivial , see @xcite . let @xmath46 , and let @xmath499 be a representative loop . pulling back the trivial bundle @xmath500 and the gauss - manin connection @xmath480 , we obtain a flat connection @xmath501 on the trivial vector bundle over the circle corresponding to the representation of @xmath502 given by @xmath503 . specifying the flat connection @xmath501 amounts to choosing a logarithm of @xmath504 . the connection @xmath501 is determined by a connection @xmath505-form @xmath506 , where the connection matrix @xmath81 corresponding to @xmath85 satisfies @xmath507 . if @xmath85 and @xmath508 are conjugate in @xmath509 , then the resulting connection matrices are conjugate , and the corresponding connections on the trivial vector bundle over the circle are isomorphic . in this sense , the connection matrix @xmath81 is determined by the homology class @xmath48 $ ] of @xmath85 . [ thm : gmcd ] let @xmath9 be a collection of @xmath18nonresonant weights . if @xmath315 and @xmath47 are compatible , then there is a commutative diagram @xmath510 thus a gauss - manin connection matrix @xmath81 is determined by a gauss - manin connection matrix @xmath511 and the surjection @xmath477 . choose a connection matrix @xmath511 so that @xmath512 . since @xmath513 is surjective by theorem [ thm : nonresprojection ] , the endomorphism @xmath514 induces an endomorphism @xmath515 , which we denote by @xmath81 . we assert that @xmath81 is a connection matrix for @xmath516 . for this , by the above discussion , it suffices to show that @xmath517 is conjugate to @xmath516 . by construction , we have @xmath518 . from this it follows that @xmath519 . by theorem [ thm : conj ] and corollary [ cor : hconj ] , up to conjugacy , we have @xmath520 . hence , @xmath517 is conjugate to @xmath516 . next we need formulas of aomoto and kita @xcite , see also @xcite , for the gauss - manin connection matrices in the case of general position arrangements . let @xmath325 , and let @xmath521 be the corresponding arrangement of type @xmath36 , a general position arrangement of @xmath1 hyperplanes in @xmath162 , with complement @xmath522 . a system of weights @xmath523 is @xmath36nonresonant if @xmath524 for each @xmath144 , @xmath525 . recall that @xmath526 . for such weights , the @xmath75**nbc * * basis for the cohomology group @xmath527 with coefficients in the corresponding local system consists of the monomials @xmath528 , where @xmath529 , @xmath530 , and @xmath531 , see example [ exm : gpbnbc ] . to state the results of aomoto and kita on the gauss - manin connection for general position arrangements , we require some notation . we use the formulation of @xcite . for @xmath532 , write @xmath533 , and let @xmath534 for each @xmath483 , @xmath535 . if @xmath536 $ ] , @xmath537 , and @xmath538 , let @xmath539 , where @xmath540 , @xmath541 , and @xmath542 . [ thm : gpgm ] for @xmath36nonresonant weights @xmath9 , the gauss - manin connection on the ( trivial ) vector bundle @xmath543 has connection @xmath505-form @xmath544 } d\log { { \sf d}}_j \otimes \omega_{{\mathcal g}}(\gamma_{\!j}),\ ] ] where the sum is over all subsets @xmath60 of @xmath22 $ ] of cardinality @xmath171 and @xmath545 denotes a @xmath505-form on @xmath546 with a simple logarithmic pole along the divisor @xmath59 . the connection matrices @xmath547 , acting on the @xmath75**nbc * * basis @xmath548 of @xmath527 , are given by the following formulas . @xmath549if @xmath550 , then @xmath551 \setminus i}{{\lambda } } _ j\right ) \eta_{i } & \text{if $ i= j'$,}\\ -\epsilon(i , j ' ) { { \lambda } } _ { i \setminus i\cap j'}\eta_{j ' } & \text{if $ |i\cap j'|=\ell-1$,}\\ 0\phantom{12345678901234567890123456789012345678 } & \text{otherwise . } \end{cases } \intertext{${\mathbf{3.}}$\quad if $ j \cap\{1,n+1\}=\{1\}$ , then } \omega_{{\mathcal g}}(\gamma_{\!j})(\eta_i)&= \begin{cases } \left(\sum_{j\in j}{{\lambda } } _ j\right ) \eta_{i } - \sum_{|i\cap i'|= \ell-1}\epsilon(i , i'){{\lambda } } _ { i\setminus i \cap i ' } \eta_{i ' } & \text{if $ i = j_1$,}\\ 0\phantom{12345678901234567890123456789012345678 } & \text{otherwise . } \end{cases } \intertext{${\mathbf{4.}}$\quad if $ j = j '' \cup \{1,n+1\}$ , where $ j''=\{j_2,\dots , j_\ell\}$ , $ 2 \le j_2 < \dots < j_\ell \le n$ , then } \omega_{{\mathcal g}}(\gamma_{\!j})(\eta_i)&= \begin{cases } -{{\lambda } } _ { i\setminus i\cap j } \eta_{i } + { { \lambda } } _ { i \setminus i\cap j}\sum_{i\cap i'=i\cap j}\epsilon(i , i ' ) \eta_{i ' } & \text{if $ j '' \subset i$,}\\ 0\phantom{12345678901234567890123456789012345678 } & \text{otherwise . } \end{cases}\end{aligned}\ ] ] we combine these formulas with theorem [ thm : lincomb ] , theorem [ thm : nonresprojection ] , and theorem [ thm : gmcd ] to obtain the main result of this paper . it is a combinatorial algorithm for calculating gauss - manin connection matrices for an arbitrary combinatorial type @xmath18 , and @xmath18nonresonant weights , in terms of those for the type @xmath36 of general position arrangements . let @xmath18 be a combinatorial type which covers the type @xmath65 . let @xmath396 be a point in @xmath66 , and @xmath90 a simple loop in @xmath25 about @xmath396 . we agree to write @xmath102 in place of @xmath92 in this situation . similarly , we write @xmath93 in place of @xmath94 . fix @xmath18nonresonant ( and hence @xmath36nonresonant ) weights @xmath9 . for @xmath95 and @xmath96 , by theorems [ thm : vanish ] and [ thm : bnbc ] , the corresponding local system cohomology groups of the fibers , @xmath97 and @xmath98 , depend only on the combinatorial types @xmath18 and @xmath36 and vanish for @xmath99 . so we write @xmath100 and @xmath101 . denote the @xmath75**nbc * * bases of these local system cohomology groups by @xmath75**nbc**@xmath552 and @xmath75**nbc**@xmath553 respectively . this notation makes it clear that a gauss - manin connection matrix depends only on the types involved and it is obtained by a combinatorial formula . [ thm : alg ] let @xmath18 be a combinatorial type which covers the type @xmath65 . let @xmath9 be a collection of @xmath18nonresonant weights , and let @xmath103 be the matrix of the surjection @xmath104 in the respective @xmath75**nbc * * bases . then a gauss - manin connection matrix @xmath102 is determined by the matrix equation @xmath105 illustrations of the algorithm provided by this result are presented in section [ sec : examples ] . [ rem : hiro ] for an arbitrary combinatorial type @xmath18 , and @xmath18nonresonant weights @xmath9 , terao @xcite shows that the gauss - manin connection on the bundle @xmath82 is determined by a connection @xmath505-form @xmath554 , where @xmath555 , each @xmath556 denotes a @xmath505-form on @xmath557 with a simple logarithmic pole along the irreducible component @xmath558 of the codimension one divisor @xmath559 , and the sum is over all such irreducible components . theorem [ thm : alg ] computes the gauss - manin connection matrices @xmath560 here . since our work is local , our results are independent of terao s global theorem . note that there are two floating inputs in theorem [ thm : alg ] : the ( stv)-condition , which ensures nonresonance , and the @xmath75**nbc * * basis , which makes the surjections explicit . should more relaxed conditions for nonresonance be discovered , and suitable cohomology bases be constructed , where the surjection may be made explicit , our methods will apply equally well in the new setting . for an arbitrary local system @xmath3 of rank greater than one on the complement of an arrangement of type @xmath18 , the results presented here need not apply , since such a local system is not in general induced by a local system on the complement of a boolean arrangement . however , if @xmath3 is an abelian local system of arbitrary rank , then @xmath3 is induced by a local system on the complement of a boolean arrangement . the methods of this paper may be applied in this generality in those instances where @xmath3 is nonresonant and a higher rank analogue of theorem [ thm : nonresprojection ] holds . in both of the above situations , analogues of the aomoto - kita formulas for @xmath36nonresonant weights , recorded in theorem [ thm : gpgm ] , would be required to express the gauss - manin connection matrices for type @xmath18 in terms of those for type @xmath36 explicitly . the moduli space @xmath25 of a combinatorial type @xmath18 of essential arrangements of @xmath1 hyperplanes in @xmath162 is of codimension one in @xmath546 if the cardinality of @xmath20 is one , @xmath561 , where @xmath562 , @xmath563 . for these types , the gauss - manin connection was determined by terao @xcite . we now sketch how these results may be recovered using the algorithm of theorem [ thm : alg ] . there are two cases to consider : @xmath564 and @xmath565 . by permuting hyperplanes , we may assume that @xmath566=\{1,\dots,\ell+1\}$ ] if @xmath564 , and that @xmath567=\{n-\ell+1,\dots , n+1\}$ ] if @xmath568 . if @xmath569 , we have @xmath570 , where @xmath571=\{2,\dots,\ell+1\}$ ] , while if @xmath565 , @xmath572 , where @xmath573=\{n-\ell+1,\dots , n\}$ ] . denote the projection @xmath574 of theorem [ thm : alg ] by @xmath575 if @xmath576 and by @xmath577 if @xmath568 . it is readily checked that these projections are given by @xmath578 p_f(\eta_i)=\sum_{j=2}^{\ell+1 } \sum_{q={\ell+2}}^n ( -1)^{j+\ell } \frac{{{\lambda } } _ j}{{{\lambda } } _ { [ \ell+1 ] } } \eta_{f_{j-1}^q } & \text{if $ i = f$,}\hfill & p_l(\eta_i)=0 \hfill & \text{if $ i = l$ , } \end{matrix}\ ] ] where @xmath579 for @xmath580 . as shown by terao @xcite , the moduli space @xmath25 is smooth , and the irreducible components of the divisor @xmath581 correspond to the combinatorial types @xmath65 which are covered by @xmath18 , and are characterized as follows : for each @xmath395 , the order of vanishing of the restriction of @xmath394 to @xmath67 along @xmath66 is equal to one , @xmath592 , see @xcite . by theorem [ thm : alg ] , a corresponding gauss - manin connection matrix @xmath593 satisfies @xmath594 where @xmath595 if @xmath576 and @xmath596 if @xmath565 . it is an exercise in linear algebra to recover terao s calculation of the connection matrix @xmath102 from @xcite . we give several small , explicit examples next , and leave this elementary , albeit lengthy , exercise in the general case to the interested reader . the arrangements @xmath609 in figure [ fig:123 ] represent types @xmath610 corresponding to irreducible components of the divisor @xmath581 . the gauss - manin connection matrices @xmath611 are determined by theorem [ thm : alg ] as follows . we conclude with an example of type @xmath287 for which @xmath625 has higher codimension in @xmath626 . let @xmath287 be the combinatorial type of the selberg arrangement @xmath5 in @xmath598 with defining polynomial @xmath627 depicted in figure [ fig : selberg ] . see @xcite for detailed studies of the gauss - manin connections arising in the context of selberg arrangements . the arrangement @xmath635 in figure [ fig : selberg ] represents one type @xmath636 covered by @xmath287 . here since lines @xmath638 , @xmath597 , and @xmath639 coincide in type @xmath636 , the order of vanishing of the restriction of @xmath640 to @xmath641 along @xmath642 is @xmath643 . for all other @xmath644 , we have @xmath592 . theorem [ thm : alg ] yields @xmath645 h. esnault , v. schechtman , v. viehweg , _ cohomology of local systems on the complement of hyperplanes _ , invent . * 109 * ( 1992 ) , 557561 . erratum , ibid . * 112 * ( 1993 ) , 447 . + http://www.ams.org/mathscinet-getitem?mr=93g:32051 [ ] v. schechtman , h. terao , a. varchenko , _ cohomology of local systems and the kac - kazhdan condition for singular vectors _ , j. pure appl . algebra * 100 * ( 1995 ) , 93102 . http://www.ams.org/mathscinet-getitem?mr=96j:32047 [ ] h. terao , _ moduli space of combinatorially equivalent arrangements of hyperplanes and logarithmic gauss - manin connections _ , topology appl . * 118 * ( 2002 ) , 255274 . 2002:08 ] a. varchenko , _ multidimensional hypergeometric functions and representation theory of lie algebras and quantum groups _ , 21 , world scientific , river edge , nj , 1995 . http://www.ams.org/mathscinet-getitem?mr=99i:32029 [ ]

we show that the magnon - exchange contribution to the single - particle and transport relaxation rates in ferromagnetic metals , which determine the thermal and electrical conductivity , respectively , at asymptotically low temperature does not obey a power law as previously thought , but rather shows an exponential temperature dependence . the reason is the splitting of the conduction band that inevitably results from a nonzero magnetization . at higher temperatures there is a sizable temperature window where the transport rate shows a temperature dependence , in accord with prior results . this window is separated from the asymptotic regime by a temperature scale that is estimated to range from tens of mk to tens of k for typical ferromagnets . we motivate and derive a very general effective theory for metallic magnets that we then use to derive these results . comparisons with existing experiments are discussed , and predictions for future experiments at low temperatures are made .

1402.3347

electronic relaxation rates contain important information about the excitations in a metallic system . the single - particle relaxation rate , @xmath1 , determines the lifetime of quasi - particles as well as the thermal conductivity @xmath2 ; the transport relaxation rate , @xmath3 , the electrical conductivity via the drude formula @xmath4 . here @xmath5 and @xmath6 are the conduction electron effective mass and number density , respectively , @xmath7 is the fermi velocity , and @xmath8 is the specific heat . there are various contributions to these relaxation rates , including those from the scattering of electrons by propagating , or particle - like , excitations . for instance , the coupling of longitudinal phonons to conduction electrons leads to the well - known bloch @xmath9-behavior of the electrical resistivity ; the corresponding effect in the single - particle relaxation rate is a @xmath10-law.@xcite in magnetically ordered phases , the coupling of the conduction electrons to any magnetic goldstone modes contributes to the relaxation rates . in isotropic heisenberg ferromagnets , the goldstone modes are the ferromagnons with a frequency - momentum relation @xmath11 . they have been found to contribute a @xmath0 term to the transport relaxation rate.@xcite in helimagnets,@xcite which have a helically modulated magnetic ground state , the corresponding goldstone mode ( the helimagnon ) has been shown to lead to a term in the electrical resistivity that is proportional to @xmath12 in the low - temperature limit.@xcite in antiferromagnets , the corresponding contribution is known to be proportional to @xmath10.@xcite these results all hold for three - dimensional systems , which is the only physical dimension in which long - range magnetic order exists . for later reference we note , however , that the various power laws quoted above are dimensionality dependent . for instance , in a generic dimension @xmath13 the contribution from ferromagnons to the resistivity is proportional to @xmath14 . in addition to the scattering by propagating excitations , there are contributions to the transport coefficients due to excitations with a continuous spectrum . the best known example is the one due to the coulomb interaction between the electrons . in simple metals it leads to a @xmath0 contribution to both the single - particle rate and the transport rate , i.e. , a lower power than the phonon contribution . however , since the relevant energy scale is the fermi energy @xmath15 or fermi temperature @xmath16 ( we use units such that @xmath17 ) , which is much larger than the debye temperature , this dominates the phonon contribution only at very low temperatures.@xcite in metals that display ferromagnetism the latter statement is not necessarily true , due to fermi surfaces that consist of multiple sheets , and the issue of both the temperature dependence and the prefactor of the coulomb contribution to the electrical resistivity is complicated . these are old questions@xcite that recently have been revisited in the context of quantum criticality and exotic metals.@xcite another example is the scattering of electrons in ferromagnets by both longitudinal magnetization fluctuations and the so - called stoner excitations in the transverse channel.@xcite the latter are dissipative , non - hydrodynamic transverse excitations in addition to the propagating spin waves . in a random - phase approximation , the contribution to the resistivity from these dissipative excitations , both longitudinal and transverse , was shown in ref . to result in a @xmath0 behavior with a prefactor that is inversely proportional to the magnetization . this is qualitatively the same behavior these authors found for the scattering by magnons , and it agrees roughly with the trend observed in fe , co , and ni.@xcite as we will see , this conclusion , as far as the magnons is concerned , is true only in a temperature window , but not at asymptotically low temperature . it should be stressed , however , that this similarity is somewhat accidental and approximation dependent even in the regime where it holds . for instance , the power law of the magnon contribution is dimensionality dependent , as mentioned above , while the contribution from the dissipative excitations is not . also , the prefactor of the former is essentially determined by the dispersion relation of the magnons , which is governed by very general principles , whereas the latter is dependent on many non - universal details . nevertheless , the fact that various contributions of very different nature to the relaxation rates show a @xmath0 temperature dependence makes the interpretation of the experimentally observed @xmath0 behavior of the electrical resistivity in many ferromagnetic materials difficult.@xcite at the same time , the electrical resistivity is a basic physical property that is very useful , for instance , for tracking and identifying magnetic phase transitions,@xcite and establishing its behavior in the ferromagnetic phase as a benchmark is important . in this paper we focus on the magnon contribution to the relaxation rates in ferromagnets and show that for this process the established result is qualitatively incorrect at asymptotically low temperatures ; instead of a @xmath0 temperature dependence , the magnon contributions to both the electrical resistivity and the thermal resistivity display an exponential behavior . a problem with the established ferromagnetic result was first noted in ref . , which showed that the results for the helimagnetic and ferromagnetic cases are not mutually consistent : if one considers the ferromagnetic limit of the helimagnetic ground state , by letting the wavelength of the helix go to infinity , one finds that the leading contribution to the relaxation rate , which would yield a power law , vanishes . what is left behind is an exponential behavior of the form@xcite 1/ _ ( t^2/)(-t_0/t ) , [ eq:1.1 ] where the temperature scale @xmath18 depends on the conduction band splitting or `` stoner gap '' @xmath19 or , equivalently , the magnetization,@xcite and on the fermi energy @xmath15 . this result is surprising , given that the relaxation rates due to magnetic goldstone modes in both helimagnets and antiferromagnets show a power - law behavior . the purpose of this paper is to discuss this problem , and to elaborate on the brief remarks that were given in appendix d of ref . . we will show that the asymptotic low - temperature behavior of both the transport relaxation rates due to magnons is indeed exponential of the form shown in eq . ( [ eq:1.1 ] ) , with @xmath20 with @xmath21 the spin - wave stiffness , which itself depends on @xmath19 , and @xmath22 and @xmath15 the fermi wave number and fermi energy , respectively . this result holds in an asymptotic regime defined by @xmath23 . however , in a sizable pre - asymptotic temperature window given by @xmath24 one recovers the @xmath0 behavior found previously . the reason for the exponential asymptotic result is the fact that , in a ferromagnet , the goldstone modes are purely transverse , and therefore couple only quasi - particles in different stoner bands . the effective electron - electron - interaction due to ferromagnon exchange therefore describes purely inter - stoner - band scattering , which leads to an activated process . in contrast , in helimagnets and antiferromagnets there is an intra - stoner - band coupling which leads to a power law . this vanishes as the characteristic wave number of the magnetic order goes to zero in the ferromagnetic limit . these results are valid for all metallic ferromagnets , whether or not the magnetism is caused by the conduction electrons themselves or by localized electrons in a different band . we will refer to such systems as `` itinerant ferromagnets '' and `` localized - moment ferromagnetes '' , respectively . in the main body of the paper we will consider a very general model that does not depend on which of these two cases is realized , and that uses only very general properties of ferromagnets that follow from symmetry arguments . a more specific stoner - type model for the case of itinerant ferromagnets is considered in an appendix . this paper is organized as follows . in sec . [ sec : ii ] we derive an effective action that describes an effective electron - electron interaction due to the exchange of ferromagnons . the effective action is valid for calculating relaxation rates to first order in the magnon propagator , and it holds for both itinerant and localized - moment ferromagnets . in sec . [ sec : iii ] we use this model to calculate the single - particle relaxation time , and in sec . [ sec : iv ] we calculate the transport relaxation time , and hence the electrical conductivity , by evaluating the pertinent kubo formula in an approximation that is equivalent to the boltzmann equation . in sec . [ sec : v ] we discuss our results . in appendix a we recall the stoner - moriya mean - field treatment of itinerant ferromagnets . in appendix b we recall the cases of electron - electron and electron - phonon scattering in non - magnetic metals , and cast them in a language that illustrates why our general method works even in the case of itinerant ferromagnets . in this section we derive and motivate an effective action that is suitable for calculating the effects of long - range ferromagnetic order , and the associated goldstone modes , on the electronic relaxation rates in a metallic ferromagnet . let @xmath25 $ ] be an action for conduction electrons in terms of fermionic spinor fields @xmath26 and @xmath27 that depend on a spin projection index @xmath28 . the electronic spin density is given by _ ( x ) = _ , |_(x)__(x ) . [ eq:2.1 ] here @xmath29 denotes the pauli matrices , and @xmath30 comprises the real - space position @xmath31 and the imaginary - time variable @xmath32 . now assume that the conduction electrons are subject to a magnetization @xmath33 of unspecified origin . the magnetization will act as an effective magnetic field that couples to the conduction electrons via a zeeman term . the action then reads s[| , ] = s_0[| , ] + dx m(x)_(x ) , [ eq:2.1 ] with @xmath34 a coupling constant that dimensionally is an energy times a volume , or an inverse density of states . in a ferromagnetic state , the magnetization has a nonzero average value that we assume to be in the 3-direction , @xmath35 . in a mean - field approximation that replaces @xmath36 by its average value the action then takes the form [ eqs:2.2 ] s_[| , ] = s_0[| , ] + dx n_,3(x ) , [ eq:2.2a ] where @xmath37 is directly proportional to the average magnetization . here we have chosen the sign of the action such that the partition function is given by z _ = d[| , ] e^s_[| , ] . [ eq:2.2b ] @xmath19 splits the conduction band into two sub - bands , one for each spin projection . we will refer to @xmath19 as the stoner gap,@xcite but we emphasize that the physical situation we are considering is much more general than the one considered in the stoner model.@xcite in particular , we do not necessarily assume that the conduction electrons themselves are the source of the magnetization . now consider fluctuations @xmath38 of the magnetization . the action , eq . ( [ eq:2.1 ] ) , then reads [ eqs:2.4 ] s[| , ] = s_[| , ] + _ dx ( x)_(x ) . [ eq:2.4a ] in addition we need an action that governs @xmath38 . if the latter is to describe the fluctuations of the physical magnetization , then this must be s _ [ ] = dxdy m_,i(x)^-1_ij(x , y)m_,j(y ) , [ eq:2.4b ] where @xmath39 is the physical magnetic susceptibility . in a ferromagnetic phase , the transverse ( @xmath40 with our choice for the magnetization direction ) components of @xmath41 contain the ferromagnons , which are the goldstone modes associated with the ferromagnetic order . the transverse part of @xmath41 is thus singular in the limit of small frequencies and wave numbers . adding eqs . ( [ eq:2.4a ] ) and ( [ eq:2.4b ] ) , and integrating out @xmath38 , we obtain a purely electronic effective action [ eqs:2.5 ] s_[| , ] = s_[| , ] + s_[| , ] , [ eq:2.5a ] with s_[| , ] = dx dy n_,i(x)^_ij(x , y)n_,j(y ) . [ eq:2.5b ] if we use only the singular , transverse , part of @xmath41 , then @xmath42 describes an effective electron - electron interaction mediated by an exchange of magnons.@xcite in order to make the effective action given by eqs . ( [ eqs:2.5 ] ) suitable for explicit calculations , we now specify @xmath43 and @xmath41 . the former in principle describes interacting electrons in a conduction band . however , the electron - electron interaction is not of any qualitative importance for our purposes , and we therefore take @xmath43 to describe noninteracting electrons with an energy - momentum relation @xmath44 . we denote the chemical potential by @xmath45 , and define @xmath46 . @xmath47 then reads [ eqs:2.6 ] s_[| , ] = _ k _ a that the magnetization splits the conduction band into two stoner bands whose fermi surfaces ( fs ) are defined by _ ( p)_p _ = 0 , [ eq:2.7 ] and we denote the density of states at the @xmath48-fermi surface and the corresponding fermi wave number by @xmath49 and @xmath50 , respectively . in the case of a parabolic band we have ^ & = & , + ^ & = & ^/2 ^ 2 . [ eq:2.7 ] the green functions for the two stoner bands are g_,(p ) = 1/(i_n - _ ( p))[eq:2.8 ] with @xmath51 ( @xmath52 integer ) a fermionic matsubara frequency . the structure of the transverse magnetic susceptibility at small frequencies and wave numbers in an isotropic ferromagnet is entirely determined by symmetry arguments.@xcite the goldstone modes of the spontaneously broken rotational symmetry in spin space are ferromagnons with a resonance frequency _ 0(k ) = d()k^2 , [ eq:2.9 ] the spin - stiffness coefficient @xmath21 vanishes as @xmath53 . it has the dimensions of a diffusion coefficient , and is given by a magnetic energy scale divided by a microscopic wave number scale squared , with the latter on the order of the fermi wave number . in the stoner - moriya mean - field theory@xcite of itinerant ferromagnets the former is given by @xmath19 , and for nearly free electrons one obtains [ eqs:2.9 ] d ( ) = /6 ^ 2 ( ) . [ eq:2.9a ] in a heisenberg spin model with exchange energy @xmath54 and lattice constant @xmath55 the corresponding result is@xcite d = ja^2 ( ) . [ eq:2.9b ] if one takes into account mode - mode coupling effects that are not included in the mean - field theory one finds that @xmath56 is a nonanalytic function of @xmath19.@xcite the transverse magnetic susceptibility can be expressed in terms of simple poles that describe circularly polarized ferromagnons , viz . , _ ( k , i ) = . [ eq:2.10 ] the coefficient @xmath57 is dimensionally an inverse volume . it vanishes as @xmath53 ; in the stoner - moriya mean - field theory it is given by ( see appendix [ app : a ] ) [ eqs:2.10 ] k ( ) = 4 ( ) . [ eq:2.10a ] in a heisenberg spin model , one has@xcite k = 2 m [ eq:2.10b ] the transverse susceptibility tensor takes the form [ eqs:2.11 ] _ t(k ) = ( cc _ + ( k ) + _ -(k ) & i[_+(k ) - _ -(k ) ] -i[_+(k ) - _ -(k ) ] & _ + ( k ) + _ -(k ) ) [ eq:2.11a ] where @xmath58 . explicitly one has for small @xmath59 and @xmath60 _ t(k ) = ( cc d()k^2 & -i(i ) i(i ) & d()k^2 ) . [ eq:2.11b ] in appendix [ app : a ] we show how this structure emerges in an explicit model calculation . the magnon exchange interaction , eq . ( [ eq:2.5b ] ) , can now be written [ eqs:2.12 ] s_[| , ] = _ , _ k n_(k)v_(k ) n_(-k ) . [ eq:2.12a ] here @xmath61 , and the effective potential is given by _ (k ) = v_(k ) + v_(-k ) [ eq:2.12b ] with v_(k ) = ( 1 - _ )^2_(k ) , [ eq:2.12c ] this effective interaction is shown diagrammatically in fig . [ fig:1 ] . -0 mm .,title="fig:",width=226 ] notice that the exchange of magnons couples only electrons with opposite spin projections , i.e. , it leads to inter - stoner - band scattering only . this is in contrast to the case of helimagnets , where there is an intra - stoner - band contribution whose prefactor is proportional to the square of the helical pitch wave number.@xcite we add a few remarks concerning the validity of this effective action . we have assumed that the conduction electrons are subject to a magnetization and magnetic fluctuations of unspecified origin whose dynamics are governed by the physical magnetic susceptibility . integrating out these fluctuations leads to an effective action that is purely electronic . since the feedback of the conduction electrons on the magnetic susceptibility has already been built into the effective action , the latter must not be used in ways that constitute , directly or indirectly , a renormalization of the susceptibility ; doing so would constitute double counting . however , it is safe to use the effective action for perturbative calculations of any observable to first order in the effective potential given by @xmath62 , and we will use it to calculate the quasiparticle and transport lifetimes to that order . we also note that the validity of this procedure is more obvious in cases where the magnetization is due to localized electrons in a band different from the conduction band than in the case of itinerant magnets . however , the coupling of the spin density to the magnetization fluctuations produced by the other electrons is still the same if all electrons are in the same band , and with the above caveats the effective action is still valid in that case . to illustrate this point we consider the ordinary fermi - liquid contribution to the electronic relaxation rate , as well as the one due to phonons , in appendix [ app : b ] , where we demonstrate that a reasoning for density fluctuations that is analogous to the one given above for magnetization fluctuations leads to the standard results for the relaxation rate in these cases . before we use the effective action to calculate the single - particle and transport relaxation rates , let us discuss the relevant energy scales and their relation to experimentally observable quantities . here we do so for the simple case of one conduction band ; in sec . [ sec : v ] we will discuss the more complicated , and more realistic , situation that arises from the presence of several bands . the most obvious fundamental magnetic energy scale is the stoner gap @xmath19 , or the closely related exchange splitting @xmath63.@xcite it can be measured by photoemission , and also obtained by band structure calculations . the smallest wave number that can be transferred by means of magnon exchange is @xmath64 . for a parabolic band this corresponds to @xmath65 , but the above expression is more general . the smallest energy that can be transferred by magnon exchange is thus t_0 = dk_0 ^ 2 d^2 ( e_/)^2 . [ eq:2.16 ] the largest momentum transfer is given by @xmath66 , and we thus have another energy scale , t_1 = 4d^2 . [ eq:2.17 ] @xmath67 one expects to be close to the exchange splitting ; within stoner theory one has @xmath68 . finally , the microscopic energy scale is given by the fermi energy @xmath15 , and and we have a hierarchy of energy scales , viz . , @xmath69 . in particular , the ratio @xmath70 is given in terms of the stoner gap in units of the microscopic energy , [ eqs:2.18 ] t_0/t_1 ( /)^2 . [ eq:2.18a ] alternatively , we can use @xmath71 as the microscopic energy scale and express the ratio @xmath70 in terms of the magnetization @xmath72 , t_0/t_1 ( m/)^2 . [ eq:2.18b ] within stoner theory this relation holds for @xmath73 , see eq . ( [ eq : a.7b ] ) , but as an order - of - magnitude estimate it is expected to hold much more generally . we finally mention that crystal - field effects break spin - rotational invariance , which gives the magnons a small gap and leads to yet another energy scale that affects the relaxation rates at very low temperatures . the magnitude of this effect is highly material dependent , and we neglect it for simplicity . we now calculate the single - particle inelastic relaxation rate due to the exchange of magnons . to linear order in the effective potential , eq . ( [ eq:2.12a ] ) yields two contributions to the electronic self energy @xmath74 , which are shown in fig . [ fig:2 ] . -0 mm for the @xmath48-spin green function.,title="fig:",width=226 ] analytically , we have _ ( p ) & = & _ k _ v_(k)g_,(p+k ) + & = & 2 ^ 2 _ k _ ( k)g_,-(p+k ) . [ eq:3.1 ] here we have defined the self energy such that the full green function @xmath75 is given by a dyson equation _ ^-1(p ) = g_,^-1(p ) - _ ( p ) . [ eq:3.2 ] now we consider the single - particle relaxation rate @xmath76 for a spin-@xmath48 quasiparticle , averaged over the fermi surface : _ ( ) = _ p ( _ ( p ) ) _ ( p , ) , [ eq:3.3 ] where @xmath77 is the spectrum of the self energy . using a spectral representation for the effective potential and performing the matsubara frequency sum in eq . ( [ eq:3.1 ] ) we find [ eqs:3.4 ] _ ( ) & = & ^- du [ n_b(u ) + n_f(u + ) ] _ & = & 2 ^ 2 ^ - _ -^ du [ n_b(u ) + n_f(u + ) ] |_(u ) , + [ eq:3.4a ] where @xmath78 and @xmath79 are the bose and fermi distribution functions , respectively . here we have defined _ (u ) & = & _ k , p ( _ ( k ) ) ( _ (p ) ) + & & _ (k - p , u ) [ eq:3.4b ] and analogously _ ( u ) = _ k , p ( _ ( k ) ) ( _ (p))_(k - p , u ) , [ eq:3.4c ] with _ ( k , u ) = ( _ 0(k ) u ) [ eq:3.4d ] the spectra of the susceptibilities @xmath80 defined in eq . ( [ eq:2.10 ] ) . we note the symmetry relation ^+ _ + ( ) = ^- _ -(- ) , [ eq:3.4e ] which follows from the symmetry properties of @xmath81 . notice that the wave vectors @xmath59 and @xmath82 in eq . ( [ eq:3.4a ] ) are pinned to different fermi surfaces as a result of the pure inter - stoner - band scattering mentioned after eq . ( [ eq:2.12b ] ) . the spectrum @xmath81 will therefore be nonzero only for frequencies t_0 u t_1 , [ eq:3.5 ] with @xmath18 and @xmath67 given by eqs . ( [ eq:2.16 ] ) and ( [ eq:2.17 ] ) . on the energy shell , @xmath83 , we obtain for the relaxation rate @xmath1 on the @xmath48-fermi surface 1/2 _ & & _ ( = 0 ) = t _ t_0/t^t_1/t + & & -40pt = te^-t_0/t & t(t / t_0 ) & ( t_1/t_0)t & . + [ eq:3.6 ] for the thermal resistivity @xmath84 this implies _ = te^-t_0/t & t(t / t_0 ) & ( t_1/t_0)t & . [ eq:3.7 ] in eq . ( [ eq:3.7 ] ) the prefactor is valid in the limit @xmath85 ; more generally there are corrections of @xmath86 . the second line in eqs . ( [ eq:3.6 ] ) and ( [ eq:3.7 ] ) is valid to leading logarithmic accuracy only . we see that at asymptotically low temperatures the relaxation rate is exponentially small , and that in the pre - asymptotic temperature window @xmath87 there is a logarithmic correction to the linear behavior . we will further discuss these results in sec . [ sec : v ] . we now turn to the transport relaxation rate , which determines the electrical resistivity . the latter is the inverse of the electrical conductivity , which is given by the kubo formula@xcite [ eqs:4.1 ] _ ij(i)=[_ij(i)-_ij(i=0 ) ] , [ eq:4.1a ] where the tensor _ ij(i ) & = & -e^2t_n_1,n_2 _ k , p v_i(k)v_j(p ) + & & -20pt _ n_1,(k)_n_1+n,(k)|_n_2,(p)_n_2- n,(p ) . + [ eq:4.1b ] is the current - current susceptibility or polarization function . here @xmath88 , and the average is to be taken with the effective action , eq . ( [ eq:2.5a ] ) . the four - fermion correlation function in eq . ( [ eq:4.1b ] ) is conveniently expressed in terms of the single - particle green function _ ( p ) = 1/(g_,(p ) - _ ( p ) ) [ eq:4.2 ] and a vector vertex function @xmath89 with components @xmath90 : _ ij(i ) & = & -ie^2 t _ i _ p , _ ( p , i ) g_(p , i - i ) + & & ^j_(p;i , i - i ) . [ eq:4.3 ] here we have assumed a quadratic dependence of @xmath44 on @xmath59 for simplicity . it is important to calculate the vertex function @xmath91 and the self energy @xmath74 in mutually consistent approximations.@xcite we use the familiar procedure that consists of a self - consistent born approximation for the self energy , which to linear order in the potential @xmath92 is represented by eq . ( [ eq:3.1 ] ) , and a ladder approximation for the vertex function , _ ( p;i , i - i ) & = & i + _ k , i^ _ v_(k - p , i ) + & & -60pt _ (k , i+i^ ) g_(k , i- i+ i^ ) + & & -40pt _ (k;i+ i^,i- i+ i^ ) . [ eq:4.4 ] we mention that umklapp processes , which are not explicitly considered here , are necessary in order to obtain a nonzero transport relaxation rate . in fact , in a galilean invariant system the electrical resistivity vanishes due to momentum conservation and the contributions contained in our approximation are cancelled by terms not included in the ladder approximation . however , the above approximation is effectively valid in the presence of umklapp processes , as is the case for coulomb scattering.@xcite if we define a scalar vertex function @xmath93 by @xmath94 , then the bethe - salpeter equation for the latter becomes _ ( p;i , i - i ) & = & 1 + _ k , i^ _ v_(p - k , i^ ) + & & -80pt g_(k , i+ i^ ) g_(k , i- i+ i^ ) + & & -40pt _ (k;i - i^,i - i - i^ ) . [ eq:4.5 ] the polarization and conductivity tensors are diagonal , @xmath95 , and the sum over matsubara frequencies in eq . ( [ eq:4.3 ] ) can be transformed into an integral along the real axis . in the limit of low temperature , the imaginary part of the self energy , which yields the relaxation rate , goes to zero as we have seen in the preceding subsection . the real part just renormalizes the fermi energy . the relevant limit is thus the one of a vanishing self energy , and in this limit the leading contributions to the integral come from terms where the frequency arguments of the two green functions lie on different sides of the real axis . in the static limit , the kubo formula for the conductivity @xmath96 , thus becomes & = & _ -^ _ p p^2 + & & _ _ ( p,+i0)^2 _ ( p;+i0,-i0 ) .[eq:4.6 ] the pole of the green function ensures that the dominant contribution from the momentum integral comes from the momenta that obey @xmath97 . furthermore , since @xmath98 scales as @xmath99 , for the leading @xmath99-dependence we can neglect all @xmath98-dependencies that do not occur in the form @xmath100 . equation ( [ eq:4.6 ] ) then reduces to [ eqs:4.7 ] = _ -^ _ n _ . [ eq:4.7a ] here @xmath101 is the density of the @xmath48-spin electrons , @xmath102 is the single - particle rate defined by eq . ( [ eq:3.3 ] ) , and _ ( ) = _ p ( _ ( p ) ) _ ( p;+i0,-i0 ) . [ eq:4.7b ] using analogous arguments we find , from eq . ( [ eq:4.5 ] ) , that @xmath103 obeys an integral equation [ eqs:4.8 ] _ ( ) & = & 1 + _ ^ du w_(u ) + & & , [ eq:4.8a ] where w_(u ) & = & _ p , k _ ( _ (k))(_(p ) ) + & & _ ( k - p , u ) k / p^2 [ eq:4.8b ] with @xmath104 the spectrum of the effective potential defined in eq . ( [ eq:2.12b ] ) . now we exploit the fact that @xmath59 and @xmath82 are pinned to the respective fermi surfaces , and use the resulting identity @xmath105\ ] ] to write w_(u ) = ( /^)^2[eq:4.9 ] with @xmath106 from eq . ( [ eq:3.4b ] ) , and ^(2)_(u ) & = & _ k , p ( _ ( k ) ) ( _ (p ) ) + & & _ (k - p , u ) . [ eq:4.10 ] note that the magnon frequency @xmath107 in eq . ( [ eq:4.10 ] ) is equal to @xmath108 on account of the spectrum , and therefore @xmath109 has an extra factor of @xmath110 compared to @xmath111 . @xmath106 determines the single - particle rate @xmath76 via eq . ( [ eq:3.4a ] ) , and we define analogously ^(2 ) _ ( ) = ^- du [ n_b(u ) + n_f(u + ) ] _ |v^(2)_(u ) [ eq:4.11 ] the integral equation for the vertex function @xmath112 now reads _ ( ) & = & 1 + ( ) ^2 du _ ^ + & & , [ eq:4.12 ] for the case of a spin - independent potential , eq . ( [ eq:4.12 ] ) reduces to the integral equation familiar from the electron - phonon scattering problem ; only the @xmath110-dependence of the kernel is different . this integral equation is usually solved in the seemingly uncontrolled approximation that replaces @xmath113 on the right - hand side by @xmath114 , turning the integral equation into an algebraic equation . in ref . two of the present authors have shown that the integral equation can be solved asymptotically exactly , that the exact solution yields a result for the conductivity that coincides with the lowest - order variational solution of the boltzmann equation , and that the simple approximation yields the same low - temperature dependence ( albeit with a different prefactor ) as the exact solution . the proof of these statements can be generalized to the current case of a two - by - two matrix equation . for the purpose of deriving the low - temperature behavior , we thus can employ the approximation , which turns eq . ( [ eq:4.12 ] ) into two coupled algebraic equations for @xmath115 . since the prefactor of the temperature dependence of the conductivity is approximation - dependent anyway , we can put @xmath83 and use the temperature - dependent rates @xmath116 and vertices @xmath117 in the kubo formula , eq . ( [ eq:4.7a ] ) . @xmath118 then obeys _ + & = & 1 + ( /^+)^2 _ -/_- , + _ - & = & 1 + ( /^-)^2 _ + /_+ . + [ eq:4.13 ] @xmath102 is given by eq . ( [ eq:3.6 ] ) , and @xmath119 is given by an analogous integral with an additional power of the frequency in the integrand . ^(2 ) & = & _ t_0/t^t_1/t dx + & & -40pt = ( 1 + ) e^-t_0/t & t / t_1 & 1 & + [ eq:4.14 ] comparing with eq . ( [ eq:3.6 ] ) we see that for asymptotically small @xmath99 , @xmath120 is proportional to @xmath76 with a small factor of proportionality @xmath121 , whereas for @xmath87 it carries an additional factor of temperature . we now solve the equations ( [ eq:4.13 ] ) . neglecting @xmath73 wherever it is not of qualitative importance , we find = . [ eq:4.15 ] equation ( [ eq:3.4e ] ) allows us to express @xmath122 entirely in terms of @xmath102 and @xmath123 , neglecting all prefactors that just give small corrections of @xmath124 to factors of @xmath125 we finally obtain a transport relaxation time [ eqs:4.16 ] _ = , [ eq:4.16a ] in terms of which the electrical conductivity is given by a drude formula = _ , [ eq:4.16b ] here @xmath126 and @xmath127 are given by eq . ( [ eq:3.6 ] ) and eq . ( [ eq:4.14 ] ) , respectively , with @xmath49 replaced by @xmath128 . note that our approximations have affected overall prefactors only , but not the relative prefactors of the four terms in the denominator in eq . ( [ eq:4.16a ] ) . comparing eq . ( [ eq:3.6 ] ) with ( [ eq:4.14 ] ) we see that , for @xmath129 , @xmath120 is proportional to @xmath76 : ^(2 ) = ( 1 + t / t_0 ) , [ eq:4.17 ] and an inspection reveals that the leading contributions among the three terms in the denominator in eq . ( [ eq:4.16a ] ) cancel , which leads to @xmath130 . at asymptotically low temperatures @xmath131 is therefore _ not _ proportional to @xmath132 , but rather carries an extra factor of @xmath133 . for the contribution to the electrical resistivity @xmath134 due to magnon exchange we finally obtain _ ( t ) = ( 4/t_1)t^2 e^-t_0/t & ( ^2/2t_1)t^2 & t & . [ eq:4.18 ] we see that in the preasymptotic temperature window @xmath87 we recover the @xmath0 behavior that was obtained in ref . , but for asymptotically low temperatures we obtain an exponentially small result that has the form of eq . ( [ eq:1.1 ] ) . we will discuss this result in the next section . to summarize our results , we have presented a very general theory of electron relaxation due to the exchange of magnons in metallic ferromagnets . the theory is valid for both itinerant ferromagnets , where the magnetization is due to the conduction electrons themselves , and for localized - moment ferromagnets , where the magnetization is due to localized spins in a different band . we have found that at asymptotically low temperatures , below a temperature scale @xmath18 , both the single - particle relaxation rate and the transport relaxation rate are exponentially small . this behavior carries over to the magnon - exchange contributions to the thermal and electrical resistivities , which are determined by these respective rates . the exponential temperature dependence is a direct consequence of the split conduction band in a metallic ferromagnet . in a preasymptotic temperature regime @xmath87 , with @xmath67 close to the exchange splitting , we recover the @xmath0 behavior of the transport rate that was found in ref . . the single - particle rate is proportional to @xmath99 in this regime . for @xmath135 the two rates both show a linear temperature dependence . we start our discussion of these results by recalling the physical reason for the exponential dependence at low temperatures . -0 mm is the stoner gap , and @xmath136 is the exchange splitting . @xmath137 and @xmath138 are the smallest and largest transferrable wave numbers , respectively . see the text for more explanation.,title="fig:",width=321 ] figure [ fig:3 ] schematically shows the split conduction band ( a ) , and the densities of states for the up ( + ) and down ( - ) spin electrons ( b ) , for the case of a spherical fermi surface . since the magnons couple only electrons with opposite spin , the smallest transferrable wave number is @xmath139 . given the magnon dispersion relation , @xmath140 , this translates into a smallest transferrable energy @xmath141 , and since the magnon stiffness coefficient @xmath21 is itself roughly proportional to @xmath136 , we have @xmath142 . for temperatures @xmath23 the relaxation rates will thus show activated behavior with an activation energy @xmath18 . the exponential behavior is multiplied by a power law that can not be captured by elementary arguments . the largest momentum transfer is given by @xmath143 , and the corresponding largest energy transfer is @xmath144 . @xmath67 is the fundamental magnetic energy scale , analogous to the debye temperature @xmath145 in the case of electron - phonon coupling . @xmath18 has no analog in the electron - phonon problem . for @xmath146 the transport - relaxation rate is small compared to the single - particle rate by a factor of @xmath147 . this is analogous to the electron - phonon case , where the corresponding factor is @xmath148 . the difference between our results and those of ueda and moriya , ref . , can be traced to the fact that these authors neglected the exchange splitting in the final stages of their calculation . as a result , they obtained a @xmath0 behavior of the transport relaxation rate at low temperatures , which in fact is valid only for temperatures larger than @xmath18 . note that this discrepancy pertains to the magnon or spin - wave contribution to the electrical resistivity only . the contributions from dissipative excitations , which we have not discussed , have been found to be unaffected by the exchange splitting and proportional to @xmath0 even at asymptotically low temperatures.@xcite for the power - law behavior at @xmath149 the quadratic spectrum of the magnons is important , and also the coupling of the electrons to the magnetic fluctuations . comparing with the case of helical magnets,@xcite we notice one important difference with respect to the latter . in either case the goldstone mode is a phase fluctuation , but in the ferromagnon case the electron spin density couples directly to the phase , whereas in the helimagnon case the coupling is to the gradient of the phase . this is because in the helimagnon case the dominant low-@xmath99 contribution to the scattering rates comes from intra - stoner - band scattering . within a given band , the phase itself has no physical meaning , and the coupling therefore involves a gradient . in the ferromagnetic case , on the other hand , we deal with inter - stoner - band scattering . the coupling therefore effectively is to the difference of two phases , which does have a physical meaning . we note in passing that this latter notion also manifests itself in a spin josephson effect , see ref . . we now turn to estimates of the values of @xmath67 and @xmath18 . to get an idea about the order of magnitude of these temperature scales , let us first consider the fictitious case of simple ( i.e. , single - conduction - band ) metals with magnetic properties as in the classic `` high - temperature '' ferromagnets nickel , cobalt , and iron . the values of the exchange splitting in these materials , as determined by photoemission , are @xmath150 ev , @xmath151 ev , and @xmath152 ev , respectively.@xcite values for the spin - stiffness coefficient @xmath21 in mev@xmath153 obtained from neutron scattering are @xmath154 for ni , @xmath155 for co , and @xmath156 for fe.@xcite with a generic value @xmath157 for the fermi wave number , and @xmath158 k for the fermi energy , eqs . ( [ eq:2.16 ] ) and ( [ eq:2.17 ] ) yield @xmath159 k for these materials , and @xmath160 mk for ni , @xmath161 k for co , and @xmath162 k for fe . estimates of the ratio @xmath70 using the relation ( [ eq:2.18b ] ) yields similar results . notice that the prefactor @xmath163 in eq . ( [ eq:4.18 ] ) is of order unity , so the prefactor of the @xmath0 behavior of the resistivity is larger than the fermi - liquid @xmath0 contribution ( see eq . ( [ eq : b.7 ] ) ) by roughly a factor of @xmath164 in a single - band model . also of interest are weak ferromagnets , such as mnsi,@xcite or ni@xmath165al , where @xmath166 mev@xmath153 ( mnsi)@xcite and @xmath167 mev@xmath153 ( ni@xmath165al),@xcite respectively . the magnetic moments , @xmath168 per formula unit for mnsi,@xcite and @xmath169 for ni@xmath165al,@xcite are about two thirds and one third , respectively , of that of ni . given the observed near - linear correlation between the magnetic moment and the exchange splitting,@xcite this suggests @xmath170 ev for mnsi,@xcite , and @xmath171 ev for ni@xmath165al . if we use again @xmath157 and @xmath158 k , this yields @xmath172 k and @xmath173 mk for mnsi , and @xmath174 k and @xmath175 mk for ni@xmath165al . in reality , all of these materials are transition metals , or compounds containing transition metals , with a complicated band structure and fermi surfaces that consist of multiple sheets . one consequence of this is that the electron - electron scattering contribution to the electrical resistivity is likely much larger than a single - band model would imply , and it has been suggested that it makes the largest contribution to the observed @xmath0 behavior at low temperatures.@xcite the reason is that different band edges have different distances from the common chemical potential , which in effect leads to different fermi temperatures . depending on whether or not the various scattering processes flip the electron spin , and whether or not they couple different sheets of the fermi surface , the relaxation rates or the relaxation times may be additive , which leads a complicated structure of the overall resistivity . in addition , there are the contributions from the dissipative spin excitations , which also are proportional to @xmath0.@xcite as a result , the low - temperature transport rate in fe , ni , and co is about 100 times larger than one would expect from the coulomb contribution in a single - band model with a single fermi temperature of about @xmath176k.@xcite @xmath67 , on the other hand , is largely unaffected by a complicated band structure : it is given by @xmath21 times the largest possible momentum transfer squared , see sec . [ subsec : ii.c ] , and in a good metal the latter is on the order of @xmath177 , with @xmath55 the lattice constant , which is close to the value of @xmath178 for a single spherical fermi surface that yields the same electron density . the estimates of the temperature scale @xmath67 , which is the magnetic analog of the debye temperature for phonons , given above are therefore model independent and depend only on the experimentally measured spin stiffness coefficient . as a result , we expect the magnon contribution to the electrical resistivity in fe , ni , and co at temperatures @xmath179 to be about an order of magnitude _ less _ than the combined contribution from the coulomb interaction and the dissipative magnetic excitations . in mnsi and ni@xmath165al @xmath67 is much lower and the magnon scattering is accordingly stronger . however , the observed prefactors of the @xmath0 term in the resistivity of mnsi and ni@xmath165al are orders of magnitude larger than even the ones in fe , ni , and co , and the same is true for the weak ferromagnet zrzn@xmath180.@xcite the prefactor @xmath163 in eq . ( [ eq:4.18 ] ) is expected to be of @xmath125 not just in model calculations , but also in real materials , since both @xmath181 and @xmath67 correlated roughly with the magnetization . given the above discussion of the relatively narrow range of plausible values of @xmath67 , we conclude that the experimental value of the prefactor of the @xmath0 term in the electrical resistivity of weak ferromagnets can not possibly be explained by electron - magnon scattering . we emphasize again , however , that these considerations do not take into account the scattering of electrons by dissipative magnetization fluctuations , which lead to a @xmath0 contribution to the resistivity even at low temperatures and whose prefactor is not as universal as that of the magnon - exchange contribution . a corresponding statement holds for the coulomb contribution . for @xmath18 the influence of the band structure is more complicated . consider the effective potential given by eqs . ( [ eq:3.4b ] - [ eq:3.4d ] ) . if the up - spin and down - spin electrons , respectively , belong to different bands with different effective masses , then there will be a lower cutoff for the frequency @xmath110 even in the limit of a vanishing stoner gap , @xmath85 . for magnon - exchange scattering between electrons in stoner - subbands of the same band , on the other hand , the structure of the calculations in secs . [ sec : iii ] and [ sec : iv ] is unchanged . we therefore expect different values of @xmath18 for the various scattering processes that involve electrons on different sheets of the fermi surface . the following picture now emerges . with decreasing temperature , contributions to the magnon - exchange part of the electronic scattering rate will sequentially freeze out as the temperature drops below a sequence of temperature scales @xmath18 . rough estimates for the lowest of these temperature scales have been given above ; estimating the higher ones requires a detailed analysis of the band structure . below this lowest @xmath18 the magnon contribution to both the transport rate and the single - particle rate will be exponentially small , leaving the coulomb contribution and the one from dissipative magnetization fluctuations as the most obvious candidates for a @xmath0 behavior . experimentally , this is expected to manifest itself in a distinct temperature dependence of the prefactor of the @xmath0 term in the electrical resistivity . it is desirable for the relevant temperature scales to be small enough that phonon contributions are negligible . in that respect , fe , ni , and co are not ideal . in mnsi , the helical nature of the magnetic phase is expected to manifest itself on the temperature scale given by @xmath18 . this leaves ni@xmath165al , or other true weak ferromagnets as the most promising candidates for observing this consequence of the exchange splitting in a metallic ferromagnet . we stress , however , that according to the above discussion the magnon contribution to the electrical resistivity in weak ferromagnets is likely dwarfed by other contributions . another possible effect of a complicated band structure is that there may be points or lines in reciprocal space where the two stoner band cross . this will weaken the exponential suppression of the relaxation rates , but the weakening will depend on the nature of the crossing . we finally mention that the interplay of quenched disorder with the scattering processes discussed above constitutes an interesting problem that is likely important for a quantitative understanding of real materials . for fairly strong disorder , @xmath182 with @xmath183 the elastic scattering time , the theory of ref . applies and it is easy to see that there is no exponential suppression of the magnon contribution to the relaxation rates at low temperature . a complete discussion of disorder effects constitutes a separate problem . we thank achim rosch for discussions . this work was supported by the nsf under grants number dmr-09 - 01952 and dmr-09 - 01907 . part of this work was performed at the aspen center for physics and supported by the nsf under grant number phys-1066293 . in this appendix we show how to recover the stoner - moriya results@xcite for itinerant ferromagnets within the present formalism . our starting point is a fermionic action s[| , ] = s_0[| , ] + dx n_(x)_(x ) , [ eq : a.1 ] with @xmath34 the spin - triplet interaction amplitude that is responsible for ferromagnetism . our notation is the same as in sec . [ sec : ii ] . a simple mean - field approximation that describes ferromagnetic order , and its coupling to the electron spin density , consists of replacing one of the spin density fields in eq . ( [ eq : a.1 ] ) by its expectation value according to _ ^2 2 _ _ - _ ^2 . [ eq : a.2 ] if we take the magnetic order to be in the 3-direction , @xmath184 , then this approximation amounts to replacing the action @xmath185 with an action @xmath47 that describes electrons with no spin - triplet interaction subject to a magnetic field of strength @xmath19 in the 3-direction : s_[| , ] = s_0[| , ] + dx n_,3(x ) . [ eq : a.3 ] the remaining question pertains to the action one should use to calculate @xmath186 ; this choice determines @xmath19 . the usual self - consistent mean - field requirement stipulates that this average be determined by @xmath47 itself : & = & _ n_,3(x ) _ d[| , ] n_,3(x ) e^s_[| , ] + & = & _ z _ . [ eq : a.4 ] for simplicity , we take @xmath43 to describe free electrons . that is , we neglect all electron - electron interactions that are not crucial for magnetism , and we assume a parabolic band ; a generalization to band electrons is straightforward . the green function corresponding to @xmath43 then is g_0(k , i_n ) = , [ eq : a.5 ] and the self - consistency condition , eq . ( [ eq : a.4 ] ) , takes the form 1 = -2 _ _ p . [ eq : a.6 ] we recognize this as the equation of state of stoner theory , with @xmath19 the stoner gap . the condition for a nonzero solution for @xmath19 is @xmath187 , and by performing the integral we find explicitly [ eqs : a.7 ] = 2 . [ eq : a.7a ] if we recall that the magnetization is given as @xmath188 ( see eqs . ( [ eq:2.1 ] , [ eq:2.2a ] ) ) , we can write this result as m = 2 . [ eq : a.7b ] the action @xmath47 contains information about the long - range order , but does not contain any ferromagnetic fluctuations . it will serve as a building block for the effective action , and we will refer to it as the `` reference ensemble '' . we now determine the spin susceptibility ^_,ij(x , y ) = n_,i(x)n_,j(y ) _ s _ , [ eq : a.8 ] associated with the reference ensemble . in terms of the reference - ensemble green function g_(k ) = _ 0 - _ 3 , [ eq : a.9 ] @xmath189 can be written _ , ij(x , y ) = - [ eq : a.10 ] where the trace is over the spin degrees of freedom . evaluating the trace , and performing a fourier transform , we find ^_,ij(k ) = ( ccc f_1(k ) & f_2(k ) & 0 + - f_2(k ) & f_1(k ) & 0 + 0 & 0 & f_3(k ) ) , [ eq : a.11 ] where [ eqs : a.12 ] f_1(k ) & = & -2_p , + [ eq : a.12a ] + f_2(k ) & = & -2i_p , + [ eq : a.12b ] + f_3(k ) & = & - 2 _ p . + [ eq : a.12c ] we note that [ eqs : a.13 ] f_1(k=0 ) & = & 1/ _ , [ eq : a.13a ] + f_2(k=0 ) & = & 0 , [ eq : a.13b ] with the first equality following from the equation of state , eq . ( [ eq : a.6 ] ) . the reference ensemble does not reflect the magnons that are the goldstone modes of the spontaneously broken symmetry in the ferromagnetic phase . to describe the magnons we need a theory of fluctuations that is consistent with the treatment of the static magnetization . quite generally , a gaussian approximation for the order - parameter fluctuations is consistent with a mean - field treatment of the order parameter itself.@xcite to determine the former we first note that the reference ensemble spin susceptibility @xmath189 corresponds to a gaussian fluctuation action [ eqs : a.14 ] _ , [ _ ] = dx dy n_,i(x)_,ij^-1(x , y)n_,j(y ) [ eq : a.14a ] that generates @xmath190 via _ , ij^(x , y ) = d [ _ ] n_,i(x)n_,j(y ) e^-a _ , [ _ ] . [ eq : a.14b ] to this we need to add the fluctuation contribution from the original spin - triplet interaction in eq . ( [ eq : a.1 ] ) . the gaussian fluctuation action then reads [ eqs : a.15 ] _ [ _ ] = dx dy n_,i(x)^-1_ij(x , y)n_,j(y ) [ eq : a.15a ] with the physical spin susceptibility @xmath191 given by ^-1_ij(x , y ) = _ , ij^-1(x , y ) - _ ij _ . [ eq : a.15b ] focusing on the transverse ( t ) channel ( @xmath192 ) , and performing a fourier transform , we have [ eqs : a.16 ] ^-1_(k , i_n ) = ( cc f_1(k , i_n)/n(k , i_n ) - _ & -f_2(k , i_n)/n(k , i_n ) + f_2(k , i_n)/n(k , i_n ) & f_1(k , i_n)/n(k , i_n ) - _ ) . [ eq : a.16a ] where n(k , i_n ) = ( f_1(k , i_n))^2 + ( f_2(k , i_n))^2 . [ eq : a.16b ] from eqs . ( [ eqs : a.13 ] ) we see that @xmath193 at zero frequency and wave number has two zero eigenvalues . these reflect the two goldstone modes . expanding to linear order in @xmath194 and to second order in @xmath59 we find explicitly [ eqs : a.17 ] ^-1_(k , i_n ) = ( cc ^2 f_k()/3 & 2i(i_n ) f_()/ + -2i(i_n ) f _ ( ) / & ^2 f_k()/3 ) , [ eq : a.17a ] where @xmath195 , @xmath196 , and f_k ( ) & = & , [ eq : a.17b ] + f _ ( ) & = & . + [ eq : a.17c ] physically , the stoner gap is always small compared to the fermi energy , and it therefore is useful to consider the limit of weak ferromagnets , @xmath197 and @xmath73 , where we have f_k(0 ) = f_(0 ) = 1 + o(^2 ) . [ eq : a.18 ] inverting eq . ( [ eq : a.17a ] ) we obtain the transverse physical spin susceptibility in the form given in eq . ( [ eq:2.11b ] ) , with the mean - field values for @xmath57 and @xmath56 as quoted in the main text . we note that the spin precession effect , which is represented by the off - diagonal matrix elements in eq . ( [ eq : a.17a ] ) and leads to the characteristic @xmath198 dispersion relation of ferromagnetic magnons , appears in a rather elementary way in this treatment of itinerant electrons . in spin models , by contrast , it emerges from a topological contribution to the action.@xcite as a further illustration of our arguments leading to an effective action for calculating relaxation rates , let us consider the well - known case of quasiparticle relaxation due to density fluctuations . to this end , we consider the very simple case of spinless , noninteracting electrons with action @xmath43 , and add a statically screened coulomb interaction s _ = _ k n(k)v_(k)n(-k ) . [ eq : b.1 ] here @xmath199 , with @xmath200 the screening wave number , and @xmath201 is the fourier transform of the electron number density @xmath202 . a finite average density is already built into @xmath43 via the chemical potential , so @xmath43 serves the purpose of the reference ensemble action @xmath47 in sec . [ sec : ii ] or appendix [ app : a ] . now we follow the logic of sec . [ subsec : ii.a ] . a number density fluctuation @xmath203 will couple to the field @xmath204 via the interaction @xmath205 to produce an action s[| , ] & = & s_0[| , ] + & & -30pt + dxdy n(x)(_x-_y)v_(x - y)n(y ) , [ eq : b.2 ] and the density fluctuations are governed by a gaussian action s_[n ] = dxdy n(x)^-1(x - y)n(y ) , [ eq : b.3 ] with @xmath191 the physical density susceptibility . integrating out the density fluctuations , we obtain an effective action [ eqs : b.4 ] s_[| , ] = s_0[| , ] + _ k n(k)v(k)n(-k ) [ eq : b.4a ] with an effective potential v(k ) = ( v_(k))^2(k ) . [ eq : b.4b ] now we calculate the single - particle relaxation rate as in sec . [ sec : iii ] . we obtain [ eqs : b.5 ] = ( = 0 ) = 2_-^ du |v(u ) , [ eq : b.5a ] where ( u ) = _ k , p ( _ k)(_p)v(k - p , u ) . [ eq : b.5b ] from eq . ( [ eq : b.4b ] ) we see that the spectrum of the potential @xmath92 is given by the spectrum of the density susceptibility , which to lowest order in the screened coulomb interaction is just the lindhard function @xmath206 . for @xmath207 , the spectrum of the latter is _ 0(k , u ) = u/ , [ eq : b.6 ] with @xmath7 the fermi velocity . for the relaxation rate due to the electron - electron interaction we thus obtain the well - known fermi - liquid result = . [ eq : b.7 ] the above derivation is similar in spirit to the arguments given in ref . . the point of this exercise is to demonstrate that our heuristic method of coupling density fluctuations to the appropriate fermion fields that we employed in sec . [ sec : ii ] still works in this case where the density fluctuations are produced by the very electrons they couple to . to put the result for the effective interaction , eq . ( [ eq : b.4a ] ) , in context , consider a bare coulomb interaction , @xmath208 , and perform an rpa resumption to produce a dynamically screened coulomb interaction v_(k ) = . [ eq : b.8 ] to linear order in the frequency , the spectrum of the effective potential @xmath92 coincides with the spectrum of @xmath209 , and @xmath92 therefore suffices to produce the leading low - temperature dependence of the relaxation rate . our effective action thus captures the leading effects of the soft modes in the system ( here , the soft particle - hole excitations that are reflected in the spectrum of the lindhard function ; in secs . [ sec : ii ] , [ sec : iii ] , the magnons ) . note that it does _ not _ suffice to produce static screening , which requires taking into account massive modes , which is why the above argument starts with a statically screened interaction . also note that the effective interaction @xmath92 is quadratic in the bare interaction @xmath205 , in accordance with fermi s golden rule . analogously , the effective interaction in sec . [ sec : ii ] , eq . ( [ eq:2.12b ] ) , is quadratic in the coupling constant @xmath34 . we also mention that the @xmath0 result , eq . ( [ eq : b.7 ] ) , holds for any short - ranged interaction , with the prefactor proportional to the potential strength squared . the above considerations assumed an electronic density fluctuation @xmath210 interacting with a density fluctuation @xmath203 created by all other electrons , in analogy with magnetization fluctuations in the case of an itinerant magnet . however , there is no reason why @xmath203 can not be a density fluctuation extraneous to the electron system , in analogy to magnetization fluctuations due to electrons in a band other than the conduction band . for instance , if @xmath203 is an ionic density fluctuation , it will still couple to @xmath210 via a statically screened coulomb interaction . equations ( [ eq : b.2 ] ) - ( [ eqs : b.5 ] ) remain formally valid , except that the susceptibility @xmath191 now describes ionic density fluctuations , i.e. , phonons . if we consider longitudinal phonons the susceptibility is the same as in a fluid and given by@xcite ( k , u ) = ^2u^2(u^2 - ^2_(k ) ) , [ eq : b.9 ] with @xmath211 the longitudinal phonon frequency . here @xmath212 is the ionic number density , @xmath213 is the longitudinal speed of sound , and @xmath214 , with @xmath92 the system volume and @xmath215 the pressure , is the compressibility . we thus have ( u ) = u^2u , [ eq : b.10 ] which leads to the familiar @xmath10 result for the single - particle scattering due to the electron - phonon interaction in metals , 1/ _ = ( 3)t^3 . [ eq : b.10 ]

in the leading order of the-expansion in hqet the dominant contribution to the semi - inclusive decays of polarized baryons into the charm strangeness mesons and is given by the partonic process . using standard values for the parameters of the process one expects a rather large branching ratio of into these two channels . in the factorization approximation the semi - inclusive decay of a polarized is governed by three unpolarized and four polarized structure functions for which we determine the nonperturbative corrections and the radiative corrections . we find that the perturbative and nonperturbative corrections amount to and , respectively . the seven structure functions can be measured through an analysis of the joint decay distributions of the process involving the polarization of the and the decays and for which we provide explicit forms . we also provide numerical results for the cabibbo suppressed semi - inclusive decays . mz - th/03 - 06 + hep - ph/0309047 + september 2003 + * * semi - inclusive decays at + including and polarization effects + m. fischer , s. groote , j.g . krner and m.c . mauser + institut fr physik , johannes - gutenberg - universitt + staudinger weg 7 , d55099 mainz , germany +

hep-ph0309047

in the leading order of the @xmath0-expansion in hqet the semi - inclusive decay @xmath14 is dominated by the partonic process @xmath15 . the basic assumption is that factorization holds for the nonleptonic decay process @xmath14 . one can then factorize the semi - inclusive decay into a current - induced @xmath16 transition and a current - induced vacuum one - meson transition . the leading order @xmath17 contribution to the @xmath16 transition is given by the partonic @xmath18 transition . there are two types of corrections to the leading order result . first there are the nonperturbative corrections which set in at @xmath6 in the heavy mass expansion . they can be estimated using the methods of the operator product expansion in hqet . second there are also the perturbative @xmath7 corrections which can be calculated using standard techniques . from a previous calculation of the corresponding decays in the mesonic sector @xmath19 one expects perturbative and nonperturbative corrections of @xmath8 @xcite and @xmath20 @xcite , respectively . when the @xmath1 is unpolarized , the decay @xmath21 is quite similar to the corresponding mesonic decay @xmath19 @xcite . in fact , to leading order in the @xmath22 expansion and to any order in the perturbative qcd corrections the two semi - inclusive decays are identical to one another . however , when the @xmath1 is polarized , there are four additional polarized structure functions that enter the decay analysis . one can thus probe four more structure functions in the semi - inclusive decay of a polarized @xmath1 than it is possible in the corresponding @xmath23-meson decay . polarized @xmath24-quarks and thereby polarized @xmath1 baryons arise quite naturally in weak decays such as @xmath25 and @xmath26 . when the polarized @xmath24-quark fragments into a @xmath1 baryon , @xmath27 of its polarization is retained @xcite . we mention that large samples of @xmath1 s are expected to be produced at the currently running @xmath28 collider tevatron 2 . in fact the first few @xmath1 s have been reconstructed by the cdf collaboration using the superior tracking capacity of their new silicon vertex trigger @xcite . in the factorization approximation the semi - inclusive decays of polarized @xmath1 baryons@xmath29 are governed by altogether seven structure functions which can be measured by an angular analysis of the decay process . we mention that there are two additional parity - violating structure functions in the decays @xmath30 which , however , can not be measured since the dominating decays of the @xmath31 are parity - conserving . five of the structure functions describe the semi - inclusive decay @xmath32 into vector mesons followed by their subsequent decay into @xmath10 and @xmath11 . the branching ratios of the @xmath31 into these two principal channels are given by @xmath33 and @xmath34 @xcite , respectively . the angular decay distribution of the semi - inclusive polarized @xmath1 decays can be obtained from the master formula ( see e.g. @xcite ) @xmath35 where @xmath36 is the density matrix of the @xmath37 which reads @xmath38 @xmath39 is the magnitude of the polarization of the @xmath1 . the @xmath40 are the helicity components of the hadronic tensor @xmath41 describing the semi - inclusive decay . the sum in eq . ( [ masterformula ] ) extends over all values of @xmath42 and @xmath43 compatible with the constraint @xmath44 ( the spin degrees of freedom of @xmath45 are being summed over ) . the polar angles @xmath46 , @xmath47 and the azimuthal angle @xmath48 are defined in fig . [ fig1 ] . because of angular momentum conservation , the second lower index in the small wigner @xmath49-function @xmath50 runs over @xmath51 for the decay @xmath10 and over @xmath52 for the decay @xmath11 . one thus obtains the angular decay distributions @xmath53 & + & \frac{3}{4 } \sin^2 \theta ( \gamma_l+\gamma_{l^p } p \cos \theta_p ) + \frac{3}{4 } \sqrt{2 } p \sin \theta_p \sin 2 \theta \cos \phi \ , \gamma_{i^p } \!\ ! \nonumber \\\end{aligned}\ ] ] and @xmath54 & + & \frac{3}{2 } \cos^2 \theta ( \gamma_l \!+\ ! \gamma_{l^p } p \cos \theta_p ) \!-\ ! \frac{3}{2 } \sqrt{2 } p \sin \theta_p \sin 2 \theta \cos \phi \ , \gamma_{i^p } \!\ ! \bigg\}. \nonumber \\\end{aligned}\ ] ] two - fold or single angle decay distributions can be obtained from eqs . ( [ angular1],[angular2 ] ) by further integration . for example , the single angle dependence on @xmath55 for both cases is given by @xmath56 & = & \frac{1}{2 } \gamma_{u+l } ( 1+\alpha_p(d_s^ * ) p \cos \theta_p ) , \end{aligned}\ ] ] where we have defined an asymmetry parameter @xmath57 . the transverse / longitudinal composition of the vector meson @xmath31 can be best determined by analyzing the @xmath58-dependence of the decay distributions after integrating over @xmath55 and @xmath48 . note that the @xmath59-dependence is different in the two decay modes . the decay distribution for @xmath60 can be obtained from the same master formula ( [ masterformula ] ) with the appropiate substitutions @xmath61 and @xmath62 . the helicity of the @xmath63 will be denoted by the symbol `` @xmath64 '' for `` scalar '' . one has @xmath65 & = & \frac{1}{2 } \gamma_s ( 1+\alpha_p(d_s ) p \cos \theta_p ) , \end{aligned}\ ] ] where we have again defined an asymmetry parameter @xmath66 . the angular coefficients @xmath67 ( @xmath68 ) appearing in the decay distributions are partial helicity rates defined by @xmath69 where the helicity structure functions @xmath70 are linear combinations of the helicity components . they read @xmath71 where , for the ease of writing , we have omitted factors of @xmath72 in the upper indices standing for the helicities of the @xmath1 . the remaining quantities appearing in ( [ partialrates ] ) are defined in sec . 3 . when the @xmath1 is unpolarized ( @xmath73 ) , or when one integrates over the angles @xmath46 and @xmath48 that describe the orientation of the polarization vector of the @xmath1 , one remains with the contributions of the three structure functions @xmath74 , @xmath75 and @xmath76 in the decay distributions in this way one recovers the decay distributions for the corresponding semi - inclusive decays of @xmath23 mesons into @xmath2 and @xmath3 treated in @xcite . as explained in sec . 2 , the decay @xmath14 involves seven structure functions which can be resolved by an angular analysis of the decay products . we begin by writing down the leading order born term contributions given by the quark level transition @xmath15 ( see fig . [ fig2]a ) . for the partial helicity rates one obtains @xmath77 where @xmath78 b_{u^{\phantom p } } & = & 2 x^2 ( 1-x^2+y^2 ) , \label{born3 } \nonumber\\[3 mm ] b_{u+l } & = & ( 1-y^2)^2+x^2(1+y^2 - 2x^2 ) , \label{born4 } \nonumber\\[3 mm ] b_{s^p } = b_{l^p } & = & \sqrt{{\lambda } } ( 1-y^2 ) , \label{born5 } \\[3 mm ] b_{u^p } & = & -2 x^2 \sqrt{{\lambda } } , \label{born6 } \nonumber\\[3 mm ] b_{(u+l)^p } & = & \sqrt{{\lambda } } ( 1-y^2 - 2x^2),\label{born7 } \nonumber\\ b_{i^p } & = & -\frac{1}{\sqrt{2 } } x \sqrt{{\lambda } } , \label{born8 } \nonumber\end{aligned}\ ] ] and where @xmath79 and @xmath80 . the kinematical factor @xmath81 is defined by @xmath82 such that @xmath83 . in eq . ( [ gammaborn ] ) , @xmath84 and @xmath85 denote the pseudoscalar and vector meson coupling constants defined by @xmath86 and @xmath87 , respectively . the kobayashi maskawa matrix element is denoted by @xmath88 , and the @xmath89 and @xmath90 are the magnitude of the three - momenta of the @xmath2 and @xmath3 in the @xmath24 rest system . the parameter @xmath91 is related to the wilson coefficients of the renormalized current - current interaction and is obtained from a combined fit of several decay modes ( @xmath92 ) @xcite . note that the structural similarity of the unpolarized and polarized rate formulae for the decay into @xmath2 and into the longitudinal @xmath3 is an accident of the born term calculation and does not persist e.g.at higher orders of @xmath93 , or for the nonperturbative contributions to the unpolarized longitudinal rate into @xmath3 to be written down later on . @xmath94 and @xmath95 determine the total @xmath96 and @xmath32 rates at the born term level , respectively . using @xmath97 and @xmath98 as in @xcite , @xmath99 , @xmath100 , @xmath101 and the central value for @xmath91 , one arrives at @xmath102 while the semi - inclusive @xmath14 rates have not been measured yet , a comparison of the born term prediction with data on the corresponding mesonic decay @xmath19 is meaningful because the born level predictions for both processes are identical . allowing for the factor @xmath103 and summing up the @xmath2 and @xmath3 modes , one arrives at a branching ratio of @xmath104 which is consistent with the measured value @xmath105 @xcite if one assumes that the above two rates saturate the semi - inclusive rate into @xmath106 . ( 0,60)*a ) * ( 0,60)*b)*(0,60)*c ) * ( 0,60)*d ) * next we turn to the @xmath7 radiative corrections . as is evident from fig . [ fig2 ] , the radiative gluon corrections connect only to the @xmath24 and @xmath108 legs of the parton decay process @xmath109 because of the conservation of colour ( see fig . 2b , 2c and 2d ) . the radiative corrections for the seven structure functions are thus identical to the corresponding radiative corrections calculated in @xcite where the process @xmath110 was considered ( including the scalar case ) keeping @xmath111 . correction to the spin @xmath112 piece of the weak current keeping both quark masses finite had been calculated before in @xcite . the @xmath7 corrections to the ( unpolarized ) spin @xmath113 piece of the weak current can also be deduced from the calculations of @xcite . ] for the @xmath7 radiative corrections one has to calculate the square of the tree - graph amplitudes figs . 2c and 2d , and the one - loop contribution fig . we concentrate on the tree graph contribution given by the squares of the tree - graph amplitudes fig . 2b and 2c which will be denoted by @xmath114 . the @xmath115 hadron tensor can be obtained from the corresponding @xmath116 hadron tensor given in @xcite by the replacements @xmath117 and @xmath118 . expression in @xcite ] one obtains @xmath119 + \nonumber \\ & & \hspace{-0.4 cm } \rule{0mm}{6 mm } + \frac{k \!\cdot\ ! p_c}{k \!\cdot\ ! p_b } \big [ ( \bar{p_b } \cdot p_b ) \big ( k^{\mu}\,p_c^{\nu}+k^{\nu}\,p_c^{\mu } - k \!\cdot\ ! p_c\,g^{\mu \nu } - i\,\epsilon^{\alpha \beta \mu \nu } k_{\alpha}\,p_{c,\beta } \big ) + \nonumber \\ & & \hspace{-0.4 cm } \rule{0mm}{6mm}- ( \bar{p}_b \!\cdot\ ! k ) \big ( ( p_b \!-\ ! k)^{\mu}\,p_b^{\nu } + ( p_b \!-\ ! k)^{\nu}\,p_c^{\mu}-(p_b \!-\ ! k ) \!\cdot\ ! p_c\,g^{\mu \nu } - i\,\epsilon^{\alpha \beta \mu \nu } ( p_b \!-\ ! k)_{\alpha}\,p_{c,\beta } \big ) \big ] + \nonumber \\ & & \hspace{-0.4 cm } \rule{0mm}{6 mm } - ( \bar{p}_b \!\cdot\ ! p_c ) \big ( k^{\mu}\,p_c^{\nu } + k^{\nu}\,p_c^{\mu}-k \!\cdot\ ! p_c\,g^{\mu \nu } - i\,\epsilon^{\alpha \beta \mu \nu } k_{\alpha}\,p_{c,\beta } \big ) + ( p_b \!\cdot\ ! p_c ) \big ( k^{\mu}\,\bar{p}_b^{\nu } + k^{\nu}\,\bar{p}_b^{\mu}-k \!\cdot\ ! \bar{p}_b\,g^{\mu \nu } \big ) \!+\kern-5pt \nonumber \\ & & \hspace{-0.4 cm } \rule{0mm}{6 mm } - ( k \!\cdot\ ! p_c ) \big ( p_b^{\mu}\,\bar{p}_b^{\nu } + p_b^{\nu}\,\bar{p}_b^{\mu}-p_b \!\cdot\ ! \bar{p}_b\,g^{\mu \nu } \big ) + ( k \!\cdot\ ! p_b ) \big ( ( p_c \!+\ ! k)^{\mu}\,\bar{p}_b^{\nu } \!+\ ! ( p_c \!+\ ! k)^{\nu}\,\bar{p}_b^{\mu } \!-\ ! ( p_c \!+\ ! k ) \!\cdot\ ! \bar{p}_b \,g^{\mu \nu } \big)+\kern-5pt \nonumber \\ & & \hspace{-0.4 cm } \rule{0mm}{6 mm } + ( k \!\cdot\ ! \bar{p}_b ) \big ( 2 p_c^{\mu}\,p_c^{\nu}-p_c \!\cdot\ ! p_c g^{\mu \nu } \big ) + i \big(\epsilon^{\alpha \beta \mu \nu}\,(k \!\cdot\ ! \bar{p}_b ) + \epsilon^{\alpha \beta \gamma \mu}\,k^{\nu } \bar{p}_{b,\gamma } - \epsilon^{\alpha \beta \gamma \nu}\,k^{\mu } \bar{p}_{b,\gamma } \big ) p_{c,\alpha}\,p_{b,\beta } + \nonumber \\ & & \hspace{-0.4 cm } \rule{0mm}{6 mm } + i \big(\epsilon^{\alpha \beta \mu \nu}\,(p_b \!\cdot\ ! \bar{p}_b ) + \epsilon^{\alpha \beta \gamma \mu}\,p^{\nu}_b \bar{p}_{b,\gamma } - \epsilon^{\alpha \beta \gamma \nu}\,p^{\mu}_b \bar{p}_{b,\gamma } \big ) k_{\alpha}\,p_{c,\beta } \bigg\}+b^{\mu \nu } \cdot \delta_{sgf } \nonumber \end{aligned}\ ] ] @xmath120 where @xmath121 is the 4-momentum of the emitted gluon . the polarization of the bottom quark is taken into account by introducing the short - hand notation @xmath122 . we have found it convenient to split the tree - graph hadron tensor into an infrared ( ir ) finite piece and an ir divergent piece given by the usual soft - gluon factor @xmath123 multiplied by the born term tensor @xmath124 @xmath125 in this way the ir singularity is isolated in the universal function @xmath123 which can be integrated by introducing a gluon mass regulator to regularize the ir singularity . the ensuing logarithmic gluon mass singularity is cancelled by the corresponding gluon mass singularity occurring in the loop contribution ( see e.g.@xcite ) . the phase - space integration of the ir convergent piece can be done without a gluon mass regulator . one first projects the convergent piece of the tree - graph tensor ( [ hadrontensor ] ) onto the seven helicity structure functions eq.(8 ) and then does the phase - space integration in the sequential order ( i ) @xmath126 ( gluon energy ) ( ii ) @xmath127 ( energy of the off - shell @xmath128 ) . the final @xmath7 answer including the one - loop contribution can be written in a very compact way by introducing combinations of dilogarithmic functions @xmath129 and @xmath130 . the contribution denoted by @xmath129 is part of the finite remainder of the born term type one - loop contribution plus the soft gluon contribution . the combinations @xmath130 appear when integrating those helicity structure functions @xmath131 and @xmath132 that are not associated with the total rate . all these functions are defined after eq.([defgammaip ] ) . we mention that we have now been able to present our results on the radiatively corrected structure functions in a much more compact form than thought possible when we wrote up @xcite . we shall present our @xmath7 results in a form where the respective born terms @xmath133 are factored out from the @xmath7 result . including the born term and the nonperturbative @xmath6 contributions to be discussed in sec . 4 we write @xmath134 and @xmath135 for @xmath136 , and @xmath137 and @xmath138 for @xmath139 . one has @xmath140 @xmath141 is the expectation value of the kinetic energy of the heavy quark in the @xmath1 baryon and @xmath142 parametrizes the spin - dependent contribution of the heavy quark in the @xmath1 baryon @xcite . to begin with , we list the reduced @xmath7 rates @xmath143 . for the reduced unpolarized and polarized scalar spin @xmath113 rates @xmath144 and @xmath145 we obtain @xmath146^{- 1 } \nonumber \\ [ 1.5 mm ] & & \times \ ; \bigg\ { \frac{2}{x^2}\,\big [ \left ( 1-x^2 \right ) \left ( 2-x^2 \right ) -\left ( 6 + 4\,x^2 + 5\,x^4 \right ) y^2 \nonumber \\ [ 1.5 mm ] & & \hspace{1cm}+\ ; \left ( 6 + 7\,x^2 \right ) y^4 - 2\,y^6 \big]\,\sqrt{\lambda}\,\ln ( y ) \nonumber \\ [ 1.5 mm ] & & + \ ; 8\,\big [ 1-x^2-\left ( 2+x^2 \right ) y^2+y^4 \big ] \ , \sqrt{\lambda}\,\ln \left ( \frac{x\,y}{\lambda } \right ) \nonumber \\ [ 1.5 mm ] & & -\ ; \frac{1}{x^2}\,\big [ \left ( 1-x^2 \right)^2 \left ( 2 + 3\,x^2 \right)-\left ( 8 - 3\,x^2 + 4\,x^4 - 3\,x^6 \right ) y^2 \nonumber \\ [ 1.5 mm ] & & \hspace{1cm}+\ ; 3 \left ( 4 + 5\,x^2 \right ) y^4 -\left ( 8 + 5\,x^2 \right ) y^6 + 2\,y^8 \big]\,\ln ( w_1 ) \nonumber \\ [ 1.5 mm ] & & + \ ; 8 \left ( 1-y^2 \right ) \big [ 1-x^2-\left ( 2+x^2 \right ) y^2+y^4 \big]\,\ln ( \eta ) \nonumber \\ [ 1.5 mm ] & & + \ ; 3\,\sqrt{\lambda}\,\big [ 3 \left ( 1-x^2 \right ) -\left ( 10 + 3\,x^2 \right ) y^2 + 3\,y^4 \big ] \bigg\}\end{aligned}\ ] ] @xmath147\,{\cal{n}}_0 + 4\,\sqrt{\lambda } \left ( 1-x^2+y^2 \right ) { \cal{n}}_4 \nonumber \\ [ 1.5 mm ] & & + \ ; \frac{2}{x^2}\,\frac{1}{1-y^2}\,\big [ 2-x^2 -\left ( 4 + 5\,x^2 \right ) y^2 + 2\,y^4 \big]\,\lambda\,\ln ( y ) + 8\,\lambda\,\ln \left ( \frac{x\,y}{\lambda } \right ) \nonumber \\ [ 1.5 mm ] & & -\ ; \frac{\sqrt{\lambda}}{x^2}\,\big [ 2 - 9\,x^2+x^4 -\left ( 4 + 3\,x^2 \right ) y^2 + 2\,y^4 \big]\,\ln ( w_1 ) + 8\,\lambda\,\ln \left [ \frac{(1+x)^2-y^2}{x } \right ] \nonumber \\ [ 1.5 mm ] & & + \ ; 4\,\big [ \left ( 1-x^2 \right ) \left ( 5 - 2\,x^2 \right ) + 2 \left ( 2-x^2 \right ) y^2 \big]\,\ln \left ( \frac{1-x}{y } \right ) \nonumber \\ [ 1.5 mm ] & & -\ ; \big [ ( 1-x)^2-y^2 \big ] \left ( 11 - 6\,x-7\,x^2 + 7\,y^2 \right ) \bigg\}\end{aligned}\ ] ] the other variables and functions appearing in eqs . ( [ defgammas ] ) and ( [ defgammasp ] ) are explained at the end of this section . for the five unpolarized and polarized reduced spin @xmath112 rates @xmath143 ( @xmath148 ) we obtain @xmath149 & & -\ ; \frac{2}{x}\,\big [ ( 1-x)^2-y^2 \big]\,\big [ ( 1-x)\,(5+x)+y^2 \big]\,{\cal{n}}_2 \nonumber \\ [ 1.5 mm ] & & -\ ; \frac{2}{x}\,\big [ ( 1+x)^2-y^2 \big]\,\big [ ( 1+x)\,(5-x)+y^2 \big]\,{\cal{n}}_3 \nonumber \\ [ 1.5 mm ] & & -\ ; \frac{2}{x^2 } \left ( 1-x^2+y^2 \right ) \left ( 1 - 2\,x^2-y^2 \right ) \sqrt{\lambda}\,\ln ( y ) + 8 \left ( 1-x^2+y^2 \right ) \sqrt{\lambda}\,\ln \left ( \frac{x\,y}{\lambda } \right ) \nonumber \\ [ 1.5 mm ] & & + \ ; \frac{1}{x^2}\,\big [ \left ( 1-x^2 \right)^2 \left ( 1 - 6\,x^2 \right ) \nonumber \\ [ 1.5 mm ] & & \hspace{1cm}-\ ; \left ( 1 + 4\,x^2 - 3\,x^4 \right ) y^2-\left ( 1 + 2\,x^2 \right ) y^4+y^6 \big]\,\ln ( w_1 ) \nonumber \\ [ 1.5 mm ] & & + \ ; 4\,\big [ 7 + 3\,x^2 -\left ( 4 - 5\,x^2 \right ) y^2 - 3\,y^4 \big]\,\ln ( \eta)-\sqrt{\lambda } \left ( 19+x^2 - 5\,y^2 \right ) \bigg\}\end{aligned}\ ] ] @xmath150^{- 1 } \ , \bigg\ { 8\,x^2 \left ( 7+x^2-y^2 \right ) { \cal{n}}_1 \nonumber \\ [ 1.5 mm ] & & + \ ; 4\,x\,\big [ ( 1-x)^2-y^2 \big]\,\big [ ( 1-x)\,(5+x ) + y^2 \big]\,{\cal{n}}_2 \nonumber \\ [ 1.5 mm ] & & + \ ; 4\,x \ ; \big [ ( 1+x)^2-y^2 \big]\,\big [ ( 1+x)\,(5-x ) + y^2 \big]\,{\cal{n}}_3 \nonumber \\ [ 1.5 mm ] & & + \ ; 2\,\big [ 1-x^2-\left ( 4 + 3\,x^2 \right ) y^2 + 3\,y^4 \big]\,\sqrt{\lambda}\,\ln ( y ) \nonumber \\ [ 1.5 mm ] & & + \ ; 8\,\big [ 1-x^2-\left ( 2+x^2 \right ) y^2+y^4 \big ] \ , \sqrt{\lambda}\,\ln \left ( \frac{x\,y}{\lambda } \right ) \nonumber \\ [ 1.5 mm ] & & -\ ; \big [ 5 \left ( 1-x^2 \right)^2-\left ( 3 + 20\,x^2-x^4 \right ) y^2+\left ( 9 - 2\,x^2 \right ) y^4+y^6 \big]\,\ln ( w_1 ) \nonumber \\ [ 1.5 mm ] & & + \ ; 8 \left ( 1+x^2-y^2 \right ) \big [ 1 - 7\,x^2-\left ( 2+x^2 \right ) y^2+y^4 \big]\,\ln ( \eta ) \nonumber \\ [ 1.5 mm ] & & + \ ; \sqrt{\lambda}\,\big [ 5 + 47\,x^2 - 4\,x^4-\left ( 22+x^2 \right ) y^2 + 5\,y^4 \big ] \bigg\}\end{aligned}\ ] ] @xmath151^{- 1 } \nonumber \\ [ 1.5 mm ] & & \times \ ; \bigg\ { -\,2\,\big [ \left ( 1-x^2 \right ) \left ( 1 - 4\,x^2 \right)+\left ( 4+x^2 \right ) y^2 - 5\,y^4 \big]\,\sqrt{\lambda}\,\ln ( y ) \nonumber \\ [ 1.5 mm ] & & + \ ; 8\,\big [ \left ( 1-x^2 \right ) \left ( 1 + 2\,x^2 \right ) -\left ( 2-x^2 \right ) y^2+y^4 \big]\,\sqrt{\lambda}\,\ln \left ( \frac{x\,y}{\lambda } \right ) \nonumber \\ [ 1.5 mm ] & & -\ ; \big [ 3 \left ( 1-x^2 \right)^2 \left ( 1 + 4\,x^2 \right ) \nonumber \\ [ 1.5 mm ] & & \hspace{1cm}-\ ; \left ( 1 + 12\,x^2 + 5\,x^4 \right ) y^2 + \left ( 11 + 2\,x^2 \right ) y^4-y^6 \big]\,\ln ( w_1 ) \nonumber \\ [ 1.5 mm ] & & + \ ; 8 \left ( 1-y^2 \right ) \big [ 1+x^2 - 4\,x^4 -\left ( 2-x^2 \right ) y^2+y^4 \big]\,\ln ( \eta ) \nonumber \\ [ 1.5 mm ] & & + \ ; \sqrt{\lambda}\,\big [ 5 + 9\,x^2 - 6\,x^4 -\left ( 22 - 9\,x^2 \right ) y^2 + 5 y^4 \big ] \bigg\}\end{aligned}\ ] ] @xmath152\,{\cal{n}}_0 + 4\,\sqrt{\lambda } \left ( 1-x^2+y^2 \right ) { \cal{n}}_4 \nonumber \\ [ 1.5 mm ] & & -\ ; \frac{2}{x^2 } \left ( 1 - 2\,x^2-y^2 \right ) \lambda\,\ln ( y)+8\,\lambda\,\ln \left ( \frac{x\,y}{\lambda } \right ) \nonumber \\ [ 1.5 mm ] & & + \ ; \frac{\sqrt{\lambda}}{x^2}\,\big [ 7 + 21\,x^2 + 2\,x^4 -\left ( 8 + 3\,x^2 \right ) y^2+y^4 \big ] \ln ( w_1 ) \nonumber \\ [ 1.5 mm ] & & + \ ; 8\,\lambda\,\ln \left [ \frac{(1+x)^2-y^2}{x } \right ] \nonumber \\ [ 1.5 mm ] & & + \ ; \frac{4}{x^2}\,\big [ \left ( 1-x^2 \right ) \left ( 3 + 14\,x^2 - 2\,x^4 \right ) \nonumber \\ [ 1.5 mm ] & & \hspace{1cm}-\ ; \left ( 6 - 7\,x^2-x^4 \right ) y^2+\left ( 3-x^2 \right ) y^4 \big]\,\ln \left ( \frac{1-x}{y } \right ) \nonumber \\ [ 1.5 mm ] & & + \ ; \frac{1}{x}\,\big [ ( 1-x)^2-y^2 \big]\,\big [ 12 - 55\,x+6\,x^2-x^3 - 3 \left ( 4+x \right ) y^2 \big ] \bigg\}\end{aligned}\ ] ] @xmath153 & & \times \ ; \bigg\ { 4\,\big [ 2 + 22\,x^2 + 11\,x^4 -\left ( 5 + 12\,x^2+x^4 \right ) y^2 + 2 \left ( 2+x^2 \right ) y^4 -y^6 \big]\,{\cal{n}}_0 \nonumber \\ [ 1.5 mm ] & & + \ ; 4 \left ( 1-y^2 \right ) \sqrt{\lambda } \left ( 1-x^2+y^2 \right ) { \cal{n}}_4 \nonumber \\ [ 1.5 mm ] & & + \ ; 2 \left ( 1 - 3\,y^2 \right ) \lambda\,\ln ( y)+8 \left ( 1-y^2 \right ) \lambda\,\ln \left ( \frac{x\,y}{\lambda } \right ) \nonumber \\ [ 1.5 mm ] & & + \ ; \sqrt{\lambda}\,\big [ 17 + 53\,x^2-\left ( 18+x^2 \right ) y^2+y^4 \big]\,\ln ( w_1 ) \nonumber \\ [ 1.5 mm ] & & + \ ; 8\,\lambda \left ( 1-y^2 \right ) \ln \left [ \frac{(1+x)^2-y^2}{x } \right ] \nonumber \\ [ 1.5 mm ] & & + \ ; 4\,\big [ \left ( 1-x^2 \right ) \left ( 11 + 24\,x^2 \right ) -\left ( 13 - 15\,x^2 \right ) y^2 + 2\,y^4 \big]\,\ln \left ( \frac{1-x}{y } \right ) \nonumber \\ [ 1.5 mm ] & & -\ ; \big [ ( 1-x)^2-y^2 \big ] \\ [ 1.5 mm ] & & \hspace{1cm}\times \ ; \big [ 15 - 22\,x+105\,x^2 -24\,x^3 + 4\,x^4-\left ( 12 - 22\,x+x^2 \right ) y^2 -3\,y^4 \big ] \bigg\ } \nonumber\end{aligned}\ ] ] @xmath154\,{\cal{n}}_0 \nonumber \\ [ 1.5 mm ] & & + \ ; 4\,\sqrt{\lambda } \left ( 1-x^2+y^2 \right ) \left ( 1 - 2\,x^2-y^2 \right ) { \cal{n}}_4 \nonumber \\ [ 1.5 mm ] & & + \ ; 2\,\lambda \left ( 3 - 4\,x^2 - 5\,y^2 \right ) \ln ( y ) + 8\,\lambda \left ( 1 - 2\,x^2-y^2 \right)\,\ln \left ( \frac{x\,y}{\lambda } \right ) \nonumber \\ [ 1.5 mm ] & & + \ ; \left ( 3-x^2+y^2 \right ) \left ( 1 + 4\,x^2-y^2 \right ) \sqrt{\lambda}\,\ln ( w_1 ) \nonumber \\ [ 1.5 mm ] & & + \ ; 8\,\lambda \left ( 1 - 2\,x^2-y^2 \right ) \ln \left [ \frac{(1+x)^2-y^2}{x } \right ] \nonumber \\ [ 1.5 mm ] & & + \ ; 4\,\big [ \left ( 1-x^2 \right ) \left ( 5 - 4\,x^2 + 4\,x^4 \right ) \nonumber \\ [ 1.5 mm ] & & \hspace{1cm}-\ ; \left ( 1-x^2 + 2\,x^4 \right ) y^2 -2 \left ( 2-x^2 \right ) y^4 \big]\,\ln \left ( \frac{1-x}{y } \right ) \nonumber \\ [ 1.5 mm ] & & -\ ; \left [ ( 1-x)^2-y^2 \right ] \\ [ 1.5 mm ] & & \hspace{1 cm } \times \ ; \big [ 15 + 2\,x-5\,x^2 - 12\,x^3 + 2\,x^4-\left ( 12 + 2\,x+7\,x^2 \right ) y^2 - 3\,y^4 \big ] \bigg\ } \nonumber\end{aligned}\ ] ] @xmath155\,{\cal{n}}_0 \nonumber \\ [ 1.5 mm ] & & + \ ; 4\,\sqrt{\lambda } \left ( 1-x^2+y^2 \right ) { \cal{n}}_4 \nonumber \\ [ 1.5 mm ] & & -\ ; \frac{1}{x^2 } \left ( 1 - 3\,x^2-y^2 \right ) \lambda \ln ( y ) + 8\,\lambda\,\ln \left ( \frac{x\,y}{\lambda } \right ) \nonumber \\ [ 1.5 mm ] & & + \ ; \frac{\sqrt{\lambda}}{2\,x^2}\,\big [ 1 + 30\,x^2 + 21\,x^4 -2 \left ( 1 + 11\,x^2 \right ) y^2+y^4 \big]\,\ln ( w_1 ) \nonumber \\ [ 1.5 mm ] & & + \ ; 8\,\lambda\,\ln \left [ \frac{(1+x)^2-y^2}{x } \right ] \nonumber \\ [ 1.5 mm ] & & + \ ; 2\,\big [ \left ( 1-x^2 \right ) \left ( 21 + 5\,x^2 \right ) -\left ( 11 - 15\,x^2 \right ) y^2 - 4\,y^4 \big]\,\ln \left ( \frac{1-x}{y } \right ) \nonumber \\ [ 1.5 mm ] & & -\ ; 2\,\big [ ( 1-x)^2-y^2 \big ] \left ( 12 - 7\,x+12\,x^2 - 9\,y^2 \right ) \bigg\}\end{aligned}\ ] ] as mentioned before the contribution denoted by @xmath129 is that part of the finite remainder of the born term type one - loop contribution plus the soft gluon contribution which contains dilogs and products or squares of logs . it is given by @xmath156 & & -\ ; \ln ( w_1)\,\ln \left ( \frac{\lambda^2\,w_3}{x^2\,y^3 } \right)-\frac{1}{2}\,\ln^2 ( w_1 ) + \ln \left [ \frac{1}{2 } \left ( 1-x^2+y^2+\sqrt{\lambda } \right ) \right ] \ln \left ( w_2\,w_3 \right ) \bigg\}\ , . \nonumber \end{aligned}\ ] ] the functions @xmath130 are defined by @xmath157 { \cal{n}}_1 & = & \mbox{li}_2 ( x\,\eta ) -\mbox{li}_2 \left ( \frac{x}{\eta } \right)+2\,\ln ( 1-\eta\,x ) \ , \ln \left [ \frac{(\eta+1)\,x}{1+x } \right]+\ln \left ( \frac{\eta}{\eta - x } \right ) \ln \left [ \frac{x^2\,(\eta-1)^2}{\eta\,(\eta - x ) } \right ] \nonumber \\ [ 6 mm ] { \cal{n}}_2 & = & \mbox{li}_2 \left [ \frac{(\eta-1)\,x}{\eta - x } \right]+\mbox{li}_2 \left [ \frac{(\eta-1)\,x}{1-x } \right]-\frac{1}{2}\,\ln^2 ( 1-x ) + \ln \left ( \frac{\eta}{\eta - x } \right ) \ln \left [ \frac{(\eta-1)\,x}{\eta - x } \right ] \nonumber \\ [ 3 mm ] & & + \ ; \ln ( 1-x)\,\ln \left ( \frac{1-x}{\eta - x } \right)+\ln ( 1-\eta\,x)\,\ln \left [ \frac{(\eta+1)\,x}{1+x } \right ] \nonumber \\ [ 6 mm ] { \cal{n}}_3 & = & \mbox{li}_2 \left ( \frac{1-\eta\,x}{1+x } \right)-\mbox{li}_2 \left [ \frac{\eta - x}{\eta\,(1+x ) } \right]-\frac{1}{2}\,\ln \left ( \frac{\eta}{\eta - x } \right ) \ln \left [ \frac{\eta\,(\eta-1)^2\,(1+x)^2}{(\eta+1)^2\,(\eta - x ) } \right ] \nonumber \\ [ 6 mm ] { \cal{n}}_4 & = & 4\,\mbox{li}_2 \left ( \frac{\eta\,\sqrt{\lambda}}{\eta - x } \right)-2\,\mbox{li}_2 \left [ \frac{(\eta-1)\,x}{1-x } \right]-2\,\mbox{li}_2 \left [ \frac{(\eta-1)\,x}{\eta - x } \right]+\mbox{li}_2 \left ( \frac{x}{\eta } \right)-\mbox{li}_2 ( x\,\eta ) \nonumber \\ [ 3 mm ] & & -\ ; \ln^2 ( 1-x)+\ln \left ( \frac{\eta}{\eta - x } \right ) \ln \left [ \frac{\eta\,(\eta+1)^2}{\eta - x } \right]+2 \ , \ln ( 1-\eta\,x)\,\ln \left [ \frac{(1-x)\,(\eta+1)}{\eta - x } \right]\ , , \nonumber\end{aligned}\ ] ] where we use the abbreviations @xmath158 \eta & = & \frac{1+x^2-y^2+\sqrt{\lambda}}{2\,x}\end{aligned}\ ] ] when one uses the operator product expansion in hqet one can determine the nonpertubative corrections to the leading partonic @xmath159 rate . the nonpertubative corrections set in at @xmath6 and arise from the kinetic energy and the spin dependent piece of the heavy quark in the heavy baryon @xcite . the strength of the kinetic and the spin dependent piece are parametrized by the expectation values of the relevant operators in the @xmath1 system and are denoted by @xmath141 and @xmath142 , respectively . we have completely recalculated the nonpertubative contributions to the seven partial rates and have found some errors in the calculation of @xcite which will be corrected in an erratum to @xcite . one has @xmath160 the nonperturbative contributions for @xmath161 and @xmath162 can be compared to the corresponding @xmath163-distributions in semi - leptonic @xmath24-decays written down in @xcite . we find agreement . for our numerical evaluation we use @xmath164 for the mean kinetic energy of the heavy quark in the @xmath1 as in @xcite . an estimate of the spin - dependent parameter has been given in @xcite with the result @xmath165 , based on an assumption that the contribution of terms arising from double insertions of the chromomagnetic operator can be neglected . a zero recoil sum rule analysis gives the constraint @xmath166 @xcite which puts the estimate of @xcite at the upper bound of the constraint . we use the value of @xcite keeping in mind that the numerical value of @xmath142 could be reduced in more realistic calculations . using @xmath167 , @xmath168 , @xmath169 , @xmath170 and @xmath171 we obtain for @xmath159 @xmath172 \hat\gamma_{u^{\phantom p } \phantom{+ l } } & = & { \phantom{-}}0.3541\,(1 - 0.1079 + 0.0255 + 0 ) , \nonumber\\[2 mm ] \hat\gamma_{l^{\phantom p } \phantom{+ l } } & = & { \phantom{-}}0.6459\,(1 - 0.1103 - 0.0341 + 0 ) , \nonumber\\[2 mm ] \hat\gamma_{u+l^{\phantom p } } & = & \phantom{+0.0000 \ , } ( 1 - 0.1095 - 0.0130 + 0 ) , \nonumber\\[2 mm ] \hat\gamma_{s^p \phantom{+ l } } & = & { \phantom{-}}0.9884\,(1 - 0.1027 - 0.0245 - 0.0087 ) , \\[2 mm ] \hat\gamma_{u^p \phantom{+ l } } & = & - 0.2646\,(1 - 0.0616 - 0.0276 - 0.0087 ) , \nonumber\\[2 mm ] \hat\gamma_{l^p \phantom{+ l } } & = & { \phantom{-}}0.6351\,(1 - 0.1043 - 0.0276 - 0.0087 ) , \nonumber\\[2 mm ] \hat\gamma_{(u+l)^p{\phantom p } } & = & { \phantom{-}}0.3705\,(1 - 0.1348 - 0.0276 - 0.0087 ) , \nonumber\\[2 mm ] \hat\gamma_{i^p \phantom{+ l } } & = & - 0.2148\,(1 - 0.0876 + 0.0059 - 0.0087 ) . \nonumber\end{aligned}\ ] ] the four entries in the round brackets correspond to the born term contribution , the @xmath7 corrections , and the nonperturbative kinetic and spin - dependent corrections in that order , as specified in eq . ( [ altogether ] ) . the reduction of the partial rates from the radiative corrections scatter around @xmath104 , where the reduction is largest for @xmath173 ( @xmath174 ) and smallest for @xmath175 ( @xmath176 ) . when normalized to the total rate , as is appropriate for density matrix elements , the corresponding density matrix elements are reduced by @xmath177 and increased by @xmath178 in magnitude by the radiative corrections , respectively . the nonperturbative corrections range from @xmath179 for the spin - dependent corrections to a maximal @xmath180 for the kinetic energy correction to @xmath181 . the nonperturbative corrections are all negative except for the kinetic energy correction to @xmath182 and @xmath183 . as specified in eqs . ( [ singleangle - star ] ) and ( [ singleangle ] ) , the asymmetry parameters @xmath184 can be measured in the semi - inclusive decays of a polarized @xmath185 into the two decay channels . for the pseudoscalar case the born term level asymmetry @xmath186 is quite close to its maximal attainable value of 1 which would be achieved for @xmath187 . the born term value is only slightly reduced to @xmath188 by the radiative and nonperturbative corrections . for the vector case the asymmetry parameter is smaller . at born term level one has @xmath189 which is reduced to @xmath190 including the radiative and nonperturbative corrections . as outlined in sec . 2 the transverse / longitudinal composition of the @xmath3 can be measured by the @xmath59-dependence in the angular decay distribution of its decay products . at the born term level the transverse / longitudinal composition is given by @xmath191 . this ratio is shifted upward by the insignificant amount of @xmath192 through the radiative corrections . adding all corrections one finds a @xmath193 enhancement in the @xmath194 ratio . next we turn to our numerical results for the cabibbo - suppressed semi - inclusive decays @xmath195 induced by the @xmath196 transitions . compared to the above cabibbo - enhanced semi - inclusive decays they are down by a factor @xmath197 , which is only slightly compensated for by a kinematical enhancement factor of @xmath198 . setting @xmath199 , i.e.@xmath187 , one has @xmath200 \hat\gamma_{u^{\phantom p } \phantom{+ l } } & = & { \phantom{-}}0.2750\,(1 - 0.1150 + 0.0285 + 0 ) , \nonumber\\[2 mm ] \hat\gamma_{l^{\phantom p } \phantom{+ l } } & = & { \phantom{-}}0.7250\,(1 - 0.1777 - 0.0267 + 0 ) , \nonumber\\[2 mm ] \hat\gamma_{u+l^{\phantom p } } & = & \phantom{+0.0000 \ , } ( 1 - 0.1605 - 0.0130 + 0 ) , \nonumber\\[2 mm ] \hat\gamma_{s^p \phantom{+ l } } & = & \phantom{+0.0000 \ , } ( 1 - 0.1745 - 0.0212 - 0.0087 ) , \\[2 mm ] \hat\gamma_{u^p \phantom{+ l } } & = & - 0.2750\,(1 - 0.1275 - 0.0212 - 0.0087 ) , \nonumber\\[2 mm ] \hat\gamma_{l^p \phantom{+ l } } & = & { \phantom{-}}0.7250\,(1 - 0.1800 - 0.0212 - 0.0087 ) , \nonumber\\[2 mm ] \hat\gamma_{(u+l)^p{\phantom p } } & = & { \phantom{-}}0.4500\,(1 - 0.2121 - 0.0212 - 0.0087 ) , \nonumber\\[2 mm ] \hat\gamma_{i^p \phantom{+ l } } & = & - 0.2233\,(1 - 0.1506 + 0.0005 - 0.0087 ) . \nonumber \end{aligned}\ ] ] in the @xmath196 case the radiative corrections and their spread are larger than in the @xmath159 case . the reduction of the partial rates from the radiative corrections now scatter around @xmath201 , where the reduction is largest for @xmath173 ( @xmath202 ) and smallest for @xmath182 ( @xmath203 ) . when normalized to the total rate , the corresponding density matrix elements are reduced by @xmath204 and increased by @xmath205 in magnitude by the radiative corrections , respectively . the dominance of the longitudinal rate is now more pronounced . at the born term level one finds @xmath206 . the ratio @xmath207 is shifted upward by @xmath208 by the radiative corrections . adding up all corrections one finds a @xmath209 upward shift for this ratio . for the asymmetry parameter one obtains @xmath210 including all corrections which shifts the uncorrected result @xmath211 downward by @xmath212 . for the asymmetry parameter @xmath213 one obtains @xmath214 which is lower than the uncorrected result of @xmath215 by @xmath193 . let us mention that our @xmath7 results on @xmath216 and @xmath217 numerically agree with the results of @xcite for both the @xmath159 and @xmath196 transitions . as a last point we want to discuss the semi - inclusive decays @xmath218 which have not been discussed so far . they are also induced by the diagrams fig . 2 when the @xmath219 transition in the upper leg is replaced by a @xmath220 transition . using @xmath221 , @xmath222 and @xmath223 one finds the born term branching fractions @xmath224 and @xmath225 . in the latter case the rate is dominated by the longitudinal contribution since @xmath226 is not far from @xmath227 where the rate would be entirely longitudinal . in fact one finds @xmath228 . it is important to note that the diagrams fig . 2 are not the only mechanisms that contribute to the semi - inclusive decays @xmath218 . additional @xmath229 and @xmath230 mesons can also be produced by fragmentation of the @xmath108-quark at the lower leg . the possibility of producing extra @xmath231 and @xmath31 mesons through fragmentation of the @xmath108-quark is ruled out for kinematic reasons . ] as concerns the @xmath230 mesons resulting from the fragmentation process they would not be polarized along their direction of flight . this lack of polarization as compared to the strong polarization of the @xmath232 mesons from the weak vertex could possibly be used to separate @xmath230 mesons coming from the two respective sources . we have calculated the perturbative @xmath7 and the nonperturbative @xmath6 corrections to the seven structure functions that can be measured in the semi - inclusive decay of a polarized @xmath1 in the process @xmath29 . we have used the factorization hypothesis to factorize the semi - inclusive decay into a current - induced @xmath16 transition and a current - induced vacuum one - meson transition . the dominant contribution to the current - induced @xmath16 transition is given by the leading order hqet transition @xmath159 . thus the semi - inclusive decays of a polarized @xmath1 offer the unique opportunity to measure seven of the nine structure functions that describe the current - induced free quark transition @xmath159 . we emphasize that there are also nonfactorizing @xmath7 contributions which have not been included in our analysis . however , the nonfactorizing @xmath7 contributions are colour suppressed and are thus expected to be small . we find that the perturbative corrections are always negative . the nonperturbative corrections are negative in most of the cases . the net effect of the corrections to the structure functions can become as large as @xmath233 for the @xmath159 transitions and can exceed @xmath233 for the @xmath196 transitions . when normalized to the total rate , as is appropriate for density matrix elements accessible to experimental measurement , the corrections become smaller but can still amount to @xmath234 . 99 r. aleksan , m. zito , a. le yaouanc , l. oliver , o. pne and j .- c . raynal , + phys . * d62 * ( 2000 ) 093017 m. fischer , j.g . krner , m.c . mauser and s. groote , + phys . * b480 * ( 2000 ) 265 f.e . close , j.g . krner , r.j.n . phillips and d.j . summers , + j.phys . * g18 * ( 1992 ) 1716 a.f . falk and m.e . peskin , phys . * d49 * ( 1994 ) 3320 i. vila for the cdf collaboration , hep - ph/0307165 review of particle physics , k. hagiwara _ et al . _ , phys . rev . * d66 * ( 2002 ) 010001 . m. fischer , s. groote , j.g . krner and m.c . mauser , + phys . * d65 * ( 2002 ) 054036 q. hokim and x .- y . pham , annals of phys . * 155 * ( 1984 ) 202 m. jeabek and j.h . khn , nucl . * b314 * ( 1989 ) 1 a. czarnecki , phys . * b252 * ( 1990 ) 467 a. denner and t. sack , nucl . * b358 * ( 1991 ) 46 a. czarnecki , m. jeabek and j.h . khn , phys . * b346 * ( 1995 ) 335 a.v . manohar and m.b . wise , phys . rev . * d49 * ( 1994 ) 1310 s. balk , j.g . krner and d. pirjol , eur . j. * c1 * ( 1998 ) 221 , + erratum to be published j.g . krner and d. pirjol , phys . rev . * d60 * ( 1999 ) 014021 a.f . falk and m. neubert , * d47 * ( 1993 ) 2982 j.g . krner and d. pirjol , phys . * b334 * ( 1994 ) 399

in this paper we present and discuss the effects of scattered light echoes ( le ) on the luminosity and spectral appearance of type ia supernovae ( sne ) . after introducing the basic concepts of le spectral synthesis , by means of le models and real observations we investigate the deviations from pure sn spectra , light and colour curves , the signatures that witness the presence of a le and the possible inferences on the extinction law . the effects on the photometric parameters and spectral features are also discussed . in particular , for the case of circumstellar dust , les are found to introduce an apparent relation between the post - maximum decline rate and the absolute luminosity which is most likely going to affect the well known pskowski - phillips relation . [ firstpage ] supernovae : general - supernovae : individual ( sn 1991 t ) - supernovae : individual ( sn 1998bu ) - supernovae : individual ( sn 2001el ) - ism : dust , extinction nebulae

astro-ph0512574

due to the crucial role played by type ia sne in modern cosmology ( see for example the review by @xcite ) , it is essential to understand the systematics which might affect the recent conclusions about the expansion rate of the universe . the availability of larger telescopes and the renewed interest for these objects as cosmological probes has opened the possibility of studying in great detail their evolution , an investigation which in the past was feasible for the closest events only . in these last few years , in fact , numerous ia have been followed with an extensive time coverage as late as one year past maximum light , leading to very interesting results ( see for example @xcite ) . in two cases , namely those of sn 1991 t @xcite and sn 1998bu @xcite , the spectra at advanced phases were clearly dominated by a phenomenon which had nothing to do with the explosion mechanism but rather with the neighboring environment . in both cases this very atypical behaviour could be explained with the contamination by sn light scattered into the line of sight by dust present in the surroundings of the explosion . the presence of such a phenomenon , also known as light echo ( hereafter le ) , was later confirmed by high spatial resolution hst imaging ( @xcite ) . other cases have been discovered for core - collapse sne like sn 1987a @xcite , sn 1993j @xcite , sn 1999ev @xcite and sn 2003gd @xcite . in the most recent years , les have been searched also in the case of historical sne @xcite . with the purpose of studying this effect and its possible applications to the analysis of the sn environments and their relation with the sn progenitors , we have started a series of papers on this topic . the first of them ( @xcite ; hereafter paper i ) was dedicated to the general characteristics of les , with particular attention to the effects of self absorption on le luminosity , spectral and colour appearance . in this and subsequent papers we will instead present and discuss applications to known and test cases . in principle , les can affect all types of sne . due to the young stellar population from which they are supposed to be originated , core - collapse sne are of course favored with respect to thermonuclear explosions . however , both due to their high homogeneity ( which makes the analysis much simpler ) and their wide use in cosmology , here we will concentrate on type ia only . in particular , the present work deals with the effects of les on their luminosity and spectra , in order both to help the observers in disentangling intrinsic features from le contamination and to understand whether les can affect the conclusions reached on their nature . the paper is organized as follows . after giving a short introduction to the basic concepts of le modeling in section [ sec : basic ] , in section [ sec : speclib ] we describe the spectral library used in our calculations . the results are then presented and discussed in section [ sec : effects ] for two different dust geometries , while section [ sec : ext ] is devoted to the implications on the extinction law . in section [ sec : discuss ] we discuss our findings which are briefly summarized in section [ sec : conclusion ] . the le phenomenon has been presented and discussed in a number of publications and we refer to them the interested reader ( see for example the recent review by @xcite and the references therein ) . the detailed description of the formation of a le spectrum has been outlined in paper i and , therefore , here we will only shortly recap the basic concepts . in single scattering approximation , the le spectrum at a given time @xmath0 can be expressed as @xmath1 where @xmath2 is the intrinsic source ( sn ) spectrum and @xmath3 is a function that describes the physical and geometrical properties of the dust . if we define @xmath4 as the optical depth to the sn along the line of sight and @xmath5 is the sn spectrum clean of le contamination , then @xmath6 if we indicate with @xmath7 the dust extinction cross section , @xmath8 the dust albedo and @xmath9 the forward scattering degree , in all those cases when there is one single cloud with dust on the line of sight ( like in the two examples of fig . [ fig : dustgeom ] ) , the formal solution of equation ( [ eq : spectrum ] ) can be written as follows ( see paper i ) : @xmath10\ ; \omega(\lambda ) \ ; c_{ext}(\lambda)\ ; e^{\tau_d(\lambda)}\ ; { \cal s}(\lambda , t)\ ] ] where @xmath11 $ ] is a time- and wavelength - dependent function that includes dust and dust geometry properties and @xmath12 is the time integrated sn spectrum , where @xmath13 is the time taken by the generic iso - delay surface to completely cross the dust cloud . the crossing time depends of course on the dust geometry and , if @xmath14 , the lower integration boundary of equation [ eq : integrated ] has to be replaced with 0 . when the sn is only mildly reddened ( @xmath151 ) , the resulting le spectrum is therefore proportional to the product between @xmath16 and the extinction cross section . when the optical depth is larger ( @xmath17 ) , single scattering description must be replaced by the more realistic _ single scattering plus attenuation _ approximation ( hereafter ssa ; see paper i ) , which holds for @xmath18 . in this more refined schema , which takes into account the le self - absorption , equation ( [ eq : ss ] ) becomes @xmath19\ ; \omega(\lambda ) \ ; c_{ext}(\lambda)\ ; e^{\tau_d(\lambda)-\tau_{eff}(\lambda , t)}\ ; { \cal s}(\lambda , t)\ ] ] where @xmath20 can be regarded as a weighted le optical depth at a given time @xmath0 . in general , one has that @xmath21 , as in the case sketched in fig . [ fig : dustgeom ] ( left panel ) , where the sn suffers a larger extinction than the le . in the cases where the sn and the le are affected roughly by the same extinction , the two effects tend to compensate and the le spectrum is again proportional to the product between the observed time integrated sn spectrum @xmath22 and the extinction cross section @xmath23 . these formal expressions are useful to understand the principles of the problem and will be used later on to explain the apparent peculiarity reported by @xcite in the extinction function ( see section [ sec : ext ] ) . nevertheless , we note that the exact ssa numerical solution is easy to implement in a code and it offers the advantage of properly taking into account the attenuation effect and the wavelength dependency of all involved quantities . while we refer the reader to paper i for more details , here we will spend some words to discuss one important geometrical effect , which is related to forward scattering and was only marginally mentioned in our previous paper . in fact , all le calculations include the so called scattering phase function @xmath24 , which describes the probability distribution of photon scattering angle @xmath25 ( see for example @xcite ) . usually , this function is parameterized , following @xcite , through the forward scattering degree @xmath9 . to illustrate the behaviour of @xmath24 , in figure [ fig : gtest ] we have plotted the results one obtains using the milky way dust mixtures modeled by @xcite for @xmath26=3.1 ( solid lines ) and @xmath26=5.5 ( dashed lines ) and for three values of the scattering angle . as one can see , the wavelength dependency is much stronger for lower values of @xmath25 , with the scattering being more efficient in the blue than in the red . therefore , when the dust is located far in front of the sn , which implies a small average scattering angle , the le spectrum is expected to be bluer than in the case where the same dust , with the same geometry , is placed close to the sn . as a consequence , the same dust cloud can produce the same extinction to the sn ( i.e. the same @xmath27 ) but different le colours , according to its position along the line of sight . the simulations show that the @xmath28 colour of the le can change by @xmath290.15 mag from a configuration where the sn is embedded in the dust cloud to one where the same cloud is placed at more than 200 lyr in front of the sn . the le spectral modeling relies on the availability of input sn spectra . for this purpose one needs to build a spectral library with a good phase coverage and time sampling , especially around maximum light . this can be understood inspecting the cumulative functions for the template light curves we discussed in paper i. these were constructed using the data of sne 1992a and 1994d , two standard , well studied and low reddening type ia ( @xcite ; @xcite ) . for @xmath30 and @xmath31 passbands , where the le is expected to be brighter , it turns out that about 90% of the radiation is emitted in 0.2 yr , while the radiation emitted in the first 10 days accounts for about 5% of total . therefore , the spectral library should include spectra from about a week before maximum to 2 - 3 months past maximum . spectra at later epochs are needed only if one is going to extend the calculations to correspondingly late phases for dust geometries with @xmath32 . we have extended the library used in paper i , which included only data from sn 1998aq @xcite , adding early and late spectra from sn 1992a ( eso key - programme on sne ) and sn 1994d @xcite , so that the data set ranges from @xmath3311 up to @xmath34406 days . irrespective of the original flux calibration , the input spectra have been re - calibrated to make their synthetic @xmath31 magnitude consistent with the @xmath31 template light curve . as for the late time spectra , we have used the available data of sne 1992a and 1994d , which are presented in figure [ fig : late ] and reach 406 days past @xmath30 maximum light . as one can see , the spectra of the two objects remain very similar , also at these late stages . in order to remove the noise in the data , we have used a multi - gaussian fit to the spectra ( solid smooth lines ) . due to the lack of data at later phases , in our calculations we have assumed that after phase + 406 there is no evolution and , from that phase on , the sn spectrum is simply scaled according to the exponential decay of the @xmath31 light curve . as a matter of fact , to our knowledge the only type ia sne with optical spectroscopy at phases later than 450 days are sne 1991 t and 1998bu , both dominated by evident les . therefore , very little is known about the optical spectral behaviour at these phases . even though there are evidences of deviations from the radioactive decay in the near - ir , as in the cases of sn 1998bu @xcite and sn 2002cx @xcite , for the optical domain things appear to be different . in fact , photometric observations of the normal ia sne 1972e @xcite , 1992a @xcite and 1996x @xcite at phases as late as 500 days indicate that the light curve follows the radioactive decay at least until these epochs . another fact that should be mentioned is that even objects which were very similar at early phases , do show differences at late phases . this is clearly illustrated in figure [ fig : compialate ] , where we have compared the spectra of four type ia at about one year past maximum light : 1992a ( [email protected] , @xcite ) , 1994d ( [email protected] , @xcite ) , 1998bu ( [email protected] , @xcite ) and 2001el ( [email protected] , @xcite ) . despite the fact that all these sne have shown very similar spectra at maximum light ( @xcite , @xcite , @xcite , @xcite ) , at late phases some differences are visible . the most pronounced one is the variable intensity of the bump at about 7300 , usually attributed to a blend of [ feii ] lines @xcite , which is almost invisible in sn 1992a and well developed in sn 2001el . other discrepancies are observed also in the blue , where the [ feii ] and [ feiii ] bumps appear to show different ratios in the various objects . in this respect , it is worthwhile noting that the spectrum of sn 1998bu is affected by a le ( see @xcite ; @xcite ) , which dominates below 4500 ( see figure [ fig : compialate ] , third panel , thin curve ) . the existence of these differences must be kept in mind , since the intrinsic properties might be mistaken with the onset of a le , even when this is not the case . using the library we have presented here and the ssa approximation ( or the monte carlo approach when required ) discussed in paper i , we have computed the le spectra and analyzed the effects these have on objects like sne 1992a and 1994d under two very different conditions , i.e. when the dust cloud is very large ( @xmath36 ) and in the opposite case when its size is very small ( @xmath37 ) . to illustrate the le effects in the large cloud case ( see fig . [ fig : dustgeom ] , left panel ) , we run a series of calculations using the ssa approximation and a simple dust geometry , placing a dust sheet perpendicular to the line of sight at @xmath39=200 lyr in front of the sn , with a thickness @xmath40=50 lyr and a particle density @xmath41=5 @xmath42 . this configuration , which implies @xmath43=0.12 ( or @xmath44=0.13 ) produces a light curve which is qualitatively similar to what was observed in 1991 t @xcite , as shown in figure [ fig : lcsheet ] . we remark that we did not make any special attempt to fit the sn 1991 t data since our purpose here is to show the general effects in a realistic situation . in this respect , it is also worth mentioning that the light curve of this slow declining object and its spectra at maximum where rather different from those we have used for our calculations ( see for example the broad light curve peak shown by 1991 t in comparison to our template in figure [ fig : lcsheet ] ) . a more detailed analysis of the sne 1991 t and 1998bu cases is going to be presented in a separate paper . as shown in figure [ fig : lcsheet ] , the light curve starts to deviate from the ordinary exponential decay ( dashed line ) at about one year after the explosion , to settle down on a very slow decline , totally driven by the le , at about two years past the explosion . between these two extremes , the observed spectrum is expected to show a gradual transition from that typical of a ia at late phases to that of a pure le . the amount of the le contamination can be judged in figure [ fig : evol ] , where we have plotted the results for @xmath0=1.2 , 1.4 and 1.6 yr from the explosion . already at the first epoch , when the total @xmath31 luminosity ( sn+le ) is only 0.2 mags brighter than that expected for the radioactive decay , the resulting spectrum is sensibly different from that of a ia at these phases , especially in the blue . due to the peak present in the le spectrum at about 4600 , the most prominent feature of the intrinsic nebular sn spectrum at 4700 , normally attributed to [ feiii ] , appears to be broadened and the region between 4000 and 4500 much bluer than in an unaffected ia . this is actually quite similar to what was seen in sn 1998bu @xcite in the spectrum obtained 381 days past @xmath30 maximum ( see also figure [ fig : compialate ] here ) , where the bump at about 4000 is unusually bright . [ sec : closedust ] the consequences of a le in the red are milder , and tend to emerge only at later phases . the dominant [ feiii ] bump gets gradually fainter compared to the other emerging features and the colour becomes bluer and bluer , reaching @xmath450.2 when the spectrum is totally dominated by the le , around 2 yr after the explosion . during the transition phase a distinguishing feature is also the appearance of caii h&k lines , reflecting their strong presence in maximum light spectra . the outcome can be different if the dust optical depth is increased , mainly due to the le self absorption ( see paper i ) . to illustrate this fact , we have run a series of calculations placing the dust sheet at @xmath39=800 lyr and increasing its optical depth to @xmath43 to about 1 ( @xmath41=40 @xmath42 ) . the distance was properly tuned to produce a le with the same observed magnitude as the one discussed earlier ( see figure [ fig : lcsheet ] , long - dashed line ) . in this case the le spectrum is much redder , having @xmath46 + 0.15 . as a matter of fact , both the sn and the le undergo roughly the same reddening , which in this case is @xmath470.35 . in general , for higher extinctions the ssa approximation becomes insufficient and a more complicated multiple scattering approach has to be performed ( see paper i ) . before moving to the next section we like to make a consideration on the late spectrum shown by sn 2001el . besides the intense bump at about 7500 , the region between 4200 and 5400 appears to be rather blue . in particular , the bump at 4400 is much more intense than in any of the other objects presented in figure [ fig : compialate ] . the synthetic colour of the observed spectrum is @xmath28=0.25 , and it reduces to @xmath28=0.07 after applying the extinction reported by @xcite for @xmath26=3.1 . this value is unusually blue when compared to the ones derived from the spectra of sn 1994d ( 0.38 ) and sn 1992a ( 0.34 ) . at a first glance this might be interpreted as the signature of an underlying le , but there are two quite important aspects that need to be considered . the first is that the synthetic magnitude deduced from the observed spectrum is @xmath31=21.24 , which is in perfect agreement with the standard luminosity decline . secondly , the spectrum drops quite fast below 4300 , contrarily to what happens for instance to 1998bu ( see figure [ fig : compialate ] ) . the simulations show that changing the optical depth of the region which is responsible for the hypothetical le does not allow to produce the drop which is seen in the blue . the conclusion is that the late time spectral appearance of sn 2001el is not caused by a le of the type considered in this section . in general , when the dust region is sufficiently extended in the direction perpendicular to the line of sight , the dust system integrates the emitted sn spectra over the whole time range from the explosion to the epoch of the observation ( @xmath36 ) , so that the resulting le spectrum is very different from that of the sn at any given epoch . therefore , it is very easy to detect its emergence , that can be difficultly mistaken with intrinsic features . an increase in the dust density would simply anticipate the time when the le emerges , change the colour of the le itself and cause a larger extinction on the sn itself , but the le characteristic spectral features will still be very well pronounced . 1.6 mm ccccccccccccc @xmath43 & @xmath48 & @xmath49 & @xmath50 & @xmath51 & @xmath52 & @xmath53 & @xmath54 & @xmath55 & @xmath56 & @xmath44 & @xmath57 & @xmath58 + & & & [ 1.45 ] & [ 1.40 ] & [ 6.98 ] & & & & + + 0.1 & 0.11 & 0.04 & 1.36 & 1.29 & 6.70 & 0.0 & @xmath330.02 & @xmath330.01 & 0.03 & 0.09 & 0.08 & + 0.01 + 0.5 & 0.54 & 0.18 & 1.16 & 1.10 & 5.91 & 1.2 & @xmath330.10 & @xmath330.05 & 0.13 & 0.44 & 0.40 & + 0.04 + 1.0 & 1.09 & 0.35 & 0.95 & 1.01 & 5.23 & 2.6 & @xmath330.24 & @xmath330.07 & 0.29 & 0.84 & 0.88 & @xmath330.04 + 2.0 & 2.17 & 0.70 & 0.55 & 0.93 & 4.29 & 3.4 & @xmath330.48 & @xmath330.16 & 0.54 & 1.69 & 1.68 & + 0.01 + 3.0 & 3.26 & 1.05 & 0.30 & 0.89 & 3.65 & 4.8 & @xmath330.75 & @xmath330.25 & 0.80 & 2.52 & 2.48 & + 0.04 + + + 0.1 & 0.11 & 0.04 & 1.38 & 1.41 & 6.94 & 0.0 & @xmath330.05 & @xmath330.02 & 0.02 & 0.05 & 0.05 & 0.00 + 0.5 & 0.54 & 0.18 & 1.30 & 1.43 & 6.79 & 1.2 & @xmath330.27 & @xmath330.09 & 0.09 & 0.28 & 0.28 & 0.00 + 1.0 & 1.09 & 0.35 & 1.24 & 1.43 & 6.63 & 2.6 & @xmath330.56 & @xmath330.14 & 0.21 & 0.53 & 0.64 & @xmath330.11 + 2.0 & 2.17 & 0.70 & 0.97 & 1.37 & 6.31 & 3.2 & @xmath331.07 & @xmath330.28 & 0.42 & 1.10 & 1.29 & @xmath330.19 + 3.0 & 3.26 & 1.05 & 0.86 & 1.33 & 6.04 & 4.4 & @xmath331.56 & @xmath330.41 & 0.64 & 1.70 & 1.99 & @xmath330.29 + + + 0.1 & 0.11 & 0.04 & 1.41 & 1.41 & 6.96 & 0.0 & @xmath330.06 & @xmath330.02 & 0.02 & 0.00 & 0.05 & 0.00 + 0.5 & 0.54 & 0.18 & 1.39 & 1.45 & 6.88 & 1.1 & @xmath330.30 & @xmath330.09 & 0.08 & 0.24 & 0.25 & @xmath330.01 + 1.0 & 1.09 & 0.35 & 1.29 & 1.48 & 6.78 & 1.5 & @xmath330.60 & @xmath330.17 & 0.18 & 0.49 & 0.55 & @xmath330.06 + 2.0 & 2.17 & 0.70 & 1.19 & 1.49 & 6.60 & 2.8 & @xmath331.17 & @xmath330.30 & 0.40 & 1.00 & 1.23 & @xmath330.23 + 3.0 & 3.26 & 1.05 & 1.03 & 1.47 & 6.46 & 3.1 & @xmath331.69 & @xmath330.44 & 0.61 & 1.57 & 1.90 & @xmath330.33 + + + + more subtle is the case when the dust system characteristic dimension is significantly smaller than @xmath59 ( see fig . [ fig : dustgeom ] , right panel ) . in fact , under these conditions , the dust system integrates the sn input spectra over times which are much shorter than the time elapsed from the explosion ( i.e. @xmath37 ) and the resulting le spectrum is much closer to the sn spectrum , making it very difficult to detect it on the basis of the spectral appearance alone . of course , in order to have this effect , the dimension of the dust cloud must be indeed small . for example , for a spherical dust cloud with radius @xmath60 and centered on the sn it is easy to show that @xmath61 . the simulations show that in order to have a le spectrum which is very similar to that of the sn at all epochs , one needs to have @xmath620.2 lyr , which implies @xmath630.1 lyr ( @xmath2910@xmath64 cm ) . since these dimensions are much smaller than those of typical interstellar dust clouds , the most natural scenario to be considered is the one of circumstellar material . as done by @xcite , @xcite and @xcite , we will consider here the simplest case , i.e. that of a spherically symmetric stellar wind in which the density is given by @xmath65 , where @xmath66 is the inner boundary of the wind , within which the density is supposed to be null . also , we assume that the wind has an outer boundary , defined by @xmath67 . under these circumstances , the optical depth along the line of sight is simply @xmath68 , from which it is clear that if @xmath69 the value of @xmath27 is practically independent from @xmath70 . as we have shown in paper i ( cf . fig . 6 ) this geometry , compared to the others we had explored , is able to produce quite bright les with a fast declining light curve . to illustrate the effects of such a dust geometry , we have run a series of calculations using the mc code we have described in paper i. in all simulations we have used the canonical milky way dust mixture with @xmath26=3.1 @xcite and we have computed the le @xmath30 and @xmath31 light curves and optical spectra for @xmath66=0.1 , 0.01 and 0.005 lyr , keeping @xmath70=1.0 lyr and for different values of @xmath43 ( 0.1 , 0.5 , 1.0 , 2.0 and 3.0 ) . two examples of the resulting curves are presented in fig . [ fig : lcwind ] for @xmath66=0.01 lyr and @xmath43=0.5 , 3.0 . the main photometric parameters obtained from the global light curves ( sn+le ) are summarized in table [ tab : models ] , where besides the the optical depth and the implied extinction and colour excess ( @xmath43 , @xmath48 and @xmath49 ) we show the usual @xmath71 , the slope at phases later than 6 months ( @xmath51 ) , the magnitude jump between maximum and 300 days past maximum ( @xmath72 ) , the time shift in the @xmath30 maximum ( @xmath53 ) , the differences in @xmath31 and @xmath28 at maximum with respect to the extincted sn ( @xmath54 , @xmath55 ) , the colour excess with respect to the unreddened sn at maximum ( @xmath56 ) , the apparent extinction obtained comparing with the intrinsic magnitude ( @xmath73 , the extinction derived from @xmath56 assuming @xmath26=3.1 ( @xmath57 ) and the deviation of the absolute magnitude at maximum with respect to the input value ( @xmath58 ) . for comparison , the corresponding values for the template light curves assumed by the models are as follows : @xmath50=1.45 mag , @xmath51=1.40 mag/100@xmath74 and @xmath52=6.98 mag . there are several conclusions that can be drawn already inspecting table [ tab : models ] . first of all , as already qualitatively noticed by @xcite , the resulting light curves become broader and the maximum is attained later then in the pure sn case ( cf . also fig . [ fig : lcwind ] ) . this has the clear effect of decreasing @xmath50 rather sensibly and the change increases with @xmath43 and @xmath66 . moreover , the global colour gets bluer than that of the purely extincted sn and this effect becomes stronger for larger @xmath27 and smaller @xmath66 . in principle , unless the values of this parameters become too deviant from the _ normal _ ones ( 0.9 @xmath751.9 , @xcite ) , they could be attributed to an intrinsic feature of the sn . nevertheless , there are other side effects that tend to betray the presence of an underlying le . in fact , the late time decay rate @xmath51 increases with @xmath27 and @xmath66 , while the magnitude jump @xmath52 grows with @xmath66 and decreases for increasing values of @xmath27 . in the case of @xmath66=0.005 lyr ( @xmath295@xmath7610@xmath77 cm ) , the global light curve around maximum light changes but its overall shape remains similar to the template light curve ( e.g. giving plausible values for @xmath50 , @xmath52 and so on ) . this has a very interesting consequence , at least for @xmath780.01 lyr . in fact , besides measuring a slower post - maximum decline rate , one would also measure a colour which is significantly bluer and a global apparent luminosity which is higher than those of the purely reddened sn . as the simulations show , these two facts combined together would lead an hypothetical observer deducing the reddening correction from the global observed colour ( and using a standard value for @xmath26 ) to derive an absolute luminosity which is brighter than the real one by an amount that can reach about 0.3 mag ( see last column of tab . [ tab : models ] ) . in conclusion , one would attribute the smaller value of @xmath50 to a larger intrinsic luminosity , as in the case of the classical phillips relation @xcite . of course this does not mean that the observed relation is produced by dust in the vicinity of the sn , since @xmath79 correlates in fact with other features like spectral line ratios @xcite . but certainly , the presence of hidden les might introduce an additional spread in the measured photometric parameters . the effects on spectra are presented in fig . [ fig : mcevol1 ] and [ fig : mcevol2 ] , where we show the time evolution in the @xmath66=0.01 lyr case for @xmath430=0.5 and 3.0 . in the low optical depth case the effect is rather mild ( the global spectrum is slightly bluer than that of the reddened sn spectrum at all phases ) , while the le influence becomes much more relevant in the high optical depth case . as a consequence of the increased le contribution to the total flux , the time integration becomes visible and this causes the spectral features to show clear deviations from the intrinsic ones . the most remarkable is the absorption profile of the si ii 6355 line around maximum light ( @xmath0=0.05 lyr ) , which is much broader . the same behaviour is shown by all lines and at all epochs . a late phases , the line peaks also appear blue - shifted with respect to the intrinsic features . moreover , as in the low density case , the global spectrum is bluer than the reddened sn spectrum at all times . in general , the strongest effect is expected around maximum light and in the weeks that immediately follow it , i.e. when the spectral evolution is fast compared with the le integration time which , in the case of @xmath66=0.01 lyr is of the order of 12 days dust density distribution , this is larger than the 7.3 days one would have in the case all the dust was located in a very thin shell with a radius of 0.01 lyr . ] . in the pre - maximum phase , the early spectra contribute quite marginally to the global flux , due to the very rapid rise in luminosity while , just after maximum , the sn flux in the preceding days is actually higher than on the current epoch , so that the contribution is indeed relevant . this is clearly displayed in fig . [ fig : mctimeseq ] , where we present the calculated evolution of the si ii region between @xmath0=0.03 and @xmath0=0.10 yrs , which correspond to @xmath337.3 and + 18.3 days past maximum light . on the first epoch , the profile of the global spectrum is already broader than in the input sn , but the deviation grows in the following days , reaching a maximum during the two weeks past maximum light . the position of the absorption through minimum appears to be red - shifted with respect to the input spectrum , producing a lower photospheric velocity estimate . during the second week , also the emission peak of the line is affected and it appears much broader , reflecting the higher velocities reached in the preceding epochs . these deviations , together with those shown by the light ( and colour ) curves , would most probably allow one to recognize the presence of a le . nevertheless , we must notice that this is true only for high optical depths , while for @xmath801 , the effects on photometric parameters and spectral features are rather mild ( see table [ tab : models ] and fig . [ fig : mcevol1 ] ) , making the detection of the le certainly much more difficult . in those cases the net effect of the le would be to alter the colours and the overall flux of the observed spectrum , with the consequences discussed by @xcite , that we will analyze in the next section . if a sn is affected by the emergence of a le , it is clear that each time one tries to derive the amount of extinction or information about the dust properties , the results will be in general misleading . as we have done in the previous section , we distinguish here the two opposite cases . when @xcite discovered the le in sn 1991 t , they noticed that , in order to reproduce the observations , they had to correct the time integrated sn spectrum @xmath81 for the effects of scattering . these were parameterized with a function of the type @xmath82 and the best fit was obtained using @xmath83=2 . as we have seen in section [ sec : basic ] , the correction one has to apply to @xmath81 includes three components : the first is related to the light scattered by the dust into the observer s line of sight ( directly proportional to @xmath23 ) , the second accounts for the reddening suffered by the sn ( proportional to @xmath84 ) while the third describes the auto - absorption within the dust cloud that generates the le itself ( proportional to @xmath85 . finally , there is a mild wavelength dependency introduced by the albedo and a more important contribution by the function @xmath86 ( see section [ sec : basic ] ) . a more convenient formulation can be obtained passing from the extinction cross section @xmath7 to the normalized extinction law @xmath87 . if we pose @xmath88 then the absorption ( in magnitudes ) can be expressed as @xmath89 and equation ( [ eq : ssa ] ) can be rewritten as @xmath90 where @xmath91 is expressed in magnitudes . in a first approximation , one can parameterize @xmath87 with a power law of the type @xmath92 . from equation ( [ eq : law1 ] ) it is clear that a fitting to the ratio between @xmath93 and @xmath22 using a power law @xmath82 would give a value for @xmath83 which is not directly related to @xmath94 , since other terms with different wavelength dependencies enter its expression . in fact , using our simulations for the perpendicular sheet with @xmath39=200 lyr we have found that for the milky way dust mixture with @xmath26=3.1 @xcite , the best fit values of @xmath83 depend mostly on the value of @xmath95 , while the dependence on @xmath96 is very mild , at least in the validity range of the ssa approximation ( @xmath971 ) . typical values for @xmath83 are 1.4 , 1.7 , 2.0 and 2.2 for @[email protected] , 0.0 , 0.2 and 0.4 respectively . it is interesting to notice that the @xmath83=2 value found by @xcite can be explained with a canonical extinction law and a small difference between the extinction suffered by the le and the sn , i.e. only 0.2 mag . the sn was certainly reddened : @xcite report in fact a total extinction @[email protected] , 0.46 of which is due to the host galaxy . if we assume @xmath98=0.46 ( or equivalently @xmath43=0.42 ) we can therefore give an estimate to @xmath99 , which turns out to be 0.26 ( i.e. @xmath100=0.24 ) . the corresponding model is presented in figure [ fig : sn91 t ] . given the fact that the synthetic spectrum has been computed using the spectra of objects like 1994d and 1998aq which are quite different from sn 1991 t , especially during the maximum light phase , the match is reasonably good . however , we must notice that the observed spectrum is also compatible with @xmath95=0 , i.e. with the sn and the le suffering the same amount of reddening ( figure [ fig : sn91 t ] , thin smooth curve ) . another example of a possibly similar situation is given by sn 1998bu . @xcite have presented a le spectrum at 670 days past maximum light , with a @xmath28 colour close to 0 , i.e. very similar to that observed for sn 1991 t @xcite . on the other hand , @xcite have reported a total extinction @xmath44=0.94 @xmath350.15 , while @xcite found @xmath44=1.0 @xmath1010.1 . using the same geometrical configuration , the best fit with the observed le spectrum is obtained for @xmath100=0.62 , which implies @xmath44=0.67 ( see figure [ fig : sn98bu ] ) . given the fact that the estimated galactic extinction in the direction of sn 1998bu is only @xmath44=0.08 @xcite , this value is about 0.2 mag lower than the average value reported by @xcite and @xcite . in this respect it is interesting to note that , due to the geometrical effect related to forward scattering that we have described in section [ sec : basic ] , if the dust is close to the sn , then a lower dust optical depth would still produce a red le spectrum . this is not because there is more reddening , but because under these circumstances the scattering efficiency does not grow at shorter wavelengths as fast as it does if the dust is located far way . this is illustrated in the insert of fig . [ fig : sn98bu ] , where we have plotted the results of a simulation where the same perpendicular sheet is placed at distance @xmath39=0 . in this case the best fit is obtained for @xmath100=0.25 , which would imply a sn extinction @xmath44=0.27 , which is significantly lower than the values reported in the literature . in conclusion , the information on the extinction and the extinction law one derives from the observed le spectrum is geometry dependent and not univocally related to the dust properties . as we have seen in sec . [ sec : closedust ] , the presence of a le produced by a dust geometry with @xmath37 is able to introduce sensible luminosity and colour changes with mild effects on the other photometric and spectroscopic characteristics , provided that @xmath102 . this , as @xcite has shown , can have an interesting consequence on the properties one would deduce for the dust . in fact , if we take the example of @xmath66=0.01 lyr and @xmath43=1.0 , from table [ tab : models ] we see that , due to the presence of a le , the global @xmath31 magnitude is 0.56 brighter than the purely extincted sn , which is affected by an extinction of @xmath48=1.09 mag . therefore , assuming a distance to the sn and using the expected absolute luminosity , one would deduce @xmath44=0.53 . then , comparing the colour at maximum with a reference value ( which here we assume to be the one of our template object ) , one would also deduce @xmath56=0.21 from which she would erroneously conclude that @xmath1032.5 . as far as the photometric properties are concerned , such an object would show normal @xmath50 ( 1.24 ) and @xmath51 ( 1.43 ) and at late phases the slight over - luminosity ( 0.3 mag ) would probably go unnoticed due to the measurement errors . therefore , at least on the basis of those parameters , it would be very difficult to recognize the presence of an underlying le . since the difference between the global luminosity and that of the pure sn is not constant in time and changes rather rapidly around maximum light ( see fig . [ fig : lcwind ] , upper right insert ) , the values of @xmath26 deduced in this way are expected to change with time in a similar way . this is illustrated in fig . [ fig : windmccol ] , where we present the results of mc simulations for @xmath66=0.01 lyr and two different values of the dust optical depth . the total - to - selective absorption ratio changes in fact very rapidly from the explosion to about 0.15 yrs , which correspond to the maximum of the intrinsic sn ( @xmath104 colour curve ( dotted curve ) . after that epoch , the ratio remains almost constant . this behaviour is qualitatively similar to the one found by @xcite ( see his fig . 2 ) and the discrepancies are probably due to the different treatments of multiple scattering=0.5 . ] . it is important to remark that the global colour curve is not obtained by a simple rigid reddening of the unextincted sn colour curve . on the contrary , the difference between bluest and reddest colour is smaller and the tail at phases later than 0.15 yrs ( @xmath2955 days ) is less steep . the effect becomes more pronounced for higher optical depths . in other words , once the observed global colour curve is corrected for the colour excess deduced at maximum , this is not going to match a reference colour curve and the late behaviour would clearly differ from the homogeneous trend found by @xcite . we reckon that this colour effect , shown also by the semi - analytical calculation of @xcite ( see his fig . 3 , lower panel ) is a good diagnostic to unveil the presence of a hidden le . at this point we must emphasize that , in all practical cases , the colour excess is deduced from the data using the colour curve . on the contrary , the extinction is usually derived from this value after assuming @xmath26 and then applied to the observed magnitude to obtain the absolute luminosity of the object ( see for example @xcite ) . following this procedure in the case of our previous example would lead to a result which is actually not far from the real one . in fact , if one adopts @xmath105=0.21 and assumes @xmath26=3.1 , this would produce an absolute magnitude which is only 0.11 mag brighter than the real value ( see table [ tab : models ] , last column ) . remarkably , this is of the same order of the correction foreseen by the phillips relation @xcite for the given @xmath71 . as we had already noticed in sec . [ sec : closedust ] , this effect is most likely going to add a spread to the observed luminosity - decline rate relation . another way of deriving information on the extinction law @xmath106 is to compare the observed spectrum @xmath107 with a reference unreddened sn spectrum @xmath108 obtained at the same phase and corrected to the same distance ( see for instance @xcite ) . if the le effects were negligible , one could in fact write @xmath109 $ ] and hence derive directly the extinction wavelength dependency . of course , when @xmath107 contains a significant contribution from the le the result is going to be misleading . this is clearly illustrated in fig . [ fig : testext ] , where we present the results obtained from our monte carlo simulations at two different epochs ( @xmath0=0.05 and 0.15 yrs ) for the case @xmath110=1.0 and @xmath66=0.01 lyr . for the first epoch , which approximately corresponds to the @xmath30 maximum light , the colour of the global spectrum is clearly redder than that of the unreddened sn ( @xmath1110.2 , see also fig . [ fig : windmccol ] ) and the derived extinction law is just slightly steeper than the input one , thus giving a smaller @xmath26 . with the exception of the siii 6355 feature , the derived curve is reasonably smooth . totally different is the case at @xmath0=0.15 , which roughly corresponds to the maximum of the @xmath28 colour curve . at this phase , in fact , the colours of the global spectrum and the unreddened sn are similar ( see also fig . [ fig : windmccol ] ) and the derived extinction law is dominated by the differences displayed by the two spectra due to the time integration effects . of course , in a real case , this could be interpreted as the presence of a le but also with a deviation from the standard spectral evolution . irrespective of the physical reason , the apparent evolution shown by the derived extinction law is a clear diagnostic signaling that additional effects are present and that @xmath26 can not be estimated with this method . given these facts , a detailed analysis of extincted ia should be able to clarify whether this mechanism is indeed at work and whether it can explain the low values of @xmath26 derived by several authors . all known cases of les in type ia sne have revealed themselves through the gradual appearance of an additive flux , during the nebular phase , in the spectral region bluewards of 4500 , where one can easily recognize the imprints of the time integrated sn spectrum . therefore , they are instances of the case a we have described in sec . [ sec : fardust ] . the spectrum discussed by @xcite is a very good example of this behaviour ( see also figure [ fig : compialate ] here ) . moreover , due to the close wavelength coincidence of the dominating [ fe iii ] nebular feature with the most prominent bump of a le spectrum , the former line is broadened . besides this , a distinguishing spectral feature of a le is the clear presence of the ca ii h&k lines which , though , at the beginning of the transition phase can be masked by the presence of a similar structure observed in pure nebular spectra at about 4000 . in general it is fairly easy to recognize the case a le contamination on the basis of the late spectra alone , especially if the optical depth is not too high , so that its spectrum is definitely blue , with @xmath1120 . in cases where the le forms in regions of high dust density , its spectrum can turn red without necessarily having the sn spectrum reddened . in this case , the le contamination is a bit more subtle and tends to leave unchanged the overall colour of the observed spectrum . however , the characteristic le bumps at about 5900 and 6500 ( see figure [ fig : evol ] ) appear quite clearly . an interesting fact that emerges from the analysis of the two known cases of sne 1991 t and 1998bu when compared to our synthetic le spectra is that clear differences are visible . as far as 1991 t is concerned , its le dominated spectrum taken at 729 days past maximum ( cfr . figure [ fig : sn91 t ] ) shows interesting features , which are directly related to its peculiar appearance during the maximum phase ( see for instance @xcite ) . in particular , we notice that the absence of the distinguishing ia feature ( i.e. the si ii 6150 ) during the pre maximum phase and its weakness in post maximum is reflected in the le spectrum . in fact , while our synthetic spectra , calculated using `` normal '' events like sne 1994d and 1998aq , do show a clear imprint of the p - cyg profile displayed by this line during the photospheric phase , this is practically absent in the data of 1991 t . another clear difference is that visible in correspondence of the caii h&k lines . again , the weakness of this feature in 1991 t is a consequence of what happened in the pre - maximum phase , when this line was totally absent ( @xcite ) . in the other case , the match between the synthetic and the observed le spectrum is fairly good but , still , the observed deviations can all be attributed to small differences in the spectral appearance of sn 1998bu with respect to sn 1994d ( see @xcite ) . all of this has an interesting consequence , i.e. that the maximum phase peculiarities can be easily seen in the le spectrum . therefore , not only one could classify an historical sn on the basis of its le spectrum alone ( disentangling for example between a type i and a type ii ) , but one could also get some more detailed information about its features during maximum light , so that it would be possible to distinguish , for instance , between a normal ia , a 1991t - like or a 1991bg - like event . this might have some relevance when traces of an historical sn explosion are found by detecting its les , as in the recent cases near the sn remnant of cas a @xcite and in lmc @xcite . provided that the le is bright enough to allow a spectroscopic observation , this would provide a powerful tool to type the sn . the detections by @xcite and @xcite lets us hope that more cases will be discovered in connection with sn remnants in our galaxy , offering a unique chance to directly classify the explosion that generated them and to establish a firm connection between the two . in principle one could envisage to use the case a le observations to derive some information about the extinction law in the host galaxy , simply comparing the sn time integrated spectrum and the le spectrum . in fact , under the conservative assumption that the le and the sn suffer the same extinction ( i.e. @xmath95=0 in equation ( [ eq : law1 ] ) ) , the ratio between @xmath93 and @xmath81 is reddening independent . unfortunately , even in these simplified circumstances , one is left with the albedo and forward scattering degree wavelength dependence . more in general , due to the way the effects of geometry , optical depth and dust properties combine together , one can not derive unambiguous information on the extinction law since very similar solutions can be obtained using different ingredients . some meaningful results might be achievable from spectroscopic observations of les if some of the dependencies can be removed by other and independent observations . of great help , in this respect , are high spatial resolution images , which can give direct information on the geometrical dust distribution . with the increased interest on type ia sne related to their usage in cosmology and the availability of larger and larger observing facilities , it is most likely that an increasing number of objects belonging to this class will be observed in the near future and followed up to even more advanced phases than it is currently done . our forecast is that these investigations will probably be disturbed by the presence of les , too faint to be detected by the typical current instrumentation . in fact , as we have shown here , a dust density of the order of @xmath41=5 @xmath42 is sufficient to produce a le compatible with those seen in sne 1991 t and 1998bu . due to the lack of le detections in the vast majority of the cases @xcite , in paper i we had concluded that the height scale of type ia must be significantly higher than that of the dust in the disk of host galaxies . indeed , a very similar conclusion was reached by @xcite on the basis of completely independent considerations . nevertheless , when one pushes the observations at later phases , due to the fast decay typical of these objects during the nebular phase ( @xmath295 mag yr@xmath113 ) , the light curve rapidly reaches levels where a le produced by a much lower densities would dominate the spectrum . for example , for the same geometry used in figure [ fig : lcsheet ] , a density of @xmath41=0.5 @xmath42 would produce a le approximately 2.5 mag fainter , implying that the transition to a le dominated spectrum would have started only 6 months later . therefore , the study of type ia at phases later than 1.5 - 2 yr , aimed for example at detecting deviations from the radioactive decay , will be most probably hindered by the emergence of a case a le . if , on the one side , it is true that there is an increasing number of instances where les of this kind are actually observed ( see also the recent case reported by @xcite and @xcite for the core - collapse sn 2003gd ) , one may in fact wonder whether and to which extent early phases could be affected , modifying for instance the light curve shape , the overall luminosity and introducing peculiarities in the spectral appearance . as we have shown in this paper , this can actually take place when the dust cloud is very close to the sn and has small dimensions , circumstances that we have generically indicated as case b. as it has been independently shown by @xcite , circumstellar dust can in principle alter the observed light and colour curves producing , for example , interesting effects on the derived extinction curve . our analysis ( see secs . [ sec : closedust ] and [ sec : caseb ] ) , run under the simplifying hypothesis that the dust is distributed according to a @xmath114 law with an inner cavity with radius @xmath66 , shows that already with moderate optical depths ( @xmath43=1 ) , the presence of dust in the immediate vicinity of a type ia ( @xmath11510@xmath116 cm ) would turn into significant changes in the photometric parameters , which would still be within the normally observed ranges . in particular , the post maximum decline rate @xmath117 would be smaller and the sn would also appear inherently brighter , due to the additional flux contribution added by the underlying le , i.e. producing an additional scatter in the pskovski - phillips relation . in this respect , we notice that there are many instances of type ia showing low decline rates ( @xmath1181 ) without displaying the spectral peculiarities typical of slow - declining sn 1991t - like events ( see for example @xcite , table 2 ) . besides influencing the photometric parameters , the time integration during the fast evolving phase around maximum light tends to produce changes in the line widths and peaks , altering the measured expansion velocities . as we have said , according to our simulations the deviations from a le - free standard behaviour would be rather subtle and could be easily misidentified as the intrinsic features of a peculiar object . whilst most of the photometric parameters tend to be mildly altered ( at least for relatively low values of the dust optical depth ) the colour evolution shows quite a clear deviation from the homogeneous behaviour shown by type ia between 30 and 90 days past maximum light @xcite . if indeed all le - free objects follow that path , then the colour curve could be a good diagnostic tool to unveil the presence of a le . as pointed out by @xcite , the presence of case b les could explain the low values for @xmath26 which have been deduced from sn observations ( see for example @xcite , @xcite , @xcite , @xcite and @xcite for the most recent examples ) . as we have shown here , this is in principle true , even though some side effects should be clearly detectable , at least for the simple dust distributions we have analyzed . according to the lower panel of fig . [ fig : windmccol ] , or to the equivalent fig . 3 of @xcite , one would expect that the deduced @xmath26 remains roughly constant ( @xmath291 ) in this phase range . nevertheless , in those plots @xmath105 is kept at the constant value attained at maximum light , while the difference in colour between the global spectrum and the sn spectrum changes with time , both in our monte carlo calculations and in the semi - analytical model presented by @xcite . in particular , when the global colour becomes equal to or even bluer than that of the unreddened sn , the extinction correction is expected to produce weird results . therefore , if one attributes the deviations from the standard colours to a different extinction law , she also has to accept the puzzling conclusion that this has to change with time . we reckon that both the colour evolution and the time variability of the derived extinction law would be sufficient to infer that the observed deviations are not due to a peculiar dust composition but rather to a le . another interesting fact worth to be mentioned is the discrepancy that one would find between the extinction derived from the equivalent width of the nai d lines and that deduced from photometric or spectroscopic comparisons with templates . in fact , while the intensity of the nai d absorption is related only to the dust present along the line of sight ( and would hence give a direct estimate of @xmath27 ) , all other methods would be affected by the le contribution . as we have seen in sec . [ sec : caseb ] , this tends to significantly decrease the amount of derived extinction . in the case of normal dust - to - gas ratios and for a simple @xmath114 spherically symmetric dust geometry , this would imply that the values deduced from the nai doublet are systematically higher than those inferred from other methods ( compare @xmath48 with @xmath44 or @xmath57 in table [ tab : models ] ) . in all this discussion we have assumed ( and required ) that there can be dust very close to the sn . in this respect there is a very important consideration one has to make . in fact , if on the one hand it is true that le effects become more subtle when the dust is closer to the sn ( and can therefore be mistaken with intrinsic properties ) , on the other there actually might be a lower limit for the sn - dust distance . in fact , for typical expansion velocity of 10@xmath119 km s@xmath113 , the ejecta would reach @xmath66=10@xmath116 cm in less than four months . this is a problem that needs to be addressed with detailed modeling , in order to see whether the interaction would be able to destroy and/or sweep away the dust and to evaluate possible effects on the observed reddening , its time evolution and so on . one should also mention that the radiation field produced by the sn explosion would probably evaporate the dust up to a certain radius . if in the case of a core - collapse event the uv flash is expected to produce a cavity in the dust with a radius of @xmath2910@xmath64 cm ( see for example @xcite ) , the lack of uv photons in a type ia might still make the dust survive the explosion . another implication that needs to be addressed is related to the fact that substantial amounts of dust should be associated with correspondingly high amounts of gas . if the gas is close enough to the sn , then some interaction is expected , like in the case of sn 2002ic @xcite . finally , what remains to be studied is the effect of asymmetric dust distributions ( disks , tori , lobes and so on ) that , among other things , would certainly produce a net continuum polarization up to several per cent ( see paper i ) . being close to the sn the polarized flux could be in fact high enough to produce detectable effects and serve as a useful diagnostic tool to identify the presence of an underlying case b le . finally , to make the scenario fully self - consistent , one should also investigate the effects of dust heating on ir re - emission , which might turn out to be another observable side - effect produced by this kind of les . detailed analysis of reddened objects will certainly clarify whether case b les are indeed taking place and this , in turn , will probably give us a few more hints about the nature of type ia progenitors . we are currently working on the highly reddened sn 2003cg @xcite , which shows clear features that could be interpreted as the signature of a case b le . the results will be reported in another paper of this series . in this paper we have analyzed and discussed the effects of les on light curves , colours and spectra of type ia sne for two different dust geometries : extended clouds ( case a ) and small circumstellar clouds ( case b ) . the main results can be summarized as follows : 1 . the effect of case a les is limited to the late phases . the exact time when the le contribution dominates the global luminosity depends on dust distance and density . 2 . there is a transition phase , during which the les contribution becomes more and more important and the global spectrum turns from a normal nebular spectrum into a pure le spectrum . 3 . during the transition phase , the overall colour tends to become bluer , spectral features broader and emission line peaks displaced with respect to a pure sn nebular spectrum . 4 . during the le - dominated phase the spectrum retains the peculiarities shown by the sn at maximum , so that the le detection in an historical sn allows to classify it with a good confidence level . 5 . the spectroscopic observation of case a les in the vicinity of sn remnants can help in establishisng a firm and direct relation between the sn type and the remnant . the information on the extinction properties one can derive from les ( both case a and b ) observations may lead to incorrect conclusions on the dust nature . case b les affect in principle all phases , without necessarily producing notable effects on the spectra . this is particularly true if the dust is confined within @xmath2910@xmath116 cm from the sn explosion and @xmath1201 . 8 . in the presence of circumstellar dust , the light curves tend to become broader with a smaller @xmath79 . the rising time increases , the late time decay becomes slower and the luminosity difference between maximum light and nebular phase tends to decrease . the sn absolute luminosity deduced from the apparent magnitude , colour excess and a standard extinction law is brighter than the real value . the difference in luminosity is of the same order of the correction foreseen by the pskowski - phillips relation . this , coupled to the fact that @xmath117 is decreased , is a possible source of noise in the luminosity - decline rate relation . 11 . as shown by @xcite , case b les modify the luminosity and colours in such a way that the apparent selective - to - total extinction ratio @xmath26 is significantly decreased . 12 . in this scenario , @xmath26 should show a rapid evolution during the first months after explosion and this , together with a non standard colour evolution , should be a simple diagnostic tool to unveil the underlying le . altavilla , g. et al . 2004 , mnras , 349 , 1344 benetti , s. et al . 2004 , mnras , 348 , 261 benetti , s. et al . 2005 , apj , 623 , 1011 boffi f.r . , sparks w.b . , macchetto f.d . , 1999 , a&as , 138 , 253 branch d. et al . 2003 , aj , 126 , 1489 bongard , s. , baron , e. , smadja , g. , branch , d. , hauschildt , p.h . , 2005 , apj , submitted , astro - ph/0512229 bowers e.j.c . et al . , 1997 , mnras , 290 , 663 cappellaro e. et al . , 2001 , apj , 549 , l215 chevalier , r. , 1986 , apj , 308,225 crotts , a.p.s . , kunkel , w.e . & mccarthy , p.j . 1989 , apj , 347 , l61 della valle m. , panagia n. , 1992 , aj , 104 , 696 draine b.t . , 2003 , apj , 598 , 1017 elias - rosa , n. et al . , 2005 , mnras , in preparation filippenko a.v . et al . , 1992 , apj , 384 , l15 garnavich p. et al . , 2001 , aas , 199 , 4701 g filippenko a.v . et al . , 1992 , apj , 384 , l15 hachingher s. , mazzali p.a . & benetti s. , 2005 , mnras , submitted hamuy m. et al . , 2003 , nature , 424 , 651 hamuy m. et al . , 2003 , nature , 424 , 651 henyey l.c . , greenstein j.l . , 1941 , aj , 93 , 70 hernandez m. et al . , 2000 , mnras , 319 , 223 jha s. et al . , 1999 , apjs , 125 , 73 kirshner , r.p . & oke , j.b . , 1975 , apj , 200 , 574 kirshner , r.p . et al . , 1993 , apj , 415 , 589 knop r.a . et al . , 2003 , apj , 598 , 102 liu j.f . , bregman j.n . , seitzer p. , 2003 , apj , 582 , 919 kozma c. , fransson c. , hillebrandt w. , travaglio c. , sollerman j. , reinecke m. , rpke f.k . , spyromilio j. , 2005 , a&a , 437 , 983 krause o. , et al . , 2005 , science , 308 , 1604 krisciunas k. , 2000 , apj , 539 , 658 krisciunas k. et al . , 2003 , aj , 125 , 166 krisciunas k. et al . , 2005 , astro - ph/0511162 leibundgut b. , 2001 , araa , 39 , 67 lira , p. , 1995 , m. s. thesis , university of chile mattila s. , lundqvist p. , sollerman j. , kozma c. , baron e. , fransson c. , leibundgut b. , nomoto k. , 2005 , a&a , 443 , 649 maund j.r . , smartt s.j . 2005 , mnras , 360 , 288 nugent p. et al . , 1995 , apj , 455 , l147 patat f. , benetti s. , cappellaro e. , danziger i.j . , della valle m. , mazzali p.a . , turatto m. 1996 , mnras , 278 , 111 patat f. , 2005 , mnras , 357 , 1161 ( * paper i * ) phillips m.m . , lira p. , suntzeff n.b . , schommer r.a . , hamuy m. , maza j. , 1999a , aj , 118 , 1766 rest a. et al . , 2005 , nature , in press , astro - ph/0510738 romaniello m. , patat f. , panagia n. , sparks w.b . , gilmozzi r. , spyromilio , j. , 2005 , apj , 629 , 250 salvo m.e . , cappellaro e. , mazzali p. , benetti s. , danziger i.j . , patat f. , turatto m. , 2001 , mnras , 321 , 254 schlegel d.j . , finkbeiner , d.p . & davis m. , 1998 , apj , 500 , 525 schmidt b.p . 1994 , apj , 434 , l19 sollerman j. et al . , 2004 , a&a , 428 , 555 sparks w.b . , 2004 , apj , 433 , 29 sparks w.b . , macchetto f. , panagia n. , boffi f.r . , branch , d. , hazen m.l . , della valle m. , 1999 , a&a , 428 , 555 spyromilio j. , gilmozzi r. , sollerman j. , leibundgut b. , fransson c. , cuby j.g . , 2004 , a&a , 426 , 547 sugerman , b.e.k . & crotts , a.p.s . , 2002 , apj,581 , l97 sugerman b.e.k . , 2003 , aj , 123 , 1939 sugerman b.e.k . , 2005 , apj , 632 , l17 suntzeff n.b . et al . , 1999 , aj , 117 , 1175 van dyk , s. , li , w. & filippenko , a. 2005 , pasp , in press , astro - ph/0508684 wang l. et al . , 2003 , apj , 591 , 1110 wang l. , 2005 , apj , in press , astro - ph/0511003

let be a closed embedding , with nonsingular . we define a constructible function on , agreeing with verdier s specialization of the constant function when is the zero - locus of a function on . our definition is given in terms of an embedded resolution of ; the independence on the choice of resolution is obtained as a consequence of the weak factorization theorem of . the main property of is a compatibility with the specialization of the chern class of the complement . with the definition adopted here , this is an easy consequence of standard intersection theory . it recovers verdier s result when is the zero - locus of a function on . our definition has a straightforward counterpart in a motivic group . the function and the corresponding chern class and motivic aspect all have natural ` monodromy ' decompositions , for for any as above . the definition also yields an expression for kai behrend s constructible function when applied to ( the singularity subscheme of ) the zero - locus of a function on .

1009.2483

[ [ section ] ] consider a family @xmath8 over the open disk , satisfying a suitable condition of local triviality over @xmath9 . in @xcite , verdier defines a ` specialization morphism ' for constructible functions , producing a function @xmath10 on the central fiber @xmath3 of the family for every constructible function @xmath11 on @xmath5 . the key property of this specialization morphism is that it commutes with the construction of chern classes of constructible functions in the sense of macpherson ( @xcite ) ; cf . theorem 5.1 in verdier s note . the specialization morphism for constructible functions is induced from a morphism at the level of constructible sheaves @xmath12 , by taking alternating sums of ranks for the corresponding complex of nearby cycles @xmath13 . the main purpose of this note is to give a more direct description of the specialization morphism ( in the algebraic category , over algebraically closed fields of characteristic @xmath14 ) , purely in terms of constructible functions and of resolution of singularities , including an elementary proof of the basic compatibility relation with chern classes . we will assume that @xmath5 is nonsingular away from @xmath3 , and focus on the case of the specialization of constant functions ; by linearity and functoriality properties , this suffices in order to determine @xmath15 in the situation considered by verdier . on the other hand , the situation we consider is more general than the specialization template recalled above : we define a constructible function @xmath2 for _ every _ proper closed subscheme @xmath3 of a variety @xmath5 ( such that @xmath1 is nonsingular ) , which agrees with verdier s specialization of the constant function @xmath4 when @xmath3 is the fiber of a morphism from @xmath5 to a nonsingular curve . the definition of @xmath2 ( definition [ maindef ] ) is straightforward , and can be summarized as follows . let @xmath16 be a proper birational morphism such that @xmath17 is nonsingular , and @xmath18 is a divisor with normal crossings and nonsingular components , and for which @xmath19 is an isomorphism . then define @xmath20 to be @xmath21 if @xmath22 is on a single component of @xmath23 of multiplicity @xmath21 , and @xmath14 otherwise ; and let @xmath2 be the push - forward of @xmath24 to @xmath3 . readers who are familiar with verdier s paper @xcite should recognize that this construction is implicit in 5 of that paper , if @xmath3 is the zero - locus of a function on @xmath5 . our contribution is limited to the realization that the weak factorization theorem of @xcite may be used to adopt this prescription as a _ definition , _ that the properties of this function follow directly from the standard apparatus of intersection theory , and that this approach extends the theory beyond the specialization situation considered by verdier ( at least in the algebraic case ) . denoting by @xmath25 the chern - schwartz - macpherson class of a constructible function , we prove the following : * theorem i. * let @xmath26 be an effective cartier divisor . then @xmath27 an expression for @xmath28 in terms of the basic ingredients needed to define @xmath2 as above may be given as soon as @xmath29 is a regular embedding ( cf . remark [ regemb ] ) . in fact , with suitable positions , theorem i holds for arbitrary closed embeddings @xmath0 ( theorem [ spethm ] ) . theorem i reproduces verdier s result when @xmath3 is a fiber of a morphism from @xmath5 to a nonsingular curve ; in that case ( but not in general ) @xmath30 may be replaced with @xmath4 , as in verdier s note . the definition of @xmath2 is clearly compatible with smooth maps , and in particular the value of @xmath2 at a point @xmath22 may be computed after restricting to an open neighborhood of @xmath22 . thus , verdier s formula for the specialization function in terms of the euler characteristic of the intersection of a nearby fiber with a ball ( 4 in @xcite ) may be used to compute @xmath2 if @xmath3 is a divisor in @xmath5 , over @xmath31 . from the definition it is clear that the function @xmath2 is birationally invariant in the following weak sense : * theorem ii . * let @xmath32 be a proper birational morphism ; let @xmath33 , and assume that @xmath34 restricts to an isomorphism @xmath35 . then @xmath36 ( here , @xmath37 is the push - forward of constructible functions . ) in fact , the whole specialization _ morphism _ commutes with arbitrary proper maps , at least in verdier s specialization situation ( @xcite , corollary 3.6 ) . it would be desirable to establish this fact for the morphism induced by @xmath2 for arbitrary @xmath3 , by the methods used in this paper . the definition summarized above yields a natural decomposition of the constructible function @xmath38 ( and hence of its chern class @xmath28 ) according to the multiplicities of some of the exceptional divisors , see remarks [ distdecfun ] and [ distdeccla ] . in the specialization situation , this decomposition matches the one induced on the milnor fiber by monodromy , as follows from the description of the latter in @xcite . as schrmann pointed out to me , an analogous description in the more general case considered here may be found in @xcite , theorem 3.2 . [ [ section-1 ] ] in the basic specialization situation , in which @xmath3 is the zero - scheme of a function @xmath39 on @xmath5 and @xmath5 is nonsingular , let @xmath40 be the _ singularity subscheme _ of @xmath3 ( i.e. , the ` critical scheme ' of @xmath39 ) . one can define a constructible function @xmath41 on @xmath3 by @xmath42 in this case ( and over @xmath31 ) , the function @xmath2 agrees with verdier s specialization function @xmath43 ( here we use notation as in @xcite , cf . especially proposition 5.1 ) . the function @xmath41 is @xmath14 outside of @xmath40 , so may be viewed as a constructible function on @xmath40 . in fact , it has been observed ( cf . e.g. , @xcite ) that the function @xmath41 is a specific linear combination of local euler obstructions , and in particular it is determined by the scheme @xmath40 and can be generalized to arbitrary schemes . kai behrend denotes this generalization @xmath44 in @xcite . the definition of @xmath2 given in this paper yields an alternative computation of @xmath41 when @xmath40 is the critical scheme of a function , and theorem ii describes the behavior through modifications along @xmath3 of this function : if @xmath32 is as in theorem ii , then @xmath45 provided that @xmath5 and @xmath46 are nonsingular , @xmath3 and hence @xmath33 are hypersurfaces , and @xmath40 , @xmath47 are their singularity subschemes . as @xmath2 is defined for arbitrary @xmath0 , there may be a generalization of ( * ) linking behrend s function and @xmath48 when @xmath3 is not necessarily a hypersurface ; it would be interesting to have statements analogous to ( * * ) , holding for more general @xmath3 . in [ wmather ] we comment on the relation between @xmath7 and the ` weighted chern - mather class ' of the singularity subscheme @xmath40 of a hypersurface @xmath3 ; the degree of this class is a donaldson - thomas type invariant ( @xcite , 4.3 ) . we also provide an explicit formula for the function @xmath41 in terms of a resolution of the hypersurface @xmath3 . it would be interesting to extend these results to the non - hypersurface case . in a different vein , j. schrmann has considered the iteration of the specialization operator over a set of generators for a complete intersection @xmath0 ( @xcite , definition 3.6 ) . it would be a natural project to compare schrmann s definition ( which depends on the order of the generators ) with our definition of @xmath2 ( which is independent of the order , and may be extended to arbitrary @xmath0 ) . schrmann also points out that the _ deformation to the normal cone _ may be used to reduce an arbitrary @xmath0 to a specialization situation ; this strategy was introduced in @xcite , and is explained in detail in @xcite , 1 . again , it would be interesting to establish the precise relation between the resulting specialization morphism and the function @xmath2 studied here . [ [ section-2 ] ] we include in [ mot ] a brief discussion of a ` motivic ' invariant @xmath6 , also defined for any closed embedding @xmath0 into a variety , still assumed for simplicity to be nonsingular outside of @xmath3 . this invariant can be defined in the quotient of the grothendieck ring of varieties by the ideal generated by the class of a torus @xmath49 , or in a more refined relative ring over @xmath3 . the definition is again extremely simple , when given in terms of a resolution in which the inverse image of @xmath3 is a divisor with normal crossings ; the proof that the invariant is well - defined also follows from the weak factorization theorem . as its constructible function counterpart , the class @xmath6 admits a natural ` monodromy ' decomposition ( although a milnor fiber is not defined in general in the situation we consider ) , see remark [ distdecmot ] . when @xmath5 is nonsingular and @xmath3 is the zero - locus of a function on @xmath5 , @xmath6 is a poor man s version of the denef - loeser motivic milnor / nearby fiber ( @xcite , 3.5 ) ; it is defined in a much coarser ring , but it carries information concerning the topological euler characteristic and some other hodge - type data . we note that @xmath6 is _ not _ the image of the limit of the _ naive _ motivic zeta function of denef - loeser , since it does carry multiplicity information , while @xmath50 discards it ( see for example @xcite , corollary 3.3 ) . the limit of the ( non - naive ) denef - loeser motivic zeta function @xmath51 encodes the multiplicity and much more as actual monodromy information , and in this sense it lifts the information carried by our @xmath6 . it would be interesting to define and study an analogous lift for more general closed embeddings @xmath0 , and possibly allowing @xmath5 to be singular along @xmath3 . the approach of @xcite could be used to unify the constructions of @xmath2 and @xmath6 given in this paper , and likely extend them to other environments , but we will not pursue such generalizations here since our aim is to keep the discussion at the simplest possible level . likewise , ` celestial ' incarnations of the milnor fiber ( in the spirit of @xcite , @xcite ) will be discussed elsewhere . [ [ section-3 ] ] in our view , the main advantages of the approach taken in this paper are the simplicity afforded by the use of the weak factorization theorem and the fact that the results have a straightforward interpretation for any closed embedding @xmath0 , whether arising from a specialization situation or not . these results hold with identical proofs over any algebraically closed field of characteristic zero . we note that the paper @xcite of van proeyen and veys also deals with arbitrary closed embeddings with nonsingular complements , as in this note . a treatment of verdier specialization over arbitrary algebraically closed fields of characteristic zero , also using only the standard apparatus of intersection theory , was given by kennedy in @xcite by relying on the lagrangian viewpoint introduced by c. sabbah @xcite . fu ( @xcite ) gives a description of verdier s specialization in terms of normal currents . we were motivated to take a new look at verdier s specialization because of applications to string - theoretic identities ( cf . @xcite , 4 ) . also , verdier specialization offers an alternative approach to the main result of @xcite . the main reason to allow @xmath5 to have singularities along @xmath3 is that this typically is the case for specializations arising from pencils of hypersurfaces in a linear system , as in these applications . see [ pencils ] for a few simpler examples illustrating this point . [ [ section-4 ] ] i am indebted to m. marcolli for helpful conversations , and i thank j. schrmann and w. veys for comments on a previous version of this paper . our schemes are separated , of finite type over an algebraically closed field @xmath52 of characteristic @xmath14 . the characteristic restriction is due to the use of resolution of singularities and the main result of @xcite , as well as the theory of chern - schwartz - macpherson classes . @xcite and @xcite for discussions of the theory in this generality . ) [ maindef ] let @xmath5 be a variety , and let @xmath26 be a closed embedding . we assume that @xmath1 is nonsingular and not empty . the constructible function @xmath2 on @xmath3 is defined as follows . * let @xmath16 be a proper birational morphism ; let @xmath18 , and @xmath53 : @xmath54^j \ar[d]_d & w \ar[d]^w \\ x \ar@{^(->}[r]^i & v } \ ] ] we assume that @xmath17 is nonsingular , @xmath23 is a divisor with normal crossings and nonsingular components @xmath55 in @xmath17 , and @xmath56 restricts to an isomorphism @xmath57 . ( such a @xmath56 exists , by resolution of singularities . ) let @xmath58 be the multiplicity of @xmath55 in @xmath23 . * we define a constructible function @xmath24 on @xmath23 by letting @xmath59 for @xmath60 , @xmath61 ( @xmath62 ) , and @xmath63 for @xmath64 , any @xmath62 . * then let @xmath65 . we remind the reader that the push - forward of constructible functions is defined as follows . for any scheme @xmath66 , denote by @xmath67 the function with value @xmath68 along @xmath66 , and @xmath14 outside of @xmath66 . if @xmath66 is a subvariety of @xmath23 , and @xmath69 , @xmath70 equals @xmath71 . by linearity , this prescription defines @xmath72 for every constructible function @xmath11 on @xmath23 . here , @xmath43 denotes the topological euler characteristic for @xmath73 ; see @xcite or @xcite for the extension to algebraically closed fields of characteristic @xmath14 . of course we have to verify that the definition of @xmath2 given in definition [ maindef ] does not depend on the choice of @xmath74 . [ spepro ] with notation as above , the function @xmath2 is independent of the choice of @xmath16 . the weak factorization theorem of @xcite reduces the verification to the following fact . * let @xmath17 , @xmath23 , @xmath24 as above ; * let @xmath75 be the blow - up of @xmath17 along a center @xmath76 that meets @xmath23 with normal crossings , and let @xmath77 ; * then @xmath78 , where @xmath79 and @xmath24 are defined by the prescription for divisors with normal crossings given in definition [ maindef ] . recall that @xmath80 meets @xmath23 with normal crossings if at each point @xmath81 of @xmath80 there is an analytic system of parameters @xmath82 for @xmath23 at @xmath81 such that @xmath80 is given by @xmath83 , and @xmath23 is given by a monomial in the @xmath84 s . the divisor @xmath77 is then a divisor with normal crossings , cf . lemma 2.4 in @xcite . the divisor @xmath85 consists of the proper transforms @xmath86 of the components @xmath55 , appearing with the same multiplicity @xmath58 , and of the exceptional divisor @xmath87 , appearing with multiplicity @xmath88 . it is clear that @xmath89 agrees with @xmath24 away from @xmath80 ; we have to verify that the functions match at all @xmath90 . the fiber of @xmath91 over @xmath81 is a projective space of dimension @xmath92 ; we have to analyze the intersection of the rest of @xmath93 ( that is , of the proper transforms @xmath86 ) with this projective space . now there are two kinds of points @xmath90 : either @xmath81 is in exactly one component @xmath55 , or it is in the intersection of several components . in the first case , @xmath86 is the unique component of @xmath93 other than @xmath87 meeting the fiber @xmath94 of @xmath87 over @xmath81 . by definition , @xmath95 on the complement of @xmath96 in @xmath86 , and @xmath97 along @xmath96 . thus , @xmath98 equals @xmath99 since @xmath100 and @xmath101 . this is as it should . if @xmath81 is in the intersection of two or more components @xmath55 , then @xmath102 , so we have to verify that @xmath103 . again there are two possibilities : either one of the components containing @xmath81 does not contain @xmath80 , and then the whole fiber @xmath100 is contained in the proper transform of that component , as well as in @xmath87 ; or all the components @xmath55 containing @xmath81 contain @xmath80 . in the first case , the value of @xmath79 is zero on the whole fiber @xmath94 , because @xmath94 is in the intersection of two components of @xmath85 ; so the equality is clear in this case . in the second case , let @xmath104 be the components of @xmath23 containing @xmath80 ; no other component of @xmath23 contains @xmath81 , by assumption . the proper transforms @xmath86 meet the fiber @xmath105 along @xmath106 hyperplanes meeting with normal crossings , @xmath107 . the value of @xmath79 along @xmath94 is then @xmath108 along the complement @xmath109 of these hyperplanes , and @xmath14 along these hyperplanes . thus , @xmath110 the proof will be complete if we show that @xmath111 ; and this is done in the elementary lemma that follows . [ elemlem ] let @xmath112 be hyperplanes in @xmath113 meeting with normal crossings , with @xmath114 , and let @xmath115 . then @xmath116 for @xmath117 , and @xmath111 for @xmath118 . the hyperplanes may be assumed to be coordinate hyperplanes , and hence @xmath109 may be described as the set of @xmath119 such that the first @xmath106 coordinates are nonzero . as the first coordinate is nonzero , we may set it to be @xmath68 , and view the rest as affine coordinates . it is then clear that @xmath120 , where @xmath121 is the @xmath68-dimensional torus . the statement is then clear , since @xmath122 . by definition , the value of @xmath2 at a point @xmath123 is @xmath124 where @xmath125 . in the complex hypersurface case , this equals the euler characteristic @xmath126 of the milnor fiber , by formula ( 2 ) in theorem 1 of @xcite . an analogous interpretation holds in general , as may be established by using theorem 3.2 in @xcite . [ distdecfun ] let @xmath127 be any function . the argument proving lemma [ spepro ] shows that one may define a constructible function @xmath128 on @xmath3 as the push - forward of the function @xmath129 with value @xmath130 on @xmath131 and @xmath14 on intersections , for @xmath23 and @xmath17 as in definition [ maindef ] . for @xmath132 identity we do not have an interpretation for @xmath128 ( or for the corresponding chern class , cf . remark [ distdeccla ] ) . letting @xmath133 be the function that is @xmath68 at @xmath134 and @xmath14 at all other integers , we have a decomposition of the identity as @xmath135 , and hence a distinguished decomposition @xmath136 the individual terms in this decomposition are clearly preserved by the morphisms considered in propositions [ birinv ] and [ smooth ] . in the hypersurface case they can be related to the monodromy action on the milnor fiber , as the nonzero contributions arise from the nonzero eigenspaces of monodromy ( cf . @xcite , theorem 4 ; and theorem 3.2 in @xcite for generalizations to the non - hypersurface case ) . from this perspective , the piece @xmath137 may be thought of as the ` unipotent part ' of @xmath2 . theorem ii from the introduction is an immediate consequence of the definition of @xmath2 : [ birinv ] let @xmath46 be a variety , and let @xmath32 be a proper morphism . let @xmath33 , and assume that @xmath34 restricts to an isomorphism @xmath35 . then @xmath138 indeed , a resolution for the pair @xmath139 as in definition [ maindef ] is also a resolution for the pair @xmath140 . another immediate consequence of the definition is the behavior with respect to smooth maps : [ smooth ] let @xmath109 be a variety , and let @xmath141 be a smooth morphism . then @xmath142 ( pull - backs of constructible functions are defined by composition . ) indeed , in this case one can construct compatible resolutions . for example , the value of @xmath2 at a point @xmath123 may be computed by restricting to an open neighborhood of @xmath22 . [ sci ] if @xmath3 and @xmath5 are both nonsingular varieties , then @xmath143 . indeed , if @xmath3 and @xmath5 are nonsingular , then we can let @xmath144 be the blow - up of @xmath5 along @xmath3 ; @xmath23 is the exceptional divisor , and the push - forward of @xmath145 equals @xmath146 because the fibers of @xmath147 are projective spaces @xmath148 . [ nonred ] the function @xmath2 depends on the scheme structure on @xmath3 . for example , let @xmath149 be the scheme defined by @xmath150 , consisting of a line @xmath151 with an embedded point at @xmath152 . then @xmath153 equals @xmath68 along @xmath154 , while @xmath155 . ( blow - up at @xmath22 , then apply definition [ maindef ] . in terms of the decomposition in remark [ distdecfun ] , @xmath153 is the sum of @xmath156 and @xmath157 . ) thus , @xmath158 , while of course @xmath159 . let @xmath3 be the reduced scheme supported on the union of three non - coplanar concurrent lines in @xmath160 ; for example , we may take @xmath161 as a defining ideal . then @xmath162 along the three components , and @xmath163 at the point of intersection @xmath22 . to see this , blow - up @xmath160 at @xmath22 first , and then along the proper transforms of the three lines , to produce a morphism @xmath164 as in definition [ maindef ] ; @xmath18 consists of @xmath165 components , one of which dominates @xmath22 . this component is a @xmath166 blown up at three points ; the complement in this component of the intersection with the three other exceptional divisors has euler characteristic @xmath14 , so the push - forward of @xmath24 equals @xmath14 at @xmath22 . if the three concurrent lines are coplanar , say with ideal @xmath167 , then the value of @xmath168 at the intersection point is @xmath169 . that the value of the function is @xmath170 away from the point of intersection is clear _ a priori _ in both cases by example [ sci ] , since as observed above we may restrict to an open set avoiding the singularity and use example [ sci ] . definition [ maindef ] has a counterpart in a quotient of the naive grothendieck group of varieties @xmath171 . recall that this is the group generated by isomorphism classes of @xmath52-varieties , modulo the relations @xmath172=[s - t]+[t]$ ] for every closed embedding @xmath173 ; setting @xmath174\cdot [ s_2 ] = [ s_1\times s_2]$ ] makes @xmath171 into a ring . denote by @xmath175 the class of @xmath176 in @xmath171 , and by @xmath177 the class of the multiplicative group of @xmath52 . we let @xmath178 denote the quotient @xmath179 . every pair @xmath0 as in definition [ maindef ] ( that is , with @xmath5 a variety and @xmath1 nonsingular ) determines a well - defined element @xmath6 of @xmath178 , as follows : * let @xmath180 be as in definition [ maindef ] ; * for every component @xmath55 of @xmath23 , let @xmath181 be the complement @xmath182 ; * then @xmath183 $ ] , where @xmath58 is the multiplicity of @xmath55 in @xmath23 . the argument given in lemma [ spepro ] proves that @xmath6 is well - defined in the quotient @xmath178 . indeed , the argument applies word - for - word to show that the class of the fiber of a blow - up over @xmath90 agrees mod @xmath184 with the class of @xmath81 ; lemma [ elemlem ] is replaced by the analogous result in @xmath178 : [ elemlem2 ] let @xmath112 be hyperplanes in @xmath113 meeting with normal crossings , with @xmath114 , and let @xmath115 . then @xmath185=1\in{{\mathscr m}}^{{\mathbb{t}}}$ ] for @xmath117 , and @xmath185=0\in{{\mathscr m}}^{{\mathbb{t}}}$ ] for @xmath118 . the proof of lemma [ elemlem ] implies this statement , as it shows that @xmath120 . it is also evident that @xmath6 satisfies the analogues of propositions [ birinv ] and [ smooth ] . since @xmath122 , the information carried by an element of @xmath178 suffices to compute topological euler characteristics , but is considerably more refined : for example , the series obtained by setting @xmath186 in the hodge - deligne polynomial can be recovered from the class in @xmath178 . let @xmath187 be a cone over a smooth plane curve @xmath188 of degree @xmath21 . blowing up at the vertex @xmath22 yields a resolution as needed in definition [ maindef ] ; the function @xmath168 is immediately computed to be @xmath68 outside of the vertex and @xmath189 at @xmath22 ; the class @xmath190 equals @xmath191+m\cdot(3-[c])\in { { \mathscr m}}^{{\mathbb{t}}}$ ] . for @xmath192 , this is not the class of a constant . if @xmath3 is the scheme of zeros of a nonzero function @xmath193 , and @xmath5 is nonsingular , then a _ motivic milnor fiber _ @xmath194 was defined and studied by j. denef and f. loeser , see @xcite , 3 . the motivic milnor fiber @xmath194 is defined in a ring @xmath195 analogous to the ring @xmath171 considered above , but localized at @xmath175 and including monodromy information . the much naiver @xmath6 generalizes to arbitrary pairs @xmath0 ( with @xmath1 nonsingular ) a small part of the information carried by the denef - loeser motivic milnor fiber in the specialization case . it is probably preferable to work in the _ relative _ grothendieck ring of varieties over @xmath3 , mod - ing out by classes of varieties @xmath196 which fiber in tori . lemma [ elemlem2 ] shows that the resulting class @xmath197 is also well - defined . for @xmath123 , the fiber of @xmath197 over @xmath22 is well - defined as a class in @xmath178 , so it has an euler characteristic , which clearly equals @xmath198 . thus , the constructible function @xmath2 may be recovered from the relative class @xmath197 . this point of view is developed fully for chern classes and more in @xcite , @xcite . @xcite , @xcite for the relative viewpoint on the denef - loeser motivic milnor fiber . ) [ distdecmot ] there is a class @xmath199 ( resp . , @xmath200 ) for every function @xmath201 , defined by @xmath202 $ ] on a resolution . in particular , there is a decomposition @xmath203 with @xmath133 as in remark [ distdecfun ] . as observed in that remark , in the hypersurface case this decomposition can be related with the monodromy decomposition . if @xmath3 is the zero scheme of a function on @xmath5 , and @xmath5 is nonsingular , then the limit of the _ naive motivic zeta function _ @xmath204 of denef - loeser ( @xcite , corollary 3.3 ) lifts @xmath205 , where @xmath206 is the constant function @xmath68 . in fact , the corresponding expression @xmath207\ ] ] ( where @xmath208 , and @xmath209 ) may be verified to be a well - defined element of @xmath171 for any @xmath0 such that @xmath1 is nonsingular , and with @xmath23 as in definition [ maindef ] . this also follows from the weak factorization theorem ; no localization is necessary . also , this expression is clearly preserved by morphisms which restrict to the identity on @xmath1 ( as in theorem ii ) . thus , @xmath205 admits a natural lift to @xmath171 in general . we do not know whether @xmath6 itself admits a natural lift to @xmath171 recall ( @xcite , @xcite , @xcite ) that every constructible function @xmath11 determines a class in the chow group , satisfying good functoriality properties and the normalization restriction of agreeing with the total chern class of the tangent bundle if applied to the constant function @xmath210 over a nonsingular variety . we call this function the _ chern - schwartz - macpherson _ ( csm ) class of @xmath11 , @xmath211 . the functoriality may be expressed as follows : if @xmath212 is a proper morphism , and @xmath11 is a constructible function on @xmath3 , then @xmath213 . the simplest instance of this property is the fact that for every complete variety @xmath3 ( nonsingular or otherwise ) , the degree of @xmath214 equals the euler characteristic of @xmath3 . we are particularly interested in the csm classes of the function @xmath2 defined in [ chidef ] , and of the function @xmath215 , with value @xmath14 on @xmath3 and @xmath68 along the complement of @xmath3 . in the situation considered by verdier ( and more generally when @xmath3 is e.g. , a cartier divisor in @xmath5 ) , there is a natural way to specialize classes defined on @xmath5 or @xmath1 to @xmath3 . we consider the following definition of the specialization of a specific class , for arbitrary @xmath0 . [ specdefspec ] let @xmath5 be a variety , @xmath0 a closed embedding , and assume that @xmath1 is nonsingular . let @xmath216 be the blow - up of @xmath5 along @xmath3 , and let @xmath217 be the exceptional divisor , @xmath218 the induced map . we define the ` specialization of @xmath219 to @xmath3 ' to be the class @xmath220 the blow - up @xmath221 in definition [ specdefspec ] may be replaced with any proper birational morphism @xmath222 dominating the blow - up and restricting to an isomorphism on @xmath223 ; this follows easily from the projection formula and the functoriality of csm classes . when @xmath3 is a cartier divisor in @xmath5 , this specialization is the ordinary pull - back , and could be defined for arbitrary classes in @xmath5 . [ spediv ] let @xmath224 be a cartier divisor . then @xmath225 . in this case @xmath226 , @xmath227 , and @xmath106 is the identity . if @xmath0 is _ not _ a cartier divisor , it is not clear how to extend definition [ specdefspec ] to arbitrary classes ; but this is not needed for the results in this paper . by lemma [ spediv ] , theorem i from the introduction follows from the following result . [ spethm ] with notation as above , @xmath228 first consider the case in which @xmath229 is nonsingular , and @xmath230 is a divisor with normal crossings and nonsingular components @xmath55 , appearing with multiplicity @xmath58 . let @xmath231 be the complement of @xmath232 in @xmath55 . by definition of @xmath24 , and by linearity of the @xmath233 operator , @xmath234 the key to the computation is the fact that the csm class of the complement of a divisor with normal crossings in a nonsingular variety is given by the chern class of a corresponding logarithmic twist of the tangent bundle ; for this , see e.g. ( * ) at the top of p. 4002 in @xcite . as @xmath181 is the complement of @xmath235 , a normal crossing divisor in @xmath55 , we have ( omitting evident pull - backs ) @xmath236 = \frac{c(tw)}{\prod_k ( 1+d_k)}\cap [ d_\ell]\quad.\ ] ] therefore @xmath237 = \frac{c(tw)}{\prod_k ( 1+d_k)}\cap [ d]\quad.\ ] ] since @xmath238 $ ] , this shows that @xmath239 where @xmath240 is the inclusion . this is the statement in the normal crossing case . now assume @xmath0 is any closed embedding , and apply ( * ) to @xmath23 , @xmath17 as in definition [ maindef ] , observing that any such @xmath144 must factor through the blow - up @xmath241 along @xmath3 . let @xmath242 , @xmath243 be the induced morphisms , so that @xmath244 , @xmath245 : @xmath246^j \ar[ld]_{d ' } \ar@{-}[d ] & & w \ar[ld]_(.6){w ' } \ar[dd]^(.6)w \\ e \ar[rr]^(.7)\iota \ar[dr]_(.4)e & \ar[d]^(.4)d & { { \widetilde{v } } } \ar[dr]_(.3){\tilde v } \\ & x \ar[rr]^(.4)i & & v } \ ] ] by the covariance of @xmath233 , we have @xmath247 where @xmath248 holds by theorem 6.2 in @xcite ( note that @xmath249 as both @xmath23 and @xmath87 are cartier divisors ) . we have @xmath250 since @xmath251 is an isomorphism off @xmath23 . finally , using proposition [ birinv ] and again the covariance of @xmath233 : @xmath252 as claimed . let @xmath253 be the line @xmath151 with embedded point @xmath22 considered in example [ nonred ] . by theorem [ spethm ] , @xmath254 + 3[p]\quad.\ ] ] ( this may of course be verified by hand using definition [ specdefspec ] . ) here , @xmath255 + 2[{{\mathbb{p}}}^1]+[{{\mathbb{p}}}^0]$ ] . restricting this class to @xmath151 gives @xmath256 + 2[p]$ ] ; the specialization @xmath257 picks up an extra @xmath258 $ ] due to the embedded point . [ regemb ] by lemma [ spediv ] , if @xmath259 embeds @xmath3 as a cartier divisor in @xmath5 , then @xmath260 acts as the pull - back to @xmath3 , so that in this case theorem [ spethm ] states that @xmath27 this is not the case if @xmath259 is a regular embedding of higher codimension . if the embedding is regular , then the exceptional divisor @xmath261 may be identified with the projective normal bundle to @xmath3 , and as such it has a universal quotient bundle @xmath262 . tracing the argument proving theorem [ spethm ] shows that @xmath263 and in fact @xmath264 for any @xmath17 , @xmath23 , etc . as in definition [ maindef ] . when @xmath0 is a regular embedding , so that the blow - up morphism @xmath265 is a local complete intersection morphism ( cf . 6.7 in @xcite ) one may be tempted to define a specialization of classes from @xmath5 to @xmath3 by pulling back to the blow - up , restricting to the exceptional divisor , and pushing forward to @xmath3 . however , the pull - back of @xmath219 may not agree with @xmath266 : the blow - up of @xmath267 at @xmath268 a point already gives a counterexample . ( corollary 4.4 in @xcite provides a condition under which a similar pull - back formula does hold . ) thus , this operation does not agree in general with the class defined in definition [ specdefspec ] . [ distdeccla ] since @xmath2 admits a distinguished decomposition @xmath269 ( see remark [ distdecfun ] ) , so does the specialization of @xmath219 : @xmath270 in the classical case of specializations along the zero set of a function , this means that every eigenspace of monodromy carries a well - defined piece of the chern class , a fact we find intriguing , and which we are not sure how to interpret for more general @xmath140 . in fact , any @xmath127 determines a class @xmath271 ; theorem [ spethm ] gives an interpretation of this class for @xmath272 identity . it would be interesting to interpret this class for more general @xmath273 . to interpret theorem [ spethm ] in terms of standard specializations , consider the template situation in which @xmath5 fibers over a nonsingular curve @xmath274 , and @xmath3 is the fiber over a marked point @xmath275 : @xmath276 \ar[d ] & v \ar[d]^v \\ \{0\ } \ar@{^(->}[r ] & t } \ ] ] further assume for simplicity that the restriction of @xmath277 to @xmath1 is a trivial fibration : @xmath278 with nonsingular ( ` general ' ) fiber @xmath279 . ( in particular , @xmath1 is nonsingular , as in [ basicca ] . ) here @xmath3 is a cartier divisor in @xmath5 , so the specialization of definition [ specdefspec ] agrees with ordinary pull - back ( lemma [ spediv ] ) . we can use this pull - back to specialize classes from the general fiber @xmath279 to the special fiber @xmath280 , as follows . the pull - back @xmath281 factors through @xmath282 via the basic exact sequence of proposition 1.8 in @xcite : indeed , for @xmath283 , @xmath284 as @xmath285 is trivial . we may then define a specialization morphism by composing this morphism with flat pull - back : @xmath286 [ specialca ] in the specialization situation just described , @xmath287)\quad.\ ] ] since the normal bundle of @xmath279 in @xmath1 is trivial , and abusing notation slightly , @xmath288=c(t(v\smallsetminus x)\cap [ v_t])$ ] . hence , by definition of @xmath15 we have @xmath289)=i^*(\tau)\quad,\ ] ] where @xmath290 is any class whose restriction to @xmath282 equals @xmath291)$ ] . we claim that @xmath292 is such a class ; it then follows that @xmath289)=i^*({{c_{\text{sm}}}}({1\hskip-3.5pt1}_{v\smallsetminus x } ) ) = \sigma_{x , v}({{c_{\text{sm}}}}({1\hskip-3.5pt1}_{v\smallsetminus x}))={{c_{\text{sm}}}}({{\psi}}_{x , v } ) \quad,\ ] ] by theorem [ spethm ] ( and lemma [ spediv ] ) . our claim is essentially immediate if @xmath5 itself is nonsingular , but as we are only assuming @xmath1 to be nonsingular we have to do a bit of work . again consider a resolution @xmath16 as in definition [ maindef ] . let @xmath293 be the restriction ( an isomorphism by hypothesis ) . also , let @xmath294 and @xmath295 be the open inclusions . thus we have the fiber diagram @xmath296 ^ -{j ' } \ar[d]_{w ' } & w \ar[d]^w \\ v\smallsetminus x \ar@{^(->}[r]^-{i ' } & v } \ ] ] with @xmath297 and @xmath298 flat , and @xmath56 , @xmath251 proper . we have @xmath299 & = w ' _ * c(t(w\smallsetminus d))\cap [ w\smallsetminus d ] \\ & \overset{(1)}= w ' _ * { j'}^ * c(tw(-\log d))\cap [ w ] \\ & \overset{(2)}= { i'}^ * w _ * c(tw(-\log d))\cap [ w ] \\ & \overset{(3)}= { i'}^ * w _ * { { c_{\text{sm}}}}({1\hskip-3.5pt1}_{w\smallsetminus d } ) \\ & = { i'}^ * { { c_{\text{sm}}}}({1\hskip-3.5pt1}_{v\smallsetminus x})\end{aligned}\ ] ] as claimed . equality ( 1 ) holds by definition of @xmath300 ; equality ( 2 ) follows from proposition 1.7 in @xcite ; equality ( 3 ) is again the computation of the csm class of the complement of a divisor with normal crossings mentioned in the proof of theorem [ spethm ] . summarizing , in the strong specialization situation detailed above , @xmath2 is the constructible function on the central fiber @xmath3 corresponding via macpherson s natural transformation to the specialization of the chern class of the general fiber . [ spegen ] the hypothesis of triviality of the family away from @xmath275 is not necessary , if specialization is interpreted appropriately . let @xmath301 be a morphism to a nonsingular curve , and let @xmath302 as above ; assume our blanket hypothesis that @xmath1 is nonsingular , but no more . as @xmath5 is nonsingular away from @xmath3 , we may assume @xmath277 is smooth , after replacing @xmath274 with a neighborhood of @xmath14 in @xmath274 . the fiber @xmath279 is then smooth for all @xmath303 in @xmath274 , and since @xmath304 is trivial , we get @xmath305=v_t\cdot ( c(t(v\smallsetminus x ) ) \cap [ v\smallsetminus x])\quad;\ ] ] since @xmath306 restricts to @xmath307 $ ] on @xmath1 , this gives @xmath308 this computation fails for @xmath309 , as the fiber @xmath3 over @xmath14 is ( possibly ) singular . however , the classes @xmath306 and @xmath219 have the same specialization to @xmath3 : indeed , their difference is supported on @xmath3 , and @xmath285 is trivial in the case considered here . thus , in this case theorem [ spethm ] gives @xmath310 comparing ( @xmath311 ) and ( @xmath312 ) , we may still view @xmath2 as the limit of the constant @xmath313 as @xmath314 . we can also consider specializations arising from maps @xmath301 , allowing @xmath274 to be a nonsingular variety of arbitrary dimension with a marked point @xmath14 , and @xmath302 a local complete intersection of codimension @xmath315 , with trivial normal bundle . we still assume @xmath1 to be nonsingular . using the formula in remark [ regemb ] , the specialization @xmath316 may be written as @xmath317 where @xmath17 , @xmath23 , @xmath147 are as in definition [ maindef ] . indeed , since @xmath285 is trivial , then @xmath318 note that in this situation @xmath319=[x]$ ] : indeed , the segre class of @xmath23 in @xmath17 pushes forward to the segre class of @xmath3 in @xmath5 , by the birational invariance of segre classes . as mentioned in the introduction , schrmann has considered specialization to a complete intersection , by iterating applications of verdier specialization ( @xcite , definition 3.6 ) . it would be interesting to establish a precise relation between schrmann s specialization and the formula given above . [ [ spesu ] ] pencils of hypersurfaces give rise to specializations , as follows . consider a pencil of hypersurfaces in a linear system in a nonsingular variety @xmath46 . let the pencil be defined by the equation @xmath320 where @xmath94 and @xmath321 are elements of the system , and @xmath322 . assume that for any @xmath323 in a neighborhood of @xmath14 , @xmath324 is nonsingular . we can interpret this datum as a specialization by letting @xmath5 be the correspondence @xmath325 this is endowed with a projection @xmath326 ; after removing a finite set of @xmath303 from @xmath176 ( that is , those @xmath303 for which the fiber @xmath327 is singular ) , we reach the standard situation considered in [ specenfi ] ( remark [ spegen ] ) : * @xmath274 is a nonsingular curve , and @xmath275 ; * @xmath301 is a surjective morphism , and @xmath302 ; * @xmath1 is nonsingular , and @xmath328 is smooth . note that @xmath5 may be singular along @xmath3 ; in fact , the singularities of @xmath5 are contained in the base locus of the pencil and in the singular locus of @xmath3 . as proven in [ mainpro ] , @xmath329 for @xmath2 as in definition [ maindef ] , and this class can be viewed as the limit as @xmath314 of the classes @xmath288 $ ] of the general fibers . for example , the degree @xmath330 equals @xmath331=\chi(v_t)$ ] , the euler characteristic of the general fiber . the following are immediate consequences of the definition . * let @xmath123 be a point at which @xmath3 ( and hence @xmath5 ) is nonsingular . then @xmath332 . * more generally , let @xmath22 be a point such that there exists a neighborhood @xmath109 of @xmath22 in @xmath5 such that @xmath109 is nonsingular , and @xmath333 is a normal crossing divisor in @xmath334 . then @xmath335 if @xmath22 is in the intersection of two or more components of @xmath3 , and @xmath336 if @xmath22 is in a single component of @xmath3 , of multiplicity @xmath21 . since @xmath332 at nonsingular points of @xmath3 , @xmath337 is supported on the singular locus of @xmath3 . this function has a compelling interpretation , see ( * ) in the introduction and [ wmather ] . the following example illustrates a typical situation . [ milnorno ] let @xmath3 be a reduced hypersurface with isolated singularities in a nonsingular variety @xmath46 , and assume a general element @xmath279 of the linear system of @xmath3 is nonsingular , and avoids the singularities of @xmath3 . then @xmath338 equals the milnor number @xmath339 of @xmath3 at @xmath22 . indeed , as the matter is local , after a resolution we may assume that @xmath3 is complete and @xmath22 is its only singularity . consider the pencil through @xmath3 and a general element @xmath279 of its linear system . then by linearity of @xmath233 , and since @xmath337 is @xmath14 away from @xmath22 , @xmath340 now @xmath341 ( as recalled in [ basicca ] ) , and @xmath342 as observed above . thus , @xmath343 and this is well - known to equal the milnor number of @xmath3 at @xmath22 ( see e.g. , @xcite , corollary 1.7 ) . the above formula is a particular case of the formula @xmath344 mentioned in the introduction , and discussed further in [ wmather ] . near points of @xmath3 away from the base locus , @xmath5 is trivially isomorphic to @xmath46 , hence it is itself nonsingular ; the specialization function is then computed directly by an embedded resolution of @xmath3 . along the base locus , @xmath5 itself may be singular , and it will usually be necessary to resolve @xmath5 first in order to apply definition [ maindef ] ( or proposition [ birinv ] ) . in the following subsections we illustrate this process in a few simple examples , for pencils of curves . let @xmath3 be a curve with an ordinary multiple point @xmath22 of multiplicity @xmath21 , and assume that this point is not in the base locus of the pencil . as pointed out above , @xmath198 may be computed by considering the embedded resolution of @xmath3 over @xmath22 . this may be schematically represented as : we have one exceptional divisor , a @xmath345 , meeting the proper transform of the curve @xmath21 times and appearing with multiplicity @xmath21 . as the euler characteristic of the complement of the @xmath21 points of intersection in the exceptional divisor is @xmath346 , we have that @xmath347 . more generally , let @xmath3 be a plane curve with an isolated singular point @xmath22 , again assumed to be away from the base locus of the pencil . if the embedded resolution of @xmath3 has exceptional divisors @xmath348 over @xmath22 , @xmath348 appears with multiplicity @xmath349 , and meets the rest of the full transform of @xmath3 at @xmath350 points , then @xmath351 indeed , each @xmath348 is a copy of @xmath345 , and the euler characteristic of the complement of @xmath350 points in @xmath345 is @xmath352 . it follows ( cf . proposition [ milnorno ] ) that the milnor number of @xmath22 is @xmath353 , yielding a quick proof of this well - known formula ( see @xcite , 8.5 , lemma 3 , for a discussion of the geometry underlying this formula over @xmath31 ) . for example , the resolution graph of an ordinary cusp is : where the numbers indicate the multiplicity of the exceptional divisors . it follows that @xmath354 at an ordinary cusp . typically , @xmath5 is _ singular _ at base points of the system at which @xmath3 is singular . subtleties in the computation of the specialization function arise precisely because of this phenomenon . again consider a pencil centered at an ordinary cusp @xmath22 , but such that the general element of the pencil is nonsingular at @xmath22 and tangent to @xmath3 at @xmath22 : a local description for the correspondence @xmath5 near @xmath22 is @xmath355 this may be viewed as a singular hypersurface in @xmath356 and is resolved by a single blow - up at @xmath357 ( as the reader may check ) . one more blow - up produces a divisor with normal crossings as needed in definition [ maindef ] , with the same resolution graph as in [ sincurves ] : but with different multiplicities , as indicated . this does not affect the value of @xmath198 : @xmath358 in fact , it is clear ` for specialization reasons ' that the value of @xmath198 at an isolated singularity @xmath22 of @xmath3 is the same whether @xmath22 is in the base locus of the pencil or not , since ( as pointed out at the beginning of this section ) the euler characteristic of @xmath3 weighted according to @xmath2 must equal the euler characteristic of the general element of the pencil , and this latter is unaffected by the intersection of @xmath3 with the general element . it is however interesting that the geometry of the resolution _ is _ affected by the base locus of the pencil : the total space of the specialization is smooth near @xmath22 if the cusp is not in the base locus ( as in [ sincurves ] ) , while it is singular if the cusp is in the base locus ( as in this subsection ) . any difference in the normal crossing resolution due to these features must compensate and produce the same value for @xmath198 . the next example will illustrate that this is not necessarily the case for non - isolated singularities . note also that the ` distinguished decomposition ' of @xmath2 ( or of its motivic counterparts ) do tell these two situations apart : for example , with notation as in remark [ distdecfun ] , @xmath359 for the cuspidal curve in [ sincurves ] ( cusp @xmath22 away from the base locus ) , while @xmath360 for the cuspidal curve in this subsection ( cusp on the base locus ) . let @xmath46 be a nonsingular surface , and @xmath361 be a ( possibly multiple , reducible ) curve . consider the pencil between @xmath3 and a nonsingular curve @xmath40 meeting a component of multiplicity @xmath362 in @xmath3 transversally at a general point @xmath22 . view this as a specialization , as explained in [ spesu ] . then @xmath332 , regardless of the multiplicity @xmath21 . to verify this , we may choose analytic coordinates @xmath363 so that @xmath22 is the origin , @xmath3 is given by @xmath364 for @xmath365 , and @xmath40 is @xmath366 ; the correspondence @xmath5 is given by @xmath367 in coordinates @xmath368 . if @xmath369 , then both @xmath3 and @xmath5 are nonsingular , and @xmath370 as seen above . if @xmath371 , @xmath5 has an @xmath372 singularity at the origin , and its resolution is classically well - known : the exceptional divisors are @xmath345s , linked according to the @xmath372 diagram , @xmath373 the only work needed here is to keep track of the multiplicities of the components in the inverse image of @xmath309 . the reader may check that the pull - back of @xmath309 in the resolution consists of the proper transform of the original central fiber , with multiplicity @xmath374 , and of a chain of smooth rational curves , with decreasing multiplicities : the contribution to @xmath198 of all but the right - most component in this diagram is @xmath14 , because the euler characteristic of the complement of @xmath170 points in @xmath345 is @xmath14 . the right - most component contributes @xmath68 , since it appears with multiplicity @xmath68 and it contributes by the euler characteristic of the complement of one point in @xmath345 , that is @xmath68 . thus @xmath332 , as claimed . note that @xmath375 at a general point of a component of multiplicity @xmath21 in @xmath3 . thus , the effect of @xmath22 being in the base locus is to erase the multiplicity information , provided that the general element @xmath40 of the pencil meets @xmath3 transversally at @xmath22 . this is in fact an instance of a general result , proven over @xmath31 in all dimensions and for arbitrarily singular @xmath3 by a. parusiski and p. pragacz in proposition 5.1 of @xcite . the proof given in this reference uses rather delicate geometric arguments ( for example , it relies on the fact that a whitney stratification satisfies thom s @xmath376 condition , @xcite ) . it would be worthwhile constructing a direct argument from the definition for @xmath2 given in this paper , and valid over any algebraically closed field of characteristic @xmath14 . [ [ section-5 ] ] let @xmath5 be a nonsingular variety , and let @xmath377 be a closed subscheme . in @xcite we have considered the _ weighted chern - mather class _ of @xmath40 , @xmath378 where @xmath379 are the supports of the components of the normal cone of @xmath40 in @xmath5 , @xmath349 is the multiplicity of the component over @xmath379 , and @xmath380 denotes the ordinary chern - mather class . also , @xmath381 is the largest dimension of a component of @xmath40 . up to a sign , @xmath382 is the same as the _ aluffi class _ of @xcite . with the definition given above , if @xmath40 is irreducible and reduced , then @xmath383 equals @xmath384 ; in particular , @xmath385 $ ] if @xmath40 is nonsingular . ( however , with this choice of sign the contribution of a component @xmath379 depends on the dimension of the largest component of @xmath40 . ) consider the case in which @xmath40 is the _ singularity subscheme _ of a hypersurface @xmath3 in @xmath5 : if @xmath39 is a local generator for the ideal of @xmath3 , then the ideal of @xmath40 in @xmath5 is locally generated by @xmath39 and the partials of @xmath39 . since @xmath7 admits a natural multiplicity decomposition ( remark [ distdeccla ] ) , so do the class @xmath383 and its degree , a donaldson - thomas type invariant ( @xcite , proposition 4.16 ) . i.e. , monodromy induces a decomposition of these invariants . the relation in proposition [ wmaprop ] is a @xmath233-counterpart of the identity @xmath388 mentioned in the introduction , and is equivalent to theorem 1.5 in @xcite . the identity amounts to the relation of @xmath41 with the euler characteristic of the perverse sheaf of vanishing cycles . it goes at least as far back as @xcite , theorem 4 ; it also follows from proposition 5.1 in @xcite , and is equation ( 4 ) in @xcite . however , these references work over @xmath31 , and it seems appropriate to indicate a proof of ( * ) which is closer in spirit to the content of this paper . by theorem [ spethm ] , @xmath389 ( cf . lemma [ spediv ] ) . thus @xmath390 \\ & = c({{\mathscr l } } ) \cap \left({{c_{\text{sm}}}}({1\hskip-3.5pt1}_x)-\frac{c(tv|_x)}{1+x}\cap [ x]\right)\end{aligned}\ ] ] with @xmath391 . the class @xmath392 $ ] is the class of the virtual tangent bundle to @xmath3 , denoted @xmath393 in @xcite . thus , @xmath394 by theorem 1.2 in @xcite , it follows that @xmath395 which is the statement . both sides of the formula in proposition [ wmaprop ] make sense in a more general situation than the case in which @xmath40 is the singularity subscheme of a hypersurface @xmath3 : the class @xmath383 is defined for arbitrary subschemes of a nonsingular variety , and definition [ maindef ] also gives a meaning to @xmath2 for any subscheme @xmath3 of a nonsingular variety . otherwise put , both sides of the identity @xmath396 admit compelling generalizations : definition [ maindef ] does not require @xmath3 to be a hypersurface , and @xmath383 is the class corresponding to a constructible function @xmath44 defined for all @xmath40 as a specific combination of local euler obstructions ( see proposition 1.5 in @xcite ) . this function has garnered some interest in donaldson - thomas theory , and is currently commonly known as _ behrend s function . _ it is tempting to guess that a statement closely related to proposition [ wmaprop ] may still hold for _ any _ @xmath40 and a suitable choice of @xmath0 with singularities along @xmath40 . ( of course @xmath397 should be replaced by @xmath398 for the right - hand - side to be supported on the singularities of @xmath3 , see example [ sci ] . ) we end with an expression for the function @xmath41 for the singularity subscheme @xmath40 of a hypersurface @xmath3 in terms of a resolution of @xmath3 . for this we need to work with @xmath399-valued constructible functions ; we do not know if a similar statement can be given over @xmath400 . assume @xmath5 is nonsingular , @xmath0 is a hypersurface , and @xmath401 is the singularity subscheme . consider a morphism @xmath402 as in definition [ maindef ] . thus , @xmath403 is a divisor @xmath23 with normal crossings and nonsingular components @xmath55 , @xmath404 ; @xmath58 denotes the multiplicity of @xmath55 in @xmath23 , and @xmath405 is the restriction of @xmath56 . as discussed above , @xmath410 . the given formula follows immediately from @xmath411 which is a restatement of definition [ maindef ] , and @xmath412 which follows from a small generalization of 4.4.3 in @xcite ( cf . theorem 2.1 and the proof of theorem 3.1 in @xcite ) . again , it is tempting to guess that a similar expression may exist for behrend s function of an arbitrary subscheme @xmath40 of a nonsingular variety @xmath5 , for a suitable choice of a corresponding pair @xmath0 . jan denef and franois loeser . geometry on arc spaces of algebraic varieties . in _ european congress of mathematics , vol . i ( barcelona , 2000 ) _ , volume 201 of _ progr . math . _ , pages 327348 . birkhuser , basel , 2001 . spcialisation de faisceaux et monodromie modre . in _ analysis and topology on singular spaces , ii , iii ( luminy , 1981 ) _ , volume 101 of _ astrisque _ , pages 332364 . france , paris , 1983 .

phantom fields with negative kinetic energy are often plagued by the vacuum quantum instability in the ultraviolet region . we present a lorentz - violating dark energy model free from this problem and show that the crossing of the cosmological constant boundary to the phantom equation of state is realized before reaching a de sitter attractor . another interesting feature is a peculiar time - dependence of the effective newton s constant ; the magnitude of this effect is naturally small but may be close to experimental limits . we also derive momentum scales of instabilities at which tachyons or ghosts appear in the infrared region around the present hubble scale and clarify the conditions under which tachyonic instabilities do not spoil homogeneity of the present / future universe .

0704.1848

the compilations of various observational data show that the universe has entered the stage of an accelerated expansion around the redshift @xmath1 @xcite . the equation of state ( eos ) parameter @xmath2 of dark energy ( de ) responsible for the acceleration of the universe has been constrained to be close to @xmath0 . however , the phantom eos ( @xmath3 ) is still allowed by observations and even favored by some analyses of the data @xcite . it is also possible that the eos of de crossed the cosmological constant boundary ( @xmath0 ) in relatively near past @xcite . the presence of the phantom corresponds to the violation of weak energy condition , the property which is generally difficult to accommodate within the framework of field theory . the simplest model which realizes the phantom eos is provided by a minimally coupled scalar field with a negative kinetic term @xcite ( see also refs . the negative kinetic energy is generally problematic because it leads to a quantum instability of the vacuum in the ultraviolet ( uv ) region @xcite : the vacuum is unstable against the catastrophic particle production of ghosts and normal ( positive energy ) fields . there have been a number of attempts to realize the phantom eos without having the pathological behaviour in the uv region . one example is scalar - tensor gravity in which a scalar field @xmath4 with a positive kinetic term is coupled to ricci scalar @xmath5 @xcite . this coupling leads to the modification of gravitational constant , but it was shown in ref . @xcite that there are some parameter regions in which a phantom effective eos is achieved without violating local gravity constraints in the present universe . another example is provided by the so - called modified gravity , including @xmath6 gravity models @xcite and the gauss - bonnet ( gb ) models @xcite . in @xmath6 models it is possible to obtain a strongly phantom effective eos , but in that case the preceding matter epoch is practically absent @xcite . for gb de models it was shown in ref . @xcite that the crossing of the cosmological constant boundary , @xmath0 , is possible , but local gravity experiments place rather strong constraints on the effective gb energy fraction @xcite . in addition , tensor perturbations are typically plagued by instabilities in the uv region if the gb term is responsible for the accelerated expansion of the universe @xcite . thus , it is generally not so easy to construct viable modified gravity models that realize the phantom effective eos without violating cosmological and local gravity constraints . the third example is the dvali - gabadadze - porrati ( dgp ) braneworld model @xcite and its extension @xcite with a gb term in the bulk , which allow for the possibility to have @xmath3 @xcite . however , it was shown in ref . @xcite that the dgp model contains a ghost mode , which casts doubts on the viability of the self - accelerating solution . while the above models more or less correspond to the modification of gravity , it was recently shown that in the einstein gravity in a lorentz - violating background the phantom eos can be achieved without any inconsistency in the uv region @xcite . in particular , in the model of ref . @xcite lorentz invariance is broken in the presence of a vector field @xmath7 which has two - derivative kinetic terms similar to those given in ref . the effect of the lorentz violation is quantified by a parameter @xmath8 , where @xmath9 is an uv cut - off scale . in analogy to ref . @xcite the vector field also has one - derivative coupling @xmath10 with a scalar field @xmath11 , where @xmath12 is a small parameter that characterizes an ir scale . in the uv region , where the spatial momentum @xmath13 is much larger than @xmath12 , ghosts , tachyons and super - luminal modes are not present . meanwhile tachyons or ghosts can appear in the ir region @xmath14 . this is not problematic provided that @xmath12 is close to the present hubble scale . in this paper we apply this lorentz - violating model to dark energy and study the cosmological dynamics in detail in the presence of mass terms in the potential , @xmath15 ( where @xmath16 ) . we show that the model has a de sitter attractor responsible for the late - time acceleration . at early times de naturally has normal eos with @xmath17 , while the phantom eos can be realized between the matter - dominated era and the final de sitter epoch . we clarify the conditions under which the cosmological constant boundary crossing to the phantom region occurs . interestingly , in a range of parameters this crossing takes place at the epoch when @xmath18 thus making the crossing potentially observable . another interesting feature of our model is the time - dependence of the effective newton s constant . it is naturally weak , but may well be comparable with current experimental limits . moreover , the effective newton s constant @xmath19 has a peculiar behaviour correlated with the deviation of @xmath2 from @xmath20 . we also derive momentum scales of instabilities of perturbations , first in minkowski spacetime . this is the extension of the work @xcite that mainly focused on the case of massless scalar ( @xmath21 ) . we show that in the uv region ( @xmath22 ) the model does not have any unhealthy states such as ghosts , tachyons or super - luminal modes . in the ir region ( @xmath14 ) tachyons or ghosts appear , depending on the momentum . finally , we study the evolution of perturbations in the cosmological background and estimate the amplitude of perturbations amplified by the tachyonic instability around the scale of the present hubble radius . the perturbations remain to be smaller than the background fields under certain restriction on the model parameters . this paper is organized as follows . in sec . [ modelsec ] we present our lorentz - violating model and derive basic equations describing spatially flat friedmann robertson walker cosmology in the presence of de , radiation and non - relativistic matter . in sec . [ dynamicssec ] the cosmological dynamics is discussed in detail analytically and numerically with an emphasis on the occurrence of a phantom phase before reaching a de sitter attractor . the time - dependence of the effective gravitational constant is also considered . in sec . [ instabilitysec ] we study the minkowski spectrum of field perturbations and clarify the properties of tachyons and ghosts in the ir region . we then discuss the tachyonic amplification of field perturbations around the present hubble scale in the cosmological background . we summarize our results in sec . [ conclusionsec ] . appendix a contains the derivation of the effective `` newtonian gravitational constant '' in our model . in appendix b we derive the fixed points of the system by rewriting the equations in autonomous form . we analyse the stability of the fixed points and show analytically that the cosmological evolution proceeds from radiation - dominated stage through matter - dominated stage to the final de sitter regime . we study a 4-dimensional lorentz - violating model whose lagrangian density includes a vector field @xmath23 and a scalar field @xmath11 : @xmath24 where @xmath25 with @xmath9 being an uv cut - off scale of the effective theory . the dimensionless parameters @xmath26 and @xmath27 are the functions of @xmath28 , and @xmath12 is a free positive parameter that characterizes an ir scale . the first two terms in ( [ lag ] ) are familiar in two - derivative theory @xcite , whereas the one - derivative term @xmath29 is introduced following the approach of ref . @xcite . we study dynamics of flat friedmann - robertson - walker ( frw ) universe @xmath30 where @xmath31 is a lapse function and @xmath32 is a scale factor . in the case of spatially homogeneous fields with @xmath33 ( @xmath34 ) , the lagrangian ( [ lag ] ) reads @xmath35 where @xmath36 , @xmath4 is the homogeneous part of the field @xmath11 and @xmath37 hereafter we study the case in which the following condition holds @xmath38 this is required to avoid a superluminal propagation in minkowski spacetime @xcite , as we will see later . throughout this paper we assume that @xmath26 and @xmath39 are of order unity . for fixed @xmath40 , the second term in the lagrangian ( [ eqn / pg5/1:lv16 ] ) has precisely the form of the einstein hilbert action specified to the flat frw metric . hence , it leads to the change of the `` cosmological '' effective planck mass @xcite @xmath41 another effective planck mass @xmath42 determines the strength of gravitational interactions at distances much shorter than the cosmological scale ; in general , these two effective planck masses are different @xcite . we show in appendix a that the `` newtonian '' planck mass in our model is given by @xmath43 both effective planck masses depend on time via @xmath44 . since the time - dependent terms in ( [ effpla ] ) and ( [ newton ] ) differ by sign only , it will be sufficient to study one of these effective masses . in what follows we concentrate on the `` newtonian '' mass ( [ newton ] ) for definiteness . in this paper we focus on the case in which the potential @xmath45 takes a separable form : @xmath46 we take into account the contributions of non - relativistic matter and radiation whose energy densities @xmath47 and @xmath48 , respectively , satisfy @xmath49 the energy density of the fields is derived by taking the derivative with respect to @xmath50 of the action @xmath51 : @xmath52_{{\cal n}=1 } = \frac{\gamma}{2 } \dot{x}^2 -\frac{3\alpha}{2}h^2 x^2 + \frac{1}{2 } \dot{\phi}^2 + v \ , . \label{page3+}\end{aligned}\ ] ] we set @xmath53 for the rest of this paper . the friedmann equation is given by @xmath54\ , , \label{fr}\end{aligned}\ ] ] where @xmath55 . the equations of motion for the homogeneous fields @xmath4 and @xmath56 are @xmath57 where @xmath58 , etc . taking the time - derivative of eq . ( [ fr ] ) and using eqs . ( [ eqx ] ) and ( [ eqp ] ) , we obtain @xmath59 where @xmath60 in what follows we assume for simplicity that @xmath26 and @xmath39 are constants , i.e. , @xmath61 . following ref . @xcite we consider the simplest potential for the fields , @xmath62 which allows for a possibility to realize a phantom phase . one way to analyse the cosmological dynamics in our model is to make use of the autonomous equations , the techniques widely used in the context of dark energy studies @xcite . this approach is presented in appendix b , where we analytically confirm that our model can lead to the sequence of radiation , matter and accelerated epochs . also , in appendix b we derive the conditions under which the de sitter solution given below is an attractor . here we first present a simpler analysis based on the slow - roll approximation . then we give numerical solutions to eqs . ( [ 8 * ] ) , ( [ 9 * ] ) , ( [ fr ] ) , ( [ eqx ] ) , ( [ eqp ] ) , exhibiting transient phantom behaviour , and study their dependence on various parameters of our model , including the initial values of the fields . one immediate point to note is that in the absence of radiation and matter , the system of equations ( [ fr ] ) , ( [ eqx ] ) , ( [ eqp ] ) has a de sitter solution , @xmath63 , for which @xmath4 and @xmath40 are also independent of time , provided that @xmath64 indeed , for constant @xmath65 and @xmath40 eqs . ( [ fr ] ) , ( [ eqx ] ) , ( [ eqp ] ) reduce to a simple algebraic system @xmath66\,,\nonumber\\ & & 3\alpha h^2=m^2\,,\nonumber\\ & & -3\epsilon hx = m^2\phi\,.\end{aligned}\ ] ] once the inequality ( [ epcon ] ) is satisfied , this system has a solution @xmath67 we will see in what follows , and elaborate in appendix b , that in a range of parameters this solution is an attractor which corresponds to the de sitter phase in asymptotic future ( hence the notation ) . in order to use this for dark energy we require that the mass scale @xmath68 is of the order of the present hubble parameter @xmath69 . then the newtonian effective planck mass , eq . ( [ newton ] ) , is given by @xmath70 in order that the change of the planck mass be small , we impose the condition @xmath71 it is worth noting that under this condition , the contribution of the field @xmath4 in the energy density dominates in the de sitter regime , @xmath72 thus , as the system approaches the de sitter attractor , the total energy density in the universe becomes determined by the scalar field energy density . another point to note is that at early times ( at the radiation - dominated epoch already ) , when the hubble parameter is large enough , the term @xmath73 in eq . ( [ eqx ] ) drives the field @xmath40 to zero , the relevant time being of the order of the hubble time . soon after that the field @xmath4 obeys the usual scalar field equation in the expanding universe , so the hubble friction freezes this field out . thus , the initial data for the interesting part of the de evolution are @xmath74 the value of @xmath75 is a free parameter of the cosmological evolution in our model . since at early times the field @xmath40 is close to zero , its effect on the evolution of the field @xmath4 is negligible . the field @xmath4 slowly rolls _ down _ its potential , and its energy density dominates over that of @xmath40 . therefore , eos for de at early times is normal , @xmath17 , with @xmath2 being close to @xmath20 . we refer to this regime as quintessence stage . as we will see below , in a range of parameters , the system eventually crosses the cosmological constant boundary @xmath0 and passes through a transient phantom phase before reaching the de sitter asymptotics ( [ phiv ] ) . the approach to the de sitter solution ( [ phiv ] ) occurs in the slow roll regime . to see how this happens , we truncate eqs . ( [ eqx ] ) and ( [ eqp ] ) to @xmath76 this truncation is legitimate provided that in addition to the usual slow - roll conditions @xmath77 and @xmath78 , the following conditions are satisfied : @xmath79 [ when writing inequalities , we always mean the absolute values of the quantities . ] note that we do not impose the condition @xmath80 unlike in ref . @xcite , since the term @xmath81 is not necessarily negligible relative to the term @xmath82 in eq . ( [ phan1 ] ) . from eq . ( [ phan1 ] ) we obtain @xmath83 where @xmath84 note that @xmath85 may be considered as a measure of the deviation from the de sitter regime ( [ phiv ] ) . substituting eq . ( [ page6 * ] ) into eq . ( [ phan2 ] ) we get the following equation @xmath86 where @xmath87 equation ( [ hdsigma ] ) shows that the field @xmath4 rolls _ up _ the potential @xmath88 for @xmath89 , i.e. , for @xmath90 this is the region in which the phantom equation of state ( @xmath3 ) is realized ; indeed , eq . ( [ rhop ] ) gives @xmath91 . another way to understand the phantom behaviour is to notice that when the system approaches the de sitter regime , the field @xmath4 dominates the energy density , see eq . ( [ page4 + ] ) , so the energy density increases as the field @xmath4 rolls up . let us find out whether the slow roll conditions ( [ con2 ] ) , ( [ con3 ] ) and ( [ con1 ] ) are indeed satisfied . making use of eq . ( [ page6 * ] ) we obtain that the condition ( [ con2 ] ) is equivalent to @xmath92 while using eqs . ( [ page6 * ] ) and ( [ hdsigma ] ) we rewrite the condition ( [ con3 ] ) as @xmath93 the second inequality ensures also the validity of the relation ( [ con1 ] ) ; this can be seen by taking the time derivative of eq . ( [ phan2 ] ) . the latter two inequalities are automatically valid at small @xmath85 , that is near the de sitter solution ( [ phiv ] ) . we conclude that the approach to the de sitter solution indeed occurs in the slow roll regime , and that the phantom phase is indeed realised provided that the relation ( [ hcondition ] ) holds . our analysis implies also that the de sitter solution ( [ phiv ] ) is an attractor : the hubble parameter slowly increases towards its de sitter value , @xmath85 decreases , and the dynamics gets frozen as @xmath94 . since the field @xmath4 dominates the energy density at the phantom slow roll stage , the condition ( [ hcondition ] ) takes a simple form @xmath95 where we made use of ( [ epco ] ) . the latter relation translates into the range of initial conditions which eventually lead to the transient phantom behaviour , @xmath96 indeed , during the radiation- and matter - dominated stages the field @xmath4 remains almost constant , and at the quintessence stage it also does not roll down much . recalling again that @xmath4 dominates the energy density , we rewrite the inequality ( [ v3 * * ] ) as @xmath97 , where @xmath98 the parameter @xmath99 may be viewed as the slow roll parameter for the field @xmath4 . indeed , one observes that @xmath100 which , together with eq . ( [ hdsigma ] ) , justifies this interpretation . from eqs . ( [ page3 + ] ) , ( [ rhop ] ) , ( [ page6 * ] ) and ( [ hdsigma ] ) it follows that during the slow roll phantom stage , the eom parameter of de is given by @xmath101 hence , the appreciable deviation from @xmath0 occurs when @xmath99 is not much smaller than unity , i.e. , when @xmath75 is appreciably smaller than @xmath102 . in the next section we confirm these expectations by numerical analysis , and also show explicitly that in a range of parameters , the cosmological evolution proceeds from radiation - dominated to matter - dominated epoch , and then to the slow roll phantom stage , before finally ending up in the de sitter regime ( [ phiv ] ) . in our numerical analysis we choose initial conditions @xmath103 with nonzero values of @xmath4 , @xmath47 and @xmath48 . this choice corresponds to the initial data ( [ * * initial ] ) . we have also tried many other initial conditions and found that the results are not sensitive to the initial values of @xmath40 , @xmath104 and @xmath105 , in accord with the discussion in the end of sec . [ sub : prelim ] . in fig . [ evo1 ] we plot the cosmological evolution for the model parameters @xmath106 , @xmath107 , @xmath108 , @xmath109 and the initial value @xmath110 . we find it convenient to present the plots in terms of the variable @xmath111 figure [ evo1 ] clearly shows that the sequence of radiation , matter and de sitter epochs can be achieved together with the @xmath0 crossing . the de eos parameter @xmath2 is nearly a constant , @xmath112 , during the radiation and matter epochs because the fields are almost frozen . at the transition era from matter domination to de domination , @xmath2 begins to grow because the kinetic energies of the fields become important ; this is the quintessence phase . however , the system soon enters the phantom phase during which the field @xmath4 rolls up the potential . hence the equation of state @xmath2 crosses the cosmological constant boundary @xmath0 and reaches a minimum value @xmath113 . the solution finally approaches the de sitter regime ( [ phiv ] ) from the phantom side . of some interest is also the eos parameter of the entire system , @xmath114 it is seen from fig . [ evo1 ] , this parameter also becomes smaller than @xmath20 soon after @xmath0 crossing . cosmological evolution for the model parameters @xmath106 , @xmath107 , @xmath115 and @xmath109 . we choose initial conditions @xmath116 , @xmath110 , and @xmath117 , @xmath118 . shown is the evolution of @xmath119 , @xmath120 , @xmath121 , @xmath2 and @xmath122 as functions of @xmath123 . note that the present epoch corresponds to @xmath124 and @xmath125 , which is denoted by a vertical line . after the cosmological constant boundary crossing , the de eos parameter @xmath2 reaches a minimum @xmath126 and then increases towards the de sitter value @xmath0 from the phantom side.,width=364,height=364 ] the pattern shown in fig . [ evo1 ] is generic in our model , provided that its parameters and initial data obey @xmath127 , @xmath128 and @xmath129 ( in fact , the inequalities here need not be strong ) . the strengths of the effects depend , of course , on the values of these parameters . in particular , the minimum value @xmath130 is related to the slow - roll parameter @xmath99 , in accord with eq . ( [ a * * ] ) . if the field @xmath4 evolves very slowly , one has @xmath97 , so @xmath130 is close to @xmath20 . on the contrary , the appreciable deviation from @xmath0 occurs if @xmath99 is not very much smaller than unity . the minimum value of the eos parameter @xmath2 of de , its value @xmath131 at the present epoch ( @xmath132 , @xmath133 ) and its maximum value as functions of @xmath134 for @xmath135 and the initial value @xmath136.,width=364,height=364 ] since the field @xmath4 is practically frozen during the radiation- and matter - dominated epochs , and evolves rather slowly later on , the dependence of the cosmological evolution on the parameters of the model and on the initial value @xmath75 can be understood , at qualitative level , by inspecting eq . ( [ a * ] ) . for the qualitative discussion of the evolution well before the asymptotic de sitter regime sets in , the parameter @xmath85 in ( [ a * ] ) may be set equal to 1 , while the value of @xmath4 may be set equal to its initial value @xmath75 . then eq . ( [ a * ] ) implies that with other parameters and @xmath75 fixed , for smaller @xmath134 one gets smaller absolute value of @xmath99 at its minimum , leading to the value of @xmath130 closer to @xmath20 . this is shown in fig . [ a ] . from fig . [ evo1 ] it is clear , however , that the minimum of @xmath2 occurs after the present epoch ( @xmath137 and @xmath138 ) . again , this is a rather generic feature of our model . therefore , instead of @xmath130 , more interesting quantities are the present value @xmath139 of the de eos parameter and also its maximum value before the cosmological constant boundary crossing . these quantities are also given in fig . overall , the behaviour shown in fig . [ evo1 ] is more pronounced at larger @xmath134 , once other parameters of solutions are kept fixed . according to eq . ( [ a * ] ) , the initial value of the field @xmath4 is also important to determine the amplitude of @xmath99 and hence @xmath2 : a smaller @xmath75 results in a stronger deviation of @xmath130 from @xmath20 . the present value @xmath139 also becomes more negative ( stronger deviating from @xmath20 ) , while the cosmological constant boundary crossing occurs earlier . from the numerical analysis we find that at relatively large values of @xmath134 the increase of @xmath68 has the opposite effect , while at smaller @xmath134 the effects due to the variation of @xmath68 are small . these properties are illustrated in figs . [ d ] and [ e ] . as discussed above , the phantom phase occurs only if the initial value of @xmath4 obeys ( [ a * * * ] ) , otherwise the de sitter attractor is approached from the quintessence side , @xmath17 . this is illustrated in fig . [ evo3 ] . the dependence of the present value @xmath139 ( i.e. , @xmath2 at @xmath124 , @xmath125 ) on the initial value of @xmath4 , for different sets of the model parameters.,width=364,height=364 ] the dependence of the redshift ( from the present epoch , @xmath124 , @xmath125 ) of the cosmological constant boundary crossing , @xmath0 , on the initial value of @xmath4.,width=364,height=364 ] cosmological evolution for the model parameters @xmath106 , @xmath107 , @xmath115 and @xmath140 . we choose initial conditions @xmath116 , @xmath141 , and @xmath117 , @xmath118 . the present epoch corresponds to @xmath142 and @xmath124 , which is denoted by a vertical line . in this case the cosmological constant boundary crossing is not realized because @xmath4 remains always larger than @xmath102.,width=345,height=345 ] at the end of this section we discuss the variation of the effective gravitational constant . according to eq . ( [ newton ] ) , the effective newton s constant that determines the interaction deep inside the horizon scale is given by @xmath143 its variation in time is conveniently expressed in terms of the following quantity @xmath144 the typical experimental and observational constraints on the variation of @xmath145 in the present universe are given by @xmath146 @xcite , which translates into the condition @xmath147 the evolution of the variation of the effective gravitational constant @xmath148 as function of @xmath149 and redshift @xmath150 . the cases ( a ) and ( b ) correspond to the model parameters and initial conditions given of figs . [ evo1 ] and [ evo3 ] , respectively . the black points represent the values at the present epoch ( @xmath151 ) . , width=336,height=336 ] the maximum value of @xmath148 as a function of @xmath134 . other parameters and initial conditions are the same as in fig . [ a ] . , width=364,height=364 ] in fig . [ gravitation ] we plot the evolution of the quantity @xmath148 for the model parameters and initial conditions given in figs . [ evo1 ] and [ evo3 ] . at the present epoch ( @xmath152 ) we obtain the values @xmath153 and @xmath154 for these two cases , respectively . comparing fig . [ gravitation ] with figs . [ evo1 ] and [ evo3 ] one observes that the variation of the gravitational constant is correlated in time with the deviation of @xmath2 from @xmath20 . this is clear from eq . ( [ eqn / pg18/1:lv16 ] ) too : the gravitational constant varies when the field @xmath40 changes in time , while the latter occurs during the transition from the matter dominated stage to the final de sitter attractor . it is precisely at this transition stage that @xmath2 substantially deviates from @xmath20 . figure [ f ] shows the maximum value of @xmath155 as a function of @xmath134 . again , the variation of the effective newton s constant is more pronounced at larger @xmath134 . this means that it is correlated with the amplitude of the deviation of @xmath2 from @xmath20 . the dependence of @xmath156 on @xmath68 and on the initial value of @xmath4 is rather weak . it is worth pointing out that as long as the deviation of the eos from @xmath0 is not so significant , the models satisfy the constraint ( [ congra ] ) , and also that our model suggests that the variation of @xmath145 is close to the present upper bound on @xmath157 . in this section the momentum scales of instabilities are present in our model . we first study dispersion relations in the minkowski space - time and then proceed to those in the frw space - time . we wish to clarify the conditions under which a tachyon or a superluminal mode appears by considering dispersion relations . we also evaluate the energy of the modes to find out a ghost state . let us consider the perturbations for the fields , @xmath158 the quadratic lagrangian for perturbations , following from the general expression ( [ lag ] ) , is @xmath159 where @xmath160 for our model ( [ poten ] ) one has @xmath161 and @xmath162 . in what follows we concentrate on this case , and assume the following relations , see ( [ epco ] ) and ( [ v3 * ] ) , @xmath163 varying the lagrangian ( [ lag2 ] ) with respect to @xmath164 , @xmath165 and @xmath166 , we obtain the equations for the field perturbations . in order to find the spectrum of the system we write the solutions in the form @xmath167 , @xmath168 and @xmath169 . the transverse mode of the vector field @xmath170 has the dispersion relation @xmath171 the three scalar modes @xmath172 , @xmath173 and @xmath174 satisfy the following equations @xmath175 expressing @xmath176 and @xmath173 in terms of @xmath174 from eqs.([spe1 ] ) , ( [ spe2 ] ) and plugging them into eq . ( [ spe3 ] ) , we find that the eigenfrequencies corresponding to three mixed states satisfy @xmath177 where @xmath178 the spectrum in the case @xmath21 was studied in ref . @xcite . our purpose here is to extend the analysis to the case of non - zero @xmath179 . denoting the solutions of eq . ( [ eqx ] ) as @xmath180 , @xmath181 and @xmath182 , we obtain the relation @xmath183 once the conditions ( [ eq / pg21/1:draft - apr6 ] ) are satisfied , then one can show that if the relation @xmath184 holds at some momentum , then the inequality ( [ ineq ] ) is satisfied for all momenta . in the limits @xmath185 and @xmath186 , we obtain the following dispersion relations , respectively . * \(a ) uv limit ( @xmath185 ) @xmath187 we see that @xmath188 and @xmath189 , @xmath190 . in all three cases the group velocities @xmath191 are less than 1 , so neither mode is superluminal at high three - momenta , provided that @xmath192 . the two - derivative terms in the lagrangian ( [ lag2 ] ) dominate in the uv limit , so there are neither ghosts nor tachyons in this limit . * \(b ) ir limit ( @xmath186 ) @xmath193\ , , \label{spectrum01}\\ \omega _ { 2}&=&\frac{m^{2}}{\alpha}\ , , \nonumber \\ \omega _ { 3}&=&\frac{\epsilon ^{2}}{\gamma } + m^{2}+\mathcal { o}(m^{2})\ , . \nonumber \ ] ] we see again that @xmath188 and @xmath189 , @xmath190 . this means that using the property ( [ ineq ] ) we can identify the modes : the first one has the behaviour ( [ spectrum1 ] ) and ( [ spectrum01 ] ) , and so on . it follows from eq . ( [ xrelation ] ) that @xmath194 never vanish . in fact the coefficients in eq . ( [ eqx ] ) are regular at all momenta , so @xmath194 are regular as well . therefore , @xmath194 never change signs and hence @xmath195 for all momenta . this means , in particular , that the second and third modes never become tachyonic . let us discuss the dangerous mode with the dispersion relation @xmath196 in some detail . the expression for the fields in each mode is @xmath197 where @xmath198 are the normalization factors . setting @xmath199 in eq . ( [ eqx ] ) , we obtain three critical momenta @xmath200\ , , \\ & & p_3 ^ 2=\frac{m^2}{\gamma}\,.\end{aligned}\ ] ] under the conditions ( [ eq / pg21/1:draft - apr6 ] ) , the critical momenta @xmath201 are approximately given by @xmath202 so that @xmath203 . the tachyonic mode ( @xmath204 ) is present for @xmath205 and @xmath206 . in order to find whether there are ghosts we calculate the energy of the modes ( [ eq / pg24/1:draft - apr6 ] ) , @xmath207\,.\end{aligned}\ ] ] for the modes with @xmath208 we have @xmath209 . for the mode with @xmath210 the energy is equal to zero at @xmath211 . while @xmath212 for @xmath213 , the energy @xmath214 changes its sign at this momentum . thus the mode with @xmath210 is a ghost for @xmath215 . we summarize the properties of the dangerous mode as follows : * \(i ) @xmath216 : healthy * \(ii ) @xmath217 : tachyon * \(iii ) @xmath218 : ghost , but not tachyon * \(iv ) @xmath219 : tachyon . unlike the case @xmath21 @xcite the tachyon is present in the deep ir region ( iv ) . to end up the discussion of the modes in minkowski space - time , we give the expressions for the minimum values of @xmath220 in the tachyonic regions , * ( ii ) : @xmath221 * ( iv ) : @xmath222 note that @xmath223 is relatively large in the region ( ii ) , so this region is the most problematic . finally we discuss the evolution of field perturbations in the frw background ( [ frw ] ) . in the cosmological context the physical momentum @xmath13 is related to the comoving momentum @xmath224 as @xmath225 . once the parameters of the model and initial data are such that the cosmological boundary crossing occurs , the present epoch ( @xmath137 ) typically corresponds to the phantom region . from fig.1 one can see that the hubble parameter does not change much during the transition from the phantom epoch to the final de sitter era . hence the present value of the hubble parameter ( @xmath69 ) is of the same order as the value @xmath226 in the de sitter asymptotics . this means that the value @xmath227 is of the same order as @xmath69 provided that @xmath39 and @xmath26 are of order unity . the tachyon appears when the momentum @xmath225 of the dangerous mode becomes smaller than @xmath228 and temporally disappears when the mode crosses the value @xmath229 . hence this instability is present for the modes which are inside the hubble radius and satisfy @xmath230 , but it is absent for the modes deep inside the hubble radius , satisfying @xmath231 . after the hubble radius crossing ( @xmath232 ) , the tachyonic instability disappears in the momentum region @xmath218 , but the tachyon appears again for @xmath233 . note that the ghost existing at @xmath218 is a not a problem because of its low energy @xcite . in what follows we discuss the evolution of field perturbations in the two tachyonic regimes . before doing that it is instructive to study the high - momentum regime that sets the initial data for the tachyonic evolution . we denote the overall amplitude of the dangerous mode as @xmath235 . since the modes are deep inside the hubble radius ( @xmath236 ) in the regime we discuss here , the field perturbation @xmath237 approximately satisfies @xmath238 where @xmath239 and @xmath240 is conformal time defined by @xmath241 . taking the asymptotic minkowski vacuum state , @xmath242 , the squared amplitude of the field perturbation @xmath166 is given by @xcite @xmath243 , one has the following estimate for the amplitude of the field perturbation at the beginning of the tachyonic instability , @xmath244 as usual , this amplitude characterizes the contribution of a logarithmic interval of momenta into @xmath245 . this interval of momenta is dangerous , as the perturbations undergo the tachyonic amplification . since the modes are still inside the hubble radius , one can neglect the gravitational effects on the `` frequency '' @xmath246 when estimating the growth of field perturbations . by the time the tachyonic amplification ends up , the amplitude of field perturbations is estimated as @xmath247 recall that @xmath248 and @xmath249 . the largest value of @xmath250 is approximately given by ( [ eq / pg25/1:draft - apr6 ] ) . substituting this value into eq . ( [ growth ] ) and recalling that the background changes slowly ( @xmath251 ) , one finds that the amplitude of the field perturbation after exit from the tachyonic regime is of order @xmath252\,,\end{aligned}\ ] ] where we used eq . ( [ varphii ] ) . recall now that @xmath253 is of the same order as @xmath254 during the phantom phase . hence the large ratio @xmath255 leads to a strong amplification of field perturbations . from ( [ phiv ] ) , the homogeneous field @xmath4 at the phantom and de sitter phase is estimated as @xmath256 . the requirement that the perturbation @xmath166 is smaller than the background field @xmath4 leads to the constraint @xmath257 < \sqrt{3\alpha \pi } \frac{hm_{\rm pl}}{\epsilon m}\,.\end{aligned}\ ] ] as an example , in the case @xmath258 gev , @xmath259 , @xmath106 and @xmath107 , we obtain the constraint @xmath260 . as long as @xmath26 and @xmath39 are of order one , the ratio @xmath255 should not be too much larger than unity . after the hubble radius crossing , the effect of the cosmic expansion can no longer be neglected when estimating the `` frequencies '' of the field perturbations . since there are no tachyonic instabilities for @xmath218 , we consider the evolution of perturbations in the region @xmath219 . in ref . @xcite the equations for the field perturbations were derived in the slow - rolling background under the condition @xmath261 . the equation for the perturbation @xmath239 is approximately given by @xmath262 here we neglected the contribution of metric perturbations on the r.h.s . of this equation . note that metric perturbations works as a back reaction effect after the field perturbation is sufficiently amplified . the growth rate of the perturbation @xmath56 is mainly determined by the terms in the parenthesis of eq . ( [ chieq ] ) rather than the backreaction of metric perturbations . the last term corresponds to the tachyonic mass term , which already appeared in minkowski spacetime [ see eq . ( [ spectrum01 ] ) ] . since the term @xmath263 is of order @xmath264 , one can estimate the ratio of the tachyonic mass relative to this gravitational term : @xmath265 if we use the de sitter value @xmath226 , this ratio is estimated as @xmath266 . hence the gravitational term @xmath263 dominates over the tachyonic mass . in the de sitter background with @xmath267 the approximate solutions to eq . ( [ chieq ] ) can be obtained by setting @xmath268 . one finds that @xmath269 for the growing solution , thereby giving @xmath270 in ref . @xcite it was shown that the physical temporal component of the vector field perturbations evolves as @xmath271 , whereas the physical spatial component of the vector field decreases as @xmath272 . the growth rate of @xmath166 and @xmath273 is small due to the condition @xmath274 . so , the second tachyonic instability is harmless for the past and present cosmological evolution . however , we notice that since the de sitter solution is a late - time attractor , the perturbations @xmath166 and @xmath273 become larger than the homogeneous background fields in the distant future . at this stage we expect that the contribution of metric perturbations can be also important . in this paper we have studied the dynamics of dark energy in a lorentz - violating model with the action given in ( [ lag ] ) . the model involves a vector field @xmath23 and a scalar field @xmath11 with mass terms @xmath68 and @xmath179 , respectively . the presence of the one - derivative term @xmath275 leads to an interesting dynamics at the ir scales larger than @xmath276 . the phantom equation of state can be realized without having ghosts , tachyons or superluminal modes in the uv region . we have taken into account the contributions of radiation and non - relativistic matter and studied the cosmological evolution of the system . interestingly , there exists a de sitter attractor solution that can be used for the late - time acceleration . the phantom regime is not an attractor , but we have found that in a range of parameters , the phantom stage occurs during the transition from the matter epoch to the final de sitter attractor . as is seen , e.g. , in fig . [ evo1 ] the equation of state parameter @xmath2 of dark energy crosses the cosmological constant boundary towards the phantom region . we clarified the conditions under which the @xmath0 crossing is realized together with the existence of the stable de sitter solution . in the model studied in this paper , the effective newton s constant is time - dependent . we have found , however , that this dependence is typically mild , though for interesting values of parameters it is close to the experimental bounds . we have also considered the field perturbations in minkowski spacetime and obtained the momentum scales of instabilities present in the ir region ( @xmath14 ) . we have found that either tachyons or ghosts appear for the spatial momenta @xmath13 smaller than @xmath228 , while in the uv region there are no unhealthy modes . in the cosmological context the presence of tachyons at the ir scales leads to the amplification of large - scale field perturbations whose wavelengths are roughly comparable to the present hubble radius . there are two tachyonic regions of spatial momenta in this model : ( a ) one is sub - horizon and its momenta are characterized by @xmath217 ; ( b ) another is super - horizon and has @xmath219 . in the region ( a ) we derived the condition under which the perturbations always remain smaller than the homogenous fields , see eq . ( [ notexceed ] ) . while the existence of the phantom phase requires that @xmath277 , the condition ( [ notexceed ] ) shows that @xmath12 can not be very much larger than @xmath68 . thus the allowed range of @xmath12 is constrained to be relatively narrow . in the tachyonic region ( b ) the growth of the perturbations is estimated as @xmath278 . since the growth rate is suppressed by the factor @xmath279 , this effect is negligible in the past and at present , though the inhomogeneities can start to dominate over the homogenous fields in the distant future . there are several issues yet to be understood . the presence of the tachyonic instability on sub - horizon scales may lead to the variation of the gravitational potential , which can be an additional source of the late - time integrated saches - wolfe effect on the cmb power spectrum . another property of this model is the peculiar time - dependence of the effective newton s constant , which may result in interesting phenomenology . the model studied in this paper is likely to belong to a wider class of lorentz - violating theories exhibiting the phantom behaviour ( see refs . @xcite for a number of lorentz - violating models ) . it would be interesting to understand how generic are the features we found in our particular model late - time de sitter attractor , transient phantom stage , time - dependent newton s constant , sub - horizon tachyons , super - horizon ghosts , etc . one more direction is to modify our model in such a way that it would be capable of describing inflationary epoch rather than the late - time acceleration . since it is known that the spectra of scalar and tensor perturbations produced during the phantom inflationary phase are typically blue - tilted @xcite , this model may give rise to some distinct features in the cmb spectrum . this work is supported by rfbr grant ( m.l . and v.r . , 05 - 02 - 17363-a ) , grant of the president of russian federation ( m.l . and v.r . , ns-7293.2006.2 ) , intas grant ( m.l . , ysf 04 - 83 - 3015 ) , grant of dynasty foundation awarded by the scientific board of icfpm ( m.l . ) , the european union through the marie curie research and training network universenet ( e.p . , mrtn - ct-2006 - 035863 ) and jsps ( s.t . , no.30318802 ) . we are going to find the effective newton s constant that determines the strength of gravitational interactions at distances shorter than all scales present in our model , including @xmath276 , @xmath280 , @xmath281 as well as the hubble distance . to this end , we neglect the last two terms in the action ( [ lag ] ) , and also neglect the time dependence of the background fields @xmath4 and @xmath40 . we also neglect the space - time curvature of the universe , and therefore consider our model in minkowski space - time . let us impose the gauge @xmath282 , where @xmath283 is the metric perturbation about the minkowski background . then the quadratic lagrangian for perturbations of metric , vector and scalar fields is readily calculated , @xmath284 \nonumber \\ & & + \frac{1}{2 } \gamma \left [ \left(\dot{b_0 } + \frac{1}{2 } x \dot{h}_{00 } \right)^2 - \left(\partial_i b_0 + \frac{1}{2 } x \dot{h}_{00 } \right)^2 \right ] \nonumber \\ & & + \frac{1}{2 } \left [ \dot{\varphi}^2 - ( \partial_i \varphi)^2\right]\,.\end{aligned}\ ] ] where @xmath285 is the background value . clearly the scalar field @xmath166 decouples in our approximation , so we will not consider it in what follows . by varying the quadratic action with respect to @xmath286 and @xmath287 , one obtains @xmath288- and @xmath289-components of the linearized energy - momentum tensor for perturbations ( note that @xmath290 ) . specifying further to scalar perturbations with @xmath291 and choosing conformal newtonian gauge , @xmath292 , @xmath293 , one obtains ( we keep the standard notation for the newtonian potential , even though the same notation was used for the original scalar field in the main text ) @xmath294 where @xmath295 and @xmath296 . the field equations for @xmath165 and @xmath297 in the absence of sources for these fields read @xmath298 now , the longitudinal ( proportional to @xmath299 ) part of the @xmath289-component of the einstein equations , in the absence of external anisotropic stresses , gives @xmath300 while the trace part and @xmath288-component are @xmath301 where @xmath302 and @xmath303 are energy density and pressure of an external source . for time - independent , pressureless source it is consistent to take all perturbations independent of time and set @xmath304 . then one finds , as usual , @xmath305 and obtains the following equation for the newtonian potential , @xmath306 thus , the effective newton s constant in the background field @xmath40 is @xmath307 this means that the effective planck mass entering the newton s law is given by ( [ newton ] ) . let us define the following dimensionless variables which are convenient for studying the dynamical system @xcite : @xmath308 then we obtain the following autonomous equations @xmath309 and @xmath310 where prime denotes the derivative with respect to @xmath311 equation ( [ fr ] ) gives the constraint @xmath312 note that the above equations are invariant under the simultaneous change of the signs of @xmath4 and @xmath40 . hence it is not restrictive to study the case of positive @xmath4 . note also that we study the case of an expanding universe with @xmath313 . the point ( b ) is also in some sense a de sitter point . it exists even in the absence of the field @xmath40 and satisfies the relation @xmath321 . to reach the solution ( b ) , the hubble parameter needs to increase towards infinity ( @xmath322 ) , and the field @xmath4 needs to diverge as well . a cosmologically viable trajectory starts from the radiation point ( d ) , connects to the matter solution ( c ) and finally approaches the de sitter point ( a ) . [ note that the initial data ( [ * * initial ] ) indeed correspond to @xmath329 as @xmath330 . ] to see that this sequence of events is indeed possible , let us study the stability of the fixed points against perturbations . let us consider linear perturbations @xmath331 . by perturbing eqs . ( [ ba1])-([ba5 ] ) we obtain @xmath332 where @xmath333 with @xmath334 the stability of the de sitter point ( a ) is important for having the late - time accelerated epoch . this depends upon the two ratios @xmath336 and @xmath337 once the parameters @xmath26 and @xmath39 are fixed . when @xmath106 and @xmath107 , for example , the parameter range of @xmath336 is determined by the ratio @xmath337 . we find that the point ( a ) is a stable attractor if the following conditions hold : for another de sitter point ( b ) , the eigenvalues are @xmath343 this means that this point is marginally stable . the zero eigenvalue comes from the perturbation equation for @xmath344 . if @xmath253 continues to increase toward the solution ( b ) , this eigenvalue actually obtains a small negative value , as can be seen from eq . ( [ delx5 ] ) . thus in such a case the point ( b ) is stable . however , we know that the phantom phase is realized only for finite field values bounded by @xmath102 , see eq . ( [ philessphia ] ) . hence it is not possible that the actual solutions approach the point ( b ) with infinite @xmath253 and @xmath4 . the matter point ( c ) has the eigenvalues @xmath345 which shows that the matter era corresponds to a saddle point with one positive eigenvalue . hence the solutions eventually repel away from this fixed point even if they temporarily approach it . a. g. riess _ et al . _ , astron . j. * 116 * , 1009 ( 1998 ) ; astron . j. * 117 * , 707 ( 1999 ) ; s. perlmutter _ et al . _ , astrophys . j. * 517 * , 565 ( 1999 ) ; p. astier _ et al . _ , astron . astrophys . * 447 * , 31 ( 2006 ) ; w. m. wood - vasey _ et al . _ , arxiv : astro - ph/0701041 . v. sahni and a. a. starobinsky , int . j. mod . d * 9 * , 373 ( 2000 ) ; v. sahni , lect . notes phys . * 653 * , 141 ( 2004 ) [ arxiv : astro - ph/0403324 ] ; s. m. carroll , living rev . rel . * 4 * , 1 ( 2001 ) ; t. padmanabhan , phys . rept . * 380 * , 235 ( 2003 ) ; p. j. e. peebles and b. ratra , rev . phys . * 75 * , 559 ( 2003 ) ; v. sahni and a. starobinsky , int . j. mod . d * 15 * 2105 ( 2006 ) ; s. nojiri and s. d. odintsov , int . j. geom . meth . mod . phys . * 4 * , 115 ( 2007 ) . a. melchiorri , l. mersini - houghton , c. j. odman and m. trodden , phys . d * 68 * , 043509 ( 2003 ) ; j. weller and a. m. lewis , mon . not . soc . * 346 * , 987 ( 2003 ) ; b. a. bassett , p. s. corasaniti and m. kunz , astrophys . j. * 617 * , l1 ( 2004 ) ; u. seljak , a. slosar and p. mcdonald , jcap * 0610 * , 014 ( 2006 ) . u. alam , v. sahni , t. d. saini and a. a. starobinsky , mon . not . soc . * 354 * , 275 ( 2004 ) ; r. lazkoz , s. nesseris and l. perivolaropoulos , jcap * 0511 * , 010 ( 2005 ) ; j. q. xia , g. b. zhao , b. feng , h. li and x. zhang , phys . rev . d * 73 * , 063521 ( 2006 ) ; s. nesseris and l. perivolaropoulos , jcap * 0701 * , 018 ( 2007 ) ; u. alam , v. sahni and a. a. starobinsky , jcap * 0702 * , 011 ( 2007 ) ; g. b. zhao , j. q. xia , b. feng and x. zhang , arxiv : astro - ph/0603621 . p. singh , m. sami and n. dadhich , phys . d * 68 * 023522 ( 2003 ) ; s. nojiri and s. d. odintsov , phys . b * 562 * , 147 ( 2003 ) ; phys . b * 565 * , 1 ( 2003 ) ; j. g. hao and x. z. li , phys . d * 67 * , 107303 ( 2003 ) ; p. f. gonzalez - diaz , phys . d * 68 * , 021303 ( 2003 ) ; l. p. chimento and r. lazkoz , phys . lett . * 91 * , 211301 ( 2003 ) ; m. p. dabrowski , t. stachowiak and m. szydlowski , phys . d * 68 * , 103519 ( 2003 ) ; l. p. chimento and r. lazkoz , phys . lett . * 91 * , 211301 ( 2003 ) ; s. tsujikawa , class.quant . grav . * 20 * , 1991 ( 2003 ) ; m. sami and a. toporensky , mod . a * 19 * , 1509 ( 2004 ) ; e. elizalde , s. nojiri and s. d. odintsov , phys . rev . d * 70 * , 043539 ( 2004 ) ; h. stefancic , phys . b * 586 * , 5 ( 2004 ) ; v. b. johri , phys . d * 70 * , 041303 ( 2004 ) ; z. k. guo , y. s. piao and y. z. zhang , phys . b * 594 * , 247 ( 2004 ) ; j. m. aguirregabiria , l. p. chimento and r. lazkoz , phys . rev . d * 70 * , 023509 ( 2004 ) ; s. nojiri , s. d. odintsov and s. tsujikawa , phys . d * 71 * , 063004 ( 2005 ) ; s. nojiri and s. d. odintsov , phys . rev . d * 72 * , 023003 ( 2005 ) ; l. perivolaropoulos , phys . rev.d * 71 * , 063503 ( 2005 ) ; t. chiba , jcap * 0503 * , 008 ( 2005 ) ; m. bouhmadi - lopez and j. a. jimenez madrid , jcap * 0505 * , 005 ( 2005 ) ; v. faraoni , class . * 22 * , 3235 ( 2005 ) ; l. p. chimento , phys . b * 633 * , 9 ( 2006 ) ; l. p. chimento and d. pavon , phys . d * 73 * , 063511 ( 2006 ) ; o. hrycyna and m. szydlowski , arxiv:0704.1651 [ hep - th ] . a. vikman , phys . d * 71 * , 023515 ( 2005 ) ; z. k. guo , y. s. piao , x. m. zhang and y. z. zhang , phys . b * 608 * , 177 ( 2005 ) ; w. hu , phys . d * 71 * , 047301 ( 2005 ) ; r. r. caldwell and m. doran , phys . d * 72 * , 043527 ( 2005 ) ; h. wei , r. g. cai and d. f. zeng , class . * 22 * , 3189 ( 2005 ) ; g. b. zhao , j. q. xia , m. li , b. feng and x. zhang , phys . rev . d * 72 * , 123515 ( 2005 ) ; s. tsujikawa , phys . d * 72 * , 083512 ( 2005 ) ; i. y. arefeva , a. s. koshelev and s. y. vernov , phys . rev . d * 72 * , 064017 ( 2005 ) ; b. mcinnes , nucl . b * 718 * , 55 ( 2005 ) ; l. p. chimento , r. lazkoz , r. maartens and i. quiros , jcap * 0609 * , 004 ( 2006 ) ; r. lazkoz and g. leon , phys . b * 638 * , 303 ( 2006 ) ; x. f. zhang and t. qiu , phys . b * 642 * , 187 ( 2006 ) ; w. zhao , phys . d * 73 * , 123509 ( 2006 ) ; h. mohseni sadjadi and m. alimohammadi , phys . d * 74 * , 043506 ( 2006 ) ; z. k. guo , y. s. piao , x. zhang and y. z. zhang , phys . d * 74 * , 127304 ( 2006 ) ; i. y. arefeva and a. s. koshelev , jhep * 0702 * , 041 ( 2007 ) ; y. f. cai , m. z. li , j. x. lu , y. s. piao , t. t. qiu and x. m. zhang , arxiv : hep - th/0701016 ; y. f. cai , t. qiu , y. s. piao , m. li and x. zhang , arxiv:0704.1090 [ gr - qc ] . j. p. uzan , phys . d * 59 * , 123510 ( 1999 ) ; l. amendola , phys . d * 60 * , 043501 ( 1999 ) ; t. chiba , phys . d * 60 * , 083508 ( 1999 ) ; n. bartolo and m. pietroni , phys . rev . d * 61 * , 023518 ( 2000 ) ; f. perrotta , c. baccigalupi and s. matarrese , phys . d * 61 * , 023507 ( 2000 ) ; a. riazuelo and j. p. uzan , phys . d * 66 * , 023525 ( 2002 ) ; d. f. torres , phys . d * 66 * , 043522 ( 2002 ) ; l. perivolaropoulos , jcap * 0510 * , 001 ( 2005 ) ; m. x. luo and q. p. su , phys . b * 626 * , 7 ( 2005 ) ; j. martin , c. schimd and j. p. uzan , phys . lett . * 96 * , 061303 ( 2006 ) . s. nojiri , s. d. odintsov and m. sasaki , phys . d * 71 * , 123509 ( 2005 ) ; m. sami , a. toporensky , p. v. tretjakov and s. tsujikawa , phys . b * 619 * , 193 ( 2005 ) ; g. calcagni , s. tsujikawa and m. sami , class . quant . * 22 * , 3977 ( 2005 ) ; b. m. n. carter and i. p. neupane , jcap * 0606 * , 004 ( 2006 ) . l. amendola , r. gannouji , d. polarski and s. tsujikawa , phys . d * 75 * , 083504 ( 2007 ) . see also l. amendola , d. polarski and s. tsujikawa , phys . lett . * 98 * , 131302 ( 2007 ) ; arxiv : astro - ph/0605384 . t. koivisto and d. f. mota , phys . b * 644 * , 104 ( 2007 ) ; s. tsujikawa and m. sami , jcap * 0701 * , 006 ( 2007 ) ; t. koivisto and d. f. mota , phys . rev . d * 75 * , 023518 ( 2007 ) ; b. m. leith and i. p. neupane , arxiv : hep - th/0702002 . a. de felice , m. hindmarsh and m. trodden , jcap * 0608 * , 005 ( 2006 ) ; g. calcagni , b. de carlos and a. de felice , nucl . b * 752 * , 404 ( 2006 ) ; z. k. guo , n. ohta and s. tsujikawa , phys . rev . d * 75 * , 023520 ( 2007 ) . l. senatore , phys . d * 71 * , 043512 ( 2005 ) ; p. creminelli , m. a. luty , a. nicolis and l. senatore , jhep * 0612 * , 080 ( 2006 ) . v. a. rubakov , theor . phys . * 149 * , 1651 ( 2006 ) [ arxiv : hep - th/0604153 ] . j. m. cline and l. valcarcel , jhep * 0403 * , 032 ( 2004 ) ; n. arkani - hamed , h. c. cheng , m. luty and j. thaler , jhep * 0507 * , 029 ( 2005 ) ; m. v. libanov and v. a. rubakov , jcap * 0509 * , 005 ( 2005 ) ; phys . rev . d * 72 * , 123503 ( 2005 ) ; s. m. carroll and j. shu , phys . rev . d * 73 * , 103515 ( 2006 ) ; h. c. cheng , m. a. luty , s. mukohyama and j. thaler , jhep * 0605 * , 076 ( 2006 ) ; s. kanno and j. soda , phys . d * 74 * , 063505 ( 2006 ) ; p. g. ferreira , b. m. gripaios , r. saffari and t. g. zlosnik , phys . d * 75 * , 044014 ( 2007 ) .

we obtain necessary optimality conditions for variational problems with a lagrangian depending on a caputo fractional derivative , a fractional and an indefinite integral . main results give fractional euler lagrange type equations and natural boundary conditions , which provide a generalization of previous results found in the literature . isoperimetric problems , problems with holonomic constraints and depending on higher - order caputo derivatives , as well as fractional lagrange problems , are considered . 49k05 , 49s05 , 26a33 , 34a08 . calculus of variations , fractional calculus , caputo derivatives , fractional necessary optimality equations .

1102.3360

in the 18th century , euler considered the problem of optimizing functionals depending not only on some unknown function @xmath0 and some derivative of @xmath0 , but also on an antiderivative of @xmath0 ( see @xcite ) . similar problems have been recently investigated in @xcite , where lagrangians containing higher - order derivatives and optimal control problems are considered . more generally , it has been shown that the results of @xcite hold on an arbitrary time scale @xcite . here we study such problems within the framework of fractional calculus . roughly speaking , a fractional calculus defines integrals and derivatives of non - integer order . let @xmath1 be a real number and @xmath2 be such that @xmath3 . here we follow @xcite and @xcite . let @xmath4\to\mathbb{r}$ ] be piecewise continuous on @xmath5 and integrable on @xmath6 $ ] . the left and right riemann liouville fractional integrals of @xmath7 of order @xmath8 are defined respectively by @xmath9 here @xmath10 is the well - known gamma function . then the left @xmath11 and right @xmath12 riemann liouville fractional derivatives of @xmath7 of order @xmath8 are defined ( if they exist ) as @xmath13 and @xmath14 the fractional derivatives and have one disadvantage when modeling real world phenomena : the fractional derivative of a constant is not zero . to eliminate this problem , one often considers fractional derivatives in the sense of caputo . let @xmath7 belong to the space @xmath15;\mathbb{r})$ ] of absolutely continuous functions . the left and right caputo fractional derivatives of @xmath7 of order @xmath8 are defined respectively by @xmath16 and @xmath17 these fractional integrals and derivatives define a rich calculus . for details see the books @xcite . here we just recall a useful property for our purposes : integration by parts . for fractional integrals , @xmath18 ( see , , ( * ? ? ? * lemma 2.7 ) ) , and for caputo fractional derivatives @xmath19_a^b\ ] ] ( see , , ( * ? ? ? * eq . ( 16 ) ) ) . in particular , for @xmath20 one has @xmath21_a^b.\ ] ] when @xmath22 , @xmath23 , @xmath24 , @xmath25 is the identity operator , and gives the classical formula of integration by parts . the fractional calculus of variations concerns finding extremizers for variational functionals depending on fractional derivatives instead of integer ones . the theory started in 1996 with the work of riewe , in order to better describe non - conservative systems in mechanics @xcite . the subject is now under strong development due to its many applications in physics and engineering , providing more accurate models of physical phenomena ( see , , @xcite ) . with respect to results on fractional variational calculus via caputo operators , we refer the reader to @xcite and references therein . our main contribution is an extension of the results presented in @xcite by considering lagrangians containing an antiderivative , that in turn depend on the unknown function , a fractional integral , and a caputo fractional derivative ( section [ sec : fundprob ] ) . transversality conditions are studied in section [ sec : natbound ] , where the variational functional @xmath26 depends also on the terminal time @xmath27 , @xmath28 , and where we obtain conditions for a pair @xmath29 to be an optimal solution to the problem . in section [ sec : isoprob ] we consider isoperimetric problems with integral constraints of the same type as the cost functionals considered in section [ sec : fundprob ] . fractional problems with holonomic constraints are considered in section [ sec : holonomic ] . the situation when the lagrangian depends on higher order caputo derivatives , , it depends on @xmath30 for @xmath31 , @xmath32 , is studied in section [ sec : higher ] , while section [ sec : fracopt ] considers fractional lagrange problems and the hamiltonian approach . in section [ sec : sufconditions ] we obtain sufficient conditions of optimization under suitable convexity assumptions on the lagrangian . we end with section [ sec : numsim ] , discussing a numerical scheme for solving the proposed fractional variational problems . the idea is to approximate fractional problems by classical ones . numerical results for two illustrative examples are described in detail . let @xmath20 and @xmath33 . the problem that we address is stated in the following way . minimize the cost functional @xmath34 where the variable @xmath35 is defined by @xmath36 subject to the boundary conditions @xmath37 we assume that the functions @xmath38 and @xmath39 are of class @xmath40 , and the trajectories @xmath41\to\mathbb{r}$ ] are absolute continuous functions , @xmath42;\mathbb{r})$ ] , such that @xmath43 and @xmath44 exist and are continuous on @xmath6 $ ] . we denote such class of functions by @xmath45;\mathbb{r})$ ] . also , to simplify , by @xmath46 $ ] and @xmath47 we denote the operators @xmath48(x)=(x , y(x),{^c_ad_x^\alpha}y(x),{_ai_x^\beta}y(x),z(x ) ) \quad \mbox{and}\quad \{y\}(x)=(x , y(x),{^c_ad_x^\alpha}y(x),{_ai_x^\beta}y(x)).\ ] ] [ elteo ] let @xmath49;\mathbb{r})$ ] be a minimizer of @xmath26 as in , subject to the boundary conditions . then , for all @xmath50 $ ] , @xmath0 is a solution of the fractional equation @xmath51(x ) + { _ xd^\alpha_b}\left ( \frac{\partial l}{\partial v}[y](x ) \right ) + { _ xi_b^\beta}\left(\frac{\partial l}{\partial w}[y](x)\right ) + \int_x^b \frac{\partial l}{\partial z}[y](t)dt\cdot \frac{\partial l}{\partial y}\{y\}(x)\\ + { _ xd^\alpha_b}\left ( \int_x^b \frac{\partial l}{\partial z}[y](t)dt \cdot \frac{\partial l}{\partial v}\{y\}(x ) \right ) + { _ xi^\beta_b}\left ( \int_x^b \frac{\partial l}{\partial z}[y](t)dt \cdot \frac{\partial l}{\partial w}\{y\}(x ) \right)=0.\end{gathered}\ ] ] let @xmath52;\mathbb{r})$ ] be such that @xmath53 , and @xmath54 be a real number with @xmath55 . if we define @xmath56 as @xmath57 , then @xmath58 . differentiating @xmath56 at @xmath59 , we get @xmath60(x)h(x ) + \frac{\partial l}{\partial v}[y](x){^c_ad^\alpha_x}h(x ) + \frac{\partial l}{\partial w}[y](x){_ai^\beta_x}h(x)\right.\\ \left.+\frac{\partial l}{\partial z}[y](x)\int_a^x\left ( \frac{\partial l}{\partial y}\{y\}(t)h(t ) + \frac{\partial l}{\partial v}\{y\}(t){^c_ad^\alpha_t}h(t ) + \frac{\partial l}{\partial w}\{y\}(t){_ai^\beta_t}h(t)\right)dt\right]dx=0.\end{gathered}\ ] ] the necessary condition follows from the next relations and the fundamental lemma of the calculus of variations ( , , @xcite ) : @xmath61(x){^c_ad^\alpha_x}h(x ) dx = \int_a^b { _ x d_b^\alpha } \left(\frac{\partial l}{\partial v}[y](x ) \right)h(x)dx + \left[{_xi_b^{1-\alpha}}\left(\frac{\partial l}{\partial v}[y](x ) \right ) h(x)\right]_a^b,\ ] ] @xmath62(x){_ai^\beta_x}h(x ) dx = \int_a^b { _ x i_b^\beta } \left(\frac{\partial l}{\partial w}[y](x ) \right)h(x)dx,\ ] ] @xmath63(x)\left(\int_a^x \frac{\partial l}{\partial y}\{y\}(t)h(t ) dt \right ) dx = \int_a^b \left ( -\frac{d}{dx}\int_x^b\frac{\partial l}{\partial z}[y](t)dt \right ) \left ( \int_a^x \frac{\partial l}{\partial y}\{y\}(t)h(t ) dt \right ) dx\\ & = \left[- \left(\int_x^b\frac{\partial l}{\partial z}[y](t)dt \right)\left(\int_a^x \frac{\partial l}{\partial y}\{y\}(t)h(t ) dt\right ) \right]_a^b + \int_a^b \left(\int_x^b\frac{\partial l}{\partial z}[y](t)dt \right ) \frac{\partial l}{\partial y}\{y\}(x)h(x ) \ , dx\\ & = \int_a^b \left(\int_x^b\frac{\partial l}{\partial z}[y](t)dt \right ) \frac{\partial l}{\partial y}\{y\}(x)h(x ) \ , dx,\end{aligned}\ ] ] @xmath63(x)\left(\int_a^x \frac{\partial l}{\partial v}\{y\}(t){^c_ad^\alpha_t}h(t ) dt \right ) dx = \int_a^b \left ( -\frac{d}{dx}\int_x^b\frac{\partial l}{\partial z}[y](t)dt \right ) \left ( \int_a^x \frac{\partial l}{\partial v}\{y\}(t){^c_ad^\alpha_t}h(t ) dt \right ) dx\\ & = \left[- \left(\int_x^b\frac{\partial l}{\partial z}[y](t)dt \right)\left(\int_a^x \frac{\partial l}{\partial v}\{y\}(t){^c_ad^\alpha_t}h(t ) dt\right ) \right]_a^b + \int_a^b \left(\int_x^b\frac{\partial l}{\partial z}[y](t)dt \right ) \frac{\partial l}{\partial v}\{y\}(x){^c_ad^\alpha_x}h(x ) \ , dx\\ & = \int_a^b { _ xd^\alpha_b}\left(\int_x^b\frac{\partial l}{\partial z}[y](t)dt \frac{\partial l}{\partial v}\{y\}(x)\right)h(x ) \ , dx + \left [ { _ xi^{1-\alpha}_b}\left(\int_x^b\frac{\partial l}{\partial z}[y](t)dt \frac{\partial l}{\partial v}\{y\}(x ) \right ) h(x ) \right]_a^b,\end{aligned}\ ] ] and @xmath64(x)\left(\int_a^x \frac{\partial l}{\partial w}\{y\}(t){_ai^\beta_t}h(t ) dt \right ) dx = \int_a^b { _ xi^\beta_b}\left(\int_x^b\frac{\partial l}{\partial z}[y](t)dt \frac{\partial l}{\partial w}\{y\}(x)\right)h(x ) \ , dx.\ ] ] the fractional euler lagrange equation involves not only fractional integrals and fractional derivatives , but also indefinite integrals . theorem [ elteo ] gives a necessary condition to determine the possible choices for extremizers . solutions to the fractional euler lagrange equation are called extremals for @xmath26 defined by . [ example : bra ] consider the functional @xmath65dx,\ ] ] where @xmath66 and @xmath67 defined on the set @xmath68;\mathbb{r } ) : \ , y(0)=0 \ , \mbox { and } \ , y(1)=1\right\}.\ ] ] let @xmath69.\ ] ] then , @xmath70 since @xmath71 for all admissible functions @xmath0 , and @xmath72 , we have that @xmath73 is a minimizer of @xmath26 . the euler lagrange equation applied to gives @xmath74 obviously , @xmath73 is a solution of the fractional differential equation . the extremizer of example [ example : bra ] is smooth on the closed interval @xmath75 $ ] . this is not always the case . as next example shows , minimizers of are not necessarily @xmath40 functions . [ example1 ] consider the following fractional variational problem : to minimize the functional @xmath76dx\ ] ] on @xmath77;\mathbb{r } ) \ , : \ , y(0)=0 \quad \mbox{and } \quad y(1)=\frac{1}{\gamma(\alpha+1)}\right\},\ ] ] where @xmath35 is given by @xmath78 since @xmath79 , we deduce easily that function @xmath80 is the global minimizer to the problem . indeed , @xmath81 for all @xmath0 , and @xmath82 . let us see that @xmath83 is an extremal for @xmath26 . the fractional euler lagrange equation becomes @xmath84 obviously , @xmath83 is a solution of equation . the minimizer of example [ example1 ] is not differentiable at 0 , as @xmath85 . however , @xmath86 and @xmath87 for any @xmath88 $ ] . if @xmath0 is a minimizer of @xmath89 subject to the boundary conditions , then @xmath0 is a solution of the fractional equation @xmath90(x ) + { _ xd^\alpha_b}\left ( \frac{\partial l}{\partial v}[y](x ) \right)=0.\ ] ] follows from theorem [ elteo ] with an @xmath91 that does not depend on @xmath92 and @xmath35 . we now derive the euler lagrange equations for functionals containing several dependent variables , , for functionals of type @xmath93 where @xmath94 and @xmath35 is defined by @xmath95 subject to the boundary conditions @xmath96 to simplify , we consider @xmath0 as the vector @xmath97 . consider a family of variations @xmath98 , where @xmath55 and @xmath99 . the boundary conditions imply that @xmath100 , for @xmath32 . the following theorem can be easily proved . [ eleqmult ] let @xmath0 be a minimizer of @xmath26 as in , subject to the boundary conditions . then , for all @xmath32 and for all @xmath50 $ ] , @xmath0 is a solution of the fractional euler lagrange equation @xmath101(x)+{_xd^\alpha_b}\left ( \frac{\partial l}{\partial v_k}[y](x ) \right)+ { _ xi^\beta_b}\left ( \frac{\partial l}{\partial w_k}[y](x ) \right)+\int_x^b \frac{\partial l}{\partial z}[y](t)dt\cdot \frac{\partial l}{\partial y_k}\{y\}(x)\\ + { _ xd^\alpha_b}\left ( \int_x^b \frac{\partial l}{\partial z}[y](t)dt\cdot \frac{\partial l}{\partial v_k}\{y\}(x ) \right ) + { _ xi^\beta_b}\left ( \int_x^b \frac{\partial l}{\partial z}[y](t)dt\cdot \frac{\partial l}{\partial w_k}\{y\}(x ) \right)=0.\end{gathered}\ ] ] in this section we consider a more general question . not only the unknown function @xmath0 is a variable in the problem , but also the terminal time @xmath27 is an unknown . for @xmath102 $ ] , consider the functional @xmath103(x)dx,\ ] ] where @xmath48(x)=(x , y(x),{^c_ad_x^\alpha}y(x),{_ai_x^\beta}y(x),z(x)).\ ] ] the problem consists in finding a pair @xmath104;\mathbb{r})\times [ a , b]$ ] for which the functional @xmath26 attains a minimum value . first we give a remark that will be used later in the proof of theorem [ teonatural ] . [ remarkintegral ] if @xmath105 is a continuous function , then ( @xcite ) @xmath106 for any @xmath66 . [ teonatural ] let @xmath29 be a minimizer of @xmath26 as in . then , for all @xmath107 $ ] , @xmath29 is a solution of the fractional equation @xmath108(x)+{_xd^\alpha_t}\left ( \frac{\partial l}{\partial v}[y](x ) \right)+{_xi_t^\beta } \left(\frac{\partial l}{\partial w}[y](x)\right ) + \int_x^t \frac{\partial l}{\partial z}[y](t)dt \cdot \frac{\partial l}{\partial y}\{y\}(x)\\ + { _ xd^\alpha_t}\left ( \int_x^t \frac{\partial l}{\partial z}[y](t)dt \cdot \frac{\partial l}{\partial v}\{y\}(x ) \right ) + { _ xi^\beta_t}\left ( \int_x^t \frac{\partial l}{\partial z}[y](t)dt \cdot \frac{\partial l}{\partial w}\{y\}(x ) \right)=0\end{gathered}\ ] ] and satisfies the transversality conditions @xmath109(x ) + \int_x^t \frac{\partial l}{\partial z}[y](t)\ , dt \cdot \frac{\partial l}{\partial v}\{y\}(x ) \right)\right]_{x = a}=0\ ] ] and @xmath110(t)=0.\ ] ] let @xmath52;\mathbb{r})$ ] be a variation , and let @xmath111 be a real number . define the function @xmath112 with @xmath55 . differentiating @xmath56 at @xmath59 , and using the same procedure as in theorem [ elteo ] , we deduce that @xmath113(t)+\int_a^t \left[\frac{\partial l}{\partial y}[y](x ) + { _ xd^\alpha_t}\left ( \frac{\partial l}{\partial v}[y](x ) \right ) + { _ xi_t^\beta}\left(\frac{\partial l}{\partial w}[y](x)\right ) + \int_x^t \frac{\partial l}{\partial z}[y](t)dt\cdot \frac{\partial l}{\partial y}\{y\}(x)\right.\\ \left.+{_xd^\alpha_t}\left ( \int_x^t \frac{\partial l}{\partial z}[y](t)dt \cdot \frac{\partial l}{\partial v}\{y\}(x)\right ) + { _ xi^\beta_t}\left ( \int_x^t \frac{\partial l}{\partial z}[y](t)dt \cdot \frac{\partial l}{\partial w}\{y\}(x ) \right)\right]h(x)dx\\ + \left [ { _ xi_t^{1-\alpha}}\left(\frac{\partial l}{\partial v}[y](x ) \right ) h(x ) \right]_a^t+\left [ { _ xi^{1-\alpha}_t } \left(\int_x^t\frac{\partial l}{\partial z}[y](t)dt \cdot \frac{\partial l}{\partial v}\{y\}(x ) \right ) h(x ) \right]_a^t=0.\end{gathered}\ ] ] the theorem follows from the arbitrariness of @xmath114 and @xmath111 . if @xmath27 is fixed , say @xmath115 , then @xmath116 and the transversality conditions reduce to @xmath117(x ) + \int_x^b \frac{\partial l}{\partial z}[y](t)\ , dt \cdot \frac{\partial l}{\partial v}\{y\}(x ) \right)\right]_a=0.\ ] ] consider the problem of minimizing the functional @xmath26 as in , but without given boundary conditions . besides equation , extremals must also satisfy @xmath118_0=0.\ ] ] again , @xmath119 given by is a solution of and . as a particular case , the following result of @xcite is deduced . [ cor : agra1 ] if @xmath0 is a minimizer of @xmath26 as in , then @xmath0 is a solution of @xmath90(x ) + { _ xd^\alpha_b}\left ( \frac{\partial l}{\partial v}[y](x ) \right)=0\ ] ] and satisfies the transversality condition @xmath120(x)\right)\right]_a=0.\ ] ] the lagrangian @xmath91 in does not depend on @xmath92 and @xmath35 , and the result follows from theorem [ teonatural ] . [ new : rem:6 ] observe that the condition @xmath120(x)\right)\right]_b=0\ ] ] is implicitly satisfied in corollary [ cor : agra1 ] ( remark [ remarkintegral ] ) . an isoperimetric problem deals with the question of optimizing a given functional under the presence of an integral constraint . this is a very old question , with its origins in the ancient greece . they where interested in determining the shape of a closed curve with a fixed length and maximum area . this problem is known as dido s problem , and is an example of an isoperimetric problem of the calculus of variations @xcite . for recent advancements on the subject we refer the reader to @xcite and references therein . in our case , within the fractional context , we state the isoperimetric problem in the following way . determine the minimizers of a given functional @xmath121 subject to the boundary conditions @xmath122 and the fractional integral constraint @xmath123 where @xmath35 is defined by @xmath124 as usual , we assume that all the functions @xmath38 , @xmath39 , and @xmath125 are of class @xmath40 . [ isopro ] let @xmath0 be a minimizer of @xmath26 as in , under the boundary conditions and isoperimetric constraint . suppose that @xmath0 is not an extremal for @xmath126 in . then there exists a constant @xmath127 such that @xmath0 is a solution of the fractional equation @xmath128(x ) + { _ xd^\alpha_b}\left ( \frac{\partial f}{\partial v}[y](x ) \right ) + { _ xi^\beta_b}\left ( \frac{\partial f}{\partial w}[y](x ) \right ) + \int_x^b \frac{\partial f}{\partial z}[y](t)dt \cdot \frac{\partial l}{\partial y}\{y\}(x)\\ + { _ xd^\alpha_b}\left ( \int_x^b \frac{\partial f}{\partial z}[y](t)dt \cdot \frac{\partial l}{\partial v}\{y\}(x)\right ) + { _ xi^\beta_b}\left ( \int_x^b \frac{\partial f}{\partial z}[y](t)dt \cdot \frac{\partial l}{\partial w}\{y\}(x ) \right)=0,\end{gathered}\ ] ] where @xmath129 , for all @xmath50 $ ] . let @xmath130 be two real numbers such that @xmath131 and @xmath132 , with @xmath133 free and @xmath134 to be determined later , and let @xmath135 and @xmath136 be two functions satisfying @xmath137 define functions @xmath56 and @xmath138 by @xmath139 and @xmath140 doing analogous calculations as in the proof of theorem [ elteo ] , one has @xmath141(x ) + { _ xd^\alpha_b}\left ( \frac{\partial g}{\partial v}[y](x ) \right ) + { _ xi^\alpha_b}\left ( \frac{\partial g}{\partial w}[y](x ) \right ) + \int_x^b \frac{\partial g}{\partial z}[y](t)dt \cdot \frac{\partial l}{\partial y}\{y\}(x)\right.\\ \left . + { _ xd^\alpha_b}\left ( \int_x^b \frac{\partial g}{\partial z}[y](t)dt \cdot \frac{\partial l}{\partial v}\{y\}(x ) \right ) + { _ xi^\beta_b}\left ( \int_x^b \frac{\partial g}{\partial z}[y](t)dt \cdot \frac{\partial l}{\partial w}\{y\}(x ) \right)\right ] h_2(x ) \ , dx.\end{gathered}\ ] ] by hypothesis , @xmath0 is not an extremal for @xmath126 and therefore there must exist a function @xmath136 for which @xmath142 since @xmath143 , by the implicit function theorem there exists a function @xmath144 , defined in some neighborhood of zero , such that @xmath145 on the other hand , @xmath56 attains a minimum value at @xmath146 when subject to the constraint . because @xmath147 , by the lagrange multiplier rule @xcite there exists a constant @xmath127 such that @xmath148 so @xmath149 differentiating @xmath56 and @xmath138 at zero , and doing the same calculations as before , we get the desired result . using the abnormal lagrange multiplier rule @xcite , the previous result can be generalized to include the case when the minimizer is an extremal of @xmath126 . let @xmath0 be a minimizer of @xmath26 as in , subject to the constraints and . then there exist two constants @xmath150 and @xmath127 , not both zero , such that @xmath0 is a solution of equation @xmath151(x)+{_xd^\alpha_b}\left ( \frac{\partial k}{\partial v}[y](x ) \right ) + { _ xi^\beta_b } \left ( \frac{\partial k}{\partial w}[y](x ) \right ) + \int_x^b \frac{\partial k}{\partial z}[y](t)dt\cdot \frac{\partial l}{\partial y}\{y\}(x)\\ + { _ xd^\alpha_b}\left ( \int_x^b \frac{\partial k}{\partial z}[y](t)dt \cdot \frac{\partial l}{\partial v}\{y\}(x)\right ) + { _ xi^\beta_b}\left ( \int_x^b \frac{\partial k}{\partial z}[y](t)dt \cdot \frac{\partial l}{\partial w}\{y\}(x ) \right)=0\end{gathered}\ ] ] for all @xmath50 $ ] , where @xmath152 . let @xmath0 be a minimizer of @xmath153 subject to the boundary conditions @xmath154 and the isoperimetric constraint @xmath155 then , there exist two constants @xmath150 and @xmath127 , not both zero , such that @xmath0 is a solution of equation @xmath156 for all @xmath50 $ ] , where @xmath152 . moreover , if @xmath0 is not an extremal for @xmath126 , then we may take @xmath157 . in this section we consider the following problem . minimize the functional @xmath158 where @xmath35 is defined by @xmath159 when restricted to the boundary conditions @xmath160 and the holonomic constraint @xmath161 as usual , here @xmath162 @xmath163 and @xmath164 are all smooth . in what follows we make use of the operator @xmath165 $ ] given by @xmath166(x ) = ( x , y_1(x),y_2(x),{^c_ad_x^\alpha}y_1(x),{^c_ad_x^\alpha}y_2(x ) , { _ ai_x^\beta}y_1(x),{_ai_x^\beta}y_2(x),z(x))\ , , \ ] ] we denote @xmath167 by @xmath168 , and the euler lagrange equation obtained in with respect to @xmath169 by @xmath170 , @xmath171 . for simplicity , we are considering functionals depending only on two functions @xmath172 and @xmath173 . theorem [ thm : hol ] is , however , easily generalized for @xmath174 variables @xmath175 . [ thm : hol ] let the pair @xmath176 be a minimizer of @xmath26 as in , subject to the constraints . if @xmath177 , then there exists a continuous function @xmath178\to\mathbb{r}$ ] such that @xmath176 is a solution of @xmath179(x ) + { _ xd^\alpha_b}\left ( \frac{\partial f}{\partial v_i}[y_1,y_2](x ) \right ) + { _ xi_b^\beta}\left(\frac{\partial f}{\partial w_i}[y_1,y_2](x)\right ) + \int_x^b \frac{\partial f}{\partial z}[y_1,y_2](t)dt\cdot \frac{\partial l}{\partial y_i}\{y_1,y_2\}(x)\\ + { _ xd^\alpha_b}\left ( \int_x^b \frac{\partial f}{\partial z}[y_1,y_2](t)dt \cdot \frac{\partial l}{\partial v_i}\{y_1,y_2\}(x ) \right ) + { _ xi^\beta_b}\left ( \int_x^b \frac{\partial f}{\partial z}[y_1,y_2](t)dt \cdot \frac{\partial l}{\partial w_i}\{y_1,y_2\}(x)\right)=0\end{gathered}\ ] ] for all @xmath50 $ ] and @xmath171 , where @xmath180(x)=l[y_1,y_2](x)-\lambda ( x ) g(x,\mathbf{y}(x))$ ] . consider a variation of the optimal solution of type @xmath181 where @xmath182 are two functions defined on @xmath6 $ ] satisfying @xmath183 and @xmath54 is a sufficiently small real parameter . since @xmath184 for all @xmath50 $ ] , we can solve equation @xmath185 with respect to @xmath136 , @xmath186 . differentiating @xmath187 at @xmath59 , and proceeding similarly as done in the proof of theorem [ elteo ] , we deduce that @xmath188 besides , since @xmath185 , differentiating at @xmath59 we get @xmath189 define the function @xmath127 on @xmath6 $ ] as @xmath190 combining and , equation can be written as @xmath191 h_1(x)\ , dx=0.\ ] ] by the arbitrariness of @xmath135 , if follows that @xmath192 define @xmath193 as @xmath194(x)=l[y_1,y_2](x)-\lambda ( x ) g(x,\mathbf{y}(x)).\ ] ] then , equations follow . in this section we consider fractional variational problems , when in presence of higher order caputo derivatives . we will restricted ourselves to the case where the orders are non integer , since the integer case is already well studied in the literature ( for a modern account see @xcite ) . let @xmath2 , @xmath195 , and @xmath196 be such that @xmath31 for @xmath32 . admissible functions @xmath0 belong to @xmath15;\mathbb{r})$ ] and are such that @xmath197 , @xmath198 , and @xmath92 exist and are continuous on @xmath6 $ ] . we denote such class of functions by @xmath199;\mathbb{r})$ ] . for @xmath200 , define the vector @xmath201 the optimization problem is the following : to minimize or maximize the functional @xmath202 @xmath203;\mathbb{r})$ ] , subject to the boundary conditions @xmath204 where @xmath205 \to \mathbb{r}$ ] is defined by @xmath124 [ thm:16 ] if @xmath203;\mathbb{r})$ ] is a minimizer of @xmath26 as in , subject to the boundary conditions , then @xmath0 is a solution of the fractional equation @xmath108(x ) + \sum_{k=1}^n{_xd^{\alpha_k}_b}\left ( \frac{\partial l}{\partial v_k}[y](x ) \right ) + { _ xi^\beta_b}\left ( \frac{\partial l}{\partial w}[y](x ) \right ) + \int_x^b \frac{\partial l}{\partial z}[y](t)dt \cdot \frac{\partial l}{\partial y}\{y\}(x)\\ + \sum_{k=1}^n{_xd^{\alpha_k}_b}\left ( \int_x^b \frac{\partial l}{\partial z}[y](t)dt \cdot \frac{\partial l}{\partial v_k}\{y\}(x)\right ) + { _ xi^\beta_b}\left ( \int_x^b \frac{\partial l}{\partial z}[y](t)dt \cdot \frac{\partial l}{\partial w}\{y\}(x ) \right)=0\end{gathered}\ ] ] for all @xmath50 $ ] , where @xmath206(x ) = \left(x , y(x),{^c_ad_x^\alpha}y(x),{_ai_x^\beta}y(x),z(x)\right)$ ] with @xmath43 as in . let @xmath207;\mathbb{r})$ ] be such that @xmath208 , for @xmath209 . define the new function @xmath56 as @xmath57 . then @xmath210(x)h(x ) + \sum_{k=1}^n \frac{\partial l}{\partial v_k}[y](x){^c_ad^{\alpha_k}_x}h(x ) + \frac{\partial l}{\partial w}[y](x){_ai^\beta_x}h(x)\right.\\ \left.+\frac{\partial l}{\partial z}[y](x)\int_a^x \left ( \frac{\partial l}{\partial y}\{y\}(t)h(t ) + \sum_{k=1}^n\frac{\partial l}{\partial v_k}\{y\}(t){^c_ad^{\alpha_k}_t}h(t ) + \frac{\partial l}{\partial w}\{y\}(t){_ai^\beta_t}h(t ) \right)dt\right]dx=0.\end{gathered}\ ] ] integrating by parts , we get that @xmath211(x){^c_ad^{\alpha_k}_x}h(x ) dx = \int_a^b { _ x d_b^{\alpha_k}}\left(\frac{\partial l}{\partial v_k}[y](x)\right)h(x)dx\\ + \sum_{m=0}^{k-1}\left[{_xd_b^{\alpha_k+m - k}}\left(\frac{\partial l}{\partial v_k}[y](x ) \right ) h^{(k-1-m)}(x)\right]_a^b = \int_a^b { _ x d_b^{\alpha_k}}\left ( \frac{\partial l}{\partial v_k}[y](x ) \right)h(x)dx\end{gathered}\ ] ] for all @xmath32 . moreover , one has @xmath62(x){_ai^\beta_x}h(x ) dx = \int_a^b { _ x i_b^\beta } \left(\frac{\partial l}{\partial w}[y](x ) \right)h(x)dx,\ ] ] @xmath64(x)\int_a^x \frac{\partial l}{\partial y}\{y\}(t)h(t ) dt \ , dx = \int_a^b \left(\int_x^b\frac{\partial l}{\partial z}[y](t)dt \right ) \frac{\partial l}{\partial y}\{y\}(x)h(x ) \ , dx,\ ] ] @xmath63(x)\left(\int_a^x \frac{\partial l}{\partial v_k}\{y\}(t){^c_ad^{\alpha_k}_t}h(t ) dt \right)dx = \int_a^b \left(\int_x^b\frac{\partial l}{\partial z}[y](t)dt \right ) \frac{\partial l}{\partial v_k}\{y\}(x){^c_ad^{\alpha_k}_x}h(x ) \ , dx\\ & = \int_a^b { _ xd^{\alpha_k}_b}\left(\int_x^b\frac{\partial l}{\partial z}[y](t)dt \frac{\partial l}{\partial v_k}\{y\}(x ) \right)h(x ) \ , dx\\ & \qquad + \sum_{m=0}^{k-1}\left[{_xd_b^{\alpha_k+m - k}}\left ( \int_x^b\frac{\partial l}{\partial z}[y](t)dt \frac{\partial l}{\partial v_k}\{y\}(x)\right ) h^{(k-1-m)}(x)\right]_a^b\\ & = \int_a^b { _ xd^{\alpha_k}_b}\left ( \int_x^b\frac{\partial l}{\partial z}[y](t)dt \frac{\partial l}{\partial v_k}\{y\}(x ) \right)h(x ) \ , dx,\end{aligned}\ ] ] and @xmath64(x)\left(\int_a^x \frac{\partial l}{\partial w}\{y\}(t){_ai^\beta_t}h(t ) dt \right)dx=\int_a^b { _ xi^\beta_b } \left(\int_x^b\frac{\partial l}{\partial z}[y](t)dt \frac{\partial l}{\partial w}\{y\}(x ) \right)h(x ) \ , dx.\ ] ] replacing these last relations into equation , and applying the fundamental lemma of the calculus of variations , we obtain the intended necessary condition . we now consider the higher - order problem without the presence of boundary conditions . [ higherordernatural ] if @xmath203;\mathbb{r})$ ] is a minimizer of @xmath26 as in , then @xmath0 is a solution of the fractional equation @xmath108(x ) + \sum_{k=1}^n{_xd^{\alpha_k}_b}\left ( \frac{\partial l}{\partial v_k}[y](x ) \right)+{_xi^\beta_b } \left ( \frac{\partial l}{\partial w}[y](x ) \right ) + \int_x^b \frac{\partial l}{\partial z}[y](t)dt \cdot \frac{\partial l}{\partial y}\{y\}(x)\\ + \sum_{k=1}^n{_xd^{\alpha_k}_b}\left ( \int_x^b \frac{\partial l}{\partial z}[y](t)dt\cdot \frac{\partial l}{\partial v_k}\{y\}(x ) \right ) + { _ xi^\beta_b}\left ( \int_x^b \frac{\partial l}{\partial z}[y](t)dt \cdot \frac{\partial l}{\partial w}\{y\}(x ) \right)=0\end{gathered}\ ] ] for all @xmath50 $ ] , and satisfies the natural boundary conditions @xmath212(x ) + \int_x^b\frac{\partial l}{\partial z}[y](t)dt \frac{\partial l}{\partial v_k}\{y\}(x)\right ) \right]_a^b=0 , \quad \mbox{for all } \quad k\in\{1,\ldots , n\}.\ ] ] the proof follows the same pattern as the proof of theorem [ thm:16 ] . since admissible functions @xmath0 are not required to satisfy given boundary conditions , the variation functions @xmath114 may take any value at the boundaries as well , and thus the condition @xmath213 is no longer imposed _ a priori_. if we consider the first variation of @xmath26 for variations @xmath114 satisfying condition , we obtain the euler lagrange equation . replacing it on the expression of the first variation , we conclude that @xmath214(x ) + \int_x^b\frac{\partial l}{\partial z}[y](t)dt \frac{\partial l}{\partial v_k}\{y\}(x ) \right)h^{(k-1-m)}(x)\right]_a^b=0.\ ] ] to obtain the transversality condition with respect to @xmath215 , for @xmath32 , we consider variations satisfying the condition @xmath216 some of the terms that appear in the natural boundary conditions are equal to zero ( remark [ remarkintegral ] and remark [ new : rem:6 ] ) . we now prove a necessary optimality condition for a fractional lagrange problem , when the lagrangian depends again on an indefinite integral . consider the cost functional defined by @xmath217 to be minimized or maximized subject to the fractional dynamical system @xmath218 and the boundary conditions @xmath219 where @xmath220 we assume the functions @xmath221 , @xmath38 , and @xmath39 , to be of class @xmath40 with respect to all their arguments . if @xmath222 , the lagrange problem reduces to the fractional variational problem studied in section [ sec : fundprob ] . an optimal solution is a pair of functions @xmath223 that minimizes @xmath26 as in , subject to the fractional dynamic equation and the boundary conditions . if @xmath223 is an optimal solution to the fractional lagrange problem , then there exists a function @xmath224 for which the triplet @xmath225 satisfies the hamiltonian system @xmath226 and the stationary condition @xmath227 where the hamiltonian @xmath228 is defined by @xmath229 and @xmath230 the result follows applying theorem [ eleqmult ] to @xmath231 with respect to @xmath0 , @xmath232 and @xmath224 . in the particular case when @xmath91 does not depend on @xmath92 and @xmath35 , we obtain ( * ? ? ? * theorem 3.5 ) . let @xmath233 be a solution of @xmath234 subject to the fractional control system @xmath235 and the boundary conditions @xmath236 and @xmath237 . define the hamiltonian by @xmath238 . then there exists a function @xmath224 for which the triplet @xmath225 fulfill the hamiltonian system @xmath239 and the stationary condition @xmath240 . to begin , let us recall the notions of convexity and concavity for @xmath40 functions of several variables . given @xmath32 and a function @xmath241 such that @xmath242 exist and are continuous for all @xmath243 , we say that @xmath244 is convex ( concave ) in @xmath245 if @xmath246 for all @xmath247 . consider the functional @xmath26 as in , and let @xmath49;\mathbb{r})$ ] be a solution of the fractional euler lagrange equation satisfying the boundary conditions . assume that @xmath91 is convex in @xmath248 . if one of the two following conditions is satisfied , 1 . @xmath249 is convex in @xmath250 and @xmath251(x ) \geq 0 $ ] for all @xmath252 $ ] ; 2 . @xmath249 is concave in @xmath250 and @xmath251(x ) \leq 0 $ ] for all @xmath252 $ ] ; then @xmath0 is a ( global ) minimizer of problem . consider @xmath114 of class @xmath45;\mathbb{r})$ ] such that @xmath253 . then , @xmath254(x)h(x ) + \frac{\partial l}{\partial v}[y](x){^c_ad^\alpha_x}h(x ) + \frac{\partial l}{\partial w}[y](x){_ai^\beta_x}h(x)\right.\\ & \quad \left . + \frac{\partial l}{\partial z}[y](x ) \int_a^x\left ( \frac{\partial l}{\partial y}\{y\}(t)h(t ) + \frac{\partial l}{\partial v}\{y\}(t){^c_ad^\alpha_t}h(t ) + \frac{\partial l}{\partial w}\{y\}(t){_ai^\beta_t}h(t)\right)dt\right]dx\\ & = \int_a^b \left [ \frac{\partial l}{\partial y}[y](x ) + { _ xd^\alpha_b}\left ( \frac{\partial l}{\partial v}[y](x ) \right ) + { _ xi_b^\beta}\left(\frac{\partial l}{\partial w}[y](x)\right ) + \int_x^b \frac{\partial l}{\partial z}[y](t)dt \cdot \frac{\partial l}{\partial y}\{y\}(x)\right.\\ & \quad \left . + { _ xd^\alpha_b}\left ( \int_x^b \frac{\partial l}{\partial z}[y](t)dt \cdot \frac{\partial l}{\partial v}\{y\}(x ) \right ) + { _ xi^\beta_b}\left ( \int_x^b \frac{\partial l}{\partial z}[y](t)dt \cdot \frac{\partial l}{\partial w}\{y\}(x ) \right)\right ] h(x)dx = 0 . \end{split}\ ] ] one can easily include the case when the boundary conditions are not given . consider functional @xmath26 as in and let @xmath49;\mathbb{r})$ ] be a solution of the fractional euler lagrange equation and the fractional natural boundary condition . assume that @xmath91 is convex in @xmath248 . if one of the two next conditions is satisfied , 1 . @xmath249 is convex in @xmath250 and @xmath251(x ) \geq 0 $ ] for all @xmath252 $ ] ; 2 . @xmath249 is concave in @xmath250 and @xmath251(x ) \leq 0 $ ] for all @xmath252 $ ] ; then @xmath0 is a ( global ) minimizer of . solving a variational problem usually means solving euler lagrange differential equations subject to some boundary conditions . it turns out that most fractional euler lagrange equations can not be solved analytically . therefore , in practical terms , numerical methods need to be developed and used in order to solve the fractional variational problems . a numerical scheme to solve fractional lagrange problems has been presented in @xcite . the method is based on approximating the problem to a set of algebraic equations using some basis functions . a more general approach can be found in @xcite that uses the oustaloup recursive approximation of the fractional derivative , and reduces the problem to an integer order ( classical ) optimal control problem . a similar approach is presented in @xcite , using an expansion formula for the left riemann liouville fractional derivative developed in @xcite . here we use a modified approximation ( see remark [ rem : ns ] ) based on the same expansion , to reduce a given fractional problem to a classical one . the expansion formula is given in the following lemma . suppose that @xmath255 $ ] , @xmath256 $ ] and @xmath257 . then the left riemann liouville fractional derivative can be expanded as @xmath258 where @xmath259,\\ c(k,\alpha)&=&\frac{1}{\gamma(2-\alpha)\gamma(\alpha-1)}\frac{\gamma(k-1+\alpha)}{(k-1)!}.\end{aligned}\ ] ] in practice , we only keep a finite number of terms in the series . we use the approximation @xmath260 for some fixed number @xmath261 , where @xmath262.\end{aligned}\ ] ] [ rem : ns ] in @xcite the authors use the fact that @xmath263 , and apply in their method the approximation @xmath264 regarding the value of @xmath265 for some values of @xmath261 ( see table [ tab ] ) , we take the first derivative in the expansion into account and keep the approximation in the form of equation . .values of @xmath265 for @xmath266 and different values of @xmath261 . [ cols="^,^,^,^,^,^,^,^",options="header " , ] we illustrate with examples [ example : bra ] and [ example1 ] how the approximation provides an accurate and efficient numerical method to solve fractional variational problems . [ ex : new : sec6 ] we obtain an approximated solution to the problem considered in example [ example : bra ] . since @xmath267 , the caputo derivative coincides with the riemann liouville derivative and we can approximate the fractional problem using . we reformulate the problem using the hamiltonian formalism by letting @xmath268 . then , @xmath269 we also include the variable @xmath270 with @xmath271 in summary , one has the following lagrange problem : @xmath272dx \longrightarrow \min \\ \begin{cases } { y'}(x ) = -ab^{-1}x^{-1}y(x)+\sum_{k=2}^{n}b^{-1}c_kx^{-k}v_k(x)+b^{-1}x^{\alpha-1}u(x)\\ { v'}_k(x ) = ( 1-k)x^{k-2}y(x),\qquad k=1,2,\ldots\\ { z'}(x ) = \left ( y(x)-x^{\alpha+1}\right)^2 \end{cases } \end{gathered}\ ] ] subject to the boundary conditions @xmath267 , @xmath273 and @xmath274 , @xmath275 setting @xmath276 , the hamiltonian is given by @xmath277+p_1(x)\left(-ab^{-1}x^{-1}y(x)+b^{-1}c_2x^{-2}v_2(x ) + b^{-1}x^{\alpha-1}u(x)\right)\\ -p_2(x)y(x)+p_3(x)\left ( y(x)-x^{\alpha+1}\right)^2.\end{gathered}\ ] ] using the classical necessary optimality condition for problem , we end up with the following two point boundary value problem : @xmath278 subject to the boundary conditions @xmath279 we solved system subject to using the matlab@xmath280 built - in function ` bvp4c ` . the resulting graph for @xmath281 , together with the corresponding value of @xmath26 , is given in figure [ fig2 ] . . ] our numerical method works well , even in the case the minimizer is not a lipschitz function . an approximated solution to the problem considered in example [ example1 ] can be obtained following exactly the same steps as in example [ ex : new : sec6 ] . recall that the minimizer to that problem is not a lipschitz function . as before , one has @xmath267 and the caputo derivative coincides with the riemann liouville derivative . we approximate the fractional problem using . let @xmath268 . then holds . in this case the variable @xmath270 satisfies @xmath282 and we approximate the fractional variational problem with the following classical one : @xmath283dx \longrightarrow \min \\ \begin{cases } { y'}(x ) = -ab^{-1}x^{-1}y(x)+\sum_{k=2}^{n}b^{-1}c_kx^{-k}v_k(x)+b^{-1}x^{\alpha-1}u(x)\\ { v'}_k(x ) = ( 1-k)x^{k-2}y(x),\qquad k=1,2,\ldots\\ { z'}(x ) = \left ( y(x)-\frac{x^\alpha}{\gamma(\alpha+1)}\right)^2 \end{cases } \end{gathered}\ ] ] subject to the boundary conditions @xmath267 , @xmath273 and @xmath274 , @xmath275 setting @xmath276 , the hamiltonian is given by @xmath284+p_1(x)\left(-ab^{-1}x^{-1}y(x)+b^{-1}c_2x^{-2}v_2(x ) + b^{-1}x^{\alpha-1}u(x)\right)\\ -p_2(x)y(x)+p_3(x)\left ( y(x)-\frac{x^\alpha}{\gamma(\alpha+1)}\right)^2.\end{gathered}\ ] ] the classical theory @xcite tell us to solve the system @xmath285 subject to boundary conditions @xmath286 as done in example [ ex : new : sec6 ] , we solved using the matlab@xmath280 built - in function ` bvp4c ` . the resulting graph for @xmath281 , together with the corresponding value of @xmath26 , is given in figure [ fig ] in contrast with the exact minimizer . . work supported by the _ portuguese foundation for science and technology _ ( fct ) , through the _ center for research and development in mathematics and applications _ ( cidma ) and the ph.d . fellowship sfrh / bd/33761/2009 ( shakoor pooseh ) . the authors are very grateful to a referee for valuable remarks and comments , which significantly contributed to the quality of the paper . r. almeida , a. b. malinowska and d. f. m. torres , a fractional calculus of variations for multiple integrals with application to vibrating string , j. math . 51 ( 2010 ) , no . 3 , 033503 , 12 pp . arxiv:1001.2722 r. almeida , a. b. malinowska and d. f. m. torres , fractional euler - lagrange differential equations via caputo derivatives , in _ fractional dynamics and control _ ( eds : d. baleanu , j. a. tenreiro machado , and a. luo ) , springer , in press . r. almeida and d. f. m. torres , necessary and sufficient conditions for the fractional calculus of variations with caputo derivatives , commun . nonlinear sci . numer . simulat . * 16 * ( 2011 ) , no . 3 , 14901500 . arxiv:1007.2937 r. a. el - nabulsi and d. f. m. torres , necessary optimality conditions for fractional action - like integrals of variational calculus with riemann - liouville derivatives of order @xmath287 , math . methods appl . sci . * 30 * ( 2007 ) , no . 15 , 19311939 . arxiv : math - ph/0702099 r. a. c. ferreira and d. f. m. torres , higher - order calculus of variations on time scales , in _ mathematical control theory and finance _ ( eds : a. sarychev , a. shiryaev , m. guerra , and m. do r. grossinho ) , 149159 , springer , berlin , 2008 . arxiv:0706.3141 r. a. c. ferreira and d. f. m. torres , isoperimetric problems of the calculus of variations on time scales , in _ nonlinear analysis and optimization ii _ ( eds : a. leizarowitz , b. s. mordukhovich , i. shafrir , and a. j. zaslavski ) , contemporary mathematics , vol . 514 , amer . soc . , providence , ri , 2010 , pp . arxiv:0805.0278 r. a. c. ferreira and d. f. m. torres , fractional @xmath114-difference equations arising from the calculus of variations , appl . anal . discrete math . * 5 * ( 2011 ) , in press . doi : 10.2298/aadm110131002f arxiv:1101.5904 a. b. malinowska and d. f. m. torres , generalized natural boundary conditions for fractional variational problems in terms of the caputo derivative , comput . math . * 59 * ( 2010 ) , no . 9 , 31103116 . arxiv:1002.3790 t. odzijewicz , a. b. malinowska and d. f. m. torres , fractional variational calculus with classical and combined caputo derivatives , nonlinear anal . ( 2011 ) , in press . doi : 10.1016/j.na.2011.01.010 arxiv:1101.2932 l. s. pontryagin , v. g. boltyanskii , r. v. gamkrelidze and e. f. mishchenko , _ the mathematical theory of optimal processes _ , translated from the russian by k. n. trirogoff ; edited by l. w. neustadt interscience publishers john wiley & sons , inc . new york , 1962 .

we propose a hilfer advection - diffusion equation of order and type , and find the power series solution by using variational iteration method . power series solutions are expressed in a form that is easy to implement numerically and in some particular cases , solutions are expressed in terms of mittag - leffler function . absolute convergence of power series solutions is proved and the sensitivity of the solutions is discussed with respect to changes in the values of different parameters . for power law initial conditions it is shown that the hilfer advection - diffusion pde gives the same solutions as the caputo and riemann - liouville advection - diffusion pde . to leading order , the fractional solution compared to the non - fractional solution increases rapidly with for at a given time ; but for this factor is weakly sensitive to . we also show that the truncation errors , arising when using the partial sum as approximate solutions , decay exponentially fast with the number of terms used . we find that for the number of terms needed is weakly sensitive to the accuracy level and to the fractional order , ; but for the required number of terms increases rapidly with the accuracy level and also with the fractional order . [ section ] [ theorem]lemma [ theorem]remark [ theorem]definition [ theorem]corollary hilfer advection - diffusion equation , analytical approximate solution , variational iteration method , mittag - leffler function , convergence of solution , numerical analysis . + 2010 mathematics subject classification . 35r11 , 35c05 , 35c10 , 35e15 , 35g25 , 65m15 and 65g99 .

1406.2024

many transport phenomenon such as the time evolution of chemical or biological species in a flow field are often modeled by partial differential equations . these pde s are of the advection - diffusion - reaction type and can be derived from mass balance , momentum balance and energy balance equations and in the case of multi - component species , we also have individual species mass balance equations , see @xcite and @xcite . in a single - phase single - component system , let us denote the scalar concentration field by @xmath10 at the position @xmath11 and at the time instant @xmath4 , whose transport is described by the non - linear advection - diffusion - reaction equation , @xmath12 where @xmath13 is the velocity field in which the scalar is transported , @xmath14 is a non - linear convective flux , @xmath15 is the scalar flux , and @xmath16 is the source / reaction term . see appendix a for details on how this eqaution is derived . many physical phenomena appearing in the studies of fluid mechanics , astrophysics , ground water flow , @xcite , meteorology , @xcite , @xcite , semiconductors , @xcite , and reactive flows , @xcite , are modeled by eq . . the nonlinear advection - diffusion equation also proves to be effective in describing the behavior of two - phase flow in oil reservoir , @xcite , non - newtonian flows , @xcite , front propagation , @xcite , traffic flow , @xcite , financial modeling , @xcite . although the above mathematical models adequately describe a lot of natural phenomena , there still exist many complex phenomena in nature which are not described adequately by these models . among them are crowded systems , such as protein diffusion within cells , @xcite , and diffusion through porous media , @xcite . so there is a need to develop new models to understand such complex phenomena . in this regard fractional calculus could be helpful in describing such complex phenomena , see @xcite , @xcite , @xcite and @xcite . for instance , in order to obtain a better understanding of anomalous diffusion caputo @xcite used fractional calculus to incorporates memory . fractional calculus continues to attract the attention of researchers in physics , biology , chemistry and other engineering sciences @xcite . the reason lies in its ability to explain the complex systems , such as , anomalous diffusion in porous media , crowding in living cells , time evolutionary processes which depend on the past history . experimental evidence of anomalous diffusion have been reported by hilfer @xcite while working on dielectric spectroscopy ( in particular , glassy formations and relaxations in polymers ) . joen et al . @xcite have reported their findings about the evidence of anomalous diffusion in living organism , and tabie et al . @xcite , during the study of crowded systems ( intra cellular transport of insulin in granules ) , have witnessed the anomalous diffusion of insulin in cells . anomalous diffusion , and transport through porous medium , can be understood as a random walk processes , @xcite , especially continuous time random walk ( ctrw ) models . conventional , brownian ctrw , is characterised by waiting times and jumps in particle location whose probability density function are gaussian and the pdf obeys the classical advection - diffusion equation . anomalous diffusion , on the other hand , possesses pdf s of waiting times and jumps which are inverse power laws , and it can be shown that such a process is described by fractional advection - diffusion equations . furthermore , in a standard diffusion process the mean square displacement , msd , @xmath17 of a particle is linearly related to the time @xmath4 , but in anomalous diffusion msd @xmath17 has a nonlinear power relationship with the time @xmath4 , that is , @xmath18 . for @xmath0 diffusion is called subdiffusion ; for @xmath19 , we have the standard diffusion ; for @xmath20 the diffusion is termed as superdiffusion . mean square displacement can be understood geometrically as the amount of space the particle has explored in the system . for more details , see metzler and klafter @xcite . they have obtained the following fractional advection - diffusion equation , @xmath21.\ ] ] where @xmath10 is a scalar field , @xmath22 is the riemann - liouville fractional derivative of order @xmath0 defined in eq . , @xmath23 and @xmath24 are called the generalized diffusion constants . note that eq . reduces to the problem considered by caputo @xcite by taking @xmath25 and replacing @xmath26 by @xmath2 ( using the same symbol ) . also note that by setting @xmath27 , @xmath28 , and @xmath29 in eq . , we obtain a time fractional diffusion equation of order @xmath30 which was considered by das in @xcite and by saha in @xcite . in the present study , we propose a similar equation to but with the riemann - liouville fractional derivative @xmath22 is replaced by the hilfer fractional derivative @xmath31 , ( defined later in eq . ) . hilfer fractional derivative @xmath31 is a sort of interpolation between the riemann - liouville fractional derivative and caputo fractional derivative , see @xcite . example of a physical system that can be modeled by hilfer fractional derivative is given by hilfer in @xcite . we interpret @xmath10 as a diffusive scalar field ( for example temperature or concentration ) , and we use the simplifying assumption @xmath32constant in eq . and obtain the following linear fractional advection - diffusion equation , @xmath33,\quad x > 0 , t > 0.\ ] ] equation is called the hilfer advection - diffusion equation of order @xmath2 and type @xmath34 ; @xmath35 represents diffusivity ; @xmath36 represents the velocity field . note that equation reduces to equation by taking @xmath37 and relabeling @xmath2 by @xmath26 . also note that in @xcite sandev considers a diffusion - reaction equation that involves hilfer fractional derivative but without the advection term . here we outline the main objectives of the present study . firstly , we find the power series solution of eq . , with @xmath38 and initial condition @xmath39 by using variational iteration method , described in section ( [ secvim ] ) . secondly , we represent the power series solution in a convenient form that is easy to use for numerical purposes , especially we give a recurrence relation for the @xmath40part in the @xmath6th term of the series . thirdly , we prove the absolute convergence of the series solution accompanied by some examples . fourthly , we analyze the behavior of the fractional solution with respect to the parameters @xmath2 , @xmath41 , and @xmath42 ; and we discuss the numerical convergence of the solutions which arise when we use the truncated series solution . finally , we examine the fractional solutions for @xmath43 and compare with the conventional solution for @xmath44 in order to elucidate trends in the solution as @xmath2 increases for different parameter values @xmath42 and @xmath41 . we have organized this paper as follows : in section [ prelim ] , we provide some basic definitions and results from fractional calculus ; in section [ secvim ] , we describe the variational iteration method to obtain the solution of the problem subject to initial condition ; in section [ case ] , we present a case study of polynomial uploading ; in section [ num ] , we discuss the numerical results and provide graphs of the solutions along with error analysis ; in section [ frvscon ] , we compare the fractional solutions with the corresponding conventional versions ; and in the last section [ conc ] , we state our conclusions of the study . in this section , we briefly discuss the importance and significance of time fractional derivatives and hilfer - composite time fractional derivative , and state some definitions and results from fractional calculus , see @xcite . it was shown by hilfer that time fractional derivatives are equivalent to infinitesimal generators of generalized time fractional evolutions , which arise in the transition from microscopic to macroscopic time scales @xcite . hilfer showed that this transition from ordinary time derivative to fractional time derivative indeed arises in physical problems @xcite . 0.2 cm * _ riemann - liouville fractional integral _ * of order @xmath2 for an absolutely integrable function @xmath45 is defined by @xmath46 when the right hand side exists . 0.2 cm * _ riemann - liouville fractional derivative _ * of order @xmath43 for an absolutely integrable function @xmath45 is defined by @xmath47 0.2 cm * _ caputo fractional derivative _ * of order @xmath43 for a function @xmath45 , whose first derivative is absolutely integrable , is defined by @xmath48 0.2 cm * _ relationship between riemann - liouville and caputo fractional derivative _ * 0.1 cm @xmath49 . \end{aligned}\end{gathered}\ ] ] 0.2 cm * _ hilfer fractional derivative _ * of order @xmath43 and type @xmath34 for an absolutely integrable function @xmath45 with respect to @xmath4 is defined by , @xmath50 [ lma1]@xcite the following fractional derivative formula holds true : @xmath51 where @xmath52 and @xmath53 . from lemma [ lma1 ] , we can easily have the following lemma . [ lma2 ] on integrating equation , we obtain @xmath54 where @xmath52 and @xmath55 . * _ remarks : _ * 0.2 cm 1 . the caputo derivative represents a type of regularization in the time domain ( origin ) for riemann - liouville derivative . 2 . hilfer fractional derivative interpolates between riemann - liouville fractional derivative and caputo fractional derivative , because if @xmath56 then hilfer fractional derivative corresponds to riemann - liouville fractional derivative and if @xmath57 then hilfer fractional derivative corresponds to caputo fractional derivative . @xmath58 is required to be finite . 4 . the three derivatives are equal if @xmath59 is continuous on @xmath60 $ ] and @xmath61 , see lemma [ lma3 ] . 0.2 cm * _ mittag - leffler function _ * 0.2 cm mittag - leffler function is the generalization of exponential function @xmath62 . 1-parameter mittag - leffler function _ * @xmath63 0.2 cm * _ 2-parameter mittag - leffler function _ * @xmath64 the motivation of studying fractional equations of form is , from one side , the hilfer generalized time fractional derivative , which combine both the derivatives , caputo and r l . it is known , from the continuous time random walk ( ctrw ) theory , that the probability density @xmath65 , in case where the characteristic waiting time diverges and the jump length variance is finite , can be obtained from the following two equivalent representations of the fractional diffusion equation @xcite @xmath66 @xmath67 in the r l and caputo sense , respectively , where @xmath41 is the generalized diffusion constant of physical dimension @xmath68 = m^2/s^\mu$ ] , and @xmath69 is the anomalous diffusion exponent . thus , if the initial conditions are properly taken into account , the caputo and riemann - liouville formulations of the time - fractional diffusion - advection equations are identical . in this section , we describe the variational iteration method , @xcite , and provide an outline for its implementation . the vim has been used extensively by several authors , see @xcite , in recent years to obtain series solutions of problems arising in different areas of applied mathematics and engineering . vim has been successfully applied to solve problems like riccati equation , heat equation , wave equation and many other problems . ji - huan he , @xcite , proposed vim to obtain the solutions of nonlinear differential equations . the method provides the solution in the form of a successive approximations that may converge to the exact solution if such a solution exists . in case where a closed form of the exact solution is not achievable , we use the truncated series , for instance , the @xmath6th partial sum of the series . vim has certain advantages over the other proposed methods like adomian decomposition method ( adm ) , @xcite , and homotopy perturbation method ( hpm ) , see @xcite . in the case of adm a lot of work has to be done to compute the adomian polynomials for nonlinear terms and in the case of hpm , a huge amount of calculation has to be done when degree of nonlinearity increases . on the other hand , no specific requirements are needed for nonlinear operators in order to use vim . for instance , hpm requires an introduction of small parameter that is sometimes difficult to incorporate in the equation or its introduction may change the physics of the problem . 0.2 cm the basic concepts and main steps for the implementation of vim are explained here . consider the following equation : @xmath70+g(x , t),\ ] ] where @xmath71 represents the hilfer fractional derivative with respect to time variable @xmath4 , and @xmath72 represents a differential operator with respect to variable @xmath11 . 0.2 cm the variational iteration method presents a correctional functional in @xmath4-direction for eq . in the form , @xmath73-g(x , t)\right)d\xi,\ ] ] with @xmath74 assumed known , where @xmath75 is a general lagrange multiplier which can be identified optimally by variational theory and @xmath76 is a restricted value that means it behaves like a constant , hence @xmath77 , where @xmath78 is the variational derivative . 0.2 cm the vim is implemented in two basic steps , see @xcite ; 1 . the determination of the lagrange multiplier @xmath75 that will be identified optimally through variational theory , and 2 . with @xmath75 determined , we substitute the result into eq . where the restriction should be omitted . 0.2 cm taking the @xmath79variation of eq . with respect to @xmath74 , we obtain @xmath80-g(x , t)\right)d\xi.\ ] ] 0.2 cm since @xmath81 and @xmath82 , we have @xmath83 0.2 cm to determine the lagrange multiplier @xmath75 , we integrate by parts the integral in eq . , and noting that variational derivative of a constant is zero , that is , @xmath84 . yields @xmath85 0.2 cm the extreme values of @xmath86 requires that @xmath87 . this means that left hand side of eq . is zero , and as a result the right hand side should be zero as well , that is , @xmath88 0.2 cm this yields the stationary conditions @xmath89 0.2 cm hence eq . becomes @xmath90-g(x , t)\right)d\xi,\ ] ] where the restriction is removed on @xmath74 . we can use eq . to obtain the successive approximation of the solution of the problem . the zeroth approximation @xmath91 can be chosen in such away that it satisfies the initial condition and the boundary conditions . appropriate selection of the zeroth approximation is necessary for the convergence of the successive approximation to the exact solution of the problem . however , we remark that beacause vim involves derivatives of all @xmath74 s inside the integral in the above equation , then vim is limited to smooth initial conditions . it is not suitable for initial conditions such as @xmath92 the latter would produce a green s function for the physical problem . nevertheless , provided smooth initial conditions can be specified then vim is often fast and very effective , as demonstrated in the case studies below . we consider the equation @xmath93,\quad x > 0 , t > 0\ ] ] with initial condition @xmath94 0.2 cm according to the variational iteration method , we consider the correctional functional in @xmath4-direction by using eq . @xmath95\right)d\xi.\ ] ] now by using eq . we obtain @xmath96\right)d\xi.\ ] ] which simplifies to @xmath97d\xi.\ ] ] starting with an initial approximation @xmath98 , we obtain a sequence of successive approximations , and the exact solution is obtained by taking the limit of the @xmath6th approximation , that is , @xmath99 we take @xmath38 in eq . , so it becomes @xmath100,\quad x > 0 , t > 0\ ] ] with the initial condition @xmath101 , for @xmath102 . we obtain the following iteration formula by using eq . @xmath103d\xi,\ ] ] with the zeroth approximation @xmath104 by taking @xmath105 in eq . and using eq . , we obtain @xmath106d\xi,\ ] ] which can be written as @xmath107 where @xmath108 by using lemma [ lma1 ] and [ lma2 ] , we obtain @xmath109 importantly , note that for the functions of type @xmath110 , where @xmath111 , the hilfer fractional derivative is independent of @xmath34 by lemma [ lma1 ] . moreover , for such functions , the caputo , reimann - liouville and hilfer derivatives are all equal . furthermore , because this is the first term in the recurrence relation , @xmath34 does not appear in any of the higher order terms ( below ) . by taking @xmath112 in eq . and using eq . , we obtain @xmath113 where @xmath114 by taking @xmath115 in eq . and using eq . , we obtain @xmath116 where @xmath117 proceeding in this way we obtain @xmath118 where @xmath119 and @xmath120 is given by the eq . . by setting @xmath121 , eq . can be written as @xmath122 by taking the limit @xmath123 of eq . we obtain @xmath124}.\ ] ] we remark , again , that the final solution above does not contain any dependency on @xmath34 . this is a consequence of the lemma [ lma1 ] . the series solution of the problem with the initial condition @xmath125 , @xmath102 , converges absolutely for all @xmath11 and @xmath4 . we denote the @xmath6th term of eq . by @xmath126}.\ ] ] applying the ratio test on the series , we obtain @xmath127}{\gamma[(n+1)(1-\alpha)+1]}\right|\\ \nonumber & \text{since } ( 1-\alpha)>0 \text { therefore } n(1-\alpha)+1>1\\ & = \left|\frac { a_{n+1}(x)}{a_{n}(x)}\kappa t^{(1-\alpha)}\right|\frac{\gamma[n(1-\alpha)+1]}{\gamma[(n+1)(1-\alpha)+1]}.\end{aligned}\ ] ] note that @xmath128 is bounded above by @xmath129 , where @xmath102 is the integer power of @xmath11 in the initial condition @xmath101 . indeed , @xmath130 , defined in eq . , is a polynomial in @xmath11 whose leading term , that is , the term with the highest power of @xmath11 is @xmath131 , and further note that degree@xmath132 for all @xmath133 . thus we can approximate @xmath130 by its leading term @xmath131 ( since all the coefficients are positive ) and therefore we obtain @xmath134 by using wendel s double inequality , see @xcite , @xmath135 for @xmath136 and @xmath137 , we deduce that @xmath138}{\gamma[(n+1)(1-\alpha)+1]}=0.\ ] ] hence , eq . gives @xmath139 thus the series solution obtained in eq . converges ( absolutely ) for all @xmath11 and @xmath4 and for all real @xmath42 . the series solution satisfies the equation with the initial condition @xmath125 , where @xmath102 . first , on differentiating eq . with respect to @xmath4 , we obtain@xmath140}\right]\\ \nonumber & = \sigma_{k=1}^{\infty } a_k(x)\frac{\kappa^k}{\gamma[k(1-\alpha)+1]}\frac{\partial}{\partial t}t^{k(1-\alpha)}\\ \nonumber & = \sigma_{k=1}^{\infty } a_k(x)\frac{\kappa^k}{k(1-\alpha)\gamma[k(1-\alpha)]}k(1-\alpha)t^{k(1-\alpha)-1}\\ & = \sigma_{k=1}^{\infty } a_k(x)\frac{\kappa^kt^{k(1-\alpha)-1}}{\gamma[k(1-\alpha)]}.\end{aligned}\ ] ] on the other hand substituting eq . in the right hand side of eq . yields @xmath141 \\ \nonumber & = \kappa\ _ 0\mathcal{d}_t^{\alpha,\beta}\left[\sigma_{k=0}^{\infty}\left\{\frac{\partial^2}{\partial x^2}a_k(x)+\frac{\partial}{\partial x}x^qa_k(x)\right\}\frac{\kappa^kt^{k(1-\alpha)}}{\gamma[k(1-\alpha)+1]}\right ] \\ \nonumber & = \kappa\sigma_{k=0}^{\infty}\left\{\frac{\partial^2}{\partial x^2}a_k(x)+\frac{\partial}{\partial x}x^qa_k(x)\right\}\ _ 0\mathcal{d}_t^{\alpha,\beta}\frac{\kappa^kt^{k(1-\alpha)}}{\gamma[k(1-\alpha)+1 ] } \\ \nonumber & = \sigma_{k=0}^{\infty}a_{k+1}(x)\frac{\kappa^{k+1}t^{(k+1)(1-\alpha)-1}}{\gamma[(k+1)(1-\alpha ) ] } \\ & = \sigma_{k=1}^{\infty}a_{k}(x)\frac{\kappa^{k}t^{k(1-\alpha)-1}}{\gamma[k(1-\alpha)]}.\end{aligned}\ ] ] the equality of eqs . and proves that @xmath13 given by eq . is the solution of problem . , eq . , ( example 1 ) for @xmath142 at @xmath143 , where @xmath144 , @xmath145 and @xmath146 , for different @xmath6 as indicated . ] , eq . , ( example 1 ) for @xmath142 at @xmath143 , where @xmath144 , @xmath145 and @xmath146 , for different @xmath6 as indicated . ] we examine the solutions for the cases @xmath144 and @xmath147 . for @xmath144 , the initial condition becomes @xmath148 , in eq . . then @xmath149 for @xmath150 and hence from eq . the solution @xmath10 is expressed as follows @xmath151^k}{\gamma[k(1-\alpha)+1]}=xe_{1-\alpha}[2\kappa t^{1-\alpha}],\ ] ] where @xmath152 , defined in eq . , is the mittag - leffler function in one parameter . the plot of the solution is shown in the fig . [ fig1 ] for the values @xmath145 and @xmath146 . note that the solution @xmath10 , for fixed @xmath4 , increases linearly with respect to variable @xmath11 and it increases exponentially with respect to variable @xmath4 , for fixed @xmath11 . 0.2 cm in order to see , how rapidly the sequence of successive approximations provided by vim converges to the exact solution , we use the @xmath6th partial sum as an approximation , @xmath153^k}{\gamma[k(1-\alpha)+1]},\ ] ] in fig . [ fig2 ] , we plot @xmath154 against @xmath4 at @xmath143 for @xmath145 and @xmath146 , for different values of @xmath6 . one can see from fig . [ fig2 ] that the solution converges by @xmath155 . later in section [ num ] , we will provide details about how many terms have to be summed up in order to obtain a given accuracy . , eq . , ( example 2 ) , for @xmath142 at @xmath143 , where @xmath156 , @xmath145 and @xmath146 , for different @xmath6 as indicated . ] , eq . , ( example 2 ) , for @xmath142 at @xmath143 , where @xmath156 , @xmath145 and @xmath146 , for different @xmath6 as indicated . ] for @xmath156 the initial condition becomes @xmath157 , in eq . . then @xmath158 for @xmath150 and hence from eq . the solution @xmath10 is expressed as follows @xmath159+e_{1-\alpha}[3\kappa t^{1-\alpha}]-e_{1-\alpha}[\kappa t^{1-\alpha}].\ ] ] the plot of the solution is shown in the fig . [ fig3 ] for @xmath145 and @xmath146 . this time , the solution @xmath10 , for fixed @xmath4 , increases quadratically with respect to variable @xmath11 and it increases exponentially with respect to variable @xmath4 , for fixed @xmath11 . 0.2 cm again , we use the @xmath6th approximation in order to see how rapidly the sequence of successive approximations provided by vim converges to the exact solution : @xmath160^k}{\gamma[k(1-\alpha)+1]}+\sigma_{k=0}^{n}\frac{[3\kappa t^{1-\alpha}]^k}{\gamma[k(1-\alpha)+1]}\\ & \quad -\sigma_{k=0}^{n}\frac{[\kappa t^{1-\alpha}]^k}{\gamma[k(1-\alpha)+1]},\end{aligned}\ ] ] and plot it for different values of @xmath6 . figure [ fig4 ] shows the plots of @xmath154 against @xmath4 for @xmath143 , @xmath145 , @xmath146 , and for different @xmath6 as indicated . this time the approximate solutions converge at @xmath161 . when @xmath162 , closed form solutions for @xmath10 becomes increasingly harder to obtain . nevertheless , for the purposes of analyzing the behavior of the solution @xmath10 we require only the dominant term in the solution . as mentioned in section [ sec4p2 ] that we can approximate @xmath130 by its leading term , that is by @xmath131 . if we replace @xmath130 by @xmath131 in eq . , we obtain the the leading term to be , @xmath163}\\ \nonumber & \approx x^p\sigma_{k=0}^{\infty } \frac{[(p+1)\kappa t^{1-\alpha}]^k}{\gamma[k(1-\alpha)+1]}\\ & \approx x^pe_{1-\alpha}[(p+1)\kappa t^{1-\alpha}].\end{aligned}\ ] ] the solution @xmath10 is thus proportional to @xmath164 at fixed t ; and it increases approximately exponentially with respect to variable @xmath4 at fixed @xmath11 . in order to further investigate the trends in the fractional solution , we compare the fractional solution to the non - fractional solution at @xmath143 and @xmath165 . at @xmath143 and @xmath165 , we get @xmath166 $ ] , and by using eq . we obtain @xmath167^k}{\gamma[k(1-\alpha)+1]}.\ ] ] in the asymptotic limit @xmath168 , we obtain @xmath169^k$ ] , which is a geometric series and it converges to @xmath170 , when @xmath171 or @xmath172 . 0.2 cm for @xmath44 , we obtain @xmath173^k}{k!}\ ] ] which converges for all @xmath41 and @xmath102 . furthermore , in eq . in the asymptotic limit @xmath168 , @xmath174 for all @xmath4 , and thus the power series converges for all @xmath4 so long as @xmath175 . in this section we discuss the general behavior of the solution of the problem with respect to different parameters and also we do some numerical analysis of the solution . first we analyze the reciprocal gamma function which appears in the series solution , that is , @xmath176}.\ ] ] the limit of this function , for a fixed @xmath177 , as @xmath178 is zero , that is , @xmath179}=0 , \quad \text { for fixed } \alpha\in(0,1).\ ] ] we want an expression for the number of terms @xmath6 needed in order to satisfy a given accuracy given by , @xmath176}\leq 10^{-\tau},\ ] ] for a given @xmath2 and tolerance level @xmath180 . for this purpose , we use the following formula , see @xcite , @xmath181 for all @xmath136 and with the optimal constants @xmath182 and @xmath183 , to obtain the required @xmath6 . table [ ta:1 ] summarizes this information . figure [ fig6 ] shows the plots of @xmath6 against @xmath180 for different values of @xmath2 . for any given @xmath2 the number of terms @xmath6 appears to scale almost linearly with @xmath180 . best linear fits were therefore obtained ; for example for @xmath184 , we obtain the best fit @xmath185 . the graph of this linear equation is the solid line shown in the fig . [ fig6 ] along with the @xmath186 prediction intervals , whereas the simulation data are plotted as the symbols . the value of @xmath187 , the coefficient of determination , is @xmath188 represents the percent of observed variability explained by the linear model , whereas the value of @xmath189 is @xmath190 which is a more realistic quantity as it accounts for the number of terms in the model . * 11c [ -0.5ex]@xmath180 & + & @xmath191 & @xmath192 & @xmath193 & @xmath194 & @xmath195 & @xmath196 & @xmath197 & @xmath198 & @xmath199 & @xmath200 + @xmath201 & @xmath202 & @xmath203 & @xmath204 & @xmath205 & @xmath206 & @xmath207 & @xmath208 & @xmath209 & @xmath210 & @xmath211 + @xmath212 & @xmath204 & @xmath205 & @xmath213 & @xmath214 & @xmath215 & @xmath208 & @xmath216 & @xmath217 & @xmath218 & @xmath219 + @xmath202 & @xmath213 & @xmath206 & @xmath207 & @xmath220 & @xmath208 & @xmath221 & @xmath222 & @xmath223 & @xmath224 & @xmath225 + @xmath204 & @xmath214 & @xmath207 & @xmath220 & @xmath208 & @xmath226 & @xmath227 & @xmath228 & @xmath229 & @xmath230 & @xmath231 + @xmath213 & @xmath207 & @xmath220 & @xmath208 & @xmath226 & @xmath209 & @xmath232 & @xmath223 & @xmath233 & @xmath234 & @xmath235 + @xmath214 & @xmath220 & @xmath208 & @xmath236 & @xmath216 & @xmath237 & @xmath238 & @xmath239 & @xmath240 & @xmath241 & @xmath242 + @xmath215 & @xmath208 & @xmath236 & @xmath221 & @xmath243 & @xmath244 & @xmath210 & @xmath245 & @xmath246 & @xmath247 & @xmath248 + 0.4 cm and @xmath6 . here , @xmath184.,title="fig : " ] + and @xmath6 . here , @xmath184.,title="fig : " ] + figure [ fig8 ] shows the plots of @xmath6 against @xmath2 for some tolerance levels , taken from the data in table [ ta:1 ] . the relationship between @xmath6 and @xmath2 is clearly non - linear . for @xmath249 the number of terms needed for a given tolerance level @xmath250 remains approximately constant ; in fact even for different tolerance levels @xmath6 appears to be approximately insensitive to @xmath2 and to @xmath180 as a rule of thumb we see that @xmath8 for @xmath5 . but for @xmath251 we see a rapid increase in the number of terms @xmath6 needed to achieve a given accuracy . it may be possible to find best - fit curves to the data plotted in fig . [ fig8 ] . for this purpose , we assume cubic polynomial fits of the form @xmath252 , where @xmath253 and @xmath254 are constants to be determined from the data using the least square method , for a tolerance level of @xmath255 , we obtain the following cubic polynomial @xmath256 . the graph of this cubic polynomial is the solid line shown in the fig . [ fig9 ] along with the @xmath186 prediction interval , whereas the simulation data are plotted symbols . the value of @xmath187 , is @xmath257 , and @xmath189 is @xmath258 . and @xmath6 . here , the tolerance level is @xmath255.,title="fig : " ] + and @xmath6 . here , the tolerance level is @xmath255.,title="fig : " ] + * 12c [ -0.5ex ] @xmath42 & [ -0.5ex ] @xmath41 & + & & @xmath191 & @xmath192 & @xmath193 & @xmath194 & @xmath195 & @xmath196 & @xmath197 & @xmath198 & @xmath199 & @xmath200 + + 1&@xmath192 & @xmath259 & @xmath260 & @xmath261 & @xmath262 & @xmath263 & @xmath264 & @xmath265 & @xmath266 & @xmath267 & @xmath268 + & @xmath193 & @xmath269 & @xmath270 & @xmath271 & @xmath272 & @xmath273 & @xmath274 & @xmath275 & @xmath276 & @xmath277 & @xmath278 + & @xmath194 & @xmath279 & @xmath280 & @xmath281 & @xmath282 & @xmath283 & @xmath284 & @xmath285 & @xmath286 & @xmath287 & @xmath288 + & @xmath195 & @xmath289 & @xmath290 & @xmath291 & @xmath292 & @xmath293 & @xmath294 & @xmath295 & @xmath296 & @xmath297 & @xmath298 + & @xmath196 & @xmath299 & @xmath300 & @xmath301 & @xmath302 & @xmath303 & @xmath304 & @xmath305 & @xmath306 & @xmath307 & @xmath308 + & @xmath197 & @xmath309 & @xmath310 & @xmath311 & @xmath312 & @xmath313 & @xmath314 & @xmath315 & @xmath316 & @xmath317 & @xmath318 + + 2&@xmath192 & @xmath319 & @xmath320 & @xmath321 & @xmath322 & @xmath323 & @xmath324 & @xmath325 & @xmath326 & @xmath327 & @xmath328 + & @xmath193 & @xmath329 & @xmath330 & @xmath331 & @xmath332 & @xmath333 & @xmath334 & @xmath335 & @xmath336 & @xmath337 & @xmath338 + & @xmath194 & @xmath339 & @xmath340 & @xmath341 & @xmath342 & @xmath343 & @xmath344 & @xmath345 & @xmath346 & @xmath347 & @xmath348 + & @xmath349 & @xmath350 & @xmath351 & @xmath352 & @xmath353 & @xmath354 & @xmath355 & @xmath356 & @xmath357 & @xmath358 & @xmath359 + & @xmath195 & @xmath360 & @xmath361 & @xmath362 & @xmath363 & @xmath364 & @xmath365 & @xmath366 & @xmath367 & @xmath368 & @xmath369 + + 3&@xmath192 & @xmath370 & @xmath371 & @xmath372 & @xmath373 & @xmath374 & @xmath375 & @xmath376 & @xmath377 & @xmath378 & @xmath379 + & @xmath380 & @xmath381 & @xmath382 & @xmath383 & @xmath384 & @xmath385 & @xmath386 & @xmath387 & @xmath388 & @xmath389 & @xmath390 + & @xmath193 & @xmath391 & @xmath392 & @xmath393 & @xmath394 & @xmath395 & @xmath396 & @xmath397 & @xmath398 & @xmath399 & @xmath400 + & @xmath401 & @xmath402 & @xmath403 & @xmath404 & @xmath405 & @xmath406 & @xmath407 & @xmath408 & @xmath409 & @xmath410 & @xmath411 + & @xmath194 & @xmath412 & @xmath413 & @xmath414 & @xmath415 & @xmath416 & @xmath417 & @xmath418 & @xmath419 & @xmath420 & @xmath421 + + 4&@xmath422 & @xmath423 & @xmath424 & @xmath425 & @xmath426 & @xmath427 & @xmath428 & @xmath429 & @xmath430 & @xmath431 & @xmath432 + & @xmath433 & @xmath434 & @xmath435 & @xmath436 & @xmath437 & @xmath438 & @xmath439 & @xmath440 & @xmath441 & @xmath442 & @xmath443 + & @xmath444 & @xmath445 & @xmath446 & @xmath447 & @xmath448 & @xmath449 & @xmath450 & @xmath451 & @xmath452 & @xmath453 & @xmath454 + & @xmath455 & @xmath456 & @xmath457 & @xmath458 & @xmath459 & @xmath460 & @xmath461 & @xmath462 & @xmath463 & @xmath464 & @xmath465 + & @xmath466 & @xmath467 & @xmath468 & @xmath469 & @xmath470 & @xmath471 & @xmath472 & @xmath473 & @xmath474 & @xmath475 & @xmath476 + 0.5 cm the solution of the problem ( with @xmath38 , and initial condition @xmath101 , @xmath477 ) is given by @xmath478},\ ] ] where @xmath479 and @xmath121 . we now examine the sensitivity of the solution @xmath10 to parameters @xmath42 , @xmath41 the diffusivity coefficient , and to @xmath2 the order of the hilfer derivative . we choose the following values : @xmath480 , @xmath481 and we select @xmath41 according to @xmath172 . we compare the values of @xmath10 at the point @xmath482 for different combinations of the above parameters . we choose three values of @xmath41 that are less than @xmath483 and one value that is equal ( approximately ) to @xmath483 and one value that is greater than @xmath483 . the data is collected in table [ ta:4 ] and it reveals that different combinations of parameter s values affect the values of @xmath10 differently . in general , we note that an increase in the values of parameters @xmath42 , @xmath41 and @xmath2 results into an increase in the values of @xmath10 . the effect of @xmath41 can be divided into two parts , first when @xmath172 and second when @xmath484 . for the first case , @xmath172 , the increase in the values of @xmath10 is not significant as compare to the second case , @xmath485 , where @xmath10 increases very rapidly with @xmath2 and @xmath41 . finally , we examine the truncation error when using the partial sum of first @xmath6 terms of the series solution given in eq . . we define the relative error as , @xmath486 where @xmath154 is the partial sum of the first @xmath6 terms . figures [ fig11 ] and [ fig12 ] show the plots of @xmath487 versus @xmath6 at the point @xmath482 for , respectively , @xmath144 ( example 1 ) , and @xmath156 ( example 2 ) . the errors fall off exponentially fast with @xmath6 , @xmath488 . it is not always possible to express the series solution in compact form like in the cases of @xmath489 and @xmath490 . for values of @xmath42 larger than @xmath147 , we approximate the exact solution by taking a very large value of @xmath6 , and then we compare this with the smaller values of @xmath6 . for example , when the initial condition is @xmath491 , @xmath492 , we take @xmath493 , and @xmath494 is the approximation to the exact solution . figure [ fig13 ] shows the plot of @xmath495 for @xmath496 . again the error falls off exponentially fast like @xmath488 . this indicates that the numerical convergence is independent of the power exponent @xmath42 . + these results show that the vim solutions for this problem convergence rapidly , and actually improves as @xmath6 increases . in this section , we compare the solutions of the fractional differential eq . with the corresponding conventional differential equation . 0.2 cm conventional version of eq . can be obtained by taking @xmath44 , see @xcite , that is , @xmath497,\quad x > 0 , t > 0.\ ] ] the solution of eq . with initial condition @xmath148 is @xmath498 and the general solution for @xmath499 is given by , section [ sec4p4 ] , @xmath500.\ ] ] figure [ fig14 ] shows the plots of the solution for different values of @xmath2 , that also includes the case @xmath44 at @xmath143 , for @xmath142 and @xmath501 . thus , the relative magnitude of the solutions compared to the conventional case can be estimated from , @xmath502 \over e^{2 \kappa t}}.\ ] ] , eq . , when @xmath143 , @xmath503 and @xmath501 . ] , eq . , when @xmath143 , @xmath503 and @xmath501 . ] asymptotic behavior of the above expression can be analyzed through the long time behavior of mittag - leffler function . by using theorem 1.3 of @xcite , we can write @xmath504 the long time behavior depends on the value of @xmath41 . if @xmath505 then @xmath506 ; if @xmath507 then @xmath508 and the fractional solution scales with the conventional solution ; if @xmath509 then @xmath510 . the solution of eq . with initial condition @xmath157 is @xmath511 where as , the fractional solution , example 2 in section [ sec4p4 ] , is @xmath512+e_{1-\alpha}[3\kappa t^{1-\alpha}]-e_{1-\alpha}[\kappa t^{1-\alpha}].\ ] ] figure [ fig15 ] shows the plots of the solution for different values of @xmath2 , that also includes the case @xmath44 which corresponds to the classical case , at @xmath143 , for @xmath142 and @xmath501 . the relative magnitude of the solutions compared to the conventional case can be estimated from , @xmath513+e_{1-\alpha}[3\kappa t^{1-\alpha}]-e_{1-\alpha}[\kappa t^{1-\alpha } ] \over x^2e^{3\kappa t}+e^{3\kappa t}-e^{\kappa t}}.\ ] ] to leading order , eq . is , @xmath514 \over e^{3\kappa t}}.\ ] ] and is easily shown that for general @xmath515 , i.e. for initial conditions @xmath101 , the corresponding relative magnitude is given by , @xmath516 \over e^{(p+1)\kappa t}};\ ] ] and the large time behaviour is , @xmath517 if @xmath518 then @xmath506 ; if @xmath519 then @xmath508 and the fractional solution scales with the conventional solution ; if @xmath520 then @xmath510 . some complex physical phenomenon such as crowded systems , and transport through porous media are not fully understood at the present time , and in order to shed new light into such phenomena a new modeling strategy has emerged in recent years which involves casting the system of interest in terms of fractional calculus . in this study the hilfer fractional advection - diffusion equation of order @xmath0 and type @xmath1 , @xmath521,\quad x > 0 , t > 0,\ ] ] with advection term @xmath522 , and power law initial conditions of the type @xmath523 for @xmath477 , was investigated numerically using the variational iteration method ( vim ) method , with a view of comparing its solution to the conventional non - fractional advection - diffusion system , and also to analyze the system numerically in order to investigate the efficiency of solving such systems numerically . for this class of initial conditions @xmath523 for @xmath477 , the above problem yields the same solutions as caputo and riemann - liouville advection - diffusion equations . however , we remark that this would be different in case if there is a source term in the problem or if the velocity also involves time variable . power series solutions were obtained , eq . . these yield closed form solutions for specific @xmath477 in terms of the mittag - leffler functions , although it becomes increasingly difficult to actually calculate it for @xmath524 . nevertheless , the leading order term can readily be obtained even for @xmath524 , which allows us to investigate the asymptotic behavior of the solutions and to carry out some numerical analysis of the method used . the power series is ( absolutely ) convergent ( for all @xmath11 , and all @xmath4 , and for @xmath102 ) . the behavior of the solution was examined by comparing the value of the solution at a fixed point , namely @xmath525 . asymptotically , the relative magnitude of the solutions is , @xmath526 which shows that the fractional solution increases approximately exponential faster than the conventional solution , as seen in figs . [ fig14 ] and [ fig15 ] . for fixed @xmath42 , the increase in @xmath525 , when @xmath527 and @xmath172 , is small ; but when @xmath251 and @xmath528 , @xmath525 increases very rapidly . for fixed @xmath2 and @xmath41 , the solution increases polynomially with @xmath11 and exponentially with @xmath4 . for the long time behaviour , as @xmath529 , we find that for all @xmath11 and for all @xmath0 , if @xmath518 then @xmath506 ; and if @xmath519 then @xmath508 and the fractional solution scales with the conventional solution ; and if @xmath520 then @xmath510 . variational iteration method ( vim ) has proved to be an efficient method for obtaining the series solution of the hilfer fractional advection - diffusion equation with the given power law initial data . truncation errors @xmath530 , arising when using the partial sum as approximate solutions , decay exponentially fast with the number of terms @xmath6 used , and then rate of convergence is independent of @xmath42 for the cases considered , @xmath531 , for @xmath532 . the number of terms @xmath6 required for a given level of accuracy for @xmath7 are relatively insensitive to both the @xmath2 and to the accuracy level required ; but for @xmath9 the number of terms increases rapidly with @xmath2 and with the accuracy level required . this threshold @xmath533 is consistent with the analysis of the solutions @xmath525 above . although these are early days in the development of fractional calculus and numerical solutions to fractional equations that describe physical systems , it is clear that numerical methods like vim will be important tools in extracting solutions of such fractional pde s in the future . future work will address the case when we have non - zero initial conditions which should yield different solutions for each @xmath34 . the authors would like to thank nstip for funding through project number 11-oil1663 - 04 . we also thank to the itc department at kfupm for providing software assistance . in a single - phase single - component system , let us denote the concentration of the scalar by @xmath10 at the position @xmath11 and at the time instant @xmath4 . first , if @xmath13 is the velocity field in which the scalar is transported , then the advection equation ( without reaction or diffusion ) is @xcite , @xmath534 second , if the concentration @xmath10 is transported by diffusion and these changes are caused by gradients in @xmath10 and the fluxes across the boundaries of the region , then the diffusion equation ( without advection or reaction ) is @xcite , @xmath535 where @xmath536 is the flux of @xmath537 at the point @xmath538 , and @xmath539 is the coefficient of diffusivity . this form in which the flux @xmath540 is proportional to the scalar gradient is called fickian diffusion . finally , there may be local changes in @xmath10 due to sources , sinks , and chemical reactions , which is described by an additional source ( reaction ) term @xmath16 . the reaction equation ( without advection or diffusion ) is @xcite , @xmath541 the general advection - diffusion - reaction equation is obtained by combining the above three effects and the overall change in the concentration @xmath10 is described by the following equation @xcite @xmath542 applications of such equations arise in many fields , such as , atmospheric chemistry , @xcite , air pollution models , @xcite , climatology , @xcite , modeling of fluid flow in homogeneous media , @xcite , catalysis , @xcite , combustion , @xcite . in a similar way nonlinear advection - diffusion - reaction equations are obtained by using nonlinear conservation laws , see @xcite , @xcite and @xcite . the nonlinear advection - diffusion - reaction equation has the form @xmath543 where @xmath544 represents the non - linear convective flux . 0.2 cm * _ laplace transform of fractional derivatives _ * 0.2 cm @xmath545:&=s^{\alpha}\widetilde{f}(s)-s^{\alpha-1}f(0^+ ) , \quad 0 < \alpha\leq 1,\\ \nonumber \text{where } \\ \nonumber f(0^+):&=\lim_{t\rightarrow 0^+}f(t).\\ \mathcal{l}[_0d_t^{\alpha}f(t);s]:&=s^{\alpha}\widetilde{f}(s)-\left(_0i_t^{1-\alpha}f\right)(0^+ ) , \quad 0<\alpha\leq 1,\\ \nonumber \text{where } \\ \nonumber \left(_0i_t^{1-\alpha}f\right)(0^+):&=\lim_{t\rightarrow 0^+}\left(_0i_t^{1-\alpha}f\right)(t)\\ \mathcal{l}\left[d_t^{\alpha,\beta}f(t);s\right]:&=s^{\alpha}\widetilde{f}(s)-s^{\beta(1-\alpha)}\left[_0i^{(1-\beta)(1-\alpha)}f(0^+)\right ] , \quad 0<\alpha<1,\\ \nonumber \text{where } \\ \nonumber \left(_0i_t^{(1-\beta)(1-\alpha)}f\right)(0^+):&=\lim_{t\rightarrow 0^+}\left(_0i_t^{(1-\beta)(1-\alpha)}f\right)(t)\end{aligned}\ ] ] w. hundsdorfer , j.verwer ; _ numerical solution of time - dependent advection - diffusion - reaction equations _ , springer - verlag , berlin , heidelberg , 2003 . bird , w.e . stewart , e.n . lightfoot ; _ transport phenomena _ , john wiley and sons , inc . , new york , 2002 . graedel , p.j . crutzen ; _ atmosphere , climate and change _ , henry holt and company , 1997 . b. sportisse ; _ air pollution modelling and simulation _ , springer - verlag , berlin , heidelberg , 2002 . t. stocker ; _ introduction to climate modelling _ , springer , new york , 2011 . k. aziz , a. settari ; _ petroleum reservoir simulation _ , applied science publishers ltd . , london , 1979 . e. cumberbatch , a. fitt ; _ mathematical modeling : case studies from industries _ , cambridge university press , uk , 2001 . i. glassman , r.a . yetter ; _ combustion _ , academic press , london , san diego , burlington , 2008 . a. kurganov , e. tadmor ; _ new high - resolution central schemes for nonlinear conservation laws and convection - diffusion equations _ , physics , 160 ( 2000 ) , 241 - 282 . a. bressan , m. lewicka , g. chen , d. wang ; _ nonlinear conservation laws and applications _ , springer science , new york , 2011 . leveque ; _ nonlinear conservation laws and finite volume methods for astrophysical fluid flow _ , computational methods for astrophysical fluid flows , springer - verlag , berlin , heidelberg , 1998 . m. khebchareon , s. saenton ; _ finite element solution for 1-d groundwater flow , advection - dispersion and interphase mass transfer : i. model development _ , thai j. math . , 3 ( 2005 ) , 223 - 240 . marchuk ; _ mathematical models in environmental problems _ , north - holland , elsevier science publisher , 1986 . holton , g.j . hakim ; _ an introduction to dynamic meteorology _ , academic press , usa , 2013 . e. s. oran , j. p. boris ; _ numerical simulation of reactive flow _ , second edition , cambridge university press , uk , 2001 . r. glowinski , j. xu ; _ numerical methods for non - newtonian fluids _ , handbook of numerical analysis , volume 16 ( 2011 ) , 1 - 801 . shuhaev , l.s . shtemenko ; _ propagation and reflection of shock waves _ , world scientific publishing , singapore , 1998 . m. treiber , a. kesting ; _ traffic flow dynamics : data , models and simulation _ , springer - verlag , berlin , heidelberg , 2013 . o. pironneau , y. achdou ; _ partial differential equations for option pricing _ , math . model . and methods in finance , special volume , handbook of numerical analysis , 2008 . jeon , v. tejedor , s. burov , e. barkai , c. selhuber - unkel , k. berg - srensen , l. oddershede , and r. metzler ; _ in vivo anomalous diffusion and weak ergodicity breaking of lipid granules _ , physical review letters 106 , no . 4 ( 2011 ) 048103 . j. sabatier , o. p. agrawal , and j. a. t. machado , _ advances in fractional calculus _ , dordrecht : springer , 2007 . tabei , s. burov , h. y. kim , a. kuznetsov , t. huynh , j. jureller , l. h. philipson , a. r. dinner , and n. f. scherer ; _ intracellular transport of insulin granules is a subordinated random walk _ , proceedings of the national academy of sciences 110 , no . 13 ( 2013 ) 4911 - 4916 . m. weiss , m. elsner , f. kartberg , t. nilsson ; _ anomalous subdiffusion is a measure for cytoplasmic crowding in living cells _ , biophysical j. , 87 ( 2004 ) , 3518 - 3524 . w. chen , h. sun , x. zhang , d. korosak ; _ anomalous diffusion modeling by fractal and fractional derivatives _ , comp . 59(5 ) , ( 2010 ) , 1754 - 1758 . k.b . oldham , j. spanier ; _ the fractional calulus _ , academic press , new york and london , 1974 . miller , b. ross ; _ an introduction to the fractional calculus and fractional differential equations _ , john wiley and sons , inc . , new york , 2003 . i. podlubny ; _ fractional differential equations _ , academic press , san diego , calfornia , usa , 1999 . k. diethelm , n.j . ford ; _ analysis of fractional differential equations _ , j. math . appl . 265 ( 2002 ) 229 - 248 . m. caputo ; _ diffusion of fluids in porous media with memory _ , geothermics , 28 ( 1999 ) 113 - 130 . s. havlin , d. ben - avraham ; _ diffusion in disordered media _ , advances in physics , 51(2002 ) , 187 - 292 . r. metzler , j. klafter ; _ the random walk s guide to anomalous diffusion : a fractional dynamics approach _ , physics reports , 339 ( 2000 ) 1 - 77 . s. das ; _ analytical solution of a fractional diffusion equation by variational iteration method _ ( 2009 ) , 483 - 487 . s. saha ray , r.k . bera ; _ analytical solution of a fractional diffusion equation by adomian decomposition method _ , comput . 174 ( 2006 ) , 329 - 336 . r. hilfer ; _ applications of fractional calculus in physics _ , world scientific publishing company , singapore , 2000 . r. hilfer ; _ experimental evidence for fractional time evolution in glass forming materials _ , chemical physics , 284 ( 2002 ) 399 - 408 . r. hilfer ; _ on fractional relaxation _ , fractals 11 ( 2003 ) 251 - 257 . r. hilfer , _ foundations of fractional dynamics : a short account _ , fractional dynamics : recent advances . world scientific , singapore 207 ( 2011 ) . t. sandev , r. metzler , z. tomovski ; _ fractional diffusion equation with a generalized riemann - liouville time fractional derivative _ , j. phys . a : math . theor . 44 ( 255203 ) ( 21pp ) , ( 2011 ) . r. metzler , and j. klafter , _ the restaurant at the end of the random walk : recent developments in the description of anomalous transport by fractional dynamics _ , journal of physics a : mathematical and general 37 , no . 31 ( 2004 ) : r161 . tomovski , t. sandev , r. metzler , and j. dubbeldam , _ generalized space time fractional diffusion equation with composite fractional time derivative _ , physica a : statistical mechanics and its applications 391 , no . 8 ( 2012 ) : 2527 - 2542 . kilbas , h.m . srivastava , j.j . trujillo ; _ theory and applications of fractional differential equations _ , north - holland , elsevier science publisher , 2006 . h. m. srivastava , . tomovski ; _ fractional calculus with an integral operator containing a generalized mittag leffler function in the kernel_ applied mathematics and computation 211.1 ( 2009 ) : 198 - 210 . k. m. furati , m. d. kassim , n. -e . tatar ; _ non - existence of global solutions for a differential equation involving hilfer fractional derivative _ , electronic journal of differential equations 2013.235 ( 2013 ) : 1 - 10 . a.m. wazwaz ; _ partial differential equations and solitary waves theory _ , springer , new york , 2009 . y. liu , x. zhao ; _ he s variational iteration method for solving convection diffusion equations _ , advanced intelligent computing theories and applications , lecture notes in computer science , springer , new york , 6215 ( 2010 ) 246 - 251 . y. molliq r. , m.s.m . noorani , i. hashim ; _ variational iteration method for fractional heat- and wave - like equations _ , nonlinear analysis : real world apllications , 10(2009 ) 1854 - 1869 . s. momani , z. odibat ; _ comparison between the homotopy perturbation method and the variational iteration method for linear fractional partial differential equations _ , , 54(2007 ) 910 - 919 . he ; _ approximate analytical solution for seepage flow with fractional derivatives in porous media _ , comput . methods appl . engrg . 167 ( 1998 ) 57 - 68 . he ; _ variational iteration method : a kind of non - linear analytical technique : some examples _ , int . j. non - linear mech . , 34 ( 1999 ) 699 - 708 . f. qi ; _ bounds for the ratio of two gamma functions _ , j. ineq . app . , volume 2010 , article i d 493058 , 84 pages . h. alzer ; _ sharp upper and lower bounds for the gamma function _ , proceedings of the royal society of edinburgh : section a mathematics , appl . ( 2009 ) , 709 - 718 .

at sufficiently low temperatures , the configurational phase space of a large spin - glass system breaks into many separated domains , each of which is referred to as a macroscopic state . the system is able to visit all spin configurations of the same macroscopic state , while it can not spontaneously jump between two different macroscopic states . ergodicity of the whole configurational phase space of the system , however , can be recovered if a temperature - annealing process is repeated an infinite number of times . in a heating - annealing cycle , the environmental temperature is first elevated to a high level and then decreased extremely slowly until a final low temperature is reached . different macroscopic states may be reached in different rounds of the annealing experiment ; while the probability of finding the system in macroscopic state decreases exponentially with the free energy of this state . for finite - connectivity spin glass systems , we use this free energy boltzmann distribution to formulate the cavity approach of mzard and parisi [ eur . phys . j. b * 20 * , 217 ( 2001 ) ] in a slightly different form . for the spin - glass model on a random regular graph of degree , the predictions of the present work agree with earlier simulational and theoretical results .

0710.1364

spin - glasses are simple models for disordered systems . they can be defined very easily in mathematical terms ; on the hand , the properties of such simple models usually are quite rich . statistical physics of spin - glasses has been studied for more than thirty years since the concept of spin - glasses was first presented by edwards and anderson @xcite in 1975 , but there are still many unsolved and heavily debated issues . in the last decades there have been a lot of theoretical investigations concerning models defined on a finite - connectivity random graph . these later models are more realistic than conventional spin - glass models ( e.g. , the sherrington - kirkpatrick model @xcite ) on a complete graph , in the sense that each spin interacts only with a finite number of other spins . as direct analytical studies of spin - glass models on three - dimensional ( regular ) lattices are still beyond reach , people hope that a deep understanding of ( mean - field ) models on finite - connectivity random graphs will shed much light on the properties of 3d systems . for a spin - glass system of very large size , it is generally believed that ergodicity is broken when the environmental temperature @xmath0 becomes lower than a certain value @xmath5 , the spin - glass transition temperature . in this low - temperature spin - glass phase , the time average of a given physical quantity no longer equals to the ensemble average . the whole configurational phase space of the system breaks into many separated domains . ergodicity is still preserved within each such configurational domains . if the system initially is in a spin configuration that belongs to domain @xmath1 of the configurational space , it will eventually visit all other configurations in this domain as time elapses . on the other hand , due to the existence of very high free energy barriers , the system is unable to transit spontaneously from one configurational space domain to another different domain . in this ergodicity - broken situation , each such configurational phase space domain is regarded as a _ macroscopic state _ or _ thermodynamic state_. the free energy @xmath6 of a macroscopic state @xmath1 is related to the microscopic configurations in @xmath1 through the following fundamental formula of statistical mechanics , @xmath7 where @xmath8 is the total number of spins in the system ; @xmath9 is the inverse temperature ( we set boltzmann s constant to unity throughout this paper ) ; and @xmath10 is the partition function for the macroscopic state @xmath1 as defined by @xmath11 \ .\ ] ] in eq . ( [ eq : partitionfunctionalpha ] ) , @xmath12 denotes a microscopic spin configuration , and @xmath13 is the total energy of this configuration . the summation in eq . ( [ eq : partitionfunctionalpha ] ) is over all those microscopic configurations belonging to macroscopic state @xmath1 . for a given spin - glass system , although we can formally write down the expressions of the partition function and free energy for a macroscopic state @xmath1 , the principal difficulty of spin - glass statistical physics is that we do not know _ a priori _ how the configurational space of the system is organized and which are the constituent spin configurations of each macroscopic state . to overcome this difficulty , one possibility is to first assume certain structural organization of the system s configurational space and then try to derive a self - consistent theory . for the sherrington - kirkpatrick model , the full - step replica - symmetry - broken ( frsb ) theory of parisi with an ultrametric organization of macroscopic states @xcite has met with great success . for finite - connectivity mean - field spin - glasses , a cavity approach was also developed by mzard and parisi @xcite . this cavity approach combined the bethe - peierls approximation @xcite for a ferromagnetic ising model with the physical picture that there is a proliferation of macroscopic states in a spin - glass system . this approach , which has been shown @xcite to be equivalent to the first - step replica - symmetry - broken ( 1rsb ) replica theory , can give very good predictions concerning the low - temperature free energy density of a system on a random graph . later on , this cavity approach was extended by mzard and parisi @xcite to the limiting case of zero temperature . the zero - temperature cavity method was applied to some hard combinatorial optimization problems with good performances ( see , e.g. , refs . these interdisciplinary applications in return also call for further understanding on the mean - field cavity approach and its possible extensions ( see , e.g. , refs . @xcite ) . for a spin - glass system at a low temperature @xmath0 , the total number of macroscopic states with a given free energy @xmath14 is denoted as @xmath15 . although it is natural to anticipate that the logarithm of @xmath15 should be an extensive quantity , the exact relationship between @xmath16 and @xmath14 is unknown and is system - dependent . inspired by parisi s frsb solution , in the cavity approach of mzard and parisi @xcite one assumes that the total number @xmath17 of macroscopic states with free energy @xmath14 diverges exponentially with @xmath14 with respect to a reference free energy @xmath18 @xcite , i.e. , @xmath19 where @xmath20 is a dimensionless constant to be determined self - consistently . the parisi parameter @xmath21 stems from the replica theory of infinite - connectivity spin - glasses @xcite , its physical meaning in the cavity framework is not transparent . for some spin - glass systems with many - body interactions , even if the exponential form of eq . ( [ eq : mp01 ] ) is valid in certain range of free energy values , the parameter @xmath21 in this equation may exceed unity . the original cavity iterative equations of ref . @xcite diverge for this situation of @xmath22 , where the statistical physical property of the spin - glass system is not determined by those macroscopic states with the globally minimal free energy density . since the cavity approach of mzard and parisi is a very useful theoretical tool in studying the low - temperature properties of many spin - glass systems , it might be helpful for us to interpret the approach from an another slightly different angle . complementary interpretations of the 1rsb cavity theory will also facilitate its further development . in this paper we demonstrate that , the 1rsb cavity formalism of mzard and parisi can be derived in an alternative way without using eq . ( [ eq : mp01 ] ) . this slightly revised cavity theory is based on the gedanken experiment of repeated temperature heating - annealing . the final macroscopic state reached at the end of an annealing process is anticipated to follow the boltzmann distribution of free energies . this theoretical approach is applied to the @xmath3 spin - glass model on a random regular graph of degree @xmath4 to test its validity . the results of the present work are in close agreement with earlier simulational and numerical results . the mathematical format of the present cavity theory is the same both for non - zero temperatures and for zero temperature . if the macroscopic states of a spin - glass system further organize into clusters of macroscopic states , it can be easily extended to take into account this situation . this paper is organized as follows . the next section introduces the @xmath3 spin - glass model and the ensemble of random regular graphs of degree @xmath23 . section [ sec : mfep ] describes the free energy boltzmann distribution of macroscopic states . in sec . [ sec : cavity - field - distribution ] the concepts of cavity field and cavity magnetization are re - introduced , and the distribution of a vertex s cavity magnetization among all the macroscopic states is calculated . various thermodynamic quantities , including the grand free energy density of the whole system and the free energy density of a macroscopic state , are calculated in sec . [ sec : densities ] and sec . [ sec : single - instance ] for an ensemble of systems and for a single system , respectively . the numerical results for the @xmath3 spin - glass model are reported and analyzed in sec . [ sec : numerics ] . we conclude the present work and discuss its possible extensions in sec . [ sec : discussion ] . the appendix gives an explicit expression for the mean energy density . in this paper we focus on just a single example , the @xmath3 spin - glass model @xcite on a random regular graph of degree @xmath23 . this model was also studied in ref . @xcite , lending us the opportunity to directly compare the results of both treatments . let us consider the graph @xmath24 which is obtained by randomly choosing with equal probability a graph from the set of all regular graphs of size @xmath8 and vertex - degree @xmath23 . @xmath24 is a random regular graph of degree @xmath23 . each of the @xmath8 vertices of @xmath24 is connected to @xmath23 other vertices , but there is no structure in the connection pattern of @xmath24 . for each edge @xmath25 of graph @xmath24 , we assign a coupling constant @xmath26 or @xmath27 with equal probability . once the coupling @xmath28 for an edge is assigned , it no longer changes . therefore , we have a _ network _ with random quenched connection pattern and random quenched coupling constants . ( in the remaining part of this paper , when we use the term @xmath29network , we mean the connection pattern of the graph plus the quenched couplings ; when we use the term @xmath29graph , we mean only the connection pattern . ) on top of such a network we define the following energy function @xmath30 where @xmath31 ( @xmath32 ) is the spin variable of vertex @xmath33 , @xmath31 can take two values , @xmath34 and @xmath35 . equation ( [ eq : energy01 ] ) is the @xmath3 spin - glass model . the total number of microscopic configurations @xmath36 in such a system is simply @xmath37 . the vertex degree of a random regular graph of degree @xmath23 is equal to that of a regular square lattice in @xmath38 dimensions . one hopes that some statistical physical properties of a spin - glass model on a random regular graph will also hold for the same model on a regular lattice . on the other hand , in a regular lattice , there exist many short loops , which make analytical calculations extremely difficult . in a random regular graph of very large size there is no such short loops . the graph is locally tree - like ; and the typical loop length in the graph scales as @xmath39 . one can exploit this absence of short loops to construct a self - consistent mean - field theory . for the benefit of later discussions , we also mention the concept of a random regular cavity graph @xmath40 @xcite . in this graph of size @xmath8 , there are @xmath41 ( @xmath42 ) vertices of degree @xmath43 and @xmath44 vertices of degree @xmath23 . a vertex of degree @xmath43 in graph @xmath40 is referred to as a _ cavity vertex_. like @xmath24 , the connection pattern of @xmath40 is also completely random ; and the quenched coupling constants for the edges of @xmath40 are also independently and equally distributed over @xmath3 . the spin - glass hamiltonian eq . ( [ eq : energy01 ] ) can also be defined for this cavity network . consider the spin - glass system eq . ( [ eq : energy01 ] ) . at high temperatures , the system is in the paramagnetic phase . each vertex of the network does not have any spontaneous magnetization , its spin fluctuates over the positive and negative directions and stays in each orientation with equal probability . the configurational phase space of the system is ergodic ( see left panel of fig . [ fig : ergodicitybreaking01 ] ) . when the temperature @xmath0 is decreased below a spin - glass transition temperature @xmath5 , the system is in the spin - glass phase . in this phase , a vertex of the system might favor one spin orientation over the opposite orientation , but this orientation preference is vertex - dependent . when the number @xmath8 of vertices is sufficiently large , the configurational phase space of this spin - glass system splits into many separated domains ( see right panel of fig . [ fig : ergodicitybreaking01 ] ) . each domain of the phase space , which corresponds to a macroscopic state or thermodynamic state of the system , contains a set of microscopic spin configurations . the system is ergodic within each macroscopic state , while ergodicity is broken at the level of macroscopic states . however , the system is able to transit from one macroscopic state to another different macroscopic state if the following temperature heating - annealing process is performed : the system is first heated to a high temperature beyond @xmath5 and waited for a long time till it reaches equilibrium . the system now stays in a high - temperature ergodic phase . afterwords , it is cooled infinitely slowly till the final low temperature @xmath0 is reached . ( a simulated - annealing idea was previously explored by kirkpatrick and co - workers @xcite to tackle hard optimization problems . ) since in the high - temperature phase the system loses memory about its prior history , at the end of the annealing process it may reach any a macroscopic state . the probability @xmath45 of the system end up being in a particular macroscopic state @xmath1 , however , is in general different for different macroscopic states . we argue in the following that , if an infinite number of the above - mentioned annealing experiment are performed , @xmath45 should depend only on the free energy @xmath6 of the macroscopic state @xmath1 . a schematic picture of ergodicity breaking in a spin - glass system . ( left ) when the temperature is high , all the microscopic configurations ( denoted by dots ) which contribute to the free energy of the system are in a single domain of the configurational phase space . ( right ) when the temperature is low enough , the microscopic configurations which contribute to the free energy of the system are contained in different domains in the configurational phase space . each domain contains a number of microscopic configurations . ergodicity is still preserved within each phase space domain , while the configuration of the system is unable to jump spontaneously from one domain to another different domain of the phase space . ] the total partition function of the system is defined as @xmath46 where @xmath13 is the total energy expression as given by eq . ( [ eq : energy01 ] ) . when ergodicity is broken , the partition function @xmath47 can be re - written as a sum over all the macroscopic states @xmath1 : @xmath48 in eq . ( [ eq : partitiondomain02 ] ) , @xmath10 and @xmath49 are the partition function and free energy of the macroscopic state @xmath1 as defined in eq . ( [ eq : partitionfunctionalpha ] ) and eq . ( [ eq : falpha ] ) , respectively . by way of repeated temperature annealing , the whole configurational space of a spin - glass system is explored ( and ergodicity is recovered ! ) . the total partition function eq . ( [ eq : partitiontotal ] ) can therefore be understood as containing all the information of the system as measured by an infinite number of temperature annealing experiments . from eq . ( [ eq : partitiondomain02 ] ) we see that each macroscopic state @xmath1 contributes a term @xmath50 to the total partition function of the system . therefore we can anticipate that , at temperature @xmath0 , the probability of the system being in macroscopic state @xmath1 at the end of an annealing experiment is given by the following distribution @xmath51 this is a free energy boltzmann distribution over all the macroscopic states @xmath1 . at this point , we introduce an artificial inverse temperature @xmath52 at the level of macroscopic states . ( such an artificial inverse temperature @xmath52 was first introduced in the early work of ref . @xcite for @xmath53 spin - glasses . ) this is better explained by constructing the following artificial single - particle system : the particle has a set of @xmath29energy levels , the @xmath29energy of level @xmath1 is equal to the free energy @xmath54 of the macroscopic state @xmath1 of the actual spin - glass system ; this one - particle system is in a heat bath with inverse temperature @xmath52 . then the partition function of the artificial system is @xmath55 the probability of this artificial system being in energy level @xmath1 is given by @xmath56 when the inverse temperature @xmath52 of the artificial system is set to @xmath57 , then the partition function @xmath58 reduces to the total partition function @xmath59 of the original spin - glass system . the partition function @xmath58 of eq . ( [ eq : newpartitionfunction ] ) can be written in another form as @xmath60 \ . \label{eq : zyb03}\ ] ] in eq . ( [ eq : zyb03 ] ) , @xmath61 is the free energy density of a macroscopic state ; the function @xmath62 is called the complexity @xcite , which is related to the total number @xmath15 of macroscopic states by the following equation @xmath63 the complexity @xmath62 , which should be non - negative , is a measure of the total number of macroscopic states with free energy density @xmath64 . when the spin - glass system is very large ( @xmath65 ) , eq . ( [ eq : zyb03 ] ) indicates that the partition function @xmath58 is contributed exclusively by those macroscopic states whose free energy density @xmath64 satisfying the equation @xmath66 equation ( [ eq : y ] ) gives an implicit relationship between the inverse temperature @xmath52 and the observed mean free energy density @xmath64 of the system . for the benefit of later discussions , we define a grand free energy @xmath67 by the following equation @xmath68\ .\ ] ] from eq . ( [ eq : zyb03 ] ) we see that the grand free energy density @xmath69 is @xmath70 where @xmath71 is the solution of eq . ( [ eq : y ] ) . in the next three sections we will study the statistical physical properties of the @xmath3 spin - glass system using @xmath72 and @xmath52 as a pair of control parameters . the free energy at the microscopic level ( within a macroscopic state ) will be calculated with the inverse temperature @xmath72 , while the grand free energy at the macroscopic state level will be calculated with the inverse temperature @xmath52 . although in the actual spin - glass system , the two inverse temperatures are identical ( @xmath73 ) , they are decoupled in the following analytical theory . this decoupling gives us extra freedom in the theoretical development . finally , to go from the artificial system to the original system , we will choose the largest value of @xmath52 in the interval of @xmath74 which satisfies the requirement that the complexity @xmath75 is non - negative . to analytically study the statistical physical property of a spin - glass model on a random graph , there are basically two different but equivalent approaches . the first one is the replica method @xcite and the second one is the cavity method @xcite . the present paper exploits the cavity approach . it corresponds to the 1rsb approximation of the replica method . basically , one assumes that macroscopic states of the spin - glass system are distributed evenly in the whole configurational phase space , and there is no further organization of the macroscopic states or further structures within each macroscopic state . consider the spin - glass system eq . ( [ eq : energy01 ] ) on a cavity network @xmath76 of @xmath8 vertices and @xmath43 cavity vertices ( see fig . [ fig : cavity01m1]a ) . let us suppose that the configurations of such a cavity system are in a macroscopic state @xmath1 . cavity vertices in a random regular cavity graph of vertex degree @xmath23 ( here @xmath77 ) . ( a ) the neighborhoods for three cavity vertices @xmath78 , @xmath79 and @xmath80 . these three vertices have vertex degree @xmath43 ( @xmath81 ) while all the other @xmath82 vertices in the graph have vertex degree @xmath23 . in a macroscopic state @xmath1 , the three cavity vertices feel the cavity fields @xmath83 , @xmath84 and @xmath85 , respectively . ( b ) a new cavity graph with @xmath86 vertices is generated by adding a new vertex @xmath33 and connecting it to @xmath43 old cavity vertices . the cavity vertex @xmath33 of the new graph feels the cavity field @xmath87 in the corresponding macroscopic state @xmath1 of the new cavity system . ] at given inverse temperature @xmath72 , the spin value @xmath88 of a cavity vertex @xmath78 fluctuates over time around certain mean value @xmath89 which depends on the macroscopic state @xmath1 . since @xmath90 is a binary variable , the marginal distribution @xmath91 of @xmath88 can be expressed in the following form @xmath92 the magnetization of the cavity vertex @xmath78 is @xmath93 from eq . ( [ eq : magnetization01 ] ) we know that @xmath83 is the magnetic field experienced by the cavity vertex @xmath78 due to the spin - spin interactions between vertex @xmath78 and its @xmath43 nearest - neighbors . we call @xmath83 the _ cavity field _ on vertex @xmath78 and @xmath89 the _ cavity magnetization _ of vertex @xmath78 . we emphasize again that @xmath83 and @xmath89 depends on the macroscopic state @xmath1 . since the connection pattern in a cavity graph with @xmath43 cavity vertices is completely random , the typical length @xmath94 of a shortest - distance path between two cavity vertices @xmath78 and @xmath79 in the cavity graph @xmath76 is a large value . actually , it can easily be shown @xcite that this distance scales logarithmically with the cavity graph size @xmath8 : @xmath95 denote @xmath96 as the joint probability distribution of the spin values @xmath97 of a group of cavity vertices in a cavity network . in a large random graph , according to eq . ( [ eq : lengthscale ] ) the shortest path length between any two randomly chosen vertices is long . therefore , as the zeroth - order approximation , one may assume that this joint probability distribution can be written as the following factorized form @xmath98 where @xmath99 is vertex @xmath78 s marginal spin value distribution as given by eq . ( [ eq : cavityfield01 ] ) . equation ( [ eq : factorization ] ) is called the bethe - peierls approximation @xcite in the literature . it assumes statistical independence among the spin states of the cavity vertices within a macroscopic state @xmath1 . ( for recent references on extensions of the bethe - peierls approximation , see refs . @xcite . ) the cavity magnetization @xmath89 of the cavity vertex @xmath78 depends on the identity of the macroscopic state @xmath1 . its value may be different in different macroscopic states . let us denote @xmath100 as the fraction of macroscopic states in which the cavity magnetization of vertex @xmath78 takes the value @xmath89 , and denote @xmath101 as the fraction of macroscopic states in which the cavity magnetization of vertex @xmath102 take the value @xmath103 , respectively . we extend the bethe - peierls approximation eq . ( [ eq : factorization ] ) to the level of macroscopic states and assume that @xmath104 equation ( [ eq : factorization - magnetization ] ) is equivalent to saying that , the fluctuations ( among all the macroscopic states ) of the cavity magnetization of two different cavity vertices are mutually independent of each other . due to the absence of short loops in a random graph , eq . ( [ eq : factorization - magnetization ] ) turns out to be a rather good approximation . to obtain a self - consistent equation for the marginal probabilities @xmath105 , we add a new vertex @xmath33 and connect it to the @xmath43 cavity vertices of @xmath76 . the quenched coupling constant @xmath28 of each newly added edge is set to be @xmath106 with equal probability . this results in a new cavity network @xmath107 of @xmath86 vertices and one single cavity vertex @xmath33 ( see fig . [ fig : cavity01m1]b ) . in the corresponding macroscopic state @xmath1 of the new cavity system @xmath107 , the cavity vertex @xmath33 feels a cavity field @xmath87 . to calculate the cavity field @xmath87 in the new cavity system , we first notice that the energy difference between the cavity network @xmath107 and the old cavity network @xmath76 is @xmath108 where @xmath109 denotes the set of @xmath43 nearest - neighbors of vertex @xmath33 in the cavity graph @xmath107 . in the macroscopic state @xmath1 , the partition function for the cavity system @xmath107 is @xmath110 \sum\limits_{\sigma_i } \exp\bigl ( \beta h_i \sigma_i \bigr ) \label{eq : zacv03m1 } \\ & = & e^{-\beta f_\alpha \bigl ( { \cal g}_k(n , k-1)\bigr ) } \prod\limits_{j\in \partial^\prime i } \biggl[\frac { \cosh \beta j_{i j } } { \cosh \beta u(j_{i j } , h_{j } ) } \biggr ] ( 2 \cosh \beta h_{i } ) \ . \label{eq : zacv03}\end{aligned}\ ] ] in going from eq . ( [ eq : zacv03m3 ] ) to eq . ( [ eq : zacv03m2 ] ) , we have used the bethe - peierls approximation eq . ( [ eq : factorization ] ) . @xmath111 is the free energy of the cavity system @xmath76 in its macroscopic state @xmath1 . in eq . ( [ eq : zacv03m1 ] ) , the quantity @xmath112 is defined as @xmath113 and the quantity @xmath87 is calculated according to @xmath114 from eq . ( [ eq : zacv03m1 ] ) we know that @xmath87 as expressed by eq . ( [ eq : hia ] ) is just the cavity field felt by vertex @xmath33 in the cavity network @xmath107 . if we know all the cavity fields on the nearest - neighbors of vertex @xmath33 , then we obtain the cavity field on vertex @xmath33 through the iterative equation ( [ eq : hia ] ) . consequently , the magnetization of the cavity vertex @xmath33 in the macroscopic state @xmath1 of the cavity system @xmath107 is calculated as @xmath115 - \prod\limits_{j\in \partial^\prime i } \bigl [ 1 - v_{i j } m_j \bigr ] } { \prod\limits_{j\in \partial^\prime i } \bigl [ 1 + v_{i j } m_j \bigr ] + \prod\limits_{j\in \partial^\prime i } \bigl [ 1 - v_{i j } m_j ] } \ , \ ] ] where @xmath116 is a shorthand notation for @xmath117 , i.e. , @xmath118 equation ( [ eq : mi ] ) is an iterative equation for the cavity magnetization @xmath119 _ within one macroscopic state @xmath1_. as we have emphasized in sec . [ subsec : single01 ] , the input cavity magnetization @xmath89 in eq . ( [ eq : mi ] ) may be different in different macroscopic states . as a consequence , the cavity magnetization @xmath119 of vertex @xmath33 does not necessarily take the same value in different macroscopic states . on the contrary , its value may fluctuate a lot among different macroscopic states of the new cavity system @xmath107 . the task is to obtain an expression for the marginal distribution @xmath120 of @xmath119 among different macroscopic states . from eq . ( [ eq : zacv03 ] ) , we know that , after the addition of the vertex @xmath33 , the free energy difference @xmath121 between the macroscopic state @xmath1 of the system @xmath107 and that of the system @xmath76 is @xmath122 + \prod\limits_{j\in \partial^\prime i } \bigl[1- v_{i j } m_j \bigr ] \biggr ) \ . \label{eq : deltaf102}\end{aligned}\ ] ] according to the free energy boltzmann distribution eq . ( [ eq : boltzmanndistribution02 ] ) , each macroscopic state is weighted by the boltzmann factor @xmath123 . after the addition of vertex @xmath33 , the total partition function of the system @xmath107 is @xmath124 \prod\limits_{j\in \partial^\prime i } \biggl [ \int { \rm d } m_j { \cal p}_j(m_j ) \biggr ] \exp(-y \delta f_1 ) \ . \label{eq : rho - m1}\end{aligned}\ ] ] we have used the factorization approximation eq . ( [ eq : factorization - magnetization ] ) in going from eq . ( [ eq : rho - m1a ] ) to eq . ( [ eq : rho - m1 ] ) . similarly , the total weight of those macroscopic states of the system @xmath107 in which the cavity magnetization of vertex @xmath33 being equal to @xmath119 is expressed as @xmath125 \times \nonumber \\ & & \prod\limits_{j\in \partial^\prime i } \biggl [ \int { \rm d } m_j { \cal p}_j(m_j ) \biggr ] \exp(-y \delta f_1 ) \delta\biggl(m_i-\frac { \prod\limits_{j\in \partial^\prime i } \bigl [ 1 + v_{i j } m_j \bigr ] - \prod\limits_{j\in \partial^\prime i } \bigl [ 1 - v_{i j } m_j \bigr ] } { \prod\limits_{j\in \partial^\prime i } \bigl [ 1 + v_{i j } m_j \bigr ] + \prod\limits_{j\in \partial^\prime i } \bigl [ 1 - v_{i j } m_j ] } \biggr ) \ . \label{eq : rho - m2}\end{aligned}\ ] ] from eq . ( [ eq : rho - m1 ] ) and eq . ( [ eq : rho - m2 ] ) we realize that , in the new system , the fraction of macroscopic states in which vertex @xmath33 bearing a cavity magnetization @xmath119 is equal to @xmath126 e^{-y \delta f_1 } \delta\biggl ( m_i - \frac { \prod\limits_{j\in \partial^\prime i } \bigl [ 1 + v_{i j } m_j \bigr ] - \prod\limits_{j\in \partial^\prime i } \bigl [ 1 - v_{i j } m_j \bigr ] } { \prod\limits_{j\in \partial^\prime i } \bigl [ 1 + v_{i j } m_j \bigr ] + \prod\limits_{j\in \partial^\prime i } \bigl [ 1 - v_{i j } m_j ] } \biggr ) \ .\ ] ] equation ( [ eq : rho ] ) is a self - consistent iterative equation for the cavity distributions @xmath127 . a steady state solution of eq . ( [ eq : rho ] ) can be obtained by population dynamics @xcite . an array of @xmath128 probability distributions @xmath129 are stored . at each step of the population dynamics , @xmath43 probability distributions are randomly chosen from this array of @xmath128 stored distributions , and a new probability distribution is generated by using eq . ( [ eq : rho ] ) . this new distribution then replaces a randomly chosen old probability distribution in the array . this iteration process is repeated many times until the population dynamics reaches a steady state . the free energy of a macroscopic state @xmath1 is defined formally by eq . ( [ eq : falpha ] ) . as we noted before , this expression is not directly applicable since we do not know what are the microscopic configurations of state @xmath1 . the cavity approach @xcite circumvents this problem by calculating the grand free energy difference between a system of @xmath8 vertices and an enlarged system of @xmath130 vertices . as demonstrated in fig . [ fig : cavity01 ] , a random network @xmath24 can be constructed from a random cavity network @xmath131 by adding @xmath43 new edges . the grand free energy difference between these two systems can be calculated . similarly , one can add two new vertices and @xmath132 new edges to change the same random cavity network @xmath133 into a random regular network @xmath134 . the grand free energy difference between these two systems can also be calculated . from the two grand free energy differences , one can obtain the grand free energy density @xmath135 for a random regular system @xmath24 . the mean free energy density of a macroscopic state and the complexity of the system can then be obtained from @xmath135 . the cavity approach to the spin - glass model eq . ( [ eq : energy01 ] ) on a random regular graph of connectivity @xmath136 . ( a ) : part of a cavity graph with @xmath137 cavity vertices . ( b ) and ( c ) : construction of a random regular graph of size @xmath8 ( b ) and @xmath130 ( c ) from the cavity graph with @xmath8 vertices and @xmath137 cavity vertices . , title="fig : " ] the cavity approach to the spin - glass model eq . ( [ eq : energy01 ] ) on a random regular graph of connectivity @xmath136 . ( a ) : part of a cavity graph with @xmath137 cavity vertices . ( b ) and ( c ) : construction of a random regular graph of size @xmath8 ( b ) and @xmath130 ( c ) from the cavity graph with @xmath8 vertices and @xmath137 cavity vertices . , title="fig : " ] the cavity approach to the spin - glass model eq . ( [ eq : energy01 ] ) on a random regular graph of connectivity @xmath136 . ( a ) : part of a cavity graph with @xmath137 cavity vertices . ( b ) and ( c ) : construction of a random regular graph of size @xmath8 ( b ) and @xmath130 ( c ) from the cavity graph with @xmath8 vertices and @xmath137 cavity vertices . , title="fig : " ] the difference @xmath139 between the configurational energy of the system on the random graph @xmath24 in fig . [ fig : cavity01]b and that of the system on the random cavity graph @xmath140 in fig . [ fig : cavity01]a is @xmath141 where @xmath142 denotes the set of @xmath43 newly added edges in going from @xmath140 to @xmath24 . each edge in set @xmath142 connects two cavity vertices of @xmath133 . following the analytical procedure of sec . [ subsec : field - distribution ] , we know that difference between the free energy of a macroscopic state @xmath1 of the system @xmath24 and that of the same macroscopic state @xmath1 of the cavity system @xmath131 is @xmath143 where @xmath144 can be understood as the free energy increase caused by the addition of an edge ( with coupling @xmath28 ) between two cavity vertices @xmath33 and @xmath78 . the grand free energy @xmath145 of the new system @xmath24 is related to the grand free energy @xmath146 of the old cavity system @xmath147 by the following equation @xmath148 where @xmath149 \\ ] ] is the increase to the grand free energy caused by adding an edge @xmath25 . the random network @xmath134 in fig . [ fig : cavity01]c is constructed from the random cavity network @xmath140 of fig . [ fig : cavity01]a by adding two vertices ( @xmath151 and @xmath152 ) . vertex @xmath151 is connected to @xmath43 cavity vertices ( @xmath153 ) of @xmath133 , and vertex @xmath152 is connected to the remaining @xmath43 cavity vertices ( @xmath154 ) of @xmath147 . vertex @xmath151 and @xmath152 are directly connected by a new edge @xmath155 , so that every vertex in the graph @xmath134 has degree @xmath23 . the energy difference between the system @xmath134 and the cavity system @xmath147 is @xmath156 the increase @xmath157 in the free energy of macroscopic state @xmath1 due to the addition of two new vertices and @xmath158 new edges can be obtained following the same procedure as given in sec . [ subsec : field - distribution ] . we find that @xmath159 in eq . ( [ eq : freeenergy03 ] ) , @xmath160 is the free energy increase caused by adding vertex @xmath151 and connecting it to the set @xmath161 of @xmath23 vertices . the explicit expression for @xmath160 is @xmath162 + \prod\limits_{j\in \partial i_0 } \bigl[1- v_{i_0 j } m_j \bigr ] \biggr ) \ . \label{eq : deltafi0}\ ] ] the expression for @xmath163 has the same form as eq . ( [ eq : deltafi0 ] ) . let us emphasize that , in eq . ( [ eq : deltafi0 ] ) , @xmath164 is the cavity magnetization of vertex @xmath152 when vertex @xmath151 is not added , i.e. , @xmath165- \prod\limits_{j_s\in \partial i_0 \backslash j_0 } \bigl [ 1 - v_{i_0 j_s } m_{j_s } \bigr ] } { \prod\limits_{j_s\in \partial i_0 \backslash j_0 } \bigl [ 1 + v_{i_0 j_s } m_{j_s } \bigr ] + \prod\limits_{j_s\in \partial i_0 \backslash j_0 } \bigl [ 1 - v_{i_0 j_s } m_{j_s } \bigr ] } \ , \label{eq : m0val } \\ m_{j_0 } & = & \frac { \prod\limits_{i_s \in \partial j_0 \backslash i_0 } \bigl [ 1 + v_{j_0 i_s } m_{i_s } \bigr]- \prod\limits_{i_s \in \partial j_0 \backslash i_0 } \bigl [ 1 - v_{j_0 i_s } m_{i_s } \bigr ] } { \prod\limits_{i_s \in \partial j_0 \backslash i_0 } \bigl [ 1 + v_{j_0 i_s } m_{i_s } \bigr ] + \prod\limits_{i_s\in \partial j_0 \backslash i_0 } \bigl [ 1 - v_{j_0 i_s } m_{i_s } \bigr ] } \ . \label{eq : m0pval}\end{aligned}\ ] ] the term @xmath166 of eq . ( [ eq : freeenergy03 ] ) is the free energy increase caused by setting up an edge between vertex @xmath151 and vertex @xmath152 ; its expression is given by eq . ( [ eq : freeenergy02 ] ) , with @xmath167 and @xmath164 being determined by eq . ( [ eq : m0val ] ) and eq . ( [ eq : m0pval ] ) , respectively . the free energy increase eq . ( [ eq : freeenergy03 ] ) can be intuitively understood as follows : since the contribution of the edge @xmath168 is counted twice in @xmath160 and @xmath163 , the free energy increase should be corrected with an edge term . with these preparations , we can calculate the total grand free energy @xmath169 of the system @xmath134 . similar to eq . ( [ eq : gfd - i ] ) , we find that @xmath170 in eq . ( [ eq : delta - g-2 - 01 ] ) , @xmath171 is the increase to the grand free energy caused by adding vertex @xmath151 and connecting it to @xmath23 vertices , with @xmath172 where @xmath173 is the cavity magnetization distribution of vertex @xmath152 in the absence of vertex @xmath151 . @xmath174 in eq . ( [ eq : delta - g-2 - 01 ] ) is calculated through eq . ( [ eq : dgij ] ) using @xmath175 and @xmath176 . the grand free energy density of the spin - glass system is @xmath177 an explicit expression for @xmath135 can be written down by applying eq . ( [ eq : gfd - i ] ) and eq . ( [ eq : delta - g-2 - 01 ] ) , which is a function of the quenched randomness in the system . the grand free energy density has the nice property of self - averaging @xcite , namely the value of @xmath69 as calculated for a typical system is equal to the averaged value of @xmath69 over many systems with different realizations of the quenched randomness in the graph connection pattern and in the edge coupling constants . when the quenched randomness is averaged out , we obtain that @xmath178 where @xmath179 and @xmath180 are calculated through eq . ( [ eq : dgi0 ] ) and eq . ( [ eq : dgij ] ) , respectively ; and an overline indicates averaging over the quenched randomness of the spin - glass system . the mean free energy density of a macroscopic state of the system is related to @xmath135 by @xmath181 and the complexity of the system at given value of the reweighting parameter @xmath52 is @xmath182 the discussion in sec . [ sec : densities ] was concerned with the typical properties of an ensemble of spin - glass systems governed by a given distribution of quenched randomness . one important advantage of the cavity approach is that , for a single instance of the quenched randomness , the thermodynamic properties can also be calculated . under the bethe - peierls approximation , the total grand free energy of a spin - glass system eq . ( [ eq : energy01 ] ) on a graph @xmath183 with couplings @xmath184 is equal to @xmath185 it is easy to see that the above equation is consistent with eq . ( [ eq : g - final ] ) . in eq . ( [ eq : gfe - single-01 ] ) , @xmath186 denotes the contribution of vertex @xmath33 and its associated edges to the total grand free energy of the system ; its expression has the same form as eq . ( [ eq : dgi0 ] ) : @xmath187 e^{-y \delta f^{(i ) } } \biggr ) \ , \ ] ] where @xmath188 is the cavity magnetization of vertex @xmath78 ( with respect to vertex @xmath33 ) in a macroscopic state @xmath1 , and @xmath189 is the probability distribution of this cavity magnetization among all the macroscopic states ; @xmath190 is the free energy contribution ( in a given macroscopic state ) of vertex @xmath33 and its associated edges , which has the same form as eq . ( [ eq : deltafi0 ] ) but with @xmath89 replaced by @xmath188 . similarly , the term @xmath191 in eq . ( [ eq : gfe - single-01 ] ) is the contribution of an edge @xmath25 to the total grand free energy of the system ; it is expressed as @xmath192 where @xmath193 is given by eq . ( [ eq : freeenergy02 ] ) but with @xmath119 replaced by @xmath194 and @xmath89 replaced by @xmath188 . the set of @xmath195 ( @xmath196 being the total number of edges in the graph @xmath183 ) probability distributions @xmath189 in eq . ( [ eq : gfe - single-01 ] ) should be carefully chosen such that the total grand free energy achieves a minimal value at given @xmath197 values . in other words , for any edge @xmath25 of the graph @xmath183 , the variation of @xmath198 with respect to both @xmath199 and @xmath200 should vanish : @xmath201 equation ( [ eq : gfe - single-04 ] ) results in the following self - consistent bethe - peierls equation for the @xmath202 s : @xmath203 e^{-y \delta f_j^{(i ) } } \delta\biggl ( m_{i\to j}-\frac { \prod\limits_{k\in \partial i \backslash j } [ 1 + v_{i k } m_{k\to i } ] - \prod\limits_{k\in \partial i \backslash j } [ 1 - v_{i k } m_{k\to i } ] } { \prod\limits_{k\in \partial i\backslash j } [ 1 + v_{i k } m_{k\to i}]+ \prod\limits_{k\in \partial i\backslash j } [ 1 - v_{i k } m_{k\to i } ] } \biggr ) \ , \ ] ] where @xmath204 + \prod\limits_{k\in \partial i\backslash j } \bigl[1- v_{i k } m_{k\to i } \bigr ] \biggr ) \ .\ ] ] @xmath205 has the same physical meaning as @xmath121 of sec . [ subsec : field - distribution ] . equation ( [ eq : gfe - single-06 ] ) is consistent with eq . ( [ eq : rho ] ) of sec . [ subsec : field - distribution ] . similar to eq . ( [ eq : f - final ] ) and eq . ( [ eq : sigma - final ] ) , the mean free energy of a macroscopic state of the sample is expressed as @xmath206 and the complexity of the system is calculated through @xmath207 because of eq . ( [ eq : gfe - single-04 ] ) , the first derivative of @xmath208 with respect to @xmath52 can easily be expressed . we find that eq . ( [ eq : fe - single-02 ] ) can be re - written as @xmath209 where @xmath210 e^{-y \delta f^{(i ) } } \delta f^{(i ) } } { \prod\limits_{j\in \partial i } \bigl[\int { \rm d } m_{j\to i } { \cal p}_{j\to i}(m_{j\to i } ) \bigr ] e^{-y \delta f^{(i ) } } } \ , \label{eq : average - dfi } \\ \bigl\langle \delta f^{(i j ) } \bigr\rangle & = & \frac{\int { \rm d } m_{j\to i } \int { \rm d } m_{i\to j } { \cal p}_{j\to i}(m_{j\to i } ) { \cal p}_{i\to j}(m_{i\to j } ) e^{-y \delta f^{(i j ) } } \delta f^{(i j ) } } { \int { \rm d } m_{j\to i } \int { \rm d } m_{i\to j } { \cal p}_{j\to i}(m_{j\to i}){\cal p}_{i\to j}(m_{i\to j } ) e^{-y \delta f^{(i j ) } } } \ . \label{eq : average - dfij}\end{aligned}\ ] ] the mean free energy of a macroscopic state is also decomposed into vertex contributions and edge contributions . according to the discussion in sec . [ sec : mfep ] , in the mean - field theory , the reweighting parameter @xmath52 should be chosen such that its value is closest to the physical inverse temperature @xmath72 . therefore , at a given value of @xmath72 , the mean free energy @xmath211 of the system is obtained by setting @xmath57 in eq . ( [ eq : fe - single-02 ] ) or eq . ( [ eq : free - energy-1rsb ] ) , provided that the complexity @xmath212 is always non - negative in the range of @xmath213 . if , on the other hand , @xmath212 changes from being positive to being negative at a point of @xmath214 , we should set @xmath52 to @xmath215 . in this later situation , the mean free energy of the system is equal to the maximal value of the grand free energy , @xmath216 . after the mean free energy @xmath211 of the spin - glass system is obtained , the mean energy of a macroscopic state is calculated through @xmath217 and the mean entropy of a macroscopic state is calculated through @xmath218 ensemble - averaged densities of mean energy and mean entropy can also be obtained from eqs . ( [ eq : mean - energy ] ) and ( [ eq : mean - entropy ] ) by averaging out the quenched randomness . an alternative derivation for the mean energy density is given in the appendix . we have applied the mean - field treatments of sec . [ sec : cavity - field - distribution ] and sec . [ sec : densities ] to the @xmath3 spin - glass model eq . ( [ eq : energy01 ] ) on a random regular graph of vertex degree @xmath4 . ( in what follows , energy is in units of @xmath219 . ) at temperature @xmath220 ( i.e. , @xmath221 ) , carrus , marinari and zuliani ( as cited in ref . @xcite ) estimated that the energy density of this system is @xmath222 . the mean - field theory of ref . @xcite also predicted the same result . this reference also predicted a mean - free energy density of @xmath223 . in the present population dynamics simulation , each probability profile @xmath120 in eq . ( [ eq : rho ] ) was represented by an vector of @xmath224 elements . a total number of @xmath128 such probability profiles were stored in the computer and they were updated using eq . ( [ eq : rho ] ) . in our work , various @xmath128 values ranging from @xmath225 to @xmath226 were used . in each elementary update of the probability distribution @xmath120 , a metropolis importance sampling @xcite with inverse temperature @xmath52 was exploited to generate a total number of @xmath227 magnetizations @xmath119 . these magnetizations were stored in an ordered set . the averaged value of the first ( second , third , @xmath228 ) @xmath229 magnetizations in this set is assigned to the first ( second , third,@xmath228 ) element of the new @xmath129 . we chose @xmath230 in our simulations . the grand free energy density @xmath69 of the @xmath3 spin - glass model on a random regular graph of vertex degree @xmath4 . the inverse temperature is fixed to @xmath221 . symbols are results obtained by population dynamics simulations using a population size of @xmath231 . the dashed line is a fit to the data using @xmath232 , with the fitting parameters @xmath233 and @xmath234 . ] mean free energy density of the @xmath3 spin - glass model on a random regular graph of vertex degree @xmath4 at @xmath221 as obtained by population dynamics with different population size @xmath128 . ( inset ) the optimal reweighting parameter @xmath235 . ] the energy density , free energy density , entropy density , and complexity as a function of the reweighting parameter @xmath52 for the @xmath3 spin - glass model on a random regular graph of vertex degree @xmath4 at inverse temperature @xmath221 . energy unit is @xmath219 . the symbols represent numerical population dynamics results using @xmath236 . the dashed line in the bottom panel is a fitting to the numerical data with @xmath237 in the range of @xmath238 . to reduce computational time , for the simulations reported in this figure , a faster but less precise numerical recipe was used to generate a new probability profile @xmath120 using eq . ( [ eq : rho ] ) . this might be the reason for the relatively large fluctuations of the mean densities with the reweighting parameter @xmath52 . ] the grand free energy density @xmath69 of the system as calculated according to eq . ( [ eq : g - final ] ) is shown in fig . [ fig : gy ] . the grand free energy density @xmath239 first increases with the reweighting parameter @xmath52 till @xmath52 reaches @xmath234 , at which point @xmath135 attains a maximal value of @xmath240 . this maximal value of @xmath135 corresponds to the best estimate of the mean free energy density of the system by the present mean - field method . it is in close agreement with the prediction of ref . we have checked that the estimated mean free energy density value is not sensitive to the population size @xmath128 ( see fig . [ fig : mfd ] ) . we have also calculated the mean energy density and mean entropy density for the model system at @xmath221 using a population size of @xmath241 , with the mean energy density being @xmath242 and the mean entropy density being @xmath243 ( see fig . [ fig : result01 ] ) . these predictions are also in good agreement with the numerical results of ref . @xcite and with the simulational work of carrus , marinari and zuliani . these good agreements of the present theoretical results with earlier simulational and numerical work suggest that the present cavity approach based on the concept of cyclic heating and annealing is feasible . as a summary , in this paper we have calculated the thermodynamic properties of a spin - glass model by combining the physical idea of repeated heating - annealing and the cavity approach of mzard and parisi @xcite . we have assumed that , during a cyclic annealing experiment , all the thermodynamic ( or macroscopic ) states of the spin - glass system at a given low temperature @xmath0 will be reachable , but with different frequencies which decrease exponentially with the free energy values of the macroscopic states . by using this free energy boltzmann distribution and by using the bethe - peierls approximation , the grand free energy of the spin - glass system can be calculated as a function of a reweighting parameter @xmath52 ; and from the knowledge of the grand free energy , the mean free energy , energy , and entropy of a macroscopic state of the system can also be obtained . for the @xmath3 spin - glass model on a random regular graph of vertex degree @xmath4 , the theoretical predictions of the present work are in good agreement with the results of earlier simulational and numerical calculations . for the @xmath3 model at @xmath221 , we found that the complexity @xmath75 becomes negative before @xmath52 reaches @xmath72 . therefore , the thermodynamic properties of the system are contributed by those macroscopic states which have the global minimal free energy density . however , it may exist other systems for which the complexity is still positive even when the reweighting parameter @xmath52 reaches @xmath72 . for such systems , the thermodynamic properties are not determined by those macroscopic states of the minimal free energy density , but by a set of @xmath29metastable macroscopic states . on the one hand , these metastable macroscopic states have a higher free energy density ; on the other hand , the number of such macroscopic states greatly exceeds the number of macroscopic states of the global minimal free energy density . in the competition between these two factors , the @xmath29metastable macroscopic states may win . then the system will reach one of these metastable macroscopic states almost surely in each round of the temperature annealing experiment . we will work on a model system with many - body interactions to check whether this is really the case . as in the work of ref . @xcite , the present theoretical treatment also assumes that the configurational space of the spin - glass system breaks into exponentially many macroscopic states , but there are no further organizations of these macroscopic states ; and it is assumed that each macroscopic state is ergodic . as pointed out in ref . @xcite , this first - step replica - symmetry - broken ( 1rsb ) cavity solution might be unstable . two types of instability is conceivable . the type i instability concerns with possible clustering of macroscopic states into @xmath29super - macroscopic states ; the type ii instability is caused by splitting of each macroscopic state into many @xmath29sub - macroscopic states . to account for the further clustering of macroscopic states appears to be relatively easy in the present mean - field framework : for each cluster of macroscopic states , a vertex has a probability profile concerning its cavity magnetization ; this probability profile is different in different clusters , and we can introduce a probability distribution of probability profiles to characterize this variation [ h. zhou , submitted to comm . ( beijing ) ] . on the hand , to take into consideration the possibility of splitting of a macroscopic state , maybe one has to introduce other reweighting parameters in the mean - field theory . this is an important issue waiting for further explorations . another important issue is related to the updating of the probability profiles through the iterative equation ( [ eq : rho ] ) or ( [ eq : gfe - single-05 ] ) . in the present paper , we used metropolis importance sampling technique to get an updated probability profile of cavity magnetization . this method appears to be rather precise but time - consuming . if faster algorithms with comparable precision could be constructed , it will be highly desirable . the numerical simulations reported in this paper were carried out at the pc clusters of the state key laboratory for scientific and engineering computing , cas , china . in the main text we did not give an explicit formula for the mean energy density @xmath244 of a spin - glass system . instead , the mean energy density was calculated from first knowing the mean free energy density of the system [ see , e.g. , eq . ( [ eq : mean - energy ] ) ] . in this appendix , we derive an explicit expression for the mean energy density @xmath244 . the assumptions used in this derivation will be clearly pointed out . consider again the two systems , @xmath133 in fig . [ fig : cavity01]a and @xmath24 in fig . [ fig : cavity01]b . ( for notational simplicity , let us hereafter referr to these two systems as system a and b , respectively . ) as was mentioned in sec . [ subsec : nton ] , for a given configuration the energy difference between these two systems is @xmath139 . when averaged over all macroscopic states , the mean total energy of the system a is @xmath245 where @xmath246 is the total free energy of macroscopic state @xmath1 of system a , and @xmath247 is the the configurational energy of system a. similarly , the mean total energy of the system b is @xmath248 \exp\bigl(-\beta h^a(\vec{\sigma } ) - \beta \delta h_2 \bigr ) } { \sum_{\vec{\sigma}\in \alpha } \exp\bigl(-\beta h^a(\vec{\sigma } ) - \beta \delta h_2 \bigr ) } } { \sum\limits_\alpha \exp\bigl(-y f_\alpha^a - y \delta f_2 \bigr ) } \ , \ ] ] where @xmath249 is defined by eq . ( [ eq : freeenergy01 ] ) . the difference of these two mean energy expressions is @xmath250 in the above equation , @xmath251 where @xmath252 in going from eq . ( [ eq : v1m1 ] ) to eq . ( [ eq : v1 ] ) we have used the bethe - peierls approximation both at the level of microscopic configurations [ eq . ( [ eq : factorization ] ) ] and at the level of macroscopic states [ eq . ( [ eq : factorization - magnetization ] ) ] . 1 . the distribution of the configurational energy of the cavity system a and that of the energy increase can be regarded as mutually independent within a macroscopic state . 2 . the distributions of the free energy of the cavity system a and that of the free energy increase can also be regarded as mutually independent . now consider the two systems , @xmath133 in fig . [ fig : cavity01]a ( system a ) and @xmath134 in fig . [ fig : cavity01]c ( system c ) . following the same analytical procedure as given in the preceding subsection , we find that , if the two assumptions ( i ) and ( ii ) listed above are valid when going from system a to system c , then the mean total energy of system c is related to that of system a by the following simple equation @xmath257 ( notice that in calculating @xmath258 through eq . ( [ eq : deij ] ) , @xmath167 should be the cavity magnetization of vertex @xmath151 in the absence of vertex @xmath259 . ) . the same analytical procedure can be applied to calculate the mean free energy density . under the assumption ( ii ) of the preceding subsection , we obtain that the mean free energy density @xmath64 is equal to @xmath261 which is just the same as eq . ( [ eq : free - energy-1rsb ] ) . based on the assumption of exponentially many macroscopic states , it was argued in ref . @xcite that , the assumption ( ii ) of the preceding subsection should be valid under the bethe - peierls approximation . in other words , the joint distribution of the free energy of the old cavity system @xmath14 and the free energy increase @xmath262 is factorized . the consistence between eq . ( [ eq : f - another - final ] ) and eq . ( [ eq : free - energy-1rsb ] ) also confirmed this point . since each macroscopic state ifself contains an exponential number of microscopic configurations , we believe that the assumption ( i ) is also valid in the bethe - peierls mean - field framework . the mean energy density of the @xmath3 spin - glass model as shown in fig . [ fig : result01 ] was calculated using eq . ( [ eq : epsilon - final ] ) , and the output was in good agreement with earlier simulation result . maybe one can further check this point by working on some simple models which are exactly solvable both analytically and computationally . the maximum matching problem as studied in refs . @xcite could be a good candidate model system .

the latest trend in studies of modern electronically and/or optically active materials is to provoke phase transformations induced by high electric fields or by short ( femtosecond ) powerful optical pulses . the systems of choice are cooperative electronic states whose broken symmetries give rise to topological defects . for typical quasi - one - dimensional architectures , those are the microscopic solitons taking from electrons the major roles as carriers of charge or spin . because of the long - range ordering , the solitons experience unusual super - long - range forces leading to a sequence of phase transitions in their ensembles : the higher - temperature transition of the confinement and the lower one of aggregation into macroscopic walls . here we present results of an extensive numerical modeling for ensembles of both neutral and charged solitons in both two- and three - dimensional systems . we suggest a specific monte carlo algorithm preserving the number of solitons , which substantially facilitates the calculations , allows to extend them to the three - dimensional case and to include the important long - range coulomb interactions . the results confirm the first confinement transition , except for a very strong coulomb repulsion , and demonstrate a pattern formation at the second transition of aggregation .

1605.04395

a new trend in controlling cooperative states of interacting electronic systems is applying either very short ( tens of femtoseconds ) powerful optical pulses or very high ( up to @xmath0 v / cm ) electric fields ( `` the electrostatic doping '' ) , see refs . [ ] . these impacts result in a very high ( up to 10% per lattice site ) concentration of excitations or charge carriers . there are convincing arguments that these states will not resemble ensembles of electrons and/or holes like for optical pumping or field - effect injection in conventional semiconductors and can include superconductivity , antiferromagnetism , ferroelectricity , charge order , charge- and spin - density waves , mott and peierls insulators . the reason is that the commonly exploited strongly correlated electronic systems show various types of symmetry breaking giving rise to degenerate ground states . the degeneracy allows for topologically nontrivial configurations exploring the possibility of traveling through different allowed ground states . their most known forms are plain domain walls , stripes , vortex lines , or dislocations , which are still macroscopic objects extending in one or two dimensions . most importantly , there are also totally localized and truly microscopic objects whose energies and quantum numbers are on the one - electron scale . these anomalous particles the solitons can determine the observable properties , which are usually ascribed to conventional electronic excitations , see refs . [ ] for early theory reviews and refs . [ ] for updates . the fact that the solitons have quantum eigenvalues ( charge or spin ) makes it possible to control and monitor their concentration . the electrostatic doping ( see the review @xcite and updates in @xcite ) should give rise to a stable @xmath1 ensemble of similarly charged kinks in a thin , sometimes atomically narrow , surface layer . the optical pumping should give rise firstly to an equal number of oppositely charged solitons , whose collisions will work to convert them secondly to an ensemble of neutral spin - carrying solitons ( which usually have lower energy than the charged ones @xcite ) . inevitably , recombination will follow , possibly via the formation of excitons as pairs of oppositely charged kinks @xcite . however , the optical emission will take long ( more than nanoseconds ) time , which can be further prolonged by the intermediate conversion to neutral spin - carrying solitons , which can recombine only via triplet channels the effect is well documented in the optics of conducting polymers @xcite . then , for a typically very fast pump - induced phase transition experiment , even the system of oppositely charged solitons , and even more of spin - carrying ones , can be treated as quasi stationary , with only slowly decreasing number of particles . this number is exactly conserved and monitored in experiments with the electrostatic doping . such ensembles of solitons are expected to have a peculiar phase diagram , with several lines of phase transitions , which are inevitably crossed in the course of the evolution or monitoring the concentration and the temperature . the study of these transitions is the main goal of this paper . in the next section of introduction , we shall summarize modern experimental and theoretical evidences in favor of existence of solitons in electronic systems . in sec.ii , we shall give a qualitative picture of phase transitions in ensembles of neutral and charged solitons . the central issue is the confinement interaction specific to solitons in contrast to conventional electrons . in sec.iii , we shall introduce the basic model , which is mapped onto the constrained ising model , and also present some analytical results . in sec.iv , we shall present the results of extensive numerical modeling , which was challenging for three - dimensional systems , particularly in presence of coulomb interactions ( ci ) . we shall observe a high - temperature transition of confinement of solitons into pairs and a particularly rich and interesting low - temperature transition of soliton aggregation into domain walls transversing the sample . sec.v is devoted to discussion and summary . appendices contain estimations for the aggregation transition for neutral and charged solitons . there are growing experimental evidences on existence of microscopic solitons and their determining role in electronic processes of quasi-1d systems : conjugated polymers ( see @xcite on theory and @xcite on experiment ) , spin - peierls chains @xcite , donor - acceptor stacks including `` electronic ferroelectrics '' @xcite , and families of the so - called `` electronic crystals '' , particularly charge density waves ( cdw ) and charge - ordered mott insulators ( see reviews @xcite ) . solitons take over band electrons in roles of primary excitations charge or spin carriers , since their activation energies are typically lower than gaps opened in the electronic spectra . the solitons feature self - trapping of electrons into mid - gap states and separation of spin and charge into spinons and holons , sometimes with their reconfinement at essentially different scales . thus , the ferroelectric charge ordering in organic conductors ( see a review @xcite ) gives access to several types of solitons observed in conductivity ( holons ) and in permittivity ( polar kinks ) , to soliton bound pairs in optics , to compound charge - spin solitons . in cdws @xcite and in surface nano - wires @xcite the individual solitons , which are the amplitude kinks , have been visually captured in stm experiments ; this is the most remarkable new achievement in proving the existence of solitons . the resolved subgap tunneling spectra @xcite recover presumably the same solitons in dynamics . the tunneling creation of soliton pairs describes nonlinear transport in cdws @xcite and polymers @xcite . the solitons can be also viewed as nucleus of the melted stripe phase in doped mott insulators or of the fflo phase in spin polarized superconductors @xcite . on this basis one can extrapolate to a picture of combined topological excitations in general strongly correlated systems : from doped antiferromagnets to strong - coupling and spin - polarized superconductors @xcite . following the above quoted experimental confirmations and theoretical results , we shall consider a system where the solitons serve as the lowest ( with respect to band electrons ) energy forms of storage of the charge or the spin . we shall restrict the study to the case of a discrete symmetry breaking , which is a very common phenomenon of a dimerization of bonds ( the family of peierls - like transitions ) or of sites ( the family of transitions with charge ordering or disproportionation ) . such a system possesses typically three types of solitons : spinless ones with charges @xmath2 and a neutral one with the spin @xmath3 . in ferroelectric systems ( see a recent review @xcite ) all solitons should carry noninteger charges . with passing of such a soliton , the order parameter changes the sign , hence the nickname `` the amplitude kink '' , or just the `` kink '' or the soliton , which we shall use in the following . cases of continuous symmetries like incommensurate cdws , spin density waves and superconductors , require special consideration . the solitons are subject to all kinds of interactions , among themselves and also with the lattice , known for conventional electrons . the important ci can be screened or not screened by external carriers and we shall consider both cases . but beyond that , there is the unusual super long - range interaction specific to the solitons as topologically nontrivial objects . the soliton is a @xmath4 domain boundary that interrupts the proper interchain arrangement . that gives rise to the confinement energy @xmath5 , with the constant confinement force @xmath6 , growing linearly with the distance between solitons @xmath7 see fig.[fig : t > t1 ] . this energy dominates at long distances even if it can be unimportant locally for a crystal of weakly interacting chains . in some special cases , the ground - state degeneracy can be lifted by an internal effect globally for the whole system . a bright example is the cis - polyacetylene @xcite , where the solitons are always confined in pairs . in cases of continuous symmetries , the ordering violated by the amplitude kink can be restored by changes in the phase of the complex order parameter @xmath8 , localized in the tails of a soliton of length @xmath9 , where @xmath10 is the temperature of the long - range ordering due to the interchain coupling . but universally , except truly @xmath4 systems like isolated atomic chains @xcite , there is a local lifting of degeneracy that comes from interchain interactions , which are responsible for establishing the long - range @xmath1 or @xmath11 ordering . namely , the interchain ordering energy @xmath12 ( per longitudinal lattice unit of the length @xmath13 ) is paid when adjacent domains at neighboring chains are not rightly correlated . the confinement interaction determines the intermediate phases and the kinetics of aggregation of nonequilibrium ( e.g. , optically induced ) domains into the long - range ordered phase . consider the influence of weak ordering interaction between chains ( @xmath1 or @xmath11 coupling ) on the state and statistical properties of the kinklike solitons @xcite . the weak interchain coupling does not affect substantially the structure of the soliton core , but below the @xmath1 or @xmath11 ordering temperature @xmath14 , its role turns out to be fundamental at large distances . since each kink separates different states of the system , the correlation between the chains is violated in its vicinity . as a result , the system loses an energy @xmath15 per site , which increases proportionally to the distance from the soliton . as we see from fig.[fig : t > t1 ] , the energy grows both with separation of two solitons on one chain and among solitons at neighboring chains , hence the tendencies to either formation of on - chain bikinks or the interchain aggregation of solitons into walls . as a result of these contradictory tendencies , as temperature lowers , the system passes through two phase transitions : the coupling of solitons into pairs at @xmath16 , and aggregation of pairs between the chains at lower @xmath17 . for a three - dimensional system , the temperature @xmath18 is a phase - transition point , below which plane domain walls appear in the system passing through the entire cross section . for a two - dimensional system , this is not the distinct phase transition ; instead a gradual increase of the transverse dimension of the paired walls takes place at @xmath19 . for a finite system , like in our modeling , there is a sample dependent temperature @xmath20 where the first wall crosses the whole sample even in @xmath21 . this qualitative picture is based upon an exact solution available for a @xmath1 system of neutral solitons with some qualitative extensions to the @xmath11 case @xcite . the case of charged solitons has been addressed in @xcite , but with a restrictive constraint : the bisoliton pairs were not supposed to move from one chain to another . here we shall present the results of unrestricted numerical modeling performed for the challenging case of a @xmath11 system , both for neutral and charged ensembles of solitons . in short , the picture of equilibrium states is the following . with lowering @xmath22 at a given concentration of kinks @xmath23 , the system passes through a sequence of two phase transitions ( at @xmath24 ) and a crossover in between ( at @xmath25 ) , which are determined by three scales of energy : @xmath12 as the local energy of the interchain ordering , the bigger @xmath26 as the nonlocal energy of soliton confinement , and the smaller @xmath27 as the characteristic energy of soliton transverse aggregation . i. @xmath28 ( see fig.[fig : t1>t > t2 ] ) . for the order parameter , this is the conventional @xmath11 ordering transition realized in a quasi-@xmath4 system . however , for the ensemble of solitons , this is a confinement transition , which takes place when the temperature decreases below the mean interchain interaction @xmath5 at the mean distance @xmath29 between the solitons . below that , at @xmath30 , individual on - chain solitons can not exist , they are confined into loose pairs which form strings of size @xmath31 with energy @xmath32 among them ( @xmath33 can be also defined by the condition @xmath34 , when confined pairs dissociate ) . with lowering @xmath22 , the pairs become progressively more confined at the thermal length @xmath35 , with rare collisions among the pairs . at @xmath36 , the pairs become tightly bound , their spacing does not fluctuate anymore . the confinement energy is reduced from the high-@xmath22 scale @xmath12 per unit cell to @xmath12 per kinks pair . actually , the pair length @xmath31 shrinks from the thermal one @xmath37 to the quantum zero point limit @xmath38 such that @xmath39 ( @xmath40 is an effective mass of the soliton ) . thus , the energy @xmath14 has two faces . for the order parameter , this is just the transition temperature @xmath41 of the second - order phase transition to the state with its nonzero mean value at @xmath42 . however , for solitons , this is the confinement transition temperature , below which they become bound into pairs the bisolitons . @xmath43 ( see fig.[fig : t < t2 ] ) . this temperature can be viewed as the onset of aggregation of solitons when first domain walls appear crossing the whole sample or forming macroscopic bubbles . as the temperature lowers beyond @xmath12 , the bisolitons start to aggregate in transverse disks , and finally at @xmath44 these disks cross the entire sample . aggregation to domain walls gains the confinement energy , which now is not lost at all the neighboring chains are always in the right arrangement . however , the entropy is lost , and this balance determines @xmath18 . the pairs still coexist with walls below @xmath18 , but with further decreasing @xmath22 they vanish providing the material for building more macroscopic walls . the @xmath18 transition can be viewed similarly to a vapor condensation in a given volume when the first appearing wetting fixes the chemical potential ( the saturation pressure ) of the gas . another analogy is the bose - einstein condensation , but in real , instead of the reciprocal , space . indeed , below @xmath12 , there is the gas of `` confined pairs '' ( bisolitons ) with energy @xmath45 , whose chemical potential @xmath46 is adjusted to maintain the given total concentration @xmath47 . when the first macroscopic domain walls appear , they serve as a reservoir of kinks fixing their chemical potential ideally at @xmath48 , hence @xmath49 . then the concentration of pairs is @xmath50 , and this number falls below the total available @xmath51 , which happens at @xmath19 . the deficit @xmath52 gives the number of solitons that have been used to build the domain walls , with @xmath53 giving the mean period of the stripe array of domain walls . notice a curious behavior : with lowering @xmath22 below @xmath18 , the mean value of the order parameter over the bulk disappears , while it is present over each cross - section . the @xmath18 transition is the one where the effective dimensionality of the system @xmath54 is reduced by @xmath55 ! we shall see this explicitly via the effective ising model with constraints . after the optical pulse has created the ensemble of solitons , their mean concentration @xmath23 starts to evolve as @xmath56 both to thermal equilibrium at a given @xmath22 and also together with the temperature @xmath57 . on this way , the system will pass through a sequence of phase transitions or at least of crossovers among different regimes which have been classified above . the trajectory is summarized in fig.[fig : phase - diag ] . and @xmath18 , the vertical dashed line shows a crossover at @xmath58 , the dashed arrowed lines give schematically the trajectory of the pumping and subsequent relaxation ( decrease of the total @xmath23 and the cooling ) . ] before the pumping , at the equilibrium ambient temperature @xmath59 , the solitons are present either as free particles , if @xmath60 , or as confined pairs if @xmath61 , where @xmath62 is the ordering transition temperature without pumping ( here @xmath63 is the soliton s core energy ) . however , in any case , the equilibrium concentration of kinks is low : @xmath64 at @xmath60 or @xmath65 at @xmath66 , because now the total big activation energy @xmath67 of the soliton is to be paid in comparison with the relatively small scale of the interchain energy @xmath12 when the number of solitons is controlled . just after the pumping , the initial concentration of kinks @xmath68 is very high and the initial temperature @xmath69 is also high ( it can even further increase at intermediate times because of the energy release from relaxation of excitations @xcite ) . suppose that @xmath70 so that we are in the disordered phase above @xmath71 the pumping has destroyed the long - range order , the solitons are not confined in pairs . with time , both @xmath56 and @xmath57 decrease , so the two sides of the last inequality move towards each other , the transition is inevitably reached at some time @xmath72 when @xmath73 . recall that this temperature is well below the thermodynamical transition temperature @xmath74 since the number of solitons is still strongly enhanced . the situation looks , at first sight , a bit less certain for the lower transition because its expected temperature @xmath75 falls with decreasing @xmath56 , unlike @xmath76 . however , @xmath77 decreases very slowly , e.g. , @xmath78 for the exponential decay of @xmath23 , which is slower than any expected decrease of @xmath57 . so the second phase transition will also happen at a certain moment @xmath79 of time , when @xmath80 , then the confined pairs of solitons will start to aggregate into macroscopic domain walls of solitons . the @xmath44 transition line will be crossed back ( via evaporation of domain walls ) in the course of returning the temperature to the ambient one and the full annihilation of excess solitons . for a generic ( no cis ) model , the configurational energy for the order parameter @xmath81 at the point @xmath82 of the chain @xmath83 is @xmath84 where @xmath85 is the interchain ordering energy per unit longitudinal length ( actually , @xmath86 is the confinement force @xmath6 , where @xmath87 is the number of nearest neighboring chains for @xmath88 ) and @xmath89 is a double - well potential with two symmetrical minima normalized to @xmath90 , which determines the two possible equivalent ground states . the soliton is a trajectory , taking the energy cost @xmath67 , which commutes between these two minima ; e.g. @xmath91 or whatever is the antisymmetric solution determined by the competition of second and third terms in ( [ h0 ] ) . it is convenient to quantize the chain length as @xmath92 in some units @xmath93 well exceeding the intrinsic width @xmath94 of the soliton ( we take @xmath93 to be equal to the quantum zero point limit size @xmath95 the minimal size of a bisoliton , see sec . [ tentative_phase_diagram ] ) and to introduce the ising spin variable as @xmath96 . then the solitons , with the linear density @xmath97 , are seen as sharp kinks , whose number per site is given by the lattice function @xmath98 such that @xcite @xmath99 where @xmath100 is the mean concentration of solitons per site , @xmath101 is their total number , @xmath102 are the dimensions of the sample in units of @xmath93 and @xmath103 , respectively . representation ( [ rho ] ) underlines the fact that a soliton is present at the site @xmath104 of the dual lattice only if @xmath105 and @xmath106 or vice versa . in the single chain limit , @xmath107 and @xmath108 play the roles of complementary order and disorder parameters ( see , e.g. , @xcite ) . the mean density can be controlled by the chemical potential @xmath109 and we arrive at the gibbs energy @xmath110 for the grand canonical ensemble of solitons @xcite : @xmath111 in a number of cases , particularly in application to doping , the solitons can possess an electric charge . if screening by external carriers is strong enough , these solitons behave as neutral ones . however , in case of intermediate or weak screening , the charges of solitons must be considered explicitly . therefore , taking into account the ci , we can add also @xcite the coulomb energy @xmath112 to the hamiltonian : @xmath113 where @xmath114 is the electron charge and @xmath115 is the dielectric constant , taken here to be isotropic . it is known that competing short - range attractive and long - range repulsive forces may result in the formation of a diverse variety of patterns @xcite . in sec . [ numerical_approach ] , we shall describe monte carlo modeling for both neutral and charged cases . it is remarkable that controlling the chemical potential , the grand canonical ensemble of solitons in a @xmath54-dimensional system can be described by the @xmath54-dimensional ising model , while the canonical ensemble is described by the stack of noninteracting @xmath116-dimensional models with an overall constraint . in this anisotropic model , only the interchain coupling constant @xmath12 has a physical origin and is frozen , while the on - chain constant @xmath117 is determined by the chemical potential . since the physical situations of interest correspond to controlling the concentration @xmath23 , then the price , and the source , of the most interesting behavior come from the self - consistency condition to invert the function @xmath118 to @xmath119 , hence traveling over a special line on the surface of ising model parameters . in this way , we can take a good advantage ( without ci ) of notion on the ising models ( full for @xmath21 and qualitative at least for @xmath120 ) . on the other hand , as we shall demonstrate in the following , a numerical procedure can be constructed which allows to directly keep the constraint on the total number of solitons . then working with the canonical ensemble we get a good advantage to deal with @xmath116-dimensional systems with only physical interchain interaction being present . consider the effective ising model with adjustable @xmath121 in @xmath54 dimensions . the concentration of solitons is always supposed to be small @xmath122 . here we find the adjusted values of @xmath123 in @xmath1 and @xmath11 for limiting cases of high and low temperatures and make estimations for @xmath76 and @xmath124 . in the high - temperature limit @xmath125 , the system is effectively one - dimensional ; the energy of one soliton is @xmath126 and the probability to find a soliton at a given point of the dual lattice is @xmath127 then @xmath128 is the effective on - chain coupling , hence , from ( [ originalh ] ) , the soliton chemical potential increases with decreasing @xmath22 . extrapolating this expression down to @xmath33 , we find an estimate for @xmath33 . for @xmath21 , we deduce from onsager s exact result @xcite @xmath129 that @xmath130 , which up to a numerical factor , agrees with the result of ref . : @xmath131 ( see also appendix a ) . for @xmath120 , we can use the approximation , where the on - chain interaction is treated exactly while the interchain one is taken into account using the mean - field theory . for the critical temperature of anisotropic @xmath11 ising model , this approach gives @xcite @xmath132 , from which we get @xmath133 . we see that both in @xmath1 and @xmath11 the ising critical temperature behaves as @xmath134 at @xmath135 , which justifies treating @xmath14 as the confinement transition . consider now the opposite limit of low temperatures @xmath136 . it looks , at first sight and wrongly , that at low temperatures , the system persists in a very simple form of one spin - ordered domain impregnated by a dilute gas of spin - reversed sites , which are our tightly bound bisolitons with the energy @xmath137 . taking into account the chemical potential and neglecting the excluded volume corrections , the concentration of bisolitons is @xmath138 since this number is fixed at @xmath139 , then lowering @xmath22 must be compensated by the decrease of @xmath117 , which is limited to be positive @xmath140 : a negative @xmath141 would switch the system to an `` antiferromagnetic '' ground state with spins alternation at each site in the chain direction , hence the infinite number of kinks . hence a new reservoir for the storage of solitons must be opened when @xmath22 falls below @xmath77 such that @xmath142 , i.e. , @xmath143which agrees with both the estimation ( [ appa_t2 ] ) for @xmath1 ( with logarithmic accuracy ) and with the exact solution ( [ t2 ] ) accessible in @xmath11 . this new reservoir can be viewed as a system of stripes ( lines in @xmath1 or planes in @xmath11 ) which cross the whole sample separating the bulk in noninteracting domains of alternating magnetization . a more thorough analysis @xcite shows that @xmath18 must be indeed a sharp phase transition at @xmath120 , while at @xmath21 it is only a crossover for growing rods of a finite extent . only for a finite @xmath1 system of width @xmath144 the rods pass through the entire cross - section of the sample at @xmath145 . estimates for @xmath123 can also be found at @xmath36 . for @xmath21 as @xmath146 and @xmath120 as @xmath147 we find that @xmath148 as ( details of derivations are given in the appendix a ) @xmath149 below @xmath44 in @xmath11 , transverse layers do not interact and @xmath117 remains @xmath150 . this picture , and its strong complication by long - range cis will be numerically verified and expanded in the next section . some analytical results for both neutral and charged systems are also given in the appendices . in this section , we consider the monte carlo ( mc ) simulation of the ensemble of solitons . we have studied the statistical properties of the canonical ensemble of solitons over broad temperature ranges in two and three dimensions for both neutral and charged solitons . in the numerical simulations , it is more efficient to keep a fixed number of solitons rather than using self - consistent @xmath151 as we did in the previous section . here then we shall work with the canonical ensemble of solitons keeping fixed their overall number , while the number of solitons at a given chain can vary ( in contrary to the more restrictive prescription employed in @xcite ) . the mc simulation was performed with the use of the standard metropolis algorithm with three types of unit movements : moving a single soliton along the chain [ fig.[solmovements]a ] , moving a solitons pair along [ fig.[solmovements]b ] , or across the chains [ fig.[solmovements]c ] . the type-(b ) movement is a superposition of two type-(a ) movements , but it is useful to consider the type-(b ) movement explicitly since it improves the low - temperature acceptance rate of the algorithm . for the case of charged solitons , we need to take into account the long - range ci , which is always a formidable task , and even more here for the system prone to pattern formation , owing to the appearance of locally noncompensated charges at a growing scale . since we are interested in the wall formation process , and for an infinite charged wall the electric potential grows linearly with the distance , then correctly imposing periodic boundary conditions for the ci becomes very important . we use the periodic version of the coulomb hamiltonian ( [ hcoulomb ] ) ( for simplicity , we put @xmath152 in the simulation ) , where the summation goes over not only all pairwise interactions within a computational cell , but also between solitons and their images and between images and the neutralizing negative background . in practice , this is done by the technique of lekner @xcite for both @xmath1 [ ] and @xmath11 [ ] cases . this technique allows to efficiently calculate the force acting on a given particle from another particle and all its images and , and by integrating the force , to obtain the effective pairwise potential . finally , we tabulate this potential for fast computations . since the ci is the long - range one , the calculation of the energy change @xmath153 at each mc - step can be time consuming . to deal with this we use two different approaches . when the mc acceptance rate is high , we employ the first ( standard ) approach : at each mc trial step we recalculate the ci energy of the shifted soliton with respect to other solitons . however , at low temperatures , when the mc acceptance rate becomes low , it turns out that this approach is ineffective , since many calculations are wasted to compute @xmath153 for rejected steps . therefore , at low temperatures ( and low mc acceptance rates ) we use another algorithm @xcite : instead of recalculating the interaction energy of the shifted soliton with every soliton in the system at each trial mc step ( computational cost of which is @xmath154 ) ) , we introduce an electric potential @xmath155 and use it to calculate the coulomb energy change : @xmath156 , the computational cost of which is only @xmath157 . however , now we have to update the potential after every accepted step at each site of the system , which costs @xmath158 . this means that this approach works better when the acceptance rate is low . combining these two approaches allows us to effectively perform simulations for the system with long - range ci at both high and low temperatures and even in @xmath11 space . in this section we shall describe our main results : the evolution of the system with lowering temperature in different regimes . we shall use several presentations : images for distributions of solitons and spins . these pictures will be further characterized by plotting the numbers of nearest spins or solitons . here we consider a system with a size @xmath159 sites and a concentration of neutral solitons @xmath160 . to demonstrate explicitly the formation and multiplication of domain walls , we shall show the patterns of soliton density ( fig.[figwalls]a , c , e ) , which is of our direct interest , and also the patterns of the reversed spins of the effective ising model ( fig.[figwalls]b , d , f ) which give a useful complementary insight . here and below , the sites with the major orientation of spins will be left as blank space , while the sites with reversed spins will be marked in black . the edges of black areas in the longitudinal ( chains ) direction indicate the positions of kinks . at higher temperatures ( @xmath161 , fig.[figwalls]a , b ) we observe an ordered state impregnated by the gas of bisolitons . when the first pair of walls condenses ( at @xmath162 , fig.[figwalls]c , d ) , the concentration of noncondensed solitons drops , then with decreasing temperature it further decreases gradually , to cure the defects in already existing walls , then it drops sharply again with the formation of the next pair of walls ( @xmath163 , fig.[figwalls]e , f ) . meanwhile , the initial pair of walls diverges loosing the mutual correlation and opening the whole domain of reversed spins in between . comparing the distributions of ising spins and solitons at different temperatures , we see that the number of reversed ising spins is not conserved ( which means that the ising magnetization can drastically drop below @xmath44 and take any value between @xmath150 and @xmath164 ) , whereas the number of solitons is preserved . interestingly , the second pair of walls nucleates in the vicinity of the first one : the incipient second pair of walls is more stable there . it happens because the movements of solitons intending to build the second wall are prohibited towards the first one , which twice reduces their escape probability , hence promoting the aggregation . for systems with smaller sizes , we can perform long enough simulations , when this transient effect vanishes . however , we believe that since our elementary soliton movements are chosen in a natural way , similar to a real time soliton dynamics , then this effect of correlated emergence of walls can take place in real systems . the obtained patterns and their evolution show up also in integrated characteristics , which are more accessible for measuring . thus , we have calculated the temperature dependencies of the ising spin magnetization and of the number of transverse neighbors . here we consider @xmath11 systems with sizes @xmath165 and @xmath166 sites for the concentration of neutral solitons @xmath167 . first , consider the ising magnetization @xmath168 dependence , which is shown in fig.[plotmvst ] . excluding the low - temperature region , this dependence is very similar to the standard plot for the ising model with the transition temperature @xmath169 . the transition is smeared due to the finite - size effects . of ising spins vs temperature ( in units of @xmath12 ) for systems with sizes @xmath170 ( circles ) and @xmath165 ( triangles ) , with @xmath160 . ] apparently , even this not very small fixed concentration of solitons does not affect much the high-@xmath22 properties , unlike the drastic effect we see at low @xmath22 . the @xmath44 phase transition happens with the dimensionality reduction of the system . from the thermodynamical point of view , when the interaction between layers vanishes ( @xmath171 ) , the magnetization must drop to @xmath150 . below we explain why it may not drop to @xmath150 in a numerical simulation . for the smaller system @xmath165 with lowering temperature , we see a sharp drop in the @xmath168 dependence at @xmath172 . this reflects the walls formation transition at @xmath44 : as it was explained in sec . [ tentative_phase_diagram ] and was demonstrated explicitly in sec . [ condensation_of_solitons_into_walls_for_neutral_solitons ] , the bisolitons aggregate into walls , then these bisoliton walls divide into single - soliton ones , and the ising spin magnetization drops . depending on a numerical experiment , @xmath173 picks a random value between @xmath150 and @xmath164 when the system freezes at @xmath174 , which is a finite - size effect associated with the finite length of the sample @xmath175 . for a sample of a macroscopic length , however , it must be @xmath176 since spin - up and spin - down domains are equally probable . for the bigger system @xmath170 , we have observed that the walls appear in pairs , with only one reversed spin per chain , and a long time is necessary for the walls to diverge , opening a growing domain of reversed spins in between . typical times of the simulation are not big enough to observe the wall splitting . this happens because in order for the bisoliton wall to divide , the system must pass through an energetically unfavorable state : when the incipient layer of new domain grows to a disk of radius @xmath177 , the energy increases by @xmath178 . if the transverse size @xmath144 is big enough , this high energy state can not be reached during the time of the simulation , and the system can not skip from one energy minimum to another with the shifted wall . therefore for such big systems only bisoliton walls are observed and the ising magnetization does not decrease at low temperatures . ) in a system @xmath165 with @xmath160 ; for the reversed spins it is shown with squares , for solitons it is shown with circles . ] second , we consider another integrated characteristics the average number of transverse neighbors , which describes the transverse correlations in the system . it is particularly interesting to do so near @xmath179 . since @xmath180 , then in the vicinity of @xmath44 only a relatively small number of reversed spins is left , being dispersed within the major domain of aligned spins . thus , in order to characterize the degree of aggregation , we calculate for each reversed ising spin ( or each soliton ) the number of its neighboring reversed spins ( or neighboring solitons ) in the transverse direction . since for the domain - wall phase the number of bulk spins is much greater than the number of spins at the interfaces , then the number of reversed spins neighbors must jump to approximately @xmath181 at @xmath179 . however for solitons , when @xmath22 decreases below @xmath44 , a substantial number of them also condenses into walls , but this number is comparable with the number of the noncondensed solitons , therefore the jump must be not that big as for the ising spins . this reasoning is confirmed by fig.[plotneighbvst8x8 ] , which shows the temperature dependence of the average number of neighbors for ising spins and for solitons . with lowering temperature , at @xmath182 , we see a sudden jump of the number of the reversed spins neighbors from @xmath183 to @xmath184 , which indicates the formation of domain walls as it is explicitly confirmed in fig.[figwalls ] . the observed transition temperature is in good agreement with the theoretical prediction ( [ t2 - 3d ] ) : @xmath185 . the corresponding plot for solitons also shows a jump , however , not that big : from @xmath183 to @xmath186 average number of neighbors . at @xmath187 , the two plots essentially coincide since almost all bisoliton pairs have the minimal size of @xmath55 site . now we consider the case of electrically charged solitons for a system of @xmath188 sites with @xmath160 . the weak ci ( according to estimates , given in the appendix b ) with the coulomb parameter @xmath189 does not affect the system qualitatively . for example , for @xmath190 , it only lowers the temperature of condensation of solitons into walls by @xmath191 down to @xmath192 therefore , in this section , we focus on a more interesting case of intermediate values of the coulomb parameter @xmath193 . in this case , the coulomb interaction is weak enough locally , so it does not prevent the binding of solitons into bisolitons , and even does not destroy the initial correlation of bisolitons at neighboring chains . however , it affects the large - scale structures such as domain walls , since for @xmath194 the wall formation becomes energetically unfavorable . in contrast to the cases of neutral solitons ( fig.[plotneighbvst8x8 ] ) and weak ci , we do not observe sharp wall formation transition for the intermediate values of the ci the temperature dependence of the average number of neighbors does not show any jumps . when @xmath195 becomes larger than @xmath196 , we still observe bisoliton walls at nonzero temperature , which now have defects ( holes ) . as @xmath22 goes to @xmath150 , these defects are grouping and can cut a wall along one of the transverse directions ( fig.[figwallscoulombwithdefects]a ) . as @xmath195 further increases , we do not observe plane walls , but only filamentary stripes , which are infinite along one transverse direction and finite along the other ( fig.[figwallscoulombwithdefects]b , recall that periodical boundary conditions are imposed ) . however , it is clear that an infinite system can not possess such infinite stripes , because they are inefficient in terms of the interchain energy @xmath12 . our interpretation for formation of lines rather than planes is that here the system demonstrates an extreme sensitivity of cis to the transverse finite - size geometry of the sample . a charged wall would create a constant electric field @xmath197 in the chains @xmath82 direction , whose repulsive force @xmath198 would oppose directly the attractive confinement force @xmath12 overpassing it at the intermediate ci , hence no stable planes could exist . however , forming the stripes ( in one transverse @xmath199 direction ) at the expense of the part of the confinement energy ( lost in the other transverse direction @xmath200 ) , the system generates a decreasing electric field @xmath201 which falls below the confinement force at sufficiently large distances , hence preserving this partial aggregation . therefore , in finite systems , the disintegration of plane walls happens via formation of linear structures . for infinite systems , we expect that these structures will further disintegrate into finite disks ( similar to described in the next paragraph ones ) . if we choose higher value of the coulomb parameter , the formation of stripes even for a finite system becomes also energetically unfavorable and we observe only disks of bisolitons ( fig.[figwallscoulombwithdefects]c ) . observed disklike formations are consistent with the analytical results presented in appendix b , where it is shown that the maximum radius of the disks is @xmath202 . here we consider a system of @xmath203 sites with @xmath204 ( the parameters were chosen in order to compensate incommensurateness effects , as explained below ) . strong cis @xmath205 affect the local bisoliton pairing . the transverse disks shrink to the minimum size of @xmath55 bisoliton . when @xmath195 further increases , these bisolitons form a wigner `` liquid '' ( fig.[fighighv1]a ) ( with a short - range order of individual solitons contrarily to a wigner crystal with a long - range order ) . it is known that the ground - state distribution of charges must form a triangular lattice in @xmath1 [ ] and a body - centered cubic lattice in @xmath11 [ ] . however , for systems with finite discretization , the commensurateness effects become very important : even small incommensurateness destroys the long - range order , while the short - range order persists @xcite . for even higher @xmath195 , the coulomb force starts to compete locally with the confinement force and the size of a bisoliton starts to grow ( fig.[fighighv1]b ) . neglecting the interaction between bisolitons we can estimate their size . a bisoliton elongates until the coulomb force is balanced by the confinement force : @xmath206 , therefore @xmath207 . this estimate holds as long as the average bisolitons size is much less than the distance between them : @xmath208 . when @xmath209 ( which happens at @xmath210 ) , this size becomes comparable to the distance between solitons , interactions between them become important . increasing the ci to the highest values , we observe that the ising order is destroyed in favor of a wigner `` liquid '' of individual solitons , rather than bisolitons ( fig.[fighighv2]a , b ) . in this section , we consider a @xmath1 system with the size of @xmath211 sites and concentration of neural solitons @xmath212 . as discussed in the sec . [ tentative_phase_diagram ] , well below the ising transition temperature @xmath33 ( for the considered system @xmath213 ) , there exists a characteristic temperature @xmath44 at which perpendicular rods start to form with their characteristic length gradually increasing with lowering @xmath22 . for our finite samples , there is also a width @xmath144 dependent temperature @xmath214 at which these rods become long enough to pass across the entire sample . figure [ ising2d ] shows the ising spin representation of the system s evolution with lowering @xmath22 . for @xmath215 , the ising disordered phase is observed ( fig.[ising2d]a ) . at @xmath216 , we observe the ising ordered phase with bound pairs of bikinks ( fig.[ising2d]b ) . these pairs shrink when @xmath22 lowers ( fig.[ising2d]c ) . then the rods of reversed spins change the predominant orientation from the longitudinal one at high temperatures to the transverse one at low @xmath22 ( fig.[ising2d]d ) . the first case indicates the regime of loosely bound on - chain pairs of kinks , while the second case indicates the transverse aggregation of tightly bound kinks . finally , at the lowest considered @xmath22 ( fig.[ising2d]e ) , the aggregated rods cross the entire sample and domains are created . sites and @xmath212 . the inset shows temperature range @xmath217 as in sec . [ numerical_results_for_3d_case ] , @xmath20 lies deeply below the ising transition temperature , only a relatively small number of reversed spins is left , being dispersed within the major domain of aligned spins ; therefore we characterize the degree of aggregation , by calculating for each soliton ( or each ising spin ) the number of its interchain neighbors . for the domain wall phase the number of reversed spins neighbors must be approximately 2 . figure [ plotneighbvst200x25nu003 ] shows the temperature dependence of the average number of neighbors of the reversed ising spins . it shows that upon cooling from the ising transition temperature ( @xmath218 ) , we arrive firstly at some characteristic temperature ( @xmath219 ) when the average number of neighbors starts to grow . on further cooling , another characteristic temperature ( @xmath220 ) is reached when the average number of neighbors suddenly changes from @xmath221 to @xmath186 , which indicates the wall formation process . figure [ plotneighbsolvst200x25nu003 ] shows the temperature dependence of the average number of soliton interchain neighbors . we see that their number increases gradually with lowering temperature and shows no peculiarities . this means that the transition at @xmath20 is due to finite - size effects and in an infinite sample we shall observe the condensation of solitons into infinite lines only at @xmath174 . sites and @xmath212 . ] here we consider a similar system , but now with charged solitons . for @xmath190 ( fig.[plotneighbvst200x25nu003_v ] ) , the temperature @xmath20 is lowered with respect to the case @xmath222 ( fig.[plotneighbvst200x25nu003 ] ) now the wall formation transition is observed at the lower @xmath223 . increasing the coulomb parameter up to @xmath224 ( fig.[plotneighbvst200x25nu003_v ] ) , we see that the wall formation transition temperature is further lowered down to @xmath225 . with increasing @xmath195 , @xmath20 eventually decreases to @xmath150 . for some critical value of the coulomb parameter , we get @xmath226 , in which case rods grow up only to a maximum size @xmath227 ( see appendix b for the details of analytical estimations ) . however , precise observation of this critical value of @xmath195 is difficult , since the acceptance rate of the algorithm exponentially vanishes to @xmath150 as @xmath228 . for strong ci @xmath229 , the behavior of the @xmath1 system is qualitatively the same as for @xmath11 . the transverse rods shrink to the minimum size of @xmath55 bisoliton , then the cis start to compete with the confinement force , so bisolitons start to elongate and , at some @xmath195 , the ising order is destroyed . for the highest values of cis a wigner `` liquid '' of individual solitons is observed , which case was studied in @xcite . sites and @xmath212 for @xmath230 ( circles ) , @xmath231 ( triangles ) , and @xmath232 ( squares ) . ] we have presented numerical and qualitative analysis of phase transitions in ensembles of solitons as they can be created and studied in experiments on optical pumping and field effect doping in systems with cooperative electronic states . for @xmath11 systems of neutral solitons , as temperature lowers , we observe two phase transitions . the first transition at @xmath33 reflects the spin ordering of the equivalent ising model . in terms of the original solitons , @xmath33 is the temperature , below which individual solitons become confined into bisoliton pairs . with further decreasing temperature , the size of a pair decreases , reaching the minimal value of @xmath55 at @xmath233 , when a gas of bisolitons forms . with further cooling of the system , the bisolitons start to aggregate into transverse disklike formations . finally , at some critical temperature @xmath44 , the second phase transition occurs : these disks cross the entire sample , domain walls are formed , and the ising magnetization drops to @xmath150 . the dimensionality of the system effectively reduces to @xmath21 . for @xmath11 systems of charged solitons , the locally small ci ( when @xmath234 ) can nevertheless affect the @xmath44 transition , where macroscopic patterns are created this happens because a large scale structure , such as a domain wall , gives rise to a high long - range electric field , which erases the gain of the confinement energy reached by the wall formation . for a macroscopic system without an external screening , even an arbitrary small coulomb parameter @xmath235 destroys the walls , only disklike formations with the maximum size @xmath236 are observed . however , if the screening is present and the screening length @xmath237 , then these domain walls still cross the entire sample ( in our numerical study the sample width @xmath144 plays a role of the screening length @xmath238 ) . for high values of the ci , the disklike formations disintegrate into separate bisolitons . when the ci is further increased , bisolitons arrange in a wigner liquid state . further , bisolitons start to elongate and when their size becomes comparable with the interpair distance , the ising order breaks and a wigner liquid of individual solitons is observed . neutral @xmath1 systems behave qualitatively similar to the @xmath11 case at high and intermediate temperatures , the ising - like transition at @xmath33 still exists . however , an important distinction is that @xmath44 is rather a crossover temperature in @xmath1 : the growing rods do not cross the entire sample for a macroscopic system . however , for a finite one , there still exists some temperature @xmath20 of domain wall formation . for charged @xmath1 system , @xmath20 lowers with the increase of the ci and , when it reaches @xmath150 , only rods of finite transverse length @xmath239 are observed even at @xmath174 . the presented results of the mc simulation for neutral solitons agree with the earlier predictions @xcite ( for both @xmath1 and @xmath11 cases ) . however , the results for charged solitons do not completely agree with the previous work @xcite , performed only for the @xmath1 case . there it was observed that with increasing @xmath195 at @xmath174 , domain walls are not destroyed by the ci , but they are rather roughened . this difference in the results occurs presumably because in @xcite the ensemble of solitons was treated with preserving number of bisolitons at each chain , rather than only globally . this limitation is overcome in our treatment which also has employed a more efficient algorithm allowing for modeling of @xmath11 systems , even for charged particles . this approach matches the condition of relaxation of a soliton system after a fast optical pumping or an impact of a strong electric field , considered in the present work : only bisolitons can jump between the chains . actually , there are electron pairs that jump while the order parameter is adjusted to their presence or absence . for the experiments with optical pumping to the gas of solitons , we predict that along the equilibration trajectory the system will experience two phase transitions : confinement of solitons at high @xmath22 and their aggregation to the stripe phase at lower @xmath22 . the presented theoretical picture should find its experimental realization . there can be traditional , already verified methods of solitons identifications ( recall the discussion and references in sec.[new_accesses_to microscopic_solitons_in_q1d_electronic_systems ] ) . not all of them , particularly the latest and most spectacular direct visualizations by the stm , can be applied in conditions of pump- or field - induced experiments , but they will serve as a preliminary assurance for the correctly chosen system . among traditional techniques , the pump - and - probe optics can trace the time evolution of solitons via their associated spectral features ; that have been already so efficiently exploited in conducting polymers @xcite . in the modern , and closer to our discussion , context of the fast optical pumping with optical probes , the solitons appeared already in studies of neutral - ionic transitions @xcite . the state of an ensemble of solitons , particularly the aggregation into regular stripes , may be traced by methods of the `` time - sliced diffraction '' . recently , the time resolution of the electron- and x - ray diffraction ( see the latest publications @xcite and references therein ) has been pushed to the subpicosecond scale , which makes it very promising to study the induced phase transitions accompanied by the structural aggregation . we mention finally that the studied here phase transitions with confinement and with segregation in the ensemble of topologically nontrivial excitations may serve as elementary illustrations to unresolved yet issues in high - energy physics ( confinement of quarks ) and cosmology ( phase transitions in the early universe ) . consider the high temperature regime @xmath240 . the exact expression for the solitons concentration at the critical temperature is @xcite @xmath241 then for the case @xmath122 , we get @xmath242 now consider the low temperature regime @xmath36 . in this limit all bisoliton pairs are shrunk to the minimal size of one reversed spin , then the magnetization is @xmath243 . using the exact expression for magnetization in the @xmath1 ising model @xcite : @xmath244^{1/8 } , \label{m2d}\ ] ] we find that @xmath245 crossover from the bikinks gas state to the state of growing transverse rods occurs when @xmath123 becomes of order of @xmath12 . then from ( [ appa_jpar2d ] ) for @xmath122 , we get with logarithmic accuracy @xmath246 to find an estimation for @xmath123 at low temperatures @xmath36 , we employ an approximation where interactions in the transverse planes are considered exactly while the weak interactions between planes are taken into account using the mean - field theory . in this approximation , the in - chain interaction energy terms per spin @xmath247 becomes @xmath248 where @xmath144 is an effective magnetic field . therefore the magnetization for @xmath11 case is simply given as the magnetization of the @xmath1 ising model in the weak effective magnetic field : @xmath249 where @xmath250 is the susceptibility of the @xmath1 ising model . using ( [ m2d ] ) and ( [ m3d ] ) , we get @xmath251 now we link the magnetization @xmath173 to the soliton concentration . since all layers are magnetized in one direction , we introduce deviations from mean magnetization @xmath252 , so that @xmath253 . if @xmath254 then @xmath148 and @xmath255 , so we can expand the latter in powers of @xmath117 : @xmath256 using ( [ rho ] ) , ( [ m3d1 ] ) , ( [ dsds ] ) we get : @xmath257 from the condition @xmath258 we re - derive the result of @xcite : @xmath259^{-1/4}\right ) } \approx \frac{8j_{\perp}}{\ln 2/\nu}. \label{t2}\ ] ] to estimate @xmath260 we assume that there is no special reasons for the denominator of ( [ jparexact ] ) being small . so , up to a factor of order of 1 , we can neglect @xmath83 in ( [ jparexact ] ) . for @xmath261 we use the results of @xcite , where it was shown that @xmath262 can be decomposed to a well convergent series @xmath263 here , @xmath264 and @xmath265 are the complete elliptic integrals of the first and the second kind , @xmath266 . for @xmath36 , @xmath267 , therefore @xmath268 , and we estimate @xmath262 as : @xmath269 expanding ( [ jparexact ] ) in the vicinity of @xmath179 and using ( [ chi - appr ] ) , we get the desired estimation : @xmath270 we see that @xmath12-dependence is contained only in @xmath44 and the slope of the line depends only on @xmath23 . here we present estimations and qualitative arguments on effects of cis . consider first the @xmath1 case . according to our modeling and following the exact results for the neutral system @xcite , we suggest that the basic units are still the straight lines ( rods ) of unseparated bikinks ; their lengths @xmath7 will be taken as dimensionless ; the physical scale is @xmath103 . the ci energy of a line of bikinks is@xmath271 where @xmath272 , @xmath273 is the screening length by remnant or external carries ; it appears for @xmath274 . our assumption of intermediate cis is that locally they are weak , i.e. , @xmath275 , but become effective for aggregates when @xmath276 is not small , i.e. , @xmath277 ( @xmath278 is the characteristic rod length ) . because of the equilibrium between rods with respect to exchange of building units the bikinks , their partial chemical potentials @xmath279 are related as @xmath280 ; here we include to the definition of @xmath281 not only the soliton energy @xmath282 but also the ci energy of the elementary pair @xmath283 , so that @xmath284 . the distribution of segments is@xmath285 the parameter @xmath286 is to be determined from the self - consistency condition @xmath287 approaching the regime of stripe formation , the sum is determined by large @xmath7 , then the summation can be approximated by integration over @xmath7 . consider , first , the case without screening , when the characteristic rod length @xmath288 . now,@xmath289 the exponent @xmath290 in ( [ int - nu ] ) is rapidly decreasing at large @xmath7 and the integral is convergent , because of the @xmath291 term , at any even positive @xmath292 . for negative @xmath292 it decreases rapidly at all @xmath7 and the sum is saturated already by @xmath293 with @xmath294 . with decreasing @xmath22 , @xmath292 increases , it reaches zero with no qualitative effect unlike the case of neutral systems , and finally becomes positive when @xmath295 acquires the rising part at @xmath296 where the maximum position given by @xmath297 is @xmath298 , which makes precise the definition of @xmath299 . the saddle - point approximation for the integral over @xmath7 gives@xmath300 neglecting pre - exponential numerical coefficients an inversion of this relation to @xmath301 and hence to @xmath302 in the leading dependence for @xmath303 yields @xmath304 it gives the saturated value of the rods length @xmath305 which is finite but large by our definition of the intermediate ci . the chemical potential saturates at the positive value . this behavior is different from the one in the neutral system in both @xmath21 and @xmath120 dimensions . if we now try to find @xmath306 , we get an equation @xmath307 which possesses only solutions @xmath308 . this means that if we use @xmath299 as a characteristic length , then there is no intermediate ci regime in @xmath1 in the unscreened case . if the screening is more efficient , such that @xmath309 , then @xmath310 saturates to @xmath311 and the cis just shift the chemical potential as @xmath312 . then we get from ( [ originalh ] ) , ( [ appa_jpar2d ] ) , and ( [ nu(mu ) ] ) @xmath313 which coincides with the appropriate limit of the exact result from @xcite under the shift @xmath314 . inversion of this relation gives the chemical potential @xmath315 of individual bisolitons as it is dictated by the reservoir of growing rods . for @xmath303 it saturates at a value @xmath316 but @xmath278 keeps growing exponentially in @xmath317 . in this case the regime of intermediate cis is determined by @xmath238 : @xmath318 . in the absence of external carriers , the screening length @xmath238 is defined self - consistently from the contribution of bisolitons and their rods . we shall do it assuming the @xmath11 media composed by noninteracting , except sharing the coulomb potential , layers that is not the case of our modeling which took a @xmath1 layer embedded into the @xmath11 space . by definition , the dimensional screening length @xmath319 is given by the derivative of the @xmath11 charge density @xmath320 over the chemical potential @xmath321 . the last relation follows from differentiation of ( [ screened ] ) . we get @xmath322 we see that the condition of the nonscreened regime @xmath323 can be satisfied only for very small @xmath23 and still at not very low @xmath22 . the most common case will be described by relations ( [ screened ] ) , which imitates the neutral system but with the up - shifted chemical potentials and the corresponding strong increase of concentration of noncondensed bikinks . the ci energy of a disk is@xmath325 where @xmath326 with @xmath327 for @xmath328 and @xmath329 for @xmath330 . the screening length @xmath331 appears for @xmath332 . assumption of intermediate cis ( which are locally weak , but prevent the transverse walls formation ) in the @xmath11 case becomes @xmath333 ( the cross - section of the sample is assumed to be @xmath334 square ) . the equilibrium relation for partial chemical potentials is held as before : @xmath335 . then the distribution of disks is @xmath336 where the first term is the confinement energy lost over the perimeter of the circle . however new items appear in the @xmath11 case , those are domain walls with the chemical potential @xmath337 . @xmath338 is to be determined from the self - consistency condition @xmath339 even in the case @xmath340 , the wall term does not contribute to the sum as long as @xmath341 . however , as @xmath22 decreases , @xmath342 also has to decrease in order to accommodate all the solitons in the system . when @xmath343 reaches the value @xmath344 , the first wall condenses , and then @xmath343 stays at @xmath150 analogously to bose - einstein condensation , but in the real space instead of the reciprocal one . first , consider the unscreened case @xmath345 . then the summation in ( [ nu(mu)3d ] ) is convergent because of the two terms in the exponent : @xmath346 as it was already without ci , and now also @xmath347 from the ci . this sum is dominated by the minimal @xmath348 if either @xmath349 or @xmath350 . recall that our definition of a moderate ci means that it is not efficient at the level of minimal distances , i.e. @xmath351 , hence only the first inequality is sufficient . then @xmath352 and the estimation for the transverse aggregation temperature @xmath353 , obtained for the neutral case , still holds here . however , when the excess number of solitons condenses to first walls , for them , @xmath324 reaches the sample width @xmath144 . then for weak cis @xmath354 when it is still unimportant hence qualitatively never important at all . in this case @xmath44 only slightly decreases by @xmath355 . for the moderate ci : @xmath356 walls do not form . we can estimate the maximum size of disks analogously to the @xmath1 case . when @xmath22 decreases , the chemical potential @xmath341 grows , then it reaches @xmath357 and continues to increase . this means that the sum ( [ nu(mu)3d ] ) is now determined by large values of @xmath324 and summation can be approximated by integration over @xmath324 , @xmath358 the exponent @xmath359 in ( [ int - nu-3d ] ) decreases even faster at large @xmath324 in comparison to the @xmath1 case , and the integral converges . @xmath360 has local extrema at points @xmath361 ( fig.[appendixb - plot ] ) . at some critical value @xmath362 the local maximum becomes positive : @xmath363 , which means that large disks of radius @xmath364 appear in the system . in order to find the corresponding critical @xmath365 we use the saddle - point approximation for the integral over @xmath324 ( only region @xmath366 contributes in this case)@xmath367 at @xmath146 , @xmath368 decreases below the fixed concentration @xmath23 , which happens at @xmath369 , then the large disks appear . now turn to the screened case . for @xmath370 , the screening is not very important : it does not affect even the largest disks structure , but only affects the interactions among them . however , for @xmath237 , the screening becomes essential , since disks can grow up to @xmath371 and cis does not prevent their further growth , so they grow into domain walls . the wall term appears in the sum ( [ nu(mu)3d ] ) when the chemical potential reaches the value @xmath372 . so , as in the @xmath1 case , at @xmath237 there is just the shift of the chemical potential @xmath373 . in conclusion , the very weak ci @xmath375 only shifts down the transition temperature @xmath44 not preventing the formation of the walls . for the intermediate ci @xmath376 , different regimes are observed depending on the screening length @xmath238 . for @xmath377 , the ci is effectively unscreened , so it prevents the wall formation and only disks of maximum size of @xmath378 are observed . for @xmath237 , the screened ci does not prevent them growing to transverse domain walls and only shifts the chemical potential . c. h. ahn , a. bhattacharya , m. di ventra , j.n . eckstein , c.d . frisbie , m.e . gershenson , a.m. goldman , i.h . inoue , j. mannhart , a.j . millis , a.f . morpurgo , d. natelson , and j .- triscone , _ rev . phys . _ * 78 * , 1185 ( 2006 ) . s. brazovskii and n. kirova , `` electron selflocalization and superstructures in quasi one - dimensional dielectrics '' in _ sov . sci . reviews _ , * v. a5 * edited by i.m . khalatnikov , harwood ac.publ . , ny , p. 99 , ( 1984 ) . s. brazovskii , microscopic solitons in correlated electronic systems : theory versus experiment . advances in theoretical physics : landau memorial conference , aip conference proceedings * 1134 * , 74 ( 2009 ) . t. huber , s.o . mariager , a. ferrer , h. schfer , j. a. johnson , s. grbel , a. lbcke , l. huber , t. kubacka , c. dornes , c. laulhe , s. ravy , g. ingold , p. beaud , j. demsar , and s.l . _ * 113 * , 026401 ( 2014 ) .

the scaling laws in an infrared conformal ( ir ) theory are dictated by the critical exponents of relevant operators . we have investigated these scaling laws at leading order in two previous papers . in this work we investigate further consequences of the scaling laws , trying to identify potential signatures that could be studied by lattice simulations . from the first derivative of the form factor we derive the behaviour of the mean charge radius of the hadronic states in the theory . we obtain which is consistent with . the mean charge radius can be used as an alternative observable to assess the size of the physical states , and hence finite size effects , in numerical simulations . furthermore , we discuss the behaviour of specific field correlators in coordinate space for the case of conformal , scale - invariant , and confining theories making use of selection rules in scaling dimensions and spin . we compute the scaling corrections to correlations functions by linearizing the renormalization group equations . we find that these correction are potentially large close to the edge of the conformal window . as an application we compute the scaling correction to the formula directly through its associated correlator as well as through the trace anomaly . the two computations are shown to be equivalent through a generalisation of the feynman - hellmann theorem for the fermion mass , and the gauge coupling . [ cols= " < " , ] 1.5 cm * conformal scaling and the size of-hadrons * 1.3 cm luigi del debbio & roman zwicky 0.5 cm _ school of physics and astronomy , university of edinburgh , edinburgh eh9 3jz , scotland _ + 0.6 cm

1306.4038

gauge theories with an infrared fixed point ( irfp ) are studied currently for building models of strongly interacting electroweak symmetry breaking @xcite . at large distances the couplings flow towards their fixed point values , and the theory becomes scale - invariant . theories with an irfp are said to lie within the _ conformal window _ , see e.g. refs . @xcite for analytical results in the perturbative regime . in the absence of supersymmetry , it is difficult to identify a fixed point in the nonperturbative regime of the theory . lattice simulations provide a first principle tool to investigate the low - energy dynamics of asymptotically free gauge theories . breaking scale invariance explicitly , e.g. by introducing a fermion mass term in the action , and studying the scaling of field correlators as the breaking parameter tends to zero , has become a common way to characterise irfps in lattice studies e.g. @xcite . a theoretical understanding of the scaling laws is a necessary tool for these analyses , and a number of useful ( hyperscaling ) relations have already been investigated in our previous work @xcite . working out these scaling relations is an interesting theoretical problem , independently of its application to the analysis of lattice data . for a recent discussion of lattice results , we refer the reader to the comprehensive review that appeared in ref . @xcite . extending our previous work on mass - deformed conformal gauge theories ( mcgt ) @xcite , we discuss here the application of the scaling laws to a number of interesting physical cases , namely the scaling of the hadron size , the scaling corrections , and the determination of selection rules for field correlators . the fact that hadrons emerge in a mcgt is a non - trivial empirical fact . at least at weak coupling this can be understood as a consequence of the fermions decoupling below the mass @xmath3 , so that the low - energy dynamics should be described by a pure yang - mills effective theory , which is believed to be of confining nature @xcite . in practical lattice simulations confinement is identified through a non - vanishing expectation value for the polyakov loop , and it is characteried by the spectrum of the bound states that determine the correlators of gauge invariant interpolating fields as in qcd . in such a theory all hadronic parameters are controlled to leading order by the coupling @xmath3 , which breaks explicitly scale invariance , and whose scaling exponent characterises the long - distance dynamics . this is clearly at odds with the behaviour observed in qcd , where chiral symmetry breaking requires the goldstone bosons to be massless in the chiral limit , while the rest of the spectrum has a finite mass , which is dictated by some typical hadronic scale . we shall refer to the bound states of an mcgt as _ @xmath3-hadrons _ in what follows . for an mcgt the properties of these @xmath3-hadrons are very different from the ones commonly encountered in qcd - like theories . being able to characterise the size of @xmath3-hadrons , and to compute the scaling of the size with the fermion mass is crucial in order to understand finite - size effects ( fse ) in the results of numerical simulations . when the volume of the lattice is not large enough to accommodate the @xmath3-hadrons , fse distort the spectrum , and may well obscure the scaling behaviour that one is trying to identify . this is an important source of systematic errors in lattice studies , and has been a major concern in the interpretation of the most recent ( and precise ) studies of the spectrum of mcgt , see e.g. refs . @xcite for a recent discussions . there has been a renewed interest recently in the existence of theories that are scale invariant without being symmetric under the full conformal group . for recent work on scale - invariant ( sft ) versus conformal field theories ( cft ) see e.g. refs . @xcite , and references therein for earlier work on the subject . it emerges from our analysis that the scaling laws for field correlators in a neighbourhood of a fixed point provide a criterion to distinguish sfts from cfts . as a consequence , we discuss the possibility of identifying the existence of a fixed point describing a cft by looking at the scaling behaviour of the correlators when the theory is deformed by a mass term . this paper is organised as follows . in sect . [ sec : confscal ] we rederive briefly the scaling laws , emphasising the features that will be useful in the rest of our study . in sect . [ sec : size ] we apply the scaling relations to form factors of conserved currents , and deduce a scaling law for the radius of the charge distribution inside the ( pseudo)scalar meson . in sect . [ sec : selection ] we use the scaling laws to formulate a criterion that allows us to distinguish a scale - invariant theory from a conformal - invariant one as well confining theories . finally in sect . [ sec : first ] , we investigate the subleading corrections to the scaling laws for generic correlation functions . the corrections are explicitly calculated for the hadronic mass in two ways and their equivalence is shown using a feynman - hellmann type relation for the gauge coupling . the relation of the charge radius to the derivative of the form factor is summarised in app . [ app : charge ] for the readers convenience . let us concentrate here on a theory with only one relevant perturbation at the irfp , whose coupling we denote by @xmath3 , and let us introduce an uv cut - off @xmath7 ; @xmath8 and @xmath9 are two local operators . the generic two - point correlator , evaluated on two arbitrary physical states @xmath10 , in the regulated bare theory : @xmath11 depends on the distance @xmath12 , the coupling @xmath3 and the scale @xmath7 . in the expression above we rescale the dimensionful coupling @xmath3 by some reference scale @xmath13 , so that the correlator depends on the dimensionless coupling @xmath14 . we denote the _ scaling dimension of the coupling _ @xmath3 by @xmath15 where @xmath16 and @xmath17 are the engineering and anomalous dimension respectively ( and clearly @xmath18 ) . we shall adopt the same conventions as in refs . @xcite , denoting by @xmath19 and @xmath20 the classical and anomalous dimensions of @xmath21 and therefore the _ scaling dimension of the operator _ @xmath22 reads : @xmath23 . for the sake of clarity , anticipating section [ sec : first ] , we shall denote by @xmath24 ( and thus @xmath25 ) any anomalous dimension at the fixed point in order to distinguish it from the one away from the fixed point . in computing the leading scaling we perform , as usual , an renormalisation group ( rg ) transformation @xmath26 : @xmath27 followed by a rescaling of all mass scales by a factor of @xmath28 , on the rhs of : @xmath29 a crucial observation is that the physical states are free of anomalous scaling @xcite . combining eqs . ( [ eq : rg - g ] ) and ( [ eq : scale - g ] ) we get @xmath30 we can exploit the arbitrariness of @xmath28 and choose it such that @xmath31 . this then implies that @xmath32 with @xmath33 , @xmath34 and @xmath35 a dimensionless function . we will use this particular form of the scaling law to derive some physical consequences in the following sections . the application to mcgts can be inferred indirectly from the caption of fig . [ fig:3-types ] . discussion of finite size effects to eq . can be found in appendix [ app : genfse ] . with @xmath36 , fall into relevant ( @xmath37 ) , irrelevant ( @xmath38 ) and marginal ( @xmath39 ) classes at the fps . in all cases there is a trivial @xmath40 ultraviolet fixed point ( uvfp ) and the @xmath41-axis corresponds to its critical surface . ( left ) the couplings @xmath42 and @xmath43 are irrelevant and relevant at the uvfp . an example of which is qcd with @xmath43 being the gauge coupling and @xmath42 the quark mass . ( middle ) non - trivial ir fixed point @xmath44 . the direction @xmath43 is relevant and irrelevant at uv and irfp respectively whereas @xmath42 is irrelevant at both fixed points . an example is ir - conformal gauge theory ( with @xmath45 ) and @xmath42 is four quark operator provided @xmath46 is really the case . ( right ) the same as before but @xmath42 is relevant at ir fixed point . examples are , assuming @xmath47 , mcgt where @xmath42 is the mass @xmath3 . another example is ir - conformal gauge theory with @xmath45 where @xmath42 is four quark operator where this time @xmath48 . ] in derivations like the one shown in the previous section it was assumed that there is only one relevant operator driving the system away from the irfp . current lattice results seem to suggest that this is indeed the case for the theories that have been investigated so far . nevertheless it might be the case that four quark operators @xmath49 that do appear for example in extended technicolor ( tc ) models , become relevant , i.e. @xmath50 . in this case a situation like the one shown in fig . [ fig:3-types ] ( right ) will apply : in the very far ir this operator will grow and drive the system away from the fixed point ; both the mass of the fermions and this additional coupling need to be tuned for the system to be on the critical surface . academically one could hope to hit a trajectory that goes directly in the uvfp for which @xmath51 , and then flow out of the uvfp along the renormalized trajectory flowing into the irfp . this would be the equivalent of finding a perfect action for the irfp . in practice , e.g. when setting the bare parameters in a simulation at finite lattice spacing , it is impossible to tune the system exactly to this point . the simple plaquette action does contain higher dimensional couplings by construction , and an infinite amount of tuning is needed to find a perfect action . thus summa summarum the study of the scaling dimension of higher dimensional operators within mcgt will remain an important topic in practice . in this section we characterise the size of hadronic states in mcgts by studying the radius of their charge distribution . the radius of the charge distribution is defined from the derivative of the form factor of the state ; the latter is defined in turn from the matrix element of the conserved vector current between hadronic states . scaling laws for the derivatives of the form factor can be deduced from the scaling laws we have obtained for the matrix elements in our previous paper @xcite . in the following let us consider a matrix element where a scalar particle @xmath52 probes a conserved vector current . on the grounds of lorentz covariance the matrix element may be parameterized as follows , which we do not specify any further at this point , might be in the flavour singlet or adjoint representation . the main point is that @xmath52 couples to it . subtle cases in the real hadronic world are @xmath53 by virtue of @xmath54-covariance ; yet @xmath55 as @xmath56 is not @xmath54-eigenstate . ] @xmath57 where @xmath58 is the momentum transfer to the current . note that the structure @xmath59 vanishes by virtue of current conservation : @xmath60 . the function @xmath61 is known as a form factor : its value at zero momentum corresponds to the charge of @xmath52 under the current @xmath62 , and its derivative corresponds to the square of the charge distribution c.f . appendix [ app : charge ] . for instance for the pion form factor in qcd , @xmath63 we wish to emphasise that is not related to the pion s special role in qcd as should be clear from the notes in the appendix [ app : charge ] . we shall later on contrast the behaviour of the pion charge radius in qcd with the charge radius of a generic @xmath3-hadron . in order to determine the scaling exponents , following the notation in @xcite , we define @xmath64 and shall assume that the derivatives exist . our main interest is to establish the behaviour of the size of the @xmath3-hadrons as a function of the relevant perturbation @xmath3 . we will proceed in two steps : ( i ) we derive the relative difference @xmath65 , and ( ii ) we determine @xmath66 . * the mass dependence of the form factor , @xmath67 for shorthand , is summarised in a scaling law akin to eq . : @xmath68 where the dots stand for higher terms in the taylor expansion . note there is no dependence on the rg - scale as the current is conserved . from eq . it is immediate to deduce @xmath69 * second we shall show @xmath70 . it follows directly from our master formula @xcite : @xmath71 where @xmath72 are physical states . we note that @xmath73 ( since @xmath62 is a conserved current ) and that @xmath74 which implies that @xmath75 . since the energy momentum vector is free from anomalous scaling it counts like its engineering dimension in the formula in the nominator of the exponent in and therefore @xmath76 ( i.e. @xmath77 ) . another way to arrive at the same result is to notice that @xmath78 is equal to the charge and since the latter can not scale with external parameters like the mass this implies that @xmath78 is independent of the mass and thus @xmath77 . putting the two results together we get : @xmath79 and for the mean charge radius squared we obtain : @xmath80 where @xmath81 denotes the mass of the hadron @xmath52 and we have used the general result @xmath82 derived for the entire hadronic spectrum in ref . thus in summary the size of the @xmath3-hadrons is inversely proportional to the hadronic mass . whereas this result does not seem surprising it is of importance for controlling fse on the lattice . whereas the scaling laws gives information on the relative size of hadrons for different values of @xmath3 it does not determine its absolute size @xmath83 . the determination of @xmath84 could then be pursued by a measurement of the slope of the form factor through eq . . using twisted boundary conditions could help in improving the momentum resolution , and hence in resolving better the slope of the form factor . the discussion of finite size effects in the context of the form factor can be found in appendix [ app : chargefse ] . it would seem that the arguments of the form factor of a scalar coupled to a conserved current ought to generalise to higher spin hadrons . the application to the analogue of the proton electromagnetic form factor should be rather straightforward . in general a more detailed analysis would necessitate the consideration of the corresponding polarisation tensors . suppose two higher spin hadrons couple to an operator @xmath22 that is not necessarily related to a physical charge . even though @xmath85 in general , we anticipate that the extension of the overlap with the operator @xmath22 is determined by based which in turn follows from generic scaling arguments . let us briefly open a parenthesis here . since @xmath86 with @xmath87 ( c.f . fig . [ fig : coupling ] for an explanation of @xmath88 ) , @xmath89 , one concludes that for @xmath90 at least for sufficiently small @xmath3 to overcome the unknown @xmath91-coefficient discussed above . other than that the hierarchy is controlled by the positivity of @xmath92 which is of course dependent on the actual gauge theory . furthermore whereas the unitarity bound implies @xmath93 no lower bound exists other than the fact that for @xmath94 the operator becomes irrelevant which goes against our working assumption as well as all results , known to the authors , in the literature . it is interesting to contrast the behaviour of the mean charge radius of the ( pseudo)scalar meson in mcgt to the one obtained for the goldstone boson in qcd . more precisely , since in both cases the masses vanish in the limit @xmath95 it is clear from a heuristic viewpoint that , for a state with sharp momentum , the particle can not be localised and therefore one expects the charge radius to diverge . the functional behaviour of the divergence is though not clear a priori . in a theory where chiral symmetry is spontaneously broken , the dynamics of the light goldstone bosons is described by chiral perturbation theory . the mean charge radius can be computed in perturbation theory , and it is found to diverge logarithmically with the pion mass @xcite . this difference suggests that the existence of a conformal fixed point could be characterised by studying the scaling of @xmath96 for the pseudoscalar meson . we wish to reemphasize @xcite that the scaling laws for the mass parameter imply that there is no remnant of the pion as a pseudo goldstone boson in a mcgt . we discuss in this section how to exploit selection rules for two - point vacuum correlators originating from scaling dimensions and spin of the quasi - primary ( to be commented on further below ) operators . in subsection [ sec : cftvssft ] we contrast these aspects from the viewpoint of distinguishing cfts from sfts ( c.f . @xcite for lecture notes on this topic ) , while in subsection [ sec : cftvsconf ] we focus on differentiating conformal from confining behaviour . consider first a scale invariant theory , and specifically ( quasi)-primary fields @xmath97 and @xmath98 with respective scaling dimension @xmath99 . in the absence of symmetry breaking , the short distance correlator obey the following selection rules : 1 . _ @xcite @xmath100 0\phantom{(x^2)^{-\alpha } } \quad \text{cft } \end{cases } \;,\ ] ] with @xmath101 . 2 . _ : _ @xmath102 0\phantom{x^\mu ( x^2)^{-(\alpha+1/2 ) } } \quad \text{cft } \end{cases } \;,\ ] ] with @xmath103 . . follows from the investigations in @xcite . the equations above state that , in order to have a non - vanishing correlator , the scaling dimension as well as the spin structure of the two operators in question have to be identical @xcite . let us add that it is the _ local _ nature of the special conformal transformations which is responsible for the selection rules quoted above . these transformations are precisely the difference between the symmetries of a cft and a sft . using eq . we get : @xmath104 0\phantom{\text{constant } } \quad \text{cft } \end{cases } \;,\ ] ] for @xmath105 such that the system is suitably close to the fixed point . more precisely for fixed @xmath106 and @xmath7 , @xmath41 ( or @xmath28 ) has to be such that the system is close to the fixed point . in general we expect the constant to be finite with the possible caveat that the correlator , which is generally not a physical observable , is affected by ir - divergences . this criterion is unfortunately of limited use for standard gauge theories . in recent years the understanding has emerged @xcite that limit cycles are the only possibility , for four - dimensional unitary quantum field theories to be scale but not conformal invariant . on the other hand limit cycles have only been found in theories with flavour dependent couplings , aka yukawa terms @xcite . these couplings are absent in the gauge theories currently studied on the lattice , and therefore it would seem that ir - conformal theories are indeed ir - conformal and not just ir - scale - invariant . let us add to this end that , currently , the only logical possibility for scale invariant theories to exist is if the theories can evade the strong version of the a - theorem at the non - perturbative level @xcite , as the latter has been shown to be valid in perturbation theory some time ago @xcite . in the previous paragraph we discussed a possible recipe for discerning theories that have only one ( flavour independent ) coupling , and that are cfts and not sfts . the selection rules can also be useful in distinguishing cfts ( ir - conformal ) from confining theories . in the following we shall assume that in a gauge theory without ir fixed points , chiral symmetry breaking and confinement occur together . for that purpose , let us analyse the dimension , and the spin selection rules for a number of example operators . we consider the case of a ) quasi - primary operators from the viewpoint of the cft candidate theory , b ) whose correlation function does not vanish by virtue of non - cft selection rules such as parity symmetry for example . 1 . _ _ scaling dimension:__. yet in the limit this mixing would disappear . a more delicate issue is the mixing with the identity , which corresponds to the disconnected part of the correlator . whereas in dimensional regularisation the mixing occurs with @xmath107 only , which is of no problem for the same reason as above , in lattice cut - off regularisation terms of the form @xmath108 and @xmath109 ( with @xmath110 with @xmath111 being the lattice spacing ) are expected to occur . whereas both have hitherto prohibited a clean extraction of the gluon condensate for instance it is the latter which seems to pose a problem for the case the discussed above . ] @xmath112 & { \cal o}_2 = \bar q q \ ; , \quad & \delta^*_{\bar q q } = & 3 -\gamma_m^ * \ ; . \end{aligned}\ ] ] we note that the correlator with vanishes to all order in perturbation theory in the massless limit as the gauge theory lagrangian is even under @xmath113 and @xmath114 whereas the correlator is odd ( since @xmath115 and @xmath116 ) . thus the correlator probes the non - perturbatve regime or more precisely chiral symmetry breaking through @xmath117 . @xmath118 & { \cal o}^\mu_3= a^{a\,\mu } = \bar q i \gamma^\mu \gamma_5 t^a q \ ; , \quad & & \delta^*_{a^{a\,\mu } } = 3 \ ; , \end{aligned}\ ] ] where @xmath119 is a @xmath120 representation matrix acting on flavour space . for the same reason as above the correlator with vanishes to all orders in perturbation theory in the massless limit . it is , however , non - vanishing in the theory with chiral symmetry breaking since the pion couples to both currents : @xmath121 where it is noted in particular that @xmath122 is finite and non - vanishing in the chiral limit for @xmath123 in a chirally broken phase . conversely @xmath124 is only non - vanishing if @xmath125 at least which is the case for the goldstone bosons only . it seems worthwhile to elaborate a bit further on this point . the correlation function assumes the following form in the chirally broken phase , @xmath126 with all other contributions to the spectrum vanishing in the chiral limit . ] . more precisely @xmath127 & = & f_\pi g_\pi \partial_\mu \left ( \frac{1}{x^2 } + { \cal o}(m_\pi \ln x ) \right ) = { \langle \bar qq \rangle } \frac{2 x^\mu}{x^4 } + { \cal o}(m ) \ ; , \end{aligned}\ ] ] where @xmath128 , for odd powers of @xmath12 and in the last equality we have made use of the gell - mann oakes renner ( gmor ) relation @xmath129 . let us briefly comment on the scaling dimensions of the operators quoted in eqs . ( [ eq : proposal1],[eq : proposal2 ] ) which were already exploited in our previous work @xcite . the scaling dimension of the gluon field strength tensor is four as it appears in the trace anomaly which is related to the physical mass . the scaling dimension of quark condensate times the mass is four , @xmath130 , for the same reason and therefore @xmath131 . the scaling dimension of @xmath132 is three because it is a partially conserved current affected only by explicit breaking . the scaling dimension of @xmath133 can be obtained from ward identities as presented in appendix b.1 of ref . it would seem worthwhile to point out that the operators quoted in eqs . ( [ eq : proposal1],[eq : proposal2 ] ) are of the ( quasi)-primary type as the non - primary operators derive from the latter through derivatives . finally in essence we get , as a replacement of eq . for the case at hand , @xmath134 0\phantom{\neq 0 } \quad \text{cft } \end{cases } \;,\ ] ] for @xmath41 such that the system is suitably close to the fixed point as previously discussed . we note that @xmath135 is known explicitly for the second example considered . ( [ eq : scale - g ] ) , and ( [ eq : scale - f ] ) show that the behaviour of @xmath136 , as @xmath137 , provides the possibility to distinguish between a cft and an sft or a confining theory . in practice one would keep @xmath12 fixed and study the behaviour of the correlator for @xmath95 . in lattice simulations one would need to work for given values of @xmath3 with sufficiently large volumes , @xmath138 , as the limits @xmath95 and @xmath139 are known not to commute . further comments can be inferred from fig . [ fig : coupling ] where the relative scales are sketched against a typical behaviour of a ( gauge ) coupling . in regard to this figure we would like to draw the readers attention to the fact that the actual value of the coupling is scheme dependent , whereas the question of whether there is a fixed point or not is scheme independent as it shows up in physical measurable quantities in terms of scaling laws . . the mass parameter @xmath3 , or equivalently the the mass scale @xmath140 ( of the hadronic bound states ) , drives the theory away from the fp . this is the picture that heuristic computations of the quark condensate suggest e.g. @xcite . note the inverse of the lattice box size has to be significantly smaller than @xmath81 in order for fse effects to be under control . more precisely as long as @xmath141 , fse are of the order @xmath142 . if this condition is not met the effects are power - like with known exponents e.g. @xcite . we should point out that we have not attempted to indicate the effect on the coupling of the actual value of @xmath143 on the curve on the graph . , width=336 ] we have already discussed the scaling of field correlators as a function of the mass @xmath3 for @xmath144 . when the coupling is not tuned to its critical value , scaling corrections appear . in this section we compute these corrections at first order in the @xmath145 . in section [ sec : linear ] we introduce the notation and discuss the linearized rg equations . in section [ sec : sc ] we compute the scaling corrections to field correlators of local operators . in section [ sec : scm ] we apply these results and compute the scaling corrections to the hadronic masses first by using the trace anomaly , and then by analysing the mass correlator . furthermore we show that the two expressions for the scaling corrections are equal by using an extension of the feynman - hellmann theorem @xcite . we assume that the bare couplings at the cut - off scale @xmath43 and @xmath146 , which correspond to the point @xmath147 in fig . [ fig:3-types](left ) with the identification @xmath148 , are chosen such that the system is on a trajectory that is close to the fixed point . we are going to linearize the rg flow equations in the deviations from the fixed point , that is to say in the variable @xmath145 , where we use the notation @xmath149 throughout this section . we shall comment on the aspects of this expansion at the end of section [ sec : discuss ] . for the beta function , the mass anomalous dimension @xmath17 , and the anomalous dimension matrix @xmath150 rather than @xmath151 in order to distinguish it from anomalous dimensions of parameters such as the mass for instance . ] of a generic set of operators @xmath152 that mix under the rg - flow , we may linearize the system around the irfp as follows : function in terms of the coupling for a system exhibiting asymptotic freedom @xmath153 and a non - trivial ir fixed point at some value @xmath154 . in section [ sec : first ] a system in the vicinity of the ir fixed point is considered as indicated in the figure.,width=336 ] @xmath155 \gamma_m & = \gamma_m^ * + \gamma_m^{(1 ) } \delta g + { \cal o } ( \delta g^2 ) \ ; , & & \nonumber \\[0.1 cm ] \gamma_{ij } & = \gamma^*_{ij } + \gamma_{ij}^{(1 ) } \delta g + { \cal o } ( \delta g^2 ) \ ; , & & ( \gamma_{ij } \equiv ( \gamma_o)_{ij } ) \;.\end{aligned}\ ] ] we have verified that in a mass independent scheme @xmath156 is universal ( scheme independent ) whereas @xmath157 are not . we remind the reader that the anomalous dimensions associated with gauge invariant operators ( such as @xmath92 ) are universal . when working with renormalized quantities we shall choose notation accordingly . the behaviour of the beta function as a function of the coupling is illustrated in fig . [ fig : one ] ; @xmath156 corresponds to the slope where the curve crosses the ir fixed point . we note that for the beta function described in fig . [ fig : one ] , the coefficient @xmath156 is positive as there are no further zeros between @xmath158 and @xmath144 . the beta function equation is easily integrated to that order , @xmath159 where @xmath7 is a uv cut - off as will become clear in the next subsection . we shall use the language of wilsonian renormalization group for which the theory is defined at some fixed uv cut - off @xmath160 is related to the uv cut - off as @xmath161 . ] . let us consider a correlation function @xmath162 , as a function of the bare parameters @xmath163 , and the uv cut - off @xmath7 . we shall denote by @xmath164 the matrix that describes the mixing of @xmath21 under renormalization . @xmath21 satisfies an rg equation ( also known as t hooft - weinberg or callan - symanzik equations ) , e.g. ref . @xcite , @xmath165 where summation over @xmath166 is implied and @xmath167 we now wish to reformulate the theory using a different uv cut - off @xmath168 @xmath169 where the parameter @xmath28 has the interpretation of a blocking factor , and @xmath170 if high - energy modes are to be integrated out . the formal solution to eq . is given by : @xmath171 where @xmath172 we assume here that we are working in a _ mass independent scheme _ , and therefore the beta function and the anomalous dimensions only depend on the gauge coupling @xmath43 . the solution ( [ eq : ansatz],[eq : rc1 ] ) is known by the name of the method of characteristics , see e.g. ref . @xcite . assuming the fixed point is in the linear regime , the three equations above can be solved to order @xmath173 : @xmath174\ , , \nonumber \\ & z_{ij}(b ) & = & \ ; \exp\left [ \gamma^ * \ln b - \frac{\gamma^{(1)}}{\beta_1 } \delta g f(b ) \right]_{ij } \;,\end{aligned}\ ] ] where we have introduced the notation @xmath175 which parameterizes the distance from the initial point in blocking space . eq . may be written using the relation as : @xmath176_{ij } o_j ( g(b ) , m b^{\gamma_m^ * } \exp\left[- \frac{\gamma_m^{(1)}}{\beta_1 } \delta g f(b ) \right ] , \lambda / b ) \nonumber \\ & = & \exp\left [ - \delta^ * \ln b + \frac{\gamma^{(1)}}{\beta_1 } \delta g f(b ) \right]_{ij } o_j ( g(b),m b^{-(1+\gamma_m^ * ) } \exp\left[- \frac{\gamma_m^{(1)}}{\beta_1 } \delta g f(b ) \right ] , \lambda ) \nonumber \;,\end{aligned}\ ] ] where in the last equality we have rescaled all dimensionful quantities by a factor @xmath28 . the matrix @xmath177 , where @xmath178 is the classical dimension of @xmath21 , yields the scaling dimensions of the operators . in order to get the @xmath179 corrections we need to expand in that variable . in our opinion this is best done from the expression in the the second line of the equation above . the last step can be done after the expansion for each individual term . furthermore , in order to avoid path ordering in coupling space , we shall assume that @xmath180 is diagonal . the corrections are parameterized as follows , @xmath181^ * + \delta g \ , o^{(1)}_i + { \cal o } ( \delta g^2 ) \right ) \;,\ ] ] where @xmath182^*f(b ) - \frac{\gamma_m^{(1)}}{\beta_1 } m^*[o_i]^*_{,m } f(b ) + [ o_i]^*_{,g } b^{-\beta_1 } \right)\ ] ] and @xmath183^*&= & o_i(g(b ) , m(b ) , \lambda / b)|_{\delta g = 0 } \ ; , \nonumber \\[0.1 cm ] [ o_i]^*_{,m } & = & \frac{\partial}{\partial m(b ) } o_i(g(b ) , m(b ) , \lambda / b)|_{\delta g = 0 } \ ; , \nonumber \\[0.1 cm ] [ o_i]^*_{,g } & = & \frac{\partial}{\partial g(b ) } o_i(g(b ) , m(b ) , \lambda / b)|_{\delta g = 0 } \nonumber \ ; , \\[0.1 cm ] m^ * & = & m(b)|_{\delta g = 0 } \ ; , \quad ( g^ * = g(b)|_{\delta g = 0 } ) \ , .\end{aligned}\ ] ] we wish to emphasise that , when @xmath184 is tuned to the fixed point coupling , @xmath185 corresponds to the leading scaling of the mass at the fixed point . the scaling corrections as a function of @xmath3 can be made explicit by rescaling all dimensionful quantities by the appropriate power of @xmath28 in the last step in eq . , and then using the arbitrariness of @xmath170 to impose : @xmath186 as a result , we obtain a scaling formula that includes the scaling corrections at first order in @xmath179 : @xmath187^ * \left ( 1 + \delta g ( a + b \,\hat m^{\frac{\beta_1}{1+\gamma_m^ * } } ) \right ) + { \cal o } ( \delta g^2 ) \ ; , \ ] ] with @xmath188^*_{,m}}{[o_i]^ * } \right\}_b \nonumber \\ b & = & \left\ { + \frac{\gamma^{(1)}_{ij } } { \beta_1 } - \frac{\gamma_m^{(1)}}{\beta_1 } m^ * \frac{[o_i]^*_{,m}}{[o_i]^ * } + \frac{[o_i]^*_{,g}}{[o_i]^ * } \right\}_b \;,\end{aligned}\ ] ] where the curly brackets with a @xmath28 superscript indicate that all physical units are to be scaled by @xmath28 , e.g. @xmath189 \}_b \stackrel{\eqref{eq : star}}{= } \ { o_i(g(b ) , m(b ) , \lambda / b)|_{\delta g = 0}\}_b \to b^{d_o } o_i(g(b ) , b m(b ) , \lambda)|_{\delta g = 0}$ ] . it is interesting to note that the scaling corrections above simplify when @xmath190 : @xmath191^*_{,g}}{[o_i]^ * } + o(\beta_1)\ , .\ ] ] this situation is expected to be realised at the lower edge of the conformal window in the banks - zaks limit . the expression yields the corrections to scaling for small fermion mass @xmath192 , while the irrelevant coupling @xmath43 is at a distance @xmath179 from the fixed point . clearly when @xmath179 vanishes , so do the scaling violations . we note that for fixed initial value @xmath43 , @xmath193 is proportional to the value of the ir fixed point coupling @xmath184 , as can be inferred from fig . [ fig:3-types ] . the linear approximation discussed here becomes therefore less reliable if the ir fixed point is at strong coupling coupling , unless @xmath43 is tuned to reduce the size of @xmath179 . note that for large @xmath179 the linear corrections tend to grow . this can be compensated by going to smaller initial masses @xmath192 since the first order ( relative ) scaling corrections are determined by the combination @xmath194 . we would like to add an important point concerning the size of the corrections at the lower edge of the conformal window . for a strong coupling fixed point , one would expect large values of @xmath195 , as well as @xmath196 , whereas the value of @xmath156 is expected to be small as the fixed point is to be lost which in turn is consistent with @xmath184 being large . moreover , unless the bare coupling @xmath43 is fine - tuned , one can expect to have rather large values of @xmath179 , driven by our ignorance in guessing the exact location of the fixed point . thus in summary the precoefficient @xmath197 should be expected to be large at the lower edge of the conformal window . on the other hand the exponent @xmath198 is then small and leads to a suppression . in the previous statements large and small are meant relative to the region away from the lower edge of the conformal window . which of the two counteracting effects dominates is unclear a priori but the argument suggests that it is important to go to small masses @xmath192 at the lower edge of the conformal window to suppress potentially large scaling corrections . this is of practical importance as many of the lattice simulations have been performed precisely at the lower edge of the conformal window in search of a theory of walking technicolor . the signs of @xmath199 and @xmath200 , in eq . , are determined by the dynamics . since in general we can not make statements about the derivative of the operators the sign of @xmath199 and @xmath200 are thus not known a priori . this is somewhat different for the hadronic masses that is to say for the operators @xmath201 and @xmath202 which is what we are going to exploit in the next section . we shall first introduce some notation and justify the formulae needed for the comparison of the two derivations of the scaling corrections to the hadronic mass in subsection [ sec : pathways ] . the following notation , @xmath203 shall prove convenient throughout this section . the subscript @xmath204 denotes the connected part of the matrix element , while @xmath205 is a physical state with definite spatial momentum and energy and @xmath206 is a ( local ) operator . above we have explicitly indicated the energy dependence of the hadronic state @xmath52 which we occasionally suppress in the remaining part of this work . note that the disconnected part of the correlator is related to the vacuum energy , that is to say the cosmological constant . as usual the lorentz invariant state normalisation is given by : @xmath207 the expectation value of the energy momentum tensor @xmath208 in a single - particle state is : @xmath209 where @xmath210 . in order to keep a compact notation we are going to extend the notation for two specific matrix elements to : @xmath211 g_{{e_h } } & \equiv & { \langle \frac{1}{g^2 } g^2 \rangle}_{e_h } \ , .\end{aligned}\ ] ] we remind the reader that the notation refers to the connected part of the matrix element only . for the discussion in this section it is convenient to use renormalized quantities . accordingly we denote the renormalized couplings by @xmath212 and @xmath213 , and the matrix elements of the renormalized operators @xmath214 and @xmath215 respectively . the renormalized coupling @xmath212 is defined as : @xmath216 en passant we note that physical quantities such as the energy momentum tensor and thus the hadronic mass do not renormalize ( i.e. @xmath217 ) . in the neighbourhood of the irfp , the renormalization constant is expanded similar to as , @xmath218 which implies : @xmath219 the trace anomaly can be written in terms of the renormalized quantities as @xcite : @xmath220 and is an rg - invariant . more precisely since @xmath221 is an rg invariant , @xmath222 inherits this property by virtue of eq . . this entails that @xmath223 . in identifying the two computation the following relations are of importance : @xmath224 the first relation is a straightforward application of the feynman - hellmann theorem and is widely used , as for instance in our previous work @xcite . the second relation is akin to a feynman - hellmann relation . it has been derived in @xcite through an rge , the trace anomaly as well as the first relation in . later it was rederived and checked in a few exactly solvable models in @xcite . let us now compute the corrections to scaling in two different ways by using results from the previous section : the corrections are obtained up to order @xmath225 and the symbol @xmath226 on other quantities denotes the linear variation in the @xmath227 variable . recall that @xmath228 ( at least in a mass independent scheme ) and @xmath229 in general and we shall therefore use notation accordingly . 1 . first we compute @xmath230 directly from the rg scaling formulae for renormalized quantities , combined with the relation : @xmath231^*_{,{\bar{g } } } b^{-\beta_1 } - \frac{{\bar{\gamma}}_m^{(1)}}{\beta_1 } { \bar{m}}^ * [ 2m_h^2]^*_{,{\bar{m } } } f(b ) \right ) + { \cal o}(\delta { \bar{g}}^2 ) \nonumber \\ & \stackrel{\eqref{eq : fh}}{=}&\delta { \bar{g } } \left ( - \frac{1}{{\bar{g}}^ * } [ { \bar{g}}_{m_h}]^ * b^{-\beta_1 } - 2 \frac{{\bar{\gamma}}_m^{(1)}}{\beta_1 } [ { \bar{q}}_{m_h}]^ * f(b ) \right ) + { \cal o}(\delta { \bar{g}}^2 ) \;. \end{aligned}\ ] ] 2 . second we compute @xmath230 through the trace anomaly : @xmath232^ * + { \bar{\gamma}}_m^{(1 ) } [ { \bar{q}}_{m_h}]^ * \right ) + \left(1 + \gamma_m^*\right ) \delta { \bar{q}}_{m_h } \ ; , \end{aligned}\ ] ] which necessitates the computation of @xmath233 . the latter is given by eq . : @xmath234^*_{,{\bar{g } } } b^{-\beta_1 } - \frac{{\bar{\gamma}}_m^{(1)}}{\beta_1}{\bar{m}}^ * [ { \bar{q}}_{m_h}]^*_{,{\bar{m } } } f(b ) \right ) \;. \end{aligned}\ ] ] the expressions , and , yield the scaling corrections as a function of @xmath235 and the mass @xmath3 . these expressions all have the same scaling exponents , yet it is not clear from these formulae that the corresponding prefactors are equal . to compare the prefactors we ought to compute @xmath236^*_{,{\bar{g}}}$ ] and @xmath236^*_{,{\bar{m}}}$ ] to leading order in @xmath237 . the latter is simply given by the leading order scaling @xmath238^*_{,{\bar{m } } } = \frac{2}{1+\gamma_m^ * } [ { \bar{q}}_{m_h}]^ * + { \cal o}(\delta { \bar{g } } ) \;,\ ] ] up to corrections which are beyond the aimed accuracy . the computation of @xmath236^*_{,{\bar{g}}}$ ] is slightly more involved ; it is obtained by differentiating @xmath239 with respect to @xmath240 using : @xmath241^ * + { \cal o}(\delta g^2)\ ] ] as well as the right hand side ( rhs ) of eq . , @xmath242^ * + { \bar{\gamma}}_m^{(1 ) } [ { \bar{q}}_{m_h}]^ * + ( 1 + \gamma_m^ * ) [ { \bar{q}}_{m_h}]^*_{,{\bar{g } } } + o(\delta { \bar{g } } ) \;,\end{aligned}\ ] ] where @xmath243 denotes differentiation with respect to @xmath240 . we have dropped the term @xmath244 from passing from the first to the second line since it is of relative order @xmath245 . by equating eqs . and we may solve for @xmath236^*_{,{\bar{g}}}$ ] and insert it into and finally into to obtain : @xmath246^ * + { \bar{\gamma}}_m^{(1 ) } [ { \bar{q}}_{m_h}]^ * \right ) + \nonumber \\[0.1 cm ] & + & \delta { \bar{g } } \big ( - 2 \frac{{\bar{\gamma}}_m^{(1)}}{\beta_1 } [ { \bar{q}}_{m_h}]^ * f(b ) \underbrace { -\frac{1}{{\bar{g}}^*}[{\bar{g}}_{m_h}]^ * b^{-\beta_1 } - b^{-\beta_1 } \left ( \frac{\beta_1}{2 { \bar{g}}^ * } [ { \bar{g}}_{m_h}]^ * + { \bar{\gamma}}_m^{(1 ) } [ { \bar{q}}_{m_h}]^*\right ) } _ { [ { \bar{q}}_{m_h}]^*_{,{\bar{g } } } } \big ) \nonumber \\[0.0 cm ] & = & \delta { \bar{g } } \left ( - \frac{1}{{\bar{g}}^ * } [ { \bar{g}}_{m_h}]^ * b^{-\beta_1 } - 2 \frac{{\bar{\gamma}}_m^{(1)}}{\beta_1 } [ { \bar{q}}_{m_h}]^ * f(b ) \right ) \ ; , \end{aligned}\ ] ] which equals eq . , as expected . we note that the second line in eq . is equal to @xmath247 at leading order . an interesting question is whether we can say something about the sign of the correction in eq . . that is to say we would like to know whether @xmath248 in @xmath249 is positive or negative . we should add that @xmath250 as can be inferred from fig . [ fig : one ] . as we are interested in the long - distance dynamics of the theory that is defined at the uv gaussian fixed point , the coupling lies in the interval @xmath251 $ ] . as previously stated @xmath252 in eq . by virtue of no zero crossings of the @xmath253-function between the uv and irfp . if the anomalous dimension increases monotonically from the uvfp @xmath254 to @xmath255 then @xmath256 , which is not compelling but to be expected . furthermore @xmath257^ * > 0 $ ] since @xmath258^ * + { \cal o}(\delta g)$ ] and @xmath259 as we have assumed @xmath3 to be a relevant direction and @xmath260 for @xmath261 . finally we see that everything depends on the sign of @xmath262 for which we can not make a definite assertion . it is well - known that naive positivity of operators , effective in quantum mechanics , is not necessarily maintained in quantum field theory . in the case at hand there is the additional complication that only the connected part of the matrix element is required . that is to say even if the total matrix element were positive the connected part might still be negative . it seems worthwhile to point out that in qcd for @xmath95 and @xmath263 eq . implies that @xmath264 indeed . summa summarum we can not say anything definite about the sign of @xmath248 as the sign of @xmath262 seems uncertain . we have explored the consequences of conformal scaling in a number of interesting cases . one of our main findings is the scaling of the radius of the @xmath3-hadrons as a function of the fermion mass . our results show that the typical size of the @xmath3-hadrons , defined from the average charge density , is a linear function of the inverse mass of the hadron . characterising the size of @xmath3-hadrons is very important in order to understand how to tame fse in numerical studies , and hence obtain reliable results from monte carlo simulations . it is worthwhile to emphasise that the dependence of the mean charge radius on the mass of the @xmath3-hadrons is radically different from the logarithmic scaling obtained in chiral perturbation theory for the goldstone boson in a chirally broken theory @xcite . the difference between the two provides yet another way to asses the difference between a conformal and a confining phase . by exploiting selection rules for scaling dimensions and spin we propose to use coordinate space correlation function , deformed by a mass term , to distinguish cfts from sfts as well as confining theories . we investigated the scaling corrections to correlation functions by linearizing the rge in the variable @xmath265 which is the distance of the initial coupling from the , presumably unknown , fixed point value . in essence this corresponds to the scaling corrections due to the ir - irrelevant coupling @xmath43 . the generic result is given in eq . and . in subsection [ sec : discuss ] we note in particular that scaling corrections can be expected to be large at the lower edge of the conformal window . this can be counteracted by going to smaller masses . we computed the scaling corrections to the hadron mass explicitly , once directly through its associated correlation function and second through the trace anomaly . the results are given in eqs . and their equivalence is made manifest in eq . . the latter was established by using the the feynman - hellmann relation for the mass and an analogous relation for the gauge coupling . the derivation of the latter is given in a separate paper @xcite . rz acknowledges the support of an advanced stfc fellowship . ldd and rz are supported by an stfc consolidated grant . we are grateful to julius kuti , jeff fortin , martin lscher and tim morris for discussion at various stages of our investigations . in this appendix we shall give a derivation of the charge radius in terms of the form factor as stated in eq . as the derivation of the latter has become sparse in modern textbooks . we shall work in minkowski - space in this section with metric signature @xmath266 . starting from the zeroth component of @xmath267 where @xmath268 and @xmath269 as usual . note for on - shell states the equality of the @xmath270-vectors @xmath271 then implies the vanishing of the @xmath272-vector @xmath273 . we define the @xmath274-dimensional fourier transform of the form factor @xmath275 for latter convenience . the scalar product with arrow vector denotes the @xmath270-dimensional scalar product . the charge of the state @xmath52 is obtained by integrating the charge density over the space @xmath276 & \stackrel{\eqref{eq : again}}{= } & 2 e_p f_+^h(0 ) \big ( \int_v d^3 x \big ) \;,\end{aligned}\ ] ] remembering the normalisation @xmath277 , setting @xmath271 on the first line and using the definition on the second line . to the more mathematical inclined reader this equation might look better if @xmath278 is not assumed before identifying @xmath279 . the latter identification leads to the first result of this appendix : @xmath280 and suggests that @xmath281 is the charge density which we shall use below . we shall define the @xmath270-dimensional laplace operator @xmath282 acting on fourier space as follows : @xmath283 we let it act on the form factor directly and through its fourier transform : @xmath284 & \stackrel{\eqref{eq : fourier}}{= } & \int \frac{d^3 x}{(2 \pi)^3 } \vec{x}^2 \hat f^h(\vec{x},q_0 ^ 2 ) e^{i \vec{x } \cdot \vec{q } } |_{q = 0 } = \int \frac{d^3 x}{(2 \pi)^3 } \vec{x}^2 \hat f^h(\vec{x},0 ) \;.\end{aligned}\ ] ] this leads , using , to the second result of this appendix : @xmath285 thus we have now justified the results quoted in eq . through and . the aim of this appendix is to present extension of presentation in the main text to include finite size effects . finite - size effects to eq . can be easily incorporated . writing explicitly the dependence of the correlators on the size @xmath143 of the physical volume , the rg equation becomes : @xmath286 the underlying assumption in the equation above is that the volume is large enough , such that a blocking transformation does not change the volume dependence . when all dimensionful quantities are rescaled by the corresponding power of the reference mass @xmath13 , eq . becomes : @xmath287 in the thermodynamic limit , @xmath288 @xmath289 where @xmath290 is a number . we discuss the modifications of the scaling laws of the form factor , relevant to the charge radius , due to finite - size effects . the form factor depends on the fermion mass , the uv cut - off , and the physical size of the lattice : @xmath291 where the hat indicates that dimensionful quantities have been rescaled by the appropriate powers of the reference mass @xmath13 . keeping @xmath7 unchanged , and performing the standard rg analysis that we used above , yields : @xmath292 expanding eq . ( [ eq : fffss ] ) in powers of @xmath293 : @xmath294 this is the same expansion obtained in eq . ( [ eq : taylor ] ) , but now the coefficients of the expansion depend on the physical size of the lattice @xmath143 . denoting the @xmath295-th derivative of the form factor by @xmath296 , and introducing the dimensionless finite - size scaling variable @xmath297 , we obtain : @xmath298 with @xmath290 a number and the dots denote the finite volume corrections . the dependence in eq . reproduces the expected mass scaling discussed before in the large - volume limit . c. t. hill and e. h. simmons , `` strong dynamics and electroweak symmetry breaking , '' phys . * 381 * ( 2003 ) 235 [ erratum - ibid . * 390 * ( 2004 ) 553 ] [ hep - ph/0203079 ] . f. sannino , `` conformal dynamics for tev physics and cosmology , '' acta phys . b * 40 * ( 2009 ) 3533 [ arxiv:0911.0931 [ hep - ph ] ] . m. a. luty and t. okui , `` conformal technicolor , '' jhep * 0609 * ( 2006 ) 070 [ hep - ph/0409274 ] . d. d. dietrich , f. sannino and k. tuominen , `` light composite higgs from higher representations versus electroweak precision measurements : predictions for cern lhc , '' phys . d * 72 * ( 2005 ) 055001 [ hep - ph/0505059 ] . w. e. caswell , `` asymptotic behavior of nonabelian gauge theories to two loop order , '' phys . lett . * 33 * ( 1974 ) 244 . t. banks and a. zaks , `` on the phase structure of vector - like gauge theories with massless fermions , '' nucl . b * 196 * ( 1982 ) 189 . t. appelquist , r. c. brower , m. i. buchoff , m. cheng , s. d. cohen , g. t. fleming , j. kiskis and m. lin _ et al . _ , `` approaching conformality with ten flavors , '' arxiv:1204.6000 [ hep - ph ] . a. cheng , a. hasenfratz , g. petropoulos and d. schaich , `` scale - dependent mass anomalous dimension from dirac eigenmodes , '' arxiv:1301.1355 [ hep - lat ] . y. aoki , t. aoyama , m. kurachi , t. maskawa , k. -i . nagai , h. ohki , e. rinaldi and a. shibata _ et al . _ , `` light composite scalar in twelve - flavor qcd on the lattice , '' arxiv:1305.6006 [ hep - lat ] . c. -j . d. lin , k. ogawa , h. ohki and e. shintani , `` lattice study of infrared behaviour in su(3 ) gauge theory with twelve massless flavours , '' jhep * 1208 * ( 2012 ) 096 [ arxiv:1205.6076 [ hep - lat ] ] . z. fodor , k. holland , j. kuti , d. nogradi , c. schroeder and c. h. wong , `` conformal finite size scaling of twelve fermion flavors , '' pos lattice * 2012 * ( 2012 ) 279 [ arxiv:1211.4238 [ hep - lat ] ] . l. del debbio and r. zwicky , `` hyperscaling relations in mass - deformed conformal gauge theories , '' phys . d * 82 * ( 2010 ) 014502 [ arxiv:1005.2371 [ hep - ph ] ] . l. del debbio , r. zwicky , `` scaling relations for the entire spectrum in mass - deformed conformal gauge theories , '' phys . * b700 * ( 2011 ) 217 - 220 . [ arxiv:1009.2894 [ hep - ph ] ] . j. giedt , `` lattice gauge theory and physics beyond the standard model , '' pos lattice * 2012 * ( 2012 ) 006 . v. a. miransky , `` dynamics in the conformal window in qcd like theories , '' phys . rev . d * 59 * ( 1999 ) 105003 [ hep - ph/9812350 ] . a. patella , l. del debbio , b. lucini , c. pica and a. rago , `` finite volume effects in su(2 ) with two adjoint fermions , '' pos lattice * 2011 * ( 2011 ) 084 [ arxiv:1111.4672 [ hep - lat ] ] . z. fodor , k. holland , j. kuti , d. nogradi , c. schroeder and c. h. wong , `` can the nearly conformal sextet gauge model hide the higgs impostor ? , '' phys . b * 718 * ( 2012 ) 657 [ arxiv:1209.0391 [ hep - lat ] ] . d. dorigoni and v. s. rychkov , `` scale invariance + unitarity @xmath299 conformal invariance ? , '' arxiv:0910.1087 [ hep - th ] . j. -f . fortin , b. grinstein and a. stergiou , `` scale without conformal invariance : an example , '' phys . b * 704 * ( 2011 ) 74 [ arxiv:1106.2540 [ hep - th ] ] . j. -f . fortin , b. grinstein and a. stergiou , `` scale without conformal invariance : theoretical foundations , '' jhep * 1207 * ( 2012 ) 025 [ arxiv:1107.3840 [ hep - th ] ] . + as with respect to later references @xmath300 throughout this paper . j. -f . fortin , b. grinstein and a. stergiou , `` limit cycles and conformal invariance , '' jhep * 1301 * ( 2013 ) 184 [ jhep * 1301 * ( 2013 ) 184 ] [ arxiv:1208.3674 [ hep - th ] ] . m. a. luty , j. polchinski and r. rattazzi , `` the @xmath111-theorem and the asymptotics of 4d quantum field theory , '' jhep * 1301 * ( 2013 ) 152 [ arxiv:1204.5221 [ hep - th ] ] . h. osborn and a. c. petkou , `` implications of conformal invariance in field theories for general dimensions , '' annals phys . * 231 * ( 1994 ) 311 [ hep - th/9307010 ] . y. nakayama , arxiv:1302.0884 [ hep - th ] . h. ginsparg , `` applied conformal field theory , '' hep - th/9108028 . i. jack and h. osborn , `` analogs for the c theorem for four - dimensional renormalizable field theories , '' nucl . phys . b * 343 * ( 1990 ) 647 . j. -f . fortin , b. grinstein and a. stergiou , `` cyclic unparticle physics , '' phys . b * 709 * ( 2012 ) 408 [ arxiv:1110.1634 [ hep - th ] ] . z. fodor , k. holland , j. kuti , d. nogradi , c. schroeder and c. h. wong , phys . b * 718 * ( 2012 ) 657 [ arxiv:1209.0391 [ hep - lat ] ] . j. gasser and h. leutwyler , `` low - energy expansion of meson form - factors , '' nucl . b * 250 * ( 1985 ) 517 . f. sannino and r. zwicky , `` unparticle and higgs as composites , '' phys . d * 79 * ( 2009 ) 015016 [ arxiv:0810.2686 [ hep - ph ] ] . s. l. adler , j. c. collins and a. duncan , `` energy - momentum - tensor trace anomaly in spin 1/2 quantum electrodynamics , '' phys . d * 15 * ( 1977 ) 1712 . j. c. collins , a. duncan and s. d. joglekar , `` trace and dilatation anomalies in gauge theories , '' phys . d * 16 * ( 1977 ) 438 . n. k. nielsen , `` the energy momentum tensor in a nonabelian quark gluon theory , '' nucl . b * 120 * ( 1977 ) 212 . p. minkowski , print-76 - 0813-bern `` on the anomalous divergence of the dilatation current in gauge theories , '' k. fujikawa , `` energy momentum tensor in quantum field theory , '' phys . d * 23 * ( 1981 ) 2262 . l. del debbio and r. zwicky , `` renormalization group , trace anomaly and feynman - hellmann theorem , '' arxiv:1306.4274 [ hep - ph ] . v. prochazka and r. zwicky , `` gluon condensates from the hamiltonian formalism , '' arxiv:1312.5495 [ hep - ph ] .

the extended galileon models possess tracker solutions with de sitter attractors along which the dark energy equation of state is constant during the matter - dominated epoch , i.e. , where is a positive constant . even with this phantom equation of state there are viable parameter spaces in which the ghosts and laplacian instabilities are absent . using the observational data of the supernovae type ia , the cosmic microwave background ( cmb ) , and baryon acoustic oscillations , we place constraints on the tracker solutions at the background level and find that the parameter is constrained to be ( 95 % cl ) in the flat universe . in order to break the degeneracy between the models we also study the evolution of cosmological density perturbations relevant to the large - scale structure ( lss ) and the integrated - sachs - wolfe ( isw ) effect in cmb . we show that , depending on the model parameters , the lss and the isw effect is either positively or negatively correlated . it is then possible to constrain viable parameter spaces further from the observational data of the isw - lss cross - correlation as well as from the matter power spectrum .

1112.1774

the main target of the dark energy research over the next few years or so is to distinguish between the @xmath3-cold - dark - matter ( @xmath3cdm ) model and dynamical models with time - varying equations of state @xmath4 . from the observational data of wmap7 combined with baryon acoustic oscillations ( bao ) @xcite and the hubble constant measurement @xcite , komatsu _ et al . _ @xcite derived the bound @xmath5 ( 68 % cl ) for the constant equation of state . adding the supernovae type ia ( sn ia ) data provides tighter constraints on @xmath4 , but still the phantom equation of state ( @xmath6 ) is allowed by the joint data analysis @xcite . this property persists for the time - varying dark energy equation of state with the parametrization such as @xmath7 @xcite , where @xmath8 is the scale factor @xcite . in the framework of general relativity ( gr ) it is generally difficult to construct theoretically consistent models of dark energy which realize @xmath6 . in quintessence @xcite with a slowly varying scalar - field potential , for example , the field equation of state is always larger than @xmath9 . a ghost field with a negative kinetic energy leads to @xmath6 @xcite , but such a field is plagued by a catastrophic instability of the vacuum associated with the spontaneous creation of ghost and photon pairs @xcite . in modified gravitational theories it is possible to realize @xmath6 without having ghosts and laplacian - type instabilities ( see refs . @xcite ) . in @xmath10 gravity , where the lagrangian @xmath11 is a function of the ricci scalar @xmath12 , the dark energy equation of state crosses the cosmological constant boundary ( @xmath13 ) @xcite for the viable models constructed to satisfy cosmological and local gravity constraints @xcite . this is also the case for the brans - dicke theory @xcite with a field potential which accommodates the chameleon mechanism @xcite to suppress the propagation of the fifth force @xcite . in modified gravity models of dark energy based on the chameleon mechanism ( including @xmath10 gravity ) , the effective potential of a scalar degree of freedom needs to be carefully designed to pass cosmological and local gravity constraints @xcite . there is another class of modified gravity models of dark energy in which a nonlinear self - interaction of a scalar degree of freedom @xmath14 can lead to the recovery of gr in a local region through the vainshtein mechanism @xcite . the representative models of this class are those based on the dvali - gabadadze - porrati ( dgp ) braneworld @xcite and the galileon gravity @xcite ( see refs . @xcite for the implementation of the vainshtein mechanism in these models ) . the nonlinear interaction of the form @xmath15 , which appears from the brane - bending mode in the dgp model @xcite , gives rise to the field equation invariant under the galilean shift @xmath16 in the flat spacetime . this was extended to more general field lagrangians satisfying the galilean symmetry in the limit of the minkowski spacetime @xcite . the cosmology based on the covariant galileon or on its modified versions has been studied by many authors @xcite . in refs . @xcite the dynamics of dark energy was investigated in the presence of the full covariant galileon lagrangian . in this model the solutions with different initial conditions converge to a common trajectory ( tracker ) . along the tracker the dark energy equation of state @xmath4 changes as @xmath17 ( radiation era ) @xmath18 @xmath19 ( matter era ) @xmath18 @xmath9 ( de sitter era ) @xcite . there exists a viable parameter space in which the ghosts and laplacian instabilities are absent . however , the joint analysis based on the observational data of sn ia , cmb , and bao shows that the tracker solution is disfavored because of the large deviation of @xmath4 from @xmath9 during the matter era @xcite . the solutions that approach the tracker only at late times are allowed from the combined data analysis @xcite . as an extension of the covariant galileon model , deffayet _ et al . _ @xcite obtained the most general lagrangian in scalar - tensor theories with second - order equations of motion . in four dimensions the corresponding lagrangian is of the form ( [ lagsum ] ) with the four functions ( [ eachlag2])-([eachlag5 ] ) given below . in fact this is equivalent to the lagrangian found by horndeski @xcite more than 3 decades ago @xcite . the conditions for the avoidance of ghosts and laplacian instabilities were recently derived in ref . @xcite in the presence of two perfect fluids ( non - relativistic matter and radiation ) . the covariant galileon corresponds to the choice @xmath20 , @xmath21 , @xmath22 , @xmath23 in eqs . ( [ eachlag2])-([eachlag5 ] ) , where @xmath24 s are dimensionless constants , @xmath25 , @xmath26 is the reduced planck mass , and @xmath27 is a constant having the dimension of mass . kimura and yamamoto @xcite studied the model with the functions @xmath20 , @xmath28 ( @xmath29 ) , @xmath30 , and @xmath31 , in which case the dark energy equation of state during the matter era is given by @xmath0 with @xmath32 . at the background level this is equivalent to the dvali - turner model @xcite , which can be consistent with the observational data for @xmath33 larger than the order of 1 . if we consider the evolution of cosmological perturbations , the lss tends to be anti - correlated with the late - time isw effect . this places the tight bound on the power @xmath33 , as @xmath34 ( 95% cl ) @xcite , in which case the dark energy equation of state is practically indistinguishable from that in the @xmath3cdm model . in ref . @xcite the present authors proposed more general extended galileon models with the functions @xmath35 , @xmath36 , @xmath37 , and @xmath38 , where the masses @xmath39 s are fixed by the hubble parameter at the late - time de sitter solution with @xmath40constant . for the powers @xmath41 , @xmath42 , @xmath43 , @xmath44 , where @xmath45 and @xmath46 are positive constants , there exists a tracker solution characterized by @xmath47constant . during the matter - dominated epoch one has @xmath0 , where @xmath48 , along the tracker . this covers the model of kimura and yamamoto @xcite as a specific case ( @xmath49 , @xmath50 , @xmath51 , @xmath52 ) . in the presence of the nonlinear field self - interactions in @xmath53 and @xmath54 , the degeneracy of the background tracker solution for given values of @xmath45 and @xmath46 is broken by considering the evolution of cosmological perturbations . hence the isw - lss anti - correlation found in refs . @xcite for @xmath55 should not be necessarily present for the models with non - zero values of @xmath56 and @xmath57 . in this paper we first place constraints on the tracker solution in the extended galileon models by using the recent observational data of sn ia , cmb , and bao . the bound on the value @xmath48 is derived from the background cosmic expansion history with / without the cosmic curvature @xmath58 . we then study the evolution of cosmological density perturbations in the presence of non - relativistic matter to break the degeneracy of the tracker solution at the background level . we will show that the lss and the isw effect are either positively or negatively correlated , depending on the parameters @xmath56 and @xmath57 . this information should be useful to distinguish between the extended galileon models with different values of @xmath56 and @xmath57 from observations . we start with the following lagrangian @xmath59 where @xmath60\,,\\ { \cal l}_{5 } & = & g_{5}(x)\ , g_{\mu\nu}\,(\nabla^{\mu}\nabla^{\nu}\phi)-(g_{5,x}/6)[(\box\phi)^{3}-3(\box\phi)\,(\nabla_{\mu}\nabla_{\nu}\phi)\,(\nabla^{\mu}\nabla^{\nu}\phi)+2(\nabla^{\mu}\nabla_{\alpha}\phi)\,(\nabla^{\alpha}\nabla_{\beta}\phi)\,(\nabla^{\beta}\nabla_{\mu}\phi)]\,.\label{eachlag5}\end{aligned}\ ] ] @xmath58 and @xmath61 ( @xmath62 ) are functions in terms of the field kinetic energy @xmath25 with @xmath63 , @xmath12 is the ricci scalar , and @xmath64 is the einstein tensor . if we allow the @xmath14-dependence for the functions @xmath58 and @xmath61 as well , the lagrangian ( [ lagsum ] ) corresponds to the most general lagrangian in scalar - tensor theories @xcite . in order to discuss models relevant to dark energy we also take into account the perfect fluids of non - relativistic matter and radiation ( with the lagrangians @xmath65 and @xmath66 respectively ) , in which case the total 4-dimensional action is given by @xmath67 in the following we focus on the extended galileon models @xcite in which @xmath58 and @xmath61 are given by @xmath68 where @xmath26 is the reduced planck mass , @xmath24 and @xmath69 ( @xmath70 ) are dimensionless constants , and @xmath39 ( @xmath70 ) are constants having the dimension of mass . in the flat universe it was shown in ref . @xcite that tracker solutions characterized by the condition @xmath71 ( @xmath72 and a dot represents a derivative with respect to cosmic time @xmath73 ) are present for @xmath74 the covariant galileon @xcite corresponds to @xmath49 and @xmath75 , i.e. @xmath76 , @xmath77 . we will extend the analysis to the general friedmann - lematre - robertson - walker ( flrw ) background with the cosmic curvature @xmath58 : @xmath78\,,\label{metric}\ ] ] where @xmath79 is the scale factor . the closed , flat , and open geometries correspond to @xmath80 , @xmath81 , and @xmath82 , respectively . for the theories given by the action ( [ action ] ) the dynamical equations of motion are @xmath83 here @xmath84 , @xmath85 , @xmath86 and @xmath87 are the energy densities of non - relativistic matter and radiation respectively , and @xmath88 where @xmath89 . from eqs . ( [ be1 ] ) and ( [ be2 ] ) we find that there exists a de sitter solution characterized by @xmath90 and @xmath91 . in order to discuss the cosmological dynamics we introduce the dimensionless variables @xcite @xmath92^{\frac{p+2q}{1 + 2q}}\,,\qquad\omega_{r}\equiv\frac{\rho_{r}}{3h^{2}m_{{\rm pl}}^{2}}\,,\label{eq : defr1r2}\ ] ] where @xmath93 , and the subscript `` ds '' represents the quantities at the de sitter solution . we relate the masses @xmath39 ( @xmath94 ) in eq . ( [ gegali ] ) with @xmath95 , as @xmath96 , @xmath97 , @xmath98 , and @xmath99 . the existence of de sitter solutions demands that the coefficients @xmath100 and @xmath101 are related with @xmath56 and @xmath57 , via @xmath102\,,\label{c3}\ ] ] where @xmath103 the density parameter of dark energy , @xmath104 , can be expressed as @xmath105 - 3\alpha(2p+4q+1)\right\ } + 4\beta(p+3q+1)\bigr].\label{eq : omde}\ ] ] from eq . ( [ be1 ] ) it follows that @xmath106 , where @xmath107 and @xmath108 . the autonomous equations for @xmath109 , @xmath110 , and @xmath111 are written in terms of @xmath109 , @xmath110 , @xmath111 , @xmath112 , @xmath113 , @xmath45 , @xmath46 . as in the case of the flat universe @xcite one can show that there is a fixed point for the differential equation of @xmath109 characterized by @xmath114 from the definition of @xmath109 in eq . ( [ eq : defr1r2 ] ) this corresponds to the tracker solution where @xmath47constant . along the tracker the autonomous equations for @xmath110 and @xmath111 are @xmath115 where a prime represents the derivative with respect to @xmath116 . combining these equations , we obtain the integrated solution @xmath117 where @xmath118 is a constant . for the theoretical consistency the parameter @xmath1 is positive @xcite . since @xmath119 , the quantity @xmath110 grows toward the value 1 at the de sitter solution . along the tracker the density parameter ( [ eq : omde ] ) is given by @xmath120 where the subscripts `` 0 '' represent the values today ( the scale factor @xmath121 , i.e. @xmath122 ) . using the relation @xmath123 as well , eq . ( [ omereq ] ) reads @xmath124 this equation can be solved as @xmath125 which is nothing but the relation @xmath106 . from eq . ( [ eq : r2omr ] ) the dark energy density parameter evolves as @xmath126 and hence @xmath127}\,.\label{eq : hr2}\ ] ] since it is not generally possible to solve eq . ( [ eq : exacts ] ) for @xmath111 in terms of @xmath128 ( apart from some specific values of @xmath1 such as @xmath129 ) , we numerically integrate eq . ( [ eq : omrn ] ) and find the expression of @xmath130 by using eqs . ( [ eq : r2omr ] ) and ( [ eq : hr2 ] ) . versus the redshift @xmath131 for the open universe with @xmath132 for @xmath49 , @xmath133 , @xmath134 , @xmath135 . the solid curve corresponds to the tracker solution with the initial conditions @xmath136 , @xmath137 , @xmath138 , @xmath139 at @xmath140 . in the cases ( a ) and ( b ) the initial conditions are chosen to be ( a ) @xmath141 , @xmath142 , @xmath143 , @xmath144 at @xmath145 , and ( b ) @xmath146 , @xmath147 , @xmath148 , @xmath149 at @xmath150 , respectively.,width=316,height=316 ] [ wdefig ] along the tracker the dark energy equation of state @xmath151 and the effective equation of state @xmath152 are given by @xmath153 in the early cosmological epoch ( @xmath154 ) these reduce to @xmath155 and @xmath156 . during the matter era in which @xmath157 it follows that @xmath158 ( for @xmath159 ) and @xmath160 . at the de sitter fixed point ( @xmath161 ) with @xmath162 one has @xmath163 . in fig . [ wdefig ] we plot the evolution of @xmath4 versus the redshift @xmath164 in the open universe with @xmath132 for the model parameters @xmath49 , @xmath133 , @xmath134 , @xmath135 ( i.e. @xmath165 ) . the tracker is shown as a solid curve , along which @xmath4 changes as @xmath166 ( radiation era ) @xmath18 @xmath167 ( matter era ) @xmath18 @xmath9 ( de sitter era ) . the effect of the cosmic curvature @xmath168 becomes important only for the late universe , which affects the luminosity distance in the sn ia observations . if the solutions start from the regime @xmath169 , the evolution of @xmath4 is different from eq . ( [ wde ] ) before they reach the tracker . for @xmath169 and @xmath154 , @xmath4 and @xmath170 are approximately given by @xmath171 in fig . [ wdefig ] we show the variation of @xmath4 for @xmath49 , @xmath133 , @xmath134 , @xmath135 with two different initial conditions satisfying @xmath169 . in both cases the density parameter @xmath168 today is @xmath132 . the cases ( a ) and ( b ) correspond to the early and late trackings , respectively . for smaller initial values of @xmath109 the tracking occurs later . as estimated by eq . ( [ wde2 ] ) , @xmath4 starts from the value @xmath172 in the deep radiation era . if the solutions do not reach the tracker during the matter era ( as in the case ( b ) in fig . [ wdefig ] ) , @xmath4 temporally approaches the value @xmath173 . in ref . @xcite it was shown that the tracker for the covariant galileon ( @xmath129 ) is disfavored from observations , but the late - time tracking solution is allowed from the data . this property comes from the fact that for the late - time tracker the deviation of @xmath4 from @xmath9 is not significant . for @xmath174 even the tracker is expected to be allowed from observations . in such cases the solutions starting from the initial conditions with @xmath169 should be also compatible with the data ( because even for @xmath129 the late - time tracking solution is allowed ) . in the following sections we will focus on the tracker solution to discuss the background observational constraints and the evolution of cosmological perturbations . in this section we place observational constraints on the tracker solution from the background cosmic expansion history . we use three data sets : 1 ) the cmb shift parameters ( wmap7 ) @xcite ; 2 ) the bao ( sdss7 ) @xcite ; 3 ) and the sn ia ( constitution ) @xcite . the total chi - square @xmath175 for all three combined data sets will be calculated on a grid representing a chosen set of available parameters . we then find the minimum on this grid , and consequently find the 1@xmath176 and 2@xmath176 contours . in order to integrate eq . ( [ eq : omrn ] ) , once a set of model parameters ( in this case not only @xmath1 , but also @xmath177 , @xmath178 , and @xmath179 ) is given , we have the choice of one initial condition , that is @xmath180 . in principle it is possible to solve eq . ( [ eq : omrn ] ) backwards for a given value of @xmath181 , but we find that the integrated results are prone to numerical instabilities . therefore , it is more convenient to obtain the expression of @xmath182 for a given @xmath183 ( chosen to be @xmath184 , where the initial redshift is @xmath185 ) , which gives @xmath186 after solving the differential equation . for given @xmath177 and the other parameters we can use eq . ( [ eq : exacts ] ) to obtain the desired value of @xmath182 . at @xmath187 , ( [ eq : exacts ] ) is written as @xmath188 since @xmath189 during the radiation domination , it is possible to solve this equation iteratively by assuming that the l.h.s . is a small correction ( indeed the r.h.s . corresponds to @xmath190 ) . at 0-th order the solution of eq . ( [ eq : exa2 - 1 ] ) is given by @xmath191 at first order we find @xmath192^{1+s}}{1+(\omega_{k,0}/\omega_{r,0})\ , e^{2n_{i}}+(\omega_{m,0}/\omega_{r,0})\ , e^{n_{i}}}\,.\ ] ] this process can be iterated up to the desired precision . in the numerical code , we employ the solution derived after the three iterations , that is @xmath193 . since the late - time de sitter background is an attractor , small differences in the initial conditions do not lead to very different final solutions . therefore we indeed find that the three iterations are sufficient to derive the parameter @xmath177 accurately . in the following we first discuss the method for carrying out the likelihood analysis to confront the tracker solution with observations and then proceed to constrain the model parameters . we use the data of the cmb shift parameters provided by wmap7 , which are related with the positions of the cmb acoustic peaks . these quantities are affected by the cosmic expansion history from the decoupling epoch to today . the redshift at the decoupling is known by means of the fitting formula of hu and sugiyama @xcite @xmath194\,[1+g_{1}\,(\omega_{m,0}h^{2})^{g_{2}}]\,,\ ] ] where @xmath195 $ ] , @xmath196 $ ] , @xmath197 $ ] , and @xmath198 corresponds to the today s density parameter of baryons . the shift of the cmb acoustic peaks can be quantified by the two shift parameters @xcite @xmath199 where @xmath200 corresponds to the sound horizon at the decoupling , given by @xmath201\}}}\,.\ ] ] note that @xmath202 is the today s value of photon energy density and @xmath203 is the comoving angular distance to the last scattering surface defined by @xmath204 $ ] . the likelihood values of @xmath205 provided by the wmap7 data @xcite are @xmath206 , @xmath207 , and @xmath208 . the chi - square associated with this measurement is @xmath209 where the inverse covariance matrix is @xmath210 we also employ the data of bao measured by the sdss7 @xcite . the redshift @xmath211 at which the baryons are released from the compton drag of photons is given by the fitting formula of eisenstein and hu @xcite : @xmath212\,,\ ] ] where @xmath213 $ ] and @xmath214 . we define the effective bao distance @xmath215^{1/3}\,,\ ] ] where @xmath216 is the diameter distance given by @xmath217\,.\ ] ] the bao observations constrain the ratio @xmath218 at particular redshifts @xmath131 , where @xmath219 is the sound horizon for @xmath220 . at @xmath221 and @xmath222 the recent observational bounds are @xmath223 and @xmath224 . the chi - square associated with the bao is evaluated as @xmath225 where the inverse covariance matrix is @xcite @xmath226 finally we consider the experimental bounds coming from the observations of the sn ia standard candles . the apparent magnitudes , together with their absolute magnitudes , can be used to generate the following chi - square @xcite @xmath227 where @xmath228 are the errors on the data , and @xmath229 is the theoretical distance modulus defined as @xmath230+\mu_{0}\,.\ ] ] here @xmath231 and @xmath232 are given , respectively , by @xmath233 in the following we will make use of the constitution sn ia data sets provided in ref . @xcite ( see also ref . @xcite ) . we now define the total chi - square as @xmath234 in the following we reduce the numerical complexity by fixing several parameters , as @xmath235 , @xmath236 , @xmath237 , and @xmath238 @xcite , where the relativistic degrees of freedom are set to be @xmath239 . then two analysis will be performed for the tracker solution : 1 ) the flat case , @xmath240 , for which two free parameters , @xmath178 and @xmath1 , are left to be varied ; 2 ) the non - flat case , for which the additional free parameter , @xmath179 , is varied as well . later on , when we compare models with different number of free parameters , we will also make use of the akaike information criterion ( aic ) method ( see e.g. , @xcite ) . for each model the aic is defined as @xmath241 where @xmath242 is the number of free parameters in the model , and @xmath243 is the minimum value for @xmath244 in the chosen parameter space . the smaller the aic the better the model . to be more precise , if the difference of @xmath244 between two different models is in the range @xmath245 , the models are considered to be equivalent , whereas if @xmath246 , the data prefer one model with respect to the other . in this case we compute the @xmath244 on a grid in the intervals @xmath247 and @xmath248 . the minimum value of @xmath244 is found to be @xmath249 for the model parameters @xmath250 and @xmath251 . then we calculate the difference of @xmath244 at each grid point , that is , @xmath252 . when @xmath253 the chi - square distribution , with two free parameters , excludes the models with those values of @xmath244 at 68% confidence level ( 1@xmath176 ) , whereas when @xmath254 those models are excluded at 95% cl ( 2@xmath176 ) . our numerical results are plotted in fig . [ obs-1 ] . even if we use the gaussian likelihood function @xmath255 , we find that the observational contours are similar to those given in fig . [ obs-1 ] . the parameters @xmath1 and @xmath178 are constrained to be @xmath256 this shows that the tracker solution with @xmath257 during the matter era can be allowed observationally . the @xmath3cdm model corresponds to the line @xmath258 in fig . [ obs-1 ] , which , as expected , is inside the 1@xmath176 contour for @xmath259 ( with @xmath260 ) . the best - fit @xmath244 for the extended galileon model is slightly smaller than that in the @xmath3cdm model . however , since @xmath261 , the observational data do not particularly favor the extended galileon model over the @xmath3cdm model . ) plane for the flat universe ( @xmath240 ) . 1@xmath176 and 2@xmath176 contours correspond to the internal and external lines , respectively.,width=432,height=316 ] [ obs-1 ] in the presence of the cosmic curvature @xmath58 we also evaluate the @xmath244 in the parameter space given by @xmath247 , @xmath248 , and @xmath263 . the minimum value of @xmath244 is found on the point @xmath264 , @xmath265 , and @xmath266 , at which @xmath267 . we evaluate the difference @xmath268 on each grid point , according to which we can exclude : a ) the values of @xmath244 at 68% cl ( which , for a system with 3 parameters , corresponds to @xmath269 ) , and b ) the values of @xmath244 at 95% cl ( i.e. @xmath270 ) . in fig . [ obs-2 ] we plot a three dimensional region ( by showing only some slices of it ) of the allowed parameter space for the tracker solution . we find that the model parameters are constrained to be @xmath271 ) space with the cosmic curvature taken into account . 1@xmath176 and 2@xmath176 contours correspond to the black and red regions , respectively.,width=432,height=336 ] the aic for the best - fit extended galileon model is @xmath272 , whereas in the best - fit @xmath3cdm model with two parameters @xmath178 and @xmath179 the aic is found to be @xmath273 with @xmath274 . since @xmath275 between the two models , they are equivalently supported by the data . however , if we compare the non - flat extended galileon model with the flat @xmath3cdm ( 1 parameter only , @xmath178 ) , then we see that @xmath276 , which states that the flat @xmath3cdm is better supported by the data . according to this criterion , the flat extended galileon tracker is favored over the non - flat case from a statistical point of view . the property of the tracker solution does not depend on the parameters @xmath112 and @xmath113 at the level of the background cosmology . if we consider cosmological perturbations , it is possible to discriminate between the models with different values of @xmath112 and @xmath113 . in order to confront the extended galileon models with the observations of lss , cmb , and weak lensing , we shall study the evolution of matter density perturbations as well as gravitational potentials . since our interest is the growth of non - relativistic matter perturbations in the late universe , we do not take into account the radiation as far as the perturbations are concerned . let us consider the perturbed metric in the longitudinal gauge about the flat flrw background @xcite @xmath277 where @xmath278 and @xmath279 are scalar metric perturbations . we perturb the scalar field as @xmath280 , and non - relativistic matter as well , in terms of the matter density perturbation @xmath281 and the scalar part of the fluid velocity @xmath282 . the density contrast of non - relativistic matter is defined as @xmath283 . we also introduce the gauge - invariant density contrast @xmath284 the full linear perturbation equations in the horndeski s most general scalar - tensor theories were derived in ref . @xcite . for the extended galileon models , in fourier space , they are given by @xmath285 where @xmath286 is a comoving wave number , and @xmath287 the readers may refer to the paper @xcite for the explicit forms of the coefficients @xmath288 , and @xmath289 . since the perturbation equations ( [ eq : psi])-([eq : mat2 ] ) are not independent , we do not need to know these unwritten coefficients to solve the equations numerically . now we are dealing with a massless scalar field , so that the mass term @xmath27 does not appear in the perturbation equations . from eqs . ( [ eq : mat1 ] ) and ( [ eq : mat2 ] ) the gauge - invariant matter perturbation ( [ gaugein ] ) obeys @xmath290 where @xmath291 . we define the effective gravitational potential @xmath292 this quantity is related to the deviation of the light rays in cmb and weak lensing observations @xcite . to quantify the difference between the two gravitational potentials @xmath279 and @xmath278 , we also introduce @xmath293 by which eq . ( [ phieff ] ) can be written as @xmath294 . for the modes deep inside the hubble radius ( @xmath295 ) we can employ the quasi - static approximation under which the dominant contributions in the perturbation equations are those including @xmath296 and @xmath297 ( or @xmath298 ) @xcite . this approximation is known to be trustable as long as the oscillating mode of the field perturbation is negligible relative to the matter - induced mode . combining eqs . ( [ eq : psi ] ) , ( [ eq : chi ] ) , and ( [ eq : dpi ] ) under this approximation , it follows that @xcite @xmath299 where @xmath300 here we introduced the bare gravitational constant @xmath301 . substituting eq . ( [ poi ] ) into eq . ( [ delmeq ] ) with the relation @xmath302 ( valid for @xmath295 ) , we obtain @xmath303 where a prime represents a derivative with respect to @xmath116 . under the quasi - static approximation the quantity @xmath304 defined in eq . ( [ eta ] ) reads @xcite @xmath305 on using eq . ( [ poi ] ) , the effective gravitational potential @xmath294 yields @xmath306 where @xmath307 the @xmath3cdm model corresponds to @xmath308 , @xmath30 , @xmath309 , in which case @xmath310 , @xmath311 , and @xmath312 . then one has @xmath313 and @xmath314 from eqs . ( [ geff ] ) and ( [ xi ] ) . in the extended galileon models the general expressions of @xmath315 and @xmath316 are quite complicated . in what follows we shall focus on the evolution of cosmological perturbations for the tracker solution ( @xmath136 ) . in the early cosmological epoch ( @xmath154 ) and during the matter domination ( @xmath317 ) , @xmath315 and @xmath316 are approximately given by @xmath318r_{2}/\delta,\label{geffap}\\ \xi & \simeq & 1+[27p(2p-1)(3\alpha^{2}p+6\beta^{2}(2p-1)+\beta(1 + 3\alpha-2(1 + 6\alpha)p))+9(24\alpha^{2}p(4p-1)\nonumber \\ & & + 2\beta(2p-1)(\beta(78p+3)-11p)+\alpha(4p(2p-1)+\beta(3 + 12(10 - 29p)p)))q+2(-45(\alpha-2\beta)\beta\nonumber \\ & & + 6(96\alpha^{2}+\alpha(10 - 321\beta)+6\beta(43\beta-2))p+4(27\alpha-54\beta-4)p^{2}+32p^{3})q^{2}+8(9(3\alpha-10\beta)(\alpha-2\beta)\nonumber \\ & & -3(27\alpha-54\beta-26)(\alpha-2\beta)p+16(1 - 3\alpha+6\beta)p^{2})q^{3}+48(\alpha-2\beta)(66\beta-8(2 + 3\beta)p+3\alpha(4p-11))q^{4}\nonumber \\ & & + 1152(\alpha-2\beta)^{2}q^{5}]r_{2}/(2\delta),\label{xiap}\end{aligned}\ ] ] where @xmath319\,.\end{aligned}\ ] ] to derive eqs . ( [ geffap ] ) and ( [ xiap ] ) we set @xmath317 and performed the taylor expansion around @xmath320 . at the de sitter solution ( @xmath321 ) , @xmath322 and @xmath316 are simply given by @xmath323 the equality of @xmath315 and @xmath316 comes from the fact that @xmath324 at @xmath321 . let us first consider the theory where @xmath325 . from eqs . ( [ geffap])-([geffde ] ) it follows that @xmath326 for @xmath49 @xcite both @xmath315 and @xmath316 diverge at the de sitter solution . in this case one has @xmath327 in the regime @xmath154 , so that @xmath315 and @xmath316 are larger than 1 for @xmath72 . the property that @xmath316 is as large as @xmath315 leads to the enhancement of the effective gravitational potential relative to the matter perturbation @xmath298 normalized by @xmath8 @xcite . this gives rise to the anti - correlation between the late - time isw effect and the lss . then the parameter @xmath46 is constrained to be @xmath328 at the 95 % confidence level @xcite , which means that @xmath4 is very close to @xmath9 along the tracker . the situation is different for @xmath329 and @xmath330 . let us consider the models with @xmath49 and @xmath133 , i.e. @xmath165 , in which case the models are compatible with the observational constraints discussed in sec . [ obsercon ] . in the regime @xmath154 , ( [ geffap ] ) and ( [ xiap ] ) read @xmath331}{20(308 - 1536\alpha+3201\beta)}r_{2}\,.\label{xi1}\end{aligned}\ ] ] at the de sitter solution eq . ( [ geffde ] ) gives @xmath332 ) plane for @xmath49 and @xmath133 . in the blue region the conditions for the avoidance of ghosts and laplacian instabilities of scalar and tensor perturbations are satisfied . the right panel shows the enlarged version of the left panel in the @xmath333 plane . the borders correspond to ( a ) @xmath334 , ( b ) @xmath335 , ( c ) @xmath336 , and ( d ) @xmath337 , respectively . taken from ref . @xcite.,title="fig:",width=336,height=249 ] ) plane for @xmath49 and @xmath133 . in the blue region the conditions for the avoidance of ghosts and laplacian instabilities of scalar and tensor perturbations are satisfied . the right panel shows the enlarged version of the left panel in the @xmath333 plane . the borders correspond to ( a ) @xmath334 , ( b ) @xmath335 , ( c ) @xmath336 , and ( d ) @xmath337 , respectively . taken from ref . @xcite.,title="fig:",width=336,height=249 ] [ validfig ] in ref . @xcite the authors clarified the viable parameter region in which the ghosts and laplacian instabilities of scalar and tensor perturbations are absent . in fig . [ validfig ] we plot the allowed parameter space in the ( @xmath338 ) plane for @xmath49 and @xmath133 . the viable region is surrounded by the four borders : ( a ) @xmath334 , ( b ) @xmath335 , ( c ) @xmath336 , and ( d ) @xmath337 . the model with @xmath325 is on the border lines [ the intersection of the lines ( a ) and ( c ) ] , in which case both @xmath315 and @xmath316 diverge at the de sitter solution . and @xmath339 versus @xmath112 for @xmath49 and @xmath133 , respectively , along the tracker solution ( @xmath136 ) at @xmath340 . the blue regions illustrate the viable parameter spaces in which the parameters @xmath112 and @xmath113 belong to the blue region in fig . [ validfig ] . the borders ( a ) , ( b ) , ( c ) , ( d ) correspond to those given in fig . [ validfig].,title="fig:",width=336,height=259 ] and @xmath339 versus @xmath112 for @xmath49 and @xmath133 , respectively , along the tracker solution ( @xmath136 ) at @xmath340 . the blue regions illustrate the viable parameter spaces in which the parameters @xmath112 and @xmath113 belong to the blue region in fig . [ validfig ] . the borders ( a ) , ( b ) , ( c ) , ( d ) correspond to those given in fig . [ validfig].,title="fig:",width=336,height=259 ] [ earlyfig ] figure [ earlyfig ] illustrates the regions for the possible values of @xmath315 in eq . ( [ geff1 ] ) and @xmath316 in eq . ( [ xi1 ] ) for @xmath49 and @xmath133 at @xmath340 . the blue regions are surrounded by the borders ( a ) , ( b ) , ( c ) , ( d ) , which correspond to those given in fig . [ validfig ] respectively . the lines ( a ) in fig . [ earlyfig ] show @xmath315 and @xmath316 for @xmath334 , along which both @xmath315 and @xmath316 decrease for larger @xmath112 . one has @xmath341 for @xmath325 , in which case our numerical simulations show that there is an anti - correlation between @xmath342 and @xmath343 for the modes deep inside the horizon . when @xmath344 the variable @xmath316 is larger than @xmath315 on the line ( a ) for the same @xmath112 , so that the anti - correlation tends to be even stronger than that for @xmath345 . on the line ( b ) in fig . [ validfig ] , i.e. @xmath335 , one has @xmath346 and @xmath347 , which are independent of the values of @xmath112 . the fact that @xmath315 is larger than @xmath316 may imply the absence of the anti - correlation between @xmath342 and @xmath343 . in fact we will show in sec . [ numerics ] that the anti - correlation tends to disappear as the model parameters approach the border line ( b ) in fig . [ validfig ] . for given @xmath112 , as we move from @xmath334 to @xmath335 , @xmath315 increases whereas @xmath316 gets smaller . for @xmath348 the viable regions for @xmath315 and @xmath316 in fig . [ earlyfig ] are surrounded by the lines ( a ) , ( c ) , and @xmath349 , whereas for @xmath350 they are surrounded by the lines ( b ) , ( d ) , and @xmath351 . versus @xmath112 for @xmath49 and @xmath133 at the de sitter solution ( @xmath352 ) . note that the parameter @xmath316 is exactly the same as @xmath315 . @xmath315 diverges at @xmath334 , so that it is unbounded from above . the borders ( b ) , ( c ) , ( d ) correspond to those given in fig . [ validfig ] , respectively.,width=364,height=268 ] [ desfig ] in fig . [ desfig ] we plot @xmath353 $ ] for the late - time de sitter solution . both @xmath315 and @xmath316 diverge on the line ( a ) in fig . [ validfig ] , including the case @xmath325 . on the line ( b ) one has @xmath354 , so that the growth rate of matter perturbations is smaller than that in the @xmath3cdm model around the future de sitter solution . the allowed values of @xmath315 and @xmath316 exist in the wide regions in fig . [ desfig ] . the present epoch corresponds to the regime between @xmath154 and @xmath161 , so we need to resort to numerical simulations to estimate the growth rate of perturbations accurately . in order to recover the general relativistic behavior in the early cosmological epoch we choose the initial conditions @xmath365 , @xmath366 , @xmath367 , and @xmath368 , in which case @xmath369 , @xmath370 , and @xmath371 are known from eq . ( [ per1])-([per3 ] ) ( where the subscript `` @xmath372 '' represents the initial values ) . for non - zero initial values of @xmath373 the field perturbation oscillates at the early stage . for the initial conditions with @xmath374 the evolution of perturbations in the low - redshift regime is hardly affected by the oscillations . this situation is similar to that found in ref . @xcite for the model with @xmath325 . ( left ) and @xmath343 ( right ) versus the redshift @xmath131 for @xmath49 , @xmath133 , @xmath134 , @xmath375 along the tracker solution . note that @xmath342 and @xmath343 are normalized by their initial values , respectively . each curve corresponds to the evolution of perturbations for the wave numbers @xmath376 , @xmath377 , @xmath378 , and @xmath379 . the dotted curves show the results obtained under the quasi - static approximation on sub - horizon scales . in this case there is an anti - correlation between @xmath342 and @xmath343 for the modes @xmath380.,title="fig:",width=307,height=297 ] ( left ) and @xmath343 ( right ) versus the redshift @xmath131 for @xmath49 , @xmath133 , @xmath134 , @xmath375 along the tracker solution . note that @xmath342 and @xmath343 are normalized by their initial values , respectively . each curve corresponds to the evolution of perturbations for the wave numbers @xmath376 , @xmath377 , @xmath378 , and @xmath379 . the dotted curves show the results obtained under the quasi - static approximation on sub - horizon scales . in this case there is an anti - correlation between @xmath342 and @xmath343 for the modes @xmath380.,title="fig:",width=307,height=297 ] [ anfig ] , but for the model parameters @xmath49 , @xmath133 , @xmath134 , @xmath135 . in this case @xmath342 and @xmath343 are not anti - correlated.,title="fig:",width=307,height=297 ] , but for the model parameters @xmath49 , @xmath133 , @xmath134 , @xmath135 . in this case @xmath342 and @xmath343 are not anti - correlated.,title="fig:",width=307,height=297 ] [ pofig ] in fig . [ anfig ] we plot the evolution of @xmath342 and @xmath343 versus the redshift @xmath381 for @xmath49 , @xmath133 , @xmath134 , and @xmath375 with several different wave numbers . in this case the model parameters are close to the border line ( a ) in fig . [ validfig ] . the solid curve in fig . [ anfig ] corresponds to the simulation for the wave number @xmath382mpc@xmath383 , where the subscript `` 0 '' represents the today s values . the numerical results of @xmath342 and @xmath343 show good agreement with those derived under the quasi - static approximation on sub - horizon scales . for the model parameters used in fig . [ anfig ] the analytic estimates ( [ geff1 ] ) and ( [ xi1 ] ) give @xmath384 and @xmath385 in the regime @xmath154 , whereas from eq . ( [ geffde2 ] ) one has @xmath386 at the de sitter solution . from fig . [ anfig ] we find that @xmath342 is anti - correlated with @xmath343 for the modes @xmath387 . this property comes from the fact that the parameter @xmath316 is always larger than 1 , while @xmath315 is smaller than @xmath316 in the regime @xmath154 . in fig . [ anfig ] the growth rates of @xmath298 and @xmath343 decrease for smaller @xmath286 . the anti - correlation between @xmath342 and @xmath343 is present for the modes @xmath380 . figure [ pofig ] shows the evolution of @xmath342 and @xmath343 for @xmath49 , @xmath133 , @xmath134 , and @xmath135 with several different wave numbers . in this case @xmath113 is smaller than that used in fig . [ anfig ] . from eqs . ( [ geff1 ] ) and ( [ xi1 ] ) one has @xmath388 and @xmath347 for @xmath154 , whereas @xmath389 at the de sitter solution . compared to the case @xmath375 , @xmath315 and @xmath316 get larger and smaller , respectively , for the same value of @xmath110 ( @xmath390 ) . this leads to the suppression of the growth of @xmath343 . from fig . [ pofig ] we find that @xmath342 and @xmath343 are positively correlated for the mode @xmath376 . for the wave numbers @xmath387 the quasi - static approximation reproduces the numerical results in high accuracy . on larger scales both @xmath298 and @xmath343 evolve more slowly , so that @xmath342 and @xmath343 are also positively correlated . for the mode @xmath376 in the regime @xmath391 with the model parameters ( i ) @xmath49 , @xmath133 , @xmath134 , @xmath375 , ( ii ) @xmath49 , @xmath133 , @xmath134 , @xmath135 , and ( iii ) @xmath49 , @xmath133 , @xmath134 , @xmath392.,width=307,height=297 ] [ gammafig ] we define growth index @xmath393 of matter perturbations , as @xcite @xmath394 in the @xmath3cdm model @xmath395 for the redshift @xmath396 @xcite . in fig . [ gammafig ] we plot the evolution of @xmath393 for @xmath49 and @xmath133 with the wave number @xmath376 in three different cases : ( i ) @xmath134 , @xmath375 , ( ii ) @xmath134 , @xmath135 , and ( iii ) @xmath134 , @xmath392 . the numerical simulations for ( i ) and ( ii ) correspond to the model parameters used in figs . [ anfig ] and [ pofig ] , respectively , whereas the model parameter in the case ( iii ) is close to the border ( b ) in fig . [ validfig ] . in the cases ( i ) , ( ii ) , ( iii ) the effective gravitational couplings in the regime @xmath154 are @xmath397 , @xmath398 , @xmath399 , respectively , whereas at the de sitter solution @xmath400 , @xmath401 , @xmath402 , respectively . for smaller @xmath113 , @xmath322 is larger in the early epoch ( @xmath154 ) , so that the deviation of @xmath393 from the value 0.55 is more significant . on the other hand , @xmath322 at the de sitter solution gets smaller for decreasing @xmath113 , which leads to the approach to the value @xmath403 around @xmath404 . while the above discussion corresponds to the case @xmath165 , we have also studied the evolution of perturbations for different values of @xmath1 and found the similar properties to those discussed above . for @xmath49 the upper bound for the allowed parameter space in the @xmath405 plane is characterized by the line @xmath334 @xcite , around which the isw - lss anti - correlation is present for @xmath406 ( i.e. for @xmath1 constrained observationally at the background level ) . as @xmath113 gets smaller for a fixed @xmath407 , the anti - correlation tends to disappear . we also found that , for @xmath49 and @xmath406 , the growth index @xmath393 varies in the range @xmath408 for @xmath396 . the isw - lss anti - correlation as well the variation of @xmath393 allows us to discriminate between the extended galileon models and the @xmath3cdm . in this paper we have studied cosmological constraints on the extended galileon models of dark energy . for the functions ( [ gegali ] ) with the powers ( [ power ] ) there exist tracker solutions along which the field equation of state @xmath4 changes as @xmath409 ( radiation era ) @xmath18 @xmath410 ( matter era ) @xmath18 @xmath9 ( de sitter era ) , where @xmath411 . unlike the case of the covariant galileon ( @xmath129 ) , @xmath4 can be close to @xmath9 during the radiation and matter eras for @xmath412 . moreover , even for @xmath6 , there are viable model parameter spaces in which the ghosts and laplacian instabilities are absent . using the recent data of sn ia , cmb , and bao , we placed observational constraints on the background tracker solutions with @xmath159 . in the flat universe we found that the model parameters are constrained to be @xmath2 and @xmath413 ( 95% cl ) from the joint data analysis . the chi - square for the best - fit case ( @xmath414 ) is slightly smaller than that in @xmath3cdm . however the difference of the aic information criteria between the two models is @xmath261 , so that the extended galileon is not particularly favored over the @xmath3cdm . we also carried out the likelihood analysis in the presence of the cosmic curvature @xmath58 and obtained the bounds @xmath415 , @xmath416 , and @xmath417 ( 95% cl ) . the tracker for the covariant galileon ( @xmath129 ) is disfavored from the data , in which case only the late - time tracking solution is allowed observationally @xcite . the background quantities for the tracker are independent of the values of @xmath112 and @xmath113 . this means that the models with different @xmath112 and @xmath113 can not be distinguished from the observational constraints derived from the background cosmic expansion history . in order to break this degeneracy we studied the evolution of cosmological perturbations in the presence of non - relativistic matter for the flat flrw background . as shown in ref . @xcite , for @xmath325 , the matter density perturbation @xmath298 divided by the scale factor @xmath8 is anti - correlated with the effective gravitational potential @xmath343 for the modes relevant to the lss . this leads to the anti - correlation between the lss and the isw effect in cmb , so that the parameter @xmath1 for the tracker is severely constrained to be @xmath418 ( 95% cl ) . for the models with @xmath329 and @xmath330 , however , the correlation between @xmath342 and @xmath343 depends on the values of @xmath112 and @xmath113 . if @xmath49 and @xmath133 ( i.e. @xmath165 ) , for example , @xmath342 and @xmath343 tend to be positively correlated for the model parameters close to the border ( b ) in fig . [ validfig ] ( @xmath335 ) , whereas they show anti - correlations for @xmath112 and @xmath113 close to the border ( a ) ( @xmath334 ) . the typical examples of the positive and negative correlations are plotted in figs . [ pofig ] and [ anfig ] , respectively . the qualitative differences between these two cases can be understood by estimating the effective gravitational coupling @xmath322 and the quantity @xmath339 derived under the quasi - static approximation on sub - horizon scales . as the model parameters approach the border ( a ) in fig . [ validfig ] , @xmath316 gets larger while @xmath322 decreases , so that the anti - correlation between @xmath342 and @xmath343 tends to be stronger . we studied the evolution of perturbations for different values of @xmath1 in the range @xmath406 and found that the basic properties for the isw - lss correlation are similar to those discussed for @xmath165 . we also estimated the growth index @xmath393 of the matter perturbation and found that , for @xmath49 and @xmath406 , it typically varies in the range @xmath408 at the redshifts for @xmath396 . hence it is also possible to distinguish between the extended galileon models and the @xmath3cdm model from the galaxy clustering . however , we expect that the tightest observational bounds on the values @xmath112 and @xmath113 should come from the isw - lss correlation . we leave such observational constraints for future works . we thank savvas nesseris for providing us a related numerical code for the likelihood analysis of cmb , bao , and sn ia . the work of a.d.f . and s.t . was supported by the grant - in - aid for scientific research fund of the jsps nos . 10271 and 30318802 . also thanks financial support for the grant - in - aid for scientific research on innovative areas ( no . 21111006 ) . y. fujii , phys . d * 26 * , 2580 ( 1982 ) ; l. h. ford , phys . rev . d * 35 * , 2339 ( 1987 ) ; c. wetterich , nucl . phys b. * 302 * , 668 ( 1988 ) ; b. ratra and j. peebles , phys . rev d * 37 * , 321 ( 1988 ) ; t. chiba , n. sugiyama and t. nakamura , mon . not . 289 * , l5 ( 1997 ) ; p. g. ferreira and m. joyce , phys . rev . lett . * 79 * , 4740 ( 1997 ) ; r. r. caldwell , r. dave and p. j. steinhardt , phys . rev . * 80 * , 1582 ( 1998 ) . r. r. caldwell , phys . b * 545 * , 23 ( 2002 ) ; p. singh , m. sami and n. dadhich , phys . d * 68 * , 023522 ( 2003 ) ; m. p. dabrowski , t. stachowiak and m. szydlowski , phys . d * 68 * , 103519 ( 2003 ) ; m. sami and a. toporensky , mod . phys . lett . a * 19 * , 1509 ( 2004 ) ; z. k. guo , y. s. piao , x. m. zhang and y. z. zhang , phys . b * 608 * , 177 ( 2005 ) ; m. z. li , b. feng and x. m. zhang , jcap * 0512 * , 002 ( 2005 ) ; h. wei , r. g. cai and d. f. zeng , class . quant . grav . * 22 * , 3189 ( 2005 ) . e. j. copeland , m. sami , s. tsujikawa , int . j. mod . phys . * d15 * , 1753 - 1936 ( 2006 ) ; k. koyama , class . quant . * 24 * , r231-r253 ( 2007 ) ; r. durrer and r. maartens , gen . grav . * 40 * , 301 - 328 ( 2008 ) ; t. p. sotiriou and v. faraoni , rev . phys . * 82 * , 451 ( 2010 ) ; a. de felice and s. tsujikawa , living rev . * 13 * , 3 ( 2010 ) ; s. tsujikawa , lect . notes phys . * 800 * , 99 - 145 ( 2010 ) ; t. clifton , p. g. ferreira , a. padilla and c. skordis , arxiv:1106.2476 [ astro-ph.co ] . l. amendola , r. gannouji , d. polarski and s. tsujikawa , phys . d * 75 * , 083504 ( 2007 ) ; a. a. starobinsky , jetp lett . * 86 * , 157 ( 2007 ) ; s. a. appleby and r. a. battye , phys . b * 654 * , 7 ( 2007 ) ; e. v. linder , phys . rev . * d80 * , 123528 ( 2009 ) . e. babichev , c. deffayet and r. ziour , int . j. mod . d * 18 * , 2147 ( 2009 ) ; r. gannouji and m. sami , phys . * d82 * , 024011 ( 2010 ) ; e. babichev , c. deffayet and g. esposito - farese , phys . d * 84 * , 061502 ( 2011 ) ; n. kaloper , a. padilla and n. tanahashi , jhep * 1110 * , 148 ( 2011 ) ; a. de felice , r. kase and s. tsujikawa , arxiv:1111.5090 [ gr - qc ] ; r. kimura , t. kobayashi and k. yamamoto , arxiv:1111.6749 [ astro-ph.co ] . c. deffayet , g. esposito - farese and a. vikman , phys . rev . * d79 * , 084003 ( 2009 ) ; c. deffayet , s. deser and g. esposito - farese , phys . rev . d * 80 * , 064015 ( 2009 ) ; c. de rham and a. j. tolley , jcap * 1005 * , 015 ( 2010 ) . n. chow and j. khoury , phys . rev . * d80 * , 024037 ( 2009 ) ; f. p. silva and k. koyama , phys . rev . d * 80 * , 121301 ( 2009 ) ; t. kobayashi , h. tashiro and d. suzuki , phys . rev . d * 81 * , 063513 ( 2010 ) ; t. kobayashi , phys . rev . * d81 * , 103533 ( 2010 ) ; a. ali , r. gannouji and m. sami , phys . rev . * d82 * , 103015 ( 2010 ) ; d. f. mota , m. sandstad and t. zlosnik , jhep * 1012 * , 051 ( 2010 ) ; a. de felice and s. tsujikawa , jcap * 1007 * , 024 ( 2010 ) ; a. de felice , s. mukohyama and s. tsujikawa , phys . rev . * d82 * , 023524 ( 2010 ) ; k. hirano and z. komiya , arxiv:1012.5451 [ astro-ph.co ] ; o. pujolas , i. sawicki and a. vikman , arxiv:1103.5360 [ hep - th ] ; c. de rham and l. heisenberg , phys . rev . * d84 * , 043503 ( 2011 ) . t. kobayashi , m. yamaguchi and j. yokoyama , phys . * 105 * , 231302 ( 2010 ) ; c. burrage , c. de rham , d. seery and a. j. tolley , jcap * 1101 * , 014 ( 2011 ) ; s. mizuno and k. koyama , phys . rev . * d82 * , 103518 ( 2010 ) ; k. hinterbichler , m. trodden and d. wesley , phys . d * 82 * , 124018 ( 2010 ) ; p. creminelli , g. damico , m. musso , j. norena and e. trincherini , jcap * 1102 * , 006 ( 2011 ) ; a. naruko and m. sasaki , class . quant . * 28 * , 072001 ( 2011 ) ; a. de felice and s. tsujikawa , jcap * 1104 * , 029 ( 2011 ) ; t. kobayashi , m. yamaguchi and j. yokoyama , phys . rev . * d83 * , 103524 ( 2011 ) ; g. goon , k. hinterbichler and m. trodden , jcap * 1112 * , 004 ( 2011 ) ; x. gao and d. a. steer , arxiv:1107.2642 [ astro-ph.co ] ; a. de felice and s. tsujikawa , phys . rev . * d84 * , 083504 ( 2011 ) . j. r. bond , g. efstathiou and m. tegmark , mon . not . * 291 * , l33 ( 1997 ) ; y. wang and p. mukherjee , phys . rev . d * 76 * , 103533 ( 2007 ) ; h. li _ et al . , _ astrophys . j. * 683 * , l1 ( 2008 ) ; j. c. b. sanchez , s. nesseris and l. perivolaropoulos , jcap * 0911 * , 029 ( 2009 ) . a. a. starobinsky , jetp lett . * 68 * , 757 ( 1998 ) ; b. boisseau , g. esposito - farese , d. polarski and a. a. starobinsky , phys . 85 * , 2236 ( 2000 ) ; g. esposito - farese and d. polarski , phys . rev . * d63 * , 063504 ( 2001 ) ; s. tsujikawa , phys . d * 76 * , 023514 ( 2007 ) ; s. tsujikawa , k. uddin and r. tavakol , phys . d * 77 * , 043007 ( 2008 ) .

we present estimates in the hubbard and heisenberg models for the spectral weight in magnetic neutron scattering experiments on the cuprates . with the aid of spin - wave theory and the time dependent gutzwiller approximation we discuss how the spectral weight is distributed among the different channels and between high and low energies . in addition to the well known total moment sum rule we discuss sum rules for each component of the dynamical structure factor tensor which are peculiar for spin 1/2 systems . the various factors that reduce the spectral weight at the relevant energies are singled out and analyzed like : shielding factors , weight at electronic energies , multimagnon process etc . although about 10% 15% of the naively expected weight is detected in experiments after consideration of these factors the missing weight is within the experimental uncertainties . a large fraction of the spectral weight is hard to detect with present experimental conditions .

cond-mat0507131

magnetic neutron scattering ( mns ) in high temperature superconducting cuprates usually detects about @xmath1 of the spectral weight dictated by a naive application of sum rules . for example the total weight in a wide range of energy and momentum in a recent experiment@xcite in la@xmath2ba@xmath3cuo@xmath4 , with @xmath5 , is @xmath6 whereas in the insulating phase the naive expectation from sum rules is that one should find @xmath7 . it is usually argued that this value should be corrected for the hole destruction of moments by a @xmath8 factor which still leaves a large fraction of spectral weight undetected . this rises various problems in the interpretation of mns . for example it has been argued that the average of the dynamical susceptibility weighted by the fourier transform of the magnetic interaction can be used to estimate the energy involved in magnetic pairing and its temperature dependence.@xcite clearly to obtain an absolute estimate the spectral weight problem needs to be sorted out first . furthermore modeling the dynamical structure factor probed by mns becomes rather problematic since sensible theoretical models do satisfy sum rules . indeed any theoretical claim of intensity agreement with the measured dynamical structure factor in absolute units needs to explain how the sum rule is satisfied or why it is violated . this is even more stringent in spin only models for which neither @xmath8 factors nor shielding corrections apply . the purpose of this work is to explain this apparent discrepancy . we provide theoretical estimates of the various factors which correct the sum rule and estimate what fraction of the spectral weight is accessible to present day experimental conditions . theoretical estimates are performed in the antiferromagnetic ( afm ) phase using the heisenberg and the hubbard model combining spin - wave theory , numerical results and the time dependent gutzwiller approximation ( tdga)@xcite and in the doped phase in the hubbard model within the tdga . apart from the mentioned @xmath8 factor we discuss the so called `` shielding factors '' due to an incomplete formation of magnetic moments . we estimate the spectral weight loss to electronic transitions at energies too high to be detectable by present day inelastic magnetic neutron scattering experiments and also the weight in multimagnon processes which is either at too high energies or is so broad in energy and momentum that it is not detectable in unpolarized neutron scattering experiments . after consideration of all these factors we arrive to the conclusion that within the experimental uncertainties the sum rule is not violated ( which is reassuring ) on the other hand a major fraction of the spectral weight is very hard to detect with present experimental conditions . the outline of the paper is at follows . in sec . [ mns ] we shortly review the theory of magnetic neutron scattering and the relevant sum rules to fix notations . this section has also a pedagogical character . apart from the well known total moment sum rule we discuss sum rules for each component of the dynamical structure factor tensor which , to the best of our knowledge , have not been applied in the present context . we also highlight some simple experimental facts that are usually assumed as granted in experimental works , like domain averages ( sec . [ da ] ) , but often overlooked in theoretical works . in sec . [ hf ] we discuss the spectral weight distribution in the undoped case and in sec . [ ahf ] we discuss the doped case . we conclude in sec . we start with a short review of magnetic neutron scattering to fix notations and discuss the sum rules that are relevant to our problem . the magnetic neutron scattering cross section is given by:@xcite @xmath9 where @xmath10 is the debye - waller factor , @xmath11 ( @xmath12 ) , @xmath13 ( @xmath14 ) is the initial ( final ) wave vector of neutrons and @xmath15barn@xmath16 . the magnetic form factor is given by @xmath17 where @xmath18 is a wannier orbital and we defined the dynamical structure factor tensor , @xmath19 is the land @xmath20-factor . for free electrons @xmath21 . in a solid a different value may be appropriate which may also depend on direction . for example ref . quotes @xmath22 in the plane and @xmath23 perpendicular to the plane for a typical cuprate . for simplicity we take an isotropic @xmath20 unless otherwise specified . @xmath24 is the fourier transform of the @xmath25 component of the spin operator and @xmath26 is the partition function . the dynamical structure factor @xmath27 obeys detailed balance @xmath28 . we will discuss sum rules for effective models of the magnetic dynamics . because effective models restrict the hilbert space sum rules turn out to be model dependent and therefore should be applied with care on modeling a real system , as discussed below . our considerations are based on the two most popular models in this context , namely the heisenberg and the hubbard model , respectively . the so called total moment sum rule is usually formulated within a heisenberg model with spin @xmath29 and reads : @xmath30 this applies to a system where magnetic ions have one or several partially filled orbitals ( for example a rare earth ion with a partially filled @xmath31-shell ) and well formed magnetic moments i.e. when double occupancy of a given orbital is negligibly . within this work we will restrict to systems where ions have only one partially filled orbital and the system can be model with a one - band hubbard model : @xmath32 here @xmath33 ( @xmath34 ) destroys ( creates ) an electron with spin @xmath35 at site @xmath36 , and @xmath37 . @xmath38 is the on - site hubbard repulsion and both nearest ( @xmath39 ) and next - nearest ( @xmath40 ) neighbor hopping has been included . most of our considerations apply also to other models where ions have only one partially filled orbital per atom , like the usual three - band hubbard model for cuprates with cu @xmath41 and o @xmath42 orbitals . we define for later use the spin autocorrelation function and the zeroth moment of the diagonal components of the dynamical structure factor as : @xmath43 when one has one orbital per site ( i.e. a spin 1/2 system ) more stringent sum rules than eq . ( [ eq : m0 ] ) apply . indeed _ each component _ of the dynamical structure factor satisfies a separate sum rule . for example from eqs . ( [ eq : sw]),([eq : srw]),([eq : m0as ] ) one finds : @xmath44 where @xmath45 with @xmath46 and we used that @xmath47 . we also defined the thermal ( @xmath48 ) and spacial averages of the orbital occupancy:@xmath49 and double occupancy : @xmath50 analogous proofs hold for the other components : @xmath51 where we took @xmath52 . ( we will occasionally restore @xmath20 below when convenient for clarity ) . ( [ eq : m0a ] ) also follows from the fact that the quantization axis in eq . ( [ eq : m0z ] ) is arbitrary . ( [ eq : m0 ] ) is valid in the heisenberg model where @xmath53 and @xmath29 is arbitrary . ( [ eq : m0a ] ) in contrast is valid for @xmath54 systems but without restriction in @xmath55 . the factor @xmath56 in eq . ( [ eq : m0a ] ) is the probability to find an atom singly occupied and reflects the fact that doubly occupied or empty atoms do not produce magnetic scattering of neutrons reducing the total scattering cross section . this is some times called `` shielding factor '' . one can also prove that the total weight of the off - diagonal components of @xmath57 adds to zero . first notice that only the symmetric part of @xmath57 contributes to eq . ( [ eq : sigma ] ) : @xmath58.\end{aligned}\ ] ] using the lehmann representation eq . ( [ eq : sw ] ) one can show that : @xmath59\nonumber\\ & = & \delta_{\alpha\beta } \mu_{b}^2 ( n-2d)\end{aligned}\ ] ] which gives us a sum rule for each component of @xmath60 . we are not aware of references quoting the sum rule eq . ( [ eq : m0ab ] ) although the result is so simple that we doubt it is original . if @xmath61 is a good quantum number @xmath62 becomes diagonal with @xmath63@xcite in this case we can write the cross section as @xmath64.\nonumber\end{aligned}\ ] ] we will restrict to systems where this expression applies . in the presence of long - range magnetic order the system shows magnetic elastic scattering . we will consider phases in which spin rotational invariance is broken with order along the @xmath65 axis ( i.e. stripes , afm , etc . ) . we assume a magnetic unit cell with @xmath66 atoms ( @xmath67 for the afm state at half - filling ) at positions @xmath68 , with @xmath69 the cell position and @xmath70 the position of the atoms within the cell ( @xmath71 ) . the vectors @xmath69 form a bravais lattice . for such a magnetic structure we have : @xmath72 where the sum is over the magnetic reciprocal basis vectors and we defined : @xmath73 and the local site - dependent magnetization @xmath74 . it is convenient to define the fluctuation operator : @xmath75 with these definitions the longitudinal structure factor can be put as : @xmath76\end{aligned}\ ] ] where for simplicity we set @xmath77 and @xmath78 is kronecker s @xmath79 whereas @xmath80 is dirac s @xmath79 . the first term in the brackets describes bragg peaks . the weight of the peaks is given by the square of the fourier transform of the magnetization inside the magnetic unit cell eq . ( [ eq : mdq ] ) . the second term describes inelastic scattering . the inelastic part ( only ! ) of the dynamical structure factor is related to the dynamical susceptibility via the fluctuation dissipation theorem . for the case discussed above of a diagonal structure factor tensor , the cross section involves the factors @xmath81 . these are polarization factors for scattering with unpolarized neutrons and are rooted in the dipolar interaction between neutrons and the electron spin . in experimental works often the polarization factors @xmath81 are included in the definition of the dynamical structure factor and an average over the orientation of domains , @xmath82 , is done : @xmath83 for a paramagnet all directions are equivalent , @xmath84 does not depend of @xmath25 and @xmath85 . ordered systems will be characterized by an order parameter that is a vector like the staggered magnetization . in general a real sample will consist of domains with different orientations of the order parameter . for a distribution of orientations that is completely isotropic in spin space @xmath86 . consider now scattering from a quasi two - dimensional ( 2d ) system with the @xmath87 direction defined perpendicular to the plane and the @xmath88 and @xmath89 directions in the plane . in quasi 2d - systems the most common experimental configuration is that the planes are perpendicular to the incident neutron beam and , depending on the energy , the component of @xmath90 perpendicular to the plane , @xmath91 may be larger than the components in the plane . in the extreme case that @xmath92 one can put [ c.f . ( [ eq : seff ] ) ] : @xmath93 this is valid regardless of the domain distribution . in order to evaluate this expression it is convenient to use a domain - dependent reference system in which the @xmath65 axis follows the ordered moment of the domain . in the case of an _ isotropic _ distribution of domains one recovers @xmath94 . that is : @xmath95+\frac23 s^{zz}(\bm{q},\omega)\ ] ] where we have grouped the transverse contribution . on the other hand , one usually deals with systems that have easy planes or easy axes . in cuprates for example the ordered moment is usually in the cu - o plane or close to that plane . for a distribution of the ordered moment which is isotropic _ within the plane _ and in the above scattering geometry @xmath96 + s^{zz}(\bm{q},\omega).\ ] ] the 1/2 factor is due to the fact that one transverse mode is perpendicular to the plane and becomes silent in this configuration . for reasons that will be clear below real experiments detect mainly the transverse structure factor . comparing eqs . ( [ eq : siso ] ) , ( [ eq : seffqz ] ) we see that in order to accurately estimate spectral weights from experiment precise information is needed on the domain distribution since the transverse component appears with different weight . the condition @xmath92 is rather extreme . instead , in the present scattering geometry , @xmath97 , @xmath98 will gradually decrease as the energy is increased . this will make the factors weighting the transverse and the longitudinal part to depend on energy . due to the larger sensibility to the transverse part in a real experiment this will lead to an apparent loss of spectral weight as the energy increases . in an ideal experiment where all components are detected with equal sensitivity one will not see a loss of spectral weight but rather a transfer of spectral weight from the longitudinal to the transverse part as the energy increases . indeed @xmath99 satisfies the sum rule @xmath100 independently of what the distribution of domains or the scattering geometry is . on analyzing experimental data one should take into account that unpolarized ins experiments magnetic and nonmagnetic scattering can not be unambiguously identified . in the insulating phase this problem can be reduced by fitting the data with a simple model like spin - wave theory . due to the lack of a simple theory in the doped phase a similar procedure is not possible . in this case the usual experimental practice is to report as `` magnetic scattering '' only those features which satisfy certain criteria like being reasonably sharp in momentum space and show a `` magnetic like '' temperature dependence and form factor dependence across different brillouin zones . all the rest is assumed to be background of unknown origin . below we analyze the distribution of weight in the insulator to get a hint of the distribution of spectral weight in the different channels and where to expect sharp and where broad features . we start by neglecting the shielding factors and estimate the spectral weights in the heisenberg model within spin - wave theory ( swt ) . we will show below how to apply these results when shielding factors are important corrections . since we are interested in gross distributions of spectral weight for simplicity we consider a heisenberg model with nearest neighbor interactions and neglect other terms like four - ring exchange . in order to write the longitudinal dynamical structure factor we introduce the following notations . for an afm the @xmath101 operator can be written within swt as : @xmath102 where @xmath103 is the number operator for holstein - primakoff bosons and @xmath104 taking the lattice constant @xmath105 . for the fourier transform @xmath106 we have : @xmath107 it is convenient to write @xmath108 with @xmath109 and define @xmath110 as the reduced sublattice magnetization . here @xmath111\ ] ] is the reduction in the ordered spin moment due to zero - point quantum fluctuations and @xmath112/2 $ ] comes from the fourier - transform of the exchanges in the square lattice . the one - magnon dispersion relation is given by : @xmath113 where @xmath114 is the superexchange constant and @xmath115 is a quantum renormalization of the one - magnon energy , near - constant over the brillouin zone in a first approximation,@xcite and estimated as 1.18 to order @xmath116 in spin - wave theory@xcite or by series expansions.@xcite with these definitions we can write the longitudinal dynamical structure factor as : @xmath117 , \nonumber\end{aligned}\ ] ] where the first term in the brackets is the bragg elastic contribution of the nel order and the second term is the inelastic contribution . the sublattice magnetization is well known to be accurately given by linear swt,@xcite therefore we do not expect significant changes on the elastic intensity if higher order corrections are included . in contrast how the inelastic part is distributed at low - energies ( a few @xmath114 s ) is expected to be sensitive to such corrections . the inelastic part can be decomposed into a sum over inelastic processes out of which the dominant term corresponds to two - magnon scattering events with intensity given by@xcite @xmath118 the two - magnons have opposite spin @xmath119 and @xmath120 such that the total spin @xmath121 is unchanged . the scattering cross - section @xmath122 depends on the wavevectors of the two magnons via@xcite @xmath123 . \label{eq : f2m}\ ] ] we have included an _ ad hoc _ intensity renormalization factor @xmath124 to be discussed below . the lineshapes in eq . ( [ eq:2 m ] ) can be evaluated by direct summation or monte - carlo methods and the result is illustrated by the dotted areas in fig . [ fig:2m_sqw ] . two - magnon events contribute a continuum band of scattering at energies above the one magnon dispersion @xmath125 . in general the intensity decreases with increasing energy and it is strongest for wavevectors near the antiferromagnetic zone center @xmath126 . fig . [ fig:2m_sofq ] shows how the energy - integrated intensity varies in the brillouin zone , generally following the same trend as the one - magnon intensity . the two - magnon signal cancels at the nuclear zone center @xmath127 as expected from first - moment sum rules for an isotropic antiferromagnet , also made apparent from eq . ( [ eq : f2 m ] ) where the cross - section for creating two identical magnons cancels , i.e. for @xmath128 such that @xmath129 , @xmath130 . [ fig:2mdos ] shows the wavevector - integrated intensity : it increases with increasing energy , linearly at low energies , then reaches a maximum at the one - magnon zone boundary and then decreases to zero at energies above twice the maximum magnon energy . in fact most of the two - magnon scatterig weight is at energies above the one - magnon cutoff @xmath131 . due to the broadness of the distribution the two - magnon spectral weight will be quite hard to detect in unpolarized neutron experiments , although we note that recent neutron experiments@xcite have observed such a high - energy continuum of excitations in the square - lattice spin-@xmath132 heisenberg antiferromagnet rb@xmath133mnf@xmath4 and the observed intensities were in agreement with neutron scattering by pairs of magnons as described by spin - wave theory . in cuprates , in addition , a large fraction of the spectral weight is at energies that are too high for neutron scattering . integrating eq . ( [ eq:2 m ] ) over energy and wavevector in a brillouin zone gives @xmath134 , whereas the bragg elastic scattering is @xmath135 . the magnetization reduction can be evaluated numerically as @xmath136 . the corresponding value of @xmath137 is shown in table [ tab : sw1 ] . comparing with the ising afm , also shown in table [ tab : sw1 ] , one sees that the effect of the transverse fluctuations is to reduce the sublattice magnetization with a concomitant transfer of spectral weight from the bragg peak to the two - magnon continuum . most of the longitudinal scattering is in this spread - out two - magnon continuum . the spin - only sum rule @xmath138 is strongly violated if intensity renormalization is neglected ( @xmath139 ) . notice that if transverse fluctuations are neglected ( ising limit ) the sum rule is exactly satisfied . we can anticipate a similar result in the hubbard model . including higher - order terms in spin - wave theory is expected to produce : _ i _ ) intensity - lowering of the two - magnon response @xcite and _ ii _ ) spread out at higher energies due to the contribution of higher multimagnon process.@xcite for simplicity we assume the intensity renormalization of the two - magnon response to be momentum independent and quantify it with the constant @xmath124 . since the bragg intensity is given accurately by swt one can get an upper bound for @xmath124 by neglecting effect _ ii _ ) and requiring that the sum rule eq . ( [ eq : m0z ] ) is fulfilled by bragg and two - magnon processes : @xmath140 . perturbative computations@xcite suggest a value not far from that suggesting that the weight in four magnon and higher multimagnon processes is not high , in agreement with numerical data.@xcite in the following to improve the values of the two - magnon intensity we tentatively adopt @xmath141 given the lack of better estimates ( table [ tab : sw1 ] ) . it is interesting to remark that the small weight at four magnon and higher multimagnon processes is in striking contrast to the response relevant for infrared experiments which show large weights instead.@xcite from a theoretical point of view , whereas magnon - magnon interactions have dramatic consequences in the shape of the spectrum of two - magnon excitation relevant for raman and ir data@xcite the effect on the two - magnon line shape relevant for magnetic neutron scattering appears much modest.@xcite this different role of interactions can be traced back to the fact that optical data probes two magnons in different sublattices whereas magnetic neutron scattering probes two magnons on the same sublattice and the interaction is dominant in the first case . the corresponding green functions at the rpa level for both cases are reported in ref . . in optical data magnon - magnon interactions are responsible for a shift of the main feature from @xmath142 to @xmath143.@xcite this effect was obtained in ref . using a high - energy approximation . one can easily see from the rpa equations for the equal - sublattice green function@xcite that at the same level of approximation the same effect does not show up in the two - magnon neutron scattering line shape . we conclude that interaction effects should be important for the intensity renormalization but they play a very different role here than the one played in optical data . .[tab : sw1 ] sublattice magnetization @xmath144 and dynamical structure factor spectral weights in units of @xmath145 . we show exact values for the ising afm and swt values for the heisenberg afm in the longitudinal channel ( @xmath65 ) and in one transverse channel ( @xmath146 ) . swt are the values neglecting all renomalization factors ( @xmath147 ) . we also include the swt intensity renormalized by the indicated @xmath26 values . the transverse contribution is split in the contribution from the magnetic zone and the nuclear zone as @xmath148 . since we are considering the heisenberg model ( @xmath53 ) a `` 1 '' in the last column implies that the sum rule is exactly satisfied within the model . [ cols="^,^,^,^,^,^",options="header " , ] the transverse dynamical structure factor is dominated by one - magnon scattering events with intensity given by@xcite @xmath149 @xmath150 is an intensity - lowering renormalization factor of the one - magnon cross - section due to zero - point fluctuations and magnon - magnon interactions , both neglected at first order in spin - wave theory ( @xmath151 , where @xmath152 is the renormalization of the transverse magnetic susceptibility , @xmath153 ) . the wavevector dependence of the one - magnon intensity is shown in fig . [ fig:2m_sofq ] . although wave - vectors @xmath154 and @xmath126 are related by symmetry the spectral weight goes to zero at wave vector @xmath127 and diverges at @xmath126 . this difference in intensity is reflected also in doped phases . integrating eq . ( [ eq : sxaf ] ) over energy and wavevector in a brillouin zone gives the one - magnon intensity for one transverse direction as @xmath155 . experimental works often restrict to the magnetic brillouin zone around @xmath126 ( dashed line in the inset of fig . [ fig:2m_sofq ] ) hereafter `` the magnetic zone '' and neglect the small weight on the magnetic brillouin zone around @xmath156 hereafter `` the nuclear zone '' . for comparison we split the one - magnon spectral weight in the two zones @xmath157 . one finds the values reported in table [ tab : sw1 ] . in this case as in the longitudinal channel the sum rule is overestimated if one neglects the intensity renormalization . estimates of the renormalization from higher - order spin - wave theory give@xcite @xmath158 at order @xmath116 , or 0.61(4 ) by series expansions.@xcite now the sum rule is underestimated . the lacking weight is expected to lay in three - magnon and higher multimagnon processes . an alternative way to estimate @xmath150 is to enforce eq . ( [ eq : m0 ] ) at large @xmath29 and extrapolate to @xmath54 . with the choice @xmath159 , the total sum rule @xmath160 [ c.f . ( [ eq : m0 ] ) ] is exhausted by elastic @xmath161 , one - magnon @xmath162 , and two - magnon scattering @xmath163 ( without renormalization ) , in units of @xmath164 . such a renormalization of the one - magnon intensity is very close to the results quoted above . strictly speaking , this argument applies only for @xmath165 since the sum rule for each channel eq . ( [ eq : m0a ] ) , which is exclusive of spin 1/2 systems , is still strongly violated . one the other hand the rapid convergence of the @xmath166 expansions often produces accurate results for @xmath54 systems as seem to be the case for @xmath167 . we now turn to the more realistic hubbard model . in this section we estimate the shielding factor for @xmath168 , within the 2d single - band hubbard model eq . ( [ eq : hubbard ] ) , to gain some insight on its impact on the sum rule eq . ( [ eq : m0ab ] ) . here we take for simplicity @xmath169 and consider the effect of varying @xmath170 . parameters more specific for the cuprates will be considered in sec . [ ahf ] . site cluster are computed with the double occupancies given in ref . . the apparent lack of extrapolation of the exact results to the non - interacting limit at @xmath171 is due to a finite size effect.,width=302 ] the shielding factor is given by the single occupancy probability @xmath172 . for a non interacting system , i.e. @xmath171 , the double occupancy is just @xmath173 so one gets @xmath174 . indeed half of the time an atom is singly occupied ( up or down ) and half of the time it is either empty or doubly occupied and hence does not produce scattering . for a paramagnetic state in the hartree - fock approximation ( hf ) one gets the same result independently of @xmath170 ( fig . [ fig:1m2d ] ) since correlations are neglected . correlations can be introduced by treating the paramagnet in the gutzwiller approximation ( ga)@xcite which leads to a reduction in double occupancy as @xmath175 becoming zero at the brinkman - rice transition point @xmath176.@xcite the corresponding value of @xmath172 is plotted also in fig . [ fig:1m2d ] . the brinkman - rice point , however , is never reached since for infinitesimal @xmath38 the paramagnetic state is unstable towards a spin - density wave ( sdw ) which can also be treated in the ga and gives the value of @xmath172 shown in fig . [ fig:1m2d ] . finally for large @xmath170 one can use a canonical transformation to map the hubbard model to a heisenberg model.@xcite it is important to realize that the energies of the low - energy excitations of both models , the `` physical '' hubbard model and the low - energy `` effective '' model coincide to leading order but the correlations functions in general do not coincide . in order to get `` physical '' correlation functions one needs to use the inverse canonical transformation to transform back the `` effective '' ground state wave function to a physical wave function . indeed within the heisenberg model the double occupancy is zero but this does not mean that the double occupancy is zero in the `` physical '' model . this is obviously very important on evaluating the right hand side of eq . ( [ eq : m0a ] ) . of course if we are interested on evaluating the sum rule _ within _ the heisenberg model it is legitimate to take @xmath53 , as done above , but if we want to compare with experiments ( or with the `` physical '' model ) one needs to compute the `` physical '' double occupancy . fortunately this is very easy in the present case because we can use a trick to avoid the back transformation . we use hellman - feynman theorem to write the double occupancy in the hubbard model as : @xmath177 with @xmath178 the ground state energy . for the latter we use the fact that for large @xmath170 it coincides with the energy of the heisenberg model where very accurate numerical estimates exist : @xmath179 and the best estimate for @xmath25 is @xmath180.@xcite here we have substituted the superexchange constant by its definition @xmath181 . since the heisenberg model has only one parameter it is clear that @xmath25 does not depend on @xmath114 and we can perform the derivative to obtain : @xmath182 the corresponding value of the single occupancy probability @xmath172 is plotted in fig . [ fig:1m2d ] and referred to as the `` @xmath183 expansion '' . this value is asymptotically exact ( within numerically accuracy@xcite ) in the large @xmath170 limit and therefore takes into account all fluctuations effects . clearly the sdw treated within the ga approximations gives a fairly good approximation for @xmath172 at large @xmath38 and interpolates smoothly to the exact result at @xmath171 so we expect it to be quite accurate in all the range of @xmath170 as can be seen also by comparing with the exact results in a @xmath184 cluster ( after ref . ) . we see that the sum rule for one diagonal component of the dynamical structure factor eq . ( [ eq : m0a ] ) changes smoothly from @xmath185 in the non - interacting case to @xmath145 in the limit of @xmath186 . it is reduced by @xmath187 ( table [ tab : hub ] ) with respect to the full moment value for @xmath188 , as relevant for cuprates.@xcite in the previous section we have evaluated the shielding factors that appear due to the finiteness of @xmath38 and reduce the total spectral weight compared to spin only models . another effect , which we study here , is that the total weight is split into a low - energy part at energies of order @xmath114 and a high - energy part at energies of order @xmath38 . present neutron scattering facilities can measure the spectra up to energies of the order of a few tenth of ev and therefore only the first part is detected . in addition we discuss how the dynamical structure factor is modified for finite @xmath38 respect to the swt result . to estimate the dynamical structure factor we use the time dependent gutzwiller approximation of refs . [ also called ga plus random - phase - approximation ( ga+rpa ) ] applied to a sdw state . we start in the next section by showing how measurements of the dispersion relation in the insulator can be used to estimate @xmath38 and @xmath189 . cuo@xmath4 after ref . and the ga+rpa result for @xmath190 and @xmath191mev . the inset shows that for @xmath192 the fit is noticeably worse . , width=302 ] in the transverse channel mns experiments reveal the spin - wave excitations of the afm.@xcite interestingly , whereas spin - wave theory on the heisenberg model eq . ( [ eq : wdqswt ] ) predicts a flat dispersion between @xmath193 and @xmath194 a substantial dispersion has been measured ( c.f . [ fig : edq ] ) . it has been argued that in cuprates corrections to the heisenberg model arising as higher orders in a @xmath183 expansion are relevant.@xcite the most important of such corrections is a term which cyclically exchanges four spins on a plaquette . a sizable value for this term has been revealed by analyzing phonon - assisted multimagnon infrared absorption@xcite and the dispersion relation@xcite shown in fig . [ fig : edq ] . in particular the dispersion between @xmath193 and @xmath194 is mainly due to this term . since the dispersion has its origin in the finiteness of @xmath183 it should show up in the transverse excitations of the hubbard model . the computation done in ga+rpa is also shown in fig . [ fig : edq ] . one obtains a very good fit of the dispersion and this provides an accurate way to estimate the strength of the repulsion . we find @xmath190 in good agreement with other estimates@xcite whereas @xmath192 gives a too flat dispersion between @xmath193 and @xmath194 ( inset ) . an equally good fit as the one show in the main panel can be achieved with @xmath190 , @xmath195 and @xmath196mev . the value of @xmath197 plays an important role in the doped phase@xcite and is close to a first principle estimate.@xcite evaluated for the hubbard model within ga+rpa for @xmath190 ( solid line ) and @xmath192 ( dashed line ) with @xmath198 for an @xmath199 sites system . the inset shows the high energy contribution of @xmath200 at energies @xmath201 . , width=302 ] in fig . [ fig : swt ] we show the momentum integrated spectral function . we see that for large @xmath170 it approaches the swt form ( fig . [ fig:2mdos ] ) . for smaller @xmath170 the maximum is not at the upper edge due to the modified dispersion relation . a small portion of the spectral weight is at high - energy due to spin - flip transitions from the lower to the upper hubbard band as shown in the inset . we will show that this effect is much larger in the longitudinal channel . integrating the spectral intensity one gets the values of @xmath202 shown in table [ tab : hub ] . ga+rpa interpolates between the extreme limits @xmath203 ( where the sum rule is exactly obeyed ) and @xmath204 where one recovers the results of linear spin - wave theory . as a consequence the sum rule in the transverse channel is increasingly more violated as @xmath170 increases with @xmath202 reaching values similar to the unrenormalized swt results ( table [ tab : sw1 ] ) for large @xmath38 . in fig . [ fig : idq ] we show the experimental intensities as a function of wave vector together with the ga+rpa result . in order to fit the experimental results we introduced an intensity renormalization @xmath205 analogous to the swt intensity renormalization factor . with @xmath206 the ga+rpa intensities are essentially equivalent to the intensities reported in fig . 3b of ref . which have been renormalized by @xmath207 _ respect to lswt_. this increase in the value of @xmath208 reflects the fact that cuprates are not in the strict heisenberg limit @xmath209 . in that limit the @xmath167 for the rpa should coincide with the value of lswt given the equivalence of the two approximations . in the extreme case of a non interacting system rpa gives exact intensities and therefore @xmath210 . as @xmath38 is decreased from infinity the rpa becomes gradually more accurate and therefore @xmath208 should increase respect to the lswt value as we indeed find . we caution that the fact that a system is not in the strict heisenberg limit does not mean by itself that it can not be described by the heisenberg model . corrections to the intensity can be introduced as explained in sec . [ sec : effect - shield - fact ] . we estimate this approach is accurate up to @xmath211 where the dispersion is still heisenberg like . for smaller @xmath38 corrections in the dispersion relation due to four spin exchange and other non heisenberg processes become obvious ( c.f . [ fig : edq ] ) . cuo@xmath4 after ref . and the ga+rpa result for @xmath190 and @xmath191mev . the ga+rpa intensity has been renormalized by a factor @xmath212 and coincides with the spin - wave result ( shown in fig . 3b of ref . ) renormalized by @xmath207 in eq . ( [ eq : sxaf ] ) . , width=302 ] the longitudinal structure factor has a bragg part and an inelastic part [ c.f . ( [ eq : szafh ] ) ] . for the sdw only one magnetic reciprocal lattice vector contributes to the sum in eq . ( [ eq : szafh ] ) namely @xmath213 ( we take the lattice constant @xmath105 ) . the bragg weight is determined by the sublattice magnetization @xmath144 . at lowest order rpa introduces longitudinal fluctuations in @xmath214 but does not correct one - body expectation values like the sublattice magnetization . ( on the contrary it introduces a correction in two - body expectation values @xcite like @xmath55 ) . therefore in the following we consider the sublattice magnetization at the ga level together with the effect of the longitudinal fluctuations at high energies in @xmath214 . the sublattice magnetization can in principle be corrected by including transverse fluctuations . we will discuss this below . at the ga we obtain the values of the sublattice magnetization shown in table [ tab : hub ] . these values lack the transverse fluctuation corrections and hence converge to the classical value , 0.5 in the large @xmath38 limit . to help intuition it is useful to distinguish between the permanent moment @xmath215 , which determines the bragg weight , and an `` intrinsic '' moment which determines the shielding factor and which we define as @xmath216 . ( notice that with this definition a free fermion system has an intrinsic moment which we denote as `` trivial '' ) . the paramagnet with small @xmath55 mentioned in sec . [ sec:1m2d ] can be considered as a system with well formed ( non - trivial ) intrinsic moments that are purely dynamical . indeed the shielding factor is close to 1 but there are no bragg peaks . in this regard the ga is more flexible than hf which is not able to produce non - trivial intrinsic moments that are not permanent . permanent moments are due to the breaking of spin - rotational symmetry . the ga for the sdw smoothly interpolates between the itinerant limit at small @xmath38 and the localized limit at large @xmath38 with a permanent moment that increases and reaches @xmath217 for @xmath218 . contrary to the paramagnetic case , quoted above , the difference between permanent and the intrinsic moment in the ga tends to be very small at large @xmath38 . thus despite the ability of the ga to distinguish the two kinds of moments this ability is not effective at large @xmath38 . in this case the permanent moment in the ga ( c.f . table [ tab : hub ] ) is reduced with respect to the fully polarized value mainly due to covalency effects . on top of this the moment will be reduced due to transverse fluctuations as discussed below . dc|cccccc u & @xmath25 & @xmath144 & & & @xmath219&@xmath172 + & & & & low & @xmath220 & & + 8 & @xmath65 & 0.43 & 0.74 & 0 & 0.15 & 0.898 & 0.889 + & @xmath146 & & 0 & 1.1 & 0.01 & 1.11 & + 10 & @xmath65 & 0.456 & 0.83 & 0 & 0.096 & 0.930 & 0.923 + & @xmath146 & & 0 & 1.196 & 0.009 & 1.205 & + 15 & @xmath65 & 0.481 & 0.927 & 0 & 0.0397 & 0.966 & 0.965 + & @xmath146 & & 0 & 1.278 & 0.002 & 1.28 & + the covalency reduced bragg peak does not exhaust the sum rule as can be seen by comparing the `` elastic '' column with the last column in table [ tab : hub ] . longitudinal fluctuations , captured by the ga+rpa approach , produce weight at energies of order @xmath38 as shown in fig . [ fig : sw ] . this is not related to two - magnon processes ( as in sec . [ sec : longitudinal - swt ] ) but to fluctuations in the length of the intrinsic moment . the weight in the high - energy part @xmath221 is also reported in table [ tab : hub ] . adding the bragg and the @xmath222 contribution one obtains the values of @xmath137 reported in table [ tab : hub ] . comparing with the last column we see that at this level of approximation the sum rule is exhausted ( actually slightly overshot ) by the bragg contribution without transverse fluctuations reductions plus the contribution at energy of order @xmath38 . the small overshot in the sum rule is due to the fact that the last column is computed at the ga level without rpa corrections , whereas @xmath137 has rpa corrections to @xmath55 . ( an analogous method was used in ref . to compute fluctuation corrections to @xmath55 ) . the @xmath137 column can be considered as an improved computation of the shielding factor . we see that the difference is very small and for practical purposes we can use the ga result . from the above results it is clear that neglecting transverse fluctuations the sum rule is well satisfied . this is analogous to the behavior found for the spin only model in the ising limit ( table [ tab : sw1 ] ) . in the ga+rpa for different values of @xmath170 in an @xmath223 sites system . , width=302 ] at this point we have weight at the bragg peak and weight at energy of order @xmath38 . how these weights get affected by transverse fluctuations ? on top of the covalency reduction transverse fluctuations reduce the permanent moment ( keeping the intrinsic moment constant ) . we have seen above ( sec . [ sec : longitudinal - swt ] ) that the reduction of the moment due to transverse fluctuations produces a concomitant large reduction of elastic spectral weight which is transfered to multimagnon processes in the longitudinal channel at energy @xmath224 . a direct computation in the hubbard model is difficult . in the next section we show how to estimate this effect in the hubbard model using the swt results . we expect that this low - energy rearrangement of spectral weights will neither affect @xmath222 nor the shielding factor computed above , which , as is clear from fig . [ fig:1m2d ] , is quite accurate at the present level of approximation . therefore we expect the present values of @xmath225 and @xmath222 ( without transverse fluctuations ) to be accurate estimates of the total spectral weight at low ( order @xmath114 ) and high ( order @xmath38 ) energies . in this section we would like to show how to use the spin wave theory results together with the results of the previous section to obtain a better distribution of the spectral weight . according to the discussion of sec [ sec:1m2d ] , to obtain the `` physical '' dynamical structure at low energies from the heisenberg model response , one should apply the canonical transformation back from the effective model to the hubbard model . since the processes involved in a @xmath183 expansion have a short spatial range and involve intermediate states at high energies , we do not expect this procedure to lead to a momentum or low - energy dependent correction . thus the `` physical '' dynamical correlation functions at low energy can be obtained from the requirement that the sum rule is satisfied as @xmath226 where @xmath227 is the dynamical correlation function of the heisenberg model and @xmath228 is the weight transfered to high energies due to the finiteness of @xmath38 computed in sec . [ sec : hub ] . alternatively a model with higher order corrections ( ring exchange , etc . ) can be taken to compute the response on the right hand side of eq . ( [ eq : hei2hub ] ) . , @xmath191mev and @xmath229mev . we show the contribution of two - magnon excitations computed in swt [ eq . ( [ eq : hei2hub ] ) ] with an effective shielding factor of @xmath230 ( table [ tab : hub ] ) and @xmath231 . we also show the contribution due to scattering across the hubbard bands computed in ga+rpa . [ fig:2mhub],width=302 ] we can consider @xmath232 in eq . ( [ eq : hei2hub ] ) as an effective shielding factor for low - energy spectral weight . one can check from the results of table [ tab : hub ] that this works for the bragg intensities neglecting transverse fluctuations . indeed we have @xmath233 where the left hand side is the bragg weight in the ga for the hubbard model without rpa correction and the right hand side is the effective shielding factor times the bragg weight in the ising limit of the spin only model ( @xmath234 ) . due to the separation of energy scales @xmath55 and @xmath222 will be insensitive to transverse fluctuations . the effect of the latter will be to transfer weight from the bragg peak to the low - energy continua without affecting the total low - energy spectral weight . thus eq . ( [ eq : esf ] ) tell us that the total low energy spectral weight ( bragg plus multi - magnon contribution ) is accurately given by the mean field value @xmath235 . as a corollary we see that a good estimate of the longitudinal effective shielding factor is given by the moment in the ga without the need of an rpa computation . i.e. the low - energy part of the longitudinal dynamical structure factor gets rescaled by the mean - field value of @xmath236 . this is consistent with the findings of hirsch and tang from numerical data@xcite if one takes into account that the hf approximation , used by hirsch and tang , gives very similar values of @xmath144 as the ga in the range of @xmath38 analyzed . in fig . [ fig:2mhub ] we show the two - magnon contribution computed in the heisenberg model and translated into a hubbard response according to eq . ( [ eq : hei2hub ] ) , together with the longitudinal response at energy @xmath237 computed directly in the hubbard model . additionally there is a bragg peak ( not shown ) with weight @xmath238 where @xmath144 is computed at mean - field level and @xmath239 takes into account the effect of transverse fluctuations . this is our best estimate for the distribution of spectral weight in the longitudinal response of the insulator . notice that the transition across the hubbard bands appears at 2ev . one should remember that on the mapping from the three - band hubbard model to the one - band hubbard model@xcite the transition across the hubbard bands of the latter represent the charge transfer transitions of the former , which according to optical data should occur close to @xmath240ev.@xcite indeed in the ionic limit , the lower hubbard band of the one - band model corresponds to o @xmath241 states and the upper hubbard band to cu @xmath242 . in la@xmath133cuo@xmath4 the dynamical structure factor can be fitted@xcite in a wide range of momentum and energy by the swt expressions eq . ( [ eq : sxaf ] ) with an intensity - lowering renormalization factor @xmath243 . we can use this result to estimate the different contributions to the moments for the _ one - magnon _ processes in la@xmath133cuo@xmath4 : @xmath244 @xmath245 and @xmath246 the last value should be interpreted as the weight for one transverse direction in the spin - wave like excitations of la@xmath133cuo@xmath4 . this is not the total weight since part of the weight will be in higher multimagnon processes . the experimental determination of the intensity lowering factor , @xmath247 neglected shielding factors . therefore the experimentally determined quantity can be put as : @xmath248 . using an effective shielding factor for @xmath190 as appropriate for la@xmath133cuo@xmath4 , @xmath249 and @xmath158 ( swt ) one obtains @xmath250 in agreement with the experiment ( inclusion of @xmath195 does not change this number appreciably ) . subtracting the observed one - magnon weight from the total expected transverse weight of @xmath251 would then leave a total of @xmath252 in three , five and higher multi - magnon processes which is within the error bars as illustrated in fig . [ fig : n2ddx ] . notice that the detected one - magnon spectral weight is significantly lower than @xmath253 , the value expected from the sum - rule neglecting the shielding factors and quantum corrections . for cu(dcoo)@xmath133.4d@xmath133o ( cftd ) rnnow _ et al._@xcite also find @xmath254 within lswt . for this compound the ratio of @xmath170 is much larger than in the cuprates and the shielding factor correction should therefore be close to 1 . since the error bars are smaller than in the cuprates this experiment shows , not surprisingly , that multimagnon processes are needed to satisfy the sum rule ( see fig . [ fig : n2ddx ] ) . the similarity for @xmath247 found for the two compounds is to some extent unexpected since there should be a difference due to the shielding factors , however , the expected difference is within the current experimental errors . notice also that @xmath20 may also differ in the two compounds . taking @xmath52 and defining the doping as @xmath255 , the shielding factor for general doping changes from @xmath256 for @xmath171 to @xmath257 for @xmath186 , the latter form being used often in experimental works . in general we expect a dependence on @xmath170 qualitatively similar to the one shown at @xmath258 ( c.f . fig . [ fig:1m2d ] ) . in order to proceed in the doped phase we need a model for the ground state . in this work we are interested in overall distributions of weights and we expect this to be to a large extent insensitive to the details of the ground state if correlations are taken reasonably well into account . this can be seen already at @xmath258 ( c.f . fig . [ fig:1m2d ] ) where we see that the paramagnet and the sdw within the ga give similar shielding factors although the nature of the ground state is completely different . detailed distributions of spectral weight as measured in ref . and computed in ref . will , of course , depend strongly on the ground state and so can be quite helpful to determine it . since those issues are beyond our present scope we restrict the description of the mean - field states to a minimum . for the specific system la@xmath2x@xmath3cuo@xmath4 ( x = sr , ba ) we use a ground state consisting of stripes as suggested by experiment.@xcite metallic stripes parallel to the cuo bond can be obtained within the ga for the hubbard model where a next - nearest neighbor hopping @xmath195 has to be implemented in order to have one doped hole per every second unit cell along the stripe @xcite ( so called `` half - filled '' stripes ) in agreement with experiments . note that a similar value for @xmath197 is predicted by first principle computations @xcite for la@xmath2sr@xmath3cuo@xmath4 . it should be mentioned that the results at half - filling reported in the previous sections are rather insensitive to the next - nearest neighbor hopping . for @xmath190 and @xmath195 we obtain an analogous dispersion than that reported in fig . [ fig : edq ] with the only difference that the value for the nearest neighbor hopping has to be scaled to @xmath196mev . this latter parameter set has been used@xcite in order to explain a recent mns experiment on ba - codoped lsco.@xcite overall spectral weights with this extended parameter set are practically the same than those reported in sec . [ sec : hubtrans ] with @xmath169 . in fig . [ fig : n2ddx ] we show the shielding factor obtained averaging ( @xmath172 ) in the mean - field solutions as a function of doping . we see that the shielding factor has a linear behavior similar to the one for the @xmath259 case . as expected we find that the shielding factor is similar for other low - energy textures ( not shown ) . for doping @xmath260 we obtain @xmath261 so there is an extra 10% reduction with respect to the undoped case . because the stripe solutions are magnetic the longitudinal structure factor has a bragg part and an inelastic part [ c.f . ( [ eq : szafh ] ) ] . since the stripe solutions are metallic , low - energy particle - hole excitations are allowed . therefore contrary to the result in the insulator ( sec . [ sec : hfhublon ] ) , already at rpa level one finds weight at magnetic energies ( @xmath262ev ) as shown in fig . [ fig : szsxsy ] . this weight however is broadly distributed in momentum space and will most likely pass unnoticed in unpolarized ins as discussed in sec . [ sec : exper ] . just as shown in fig . [ fig:2mhub ] for the insulator a more elaborate computation , taking into account longitudinal fluctuations , will show in addition weight at the two - magnon excitations . since the inelastic longitudinal component due to multimagnon scattering is already quite featureless in the insulator ( sec . [ sec : longitudinal - swt ] and [ sec : hfhublon ] ) it is reasonably to expect that it will be even more featureless in the doped phase . we conclude that within standard protocols ( sec . [ sec : exper ] ) to first approximation all the inelastic longitudinal spectral weight will be assigned to the background . and @xmath195 for the sdw at @xmath258 and bond - centered metallic stripes solutions as a function of doping ( red line ) and for @xmath259 ( dotted blue line ) . for the zeroth moment mbz means that the integration was restricted to the magnetic brillouin zone . la@xmath133cuo@xmath4@xcite and cftd@xcite label the moments estimated in insulators ( sec . [ sec : col ] ) and la@xmath2ba@xmath3cuo@xmath4@xcite labels the experimental value of @xmath263 in the magnetic brillouin zone , corrected by the polarization factor . in the last case the error bars reflect the two extreme possibilities for the polarization factor 2/3 and 1/2 ( c.f . [ mns ] ) . the relative error should be larger than at zero doping where the error bars have the usual sense . finally we show the zeroth moment of the ga+rpa transverse spectra at @xmath260 in the whole brillouin zone ( + ) and in the mbz @xmath264 . the two lower @xmath264 include a @xmath212 renormalization factor and for the lower one the energy integration was restricted to energies smaller than 0.23ev . [ fig : n2ddx],width=302 ] in a @xmath265 site cluster for @xmath190 and @xmath195 and @xmath196mev . only the inelastic part is shown and we neglect the intensity renormalization ( @xmath147 ) . the ground state consist of 4 sc stripe running along the short dimension . , width=302 ] for the specific case of a striped ground state we can also estimate the elastic spectral weight in the longitudinal channel . for doping @xmath260 and an array of @xmath266 sc stripes we find elastic peaks at @xmath267 and @xmath268 with @xmath269 and very small weight , @xmath270 , in the higher harmonics @xmath271 . it was shown in the insulator that if one neglects transverse fluctuations the sum rule in the longitudinal channel is exhausted with a good degree of approximation by the bragg weight plus the inelastic high energy weight . is there an analogous behavior in the doped phase ? the elastic weight neglecting transverse fluctuations is @xmath272 and the inelastic longitudinal spectral weight is @xmath273 . this includes low - energy particle - hole excitations and weight due to transitions across the hubbard bands . the total weight in the longitudinal channel is @xmath274 to be compared with a shielding factor of 0.79 . in this case we find an overstrike of the sum rule which is more sever than in the insulator and which is even larger than the maximum allowed value @xmath275 . this is because rpa becomes less accurate in the metallic phase due to the presence of small energy denominators . the violation however is not as severe as for the transverse channels as reported below . the elastic weight neglecting transverse fluctuations is roughly half of the one found in the insulator ( @xmath276 ) . just as in the insulator a large fraction of this weight ( at least 60% ) will be transfered to multimagnon excitations if transverse fluctuations are taken into account . it is conceivable that all the weight in the bragg peak is transfered in which case the system becomes quantum critical or quantum disordered.@xcite the transverse component of the dynamical structure factor shows sharp features due to the propagating spin - wave like modes of the stripes . in addition , since the stripes are metallic , a fraction of the spectral weight is in a broad particle hole continuum . the latter features however are mainly located in the nuclear brillouin zone ( nbz).@xcite in the magnetic brillouin zone ( mbz ) we find mainly sharp propagating collective modes which are good candidates to be easily detected . these collective modes are analogous to the spin - wave modes of the insulator and therefore we expect that their spectral weight will be overestimated in rpa and should be corrected by a quantum renormalization @xmath277 as in the insulator . @xmath277 may be smaller than in the insulator due to stronger quantum fluctuations . on the other hand for a heavily doped system one should recover the noninteracting value @xmath278 . since doping is small we tentatively take the same value as in the insulator @xmath279 . using @xmath190 , @xmath195 , @xmath196ev and @xmath280 , in a @xmath281 sites lattice , together with the experimental data for la@xmath2ba@xmath3cuo@xmath4 @xmath5@xcite . the lower panel shows the partial sum rule weight . @xmath282 , title="fig:",width=302 ] using @xmath190 , @xmath195 , @xmath196ev and @xmath280 , in a @xmath281 sites lattice , together with the experimental data for la@xmath2ba@xmath3cuo@xmath4 @xmath5@xcite . the lower panel shows the partial sum rule weight . @xmath282 , title="fig:",width=302 ] the momentum distribution of the transverse spectral weight has been shown in ref . . [ fig : n2ddx ] we show the transverse integrated spectral weight compared with the experimental results . at zero doping the spectral weight coincides with the theoretical value ( not shown for clarity ) as discussed in sec . [ hf ] . at finite doping the weight in the mbz decreases roughly as the shielding factor as can be seen by comparing the points labeled la@xmath133cuo@xmath4(mbz ) and rpa@xmath283(mbz ) . comparing fig . [ fig : swt ] and fig . [ fig : szsxsy ] we see that whereas in the insulator almost all of the spectral weight is at magnetic energies in the doped phase a large fraction of the spectral weight is at intermediate energies not accessibly to mns . in other words doping induces a transfer of spectral weight from low energies to intermediate energies in this channel . interestingly optical spectra shows that doping also generates structure at the same energies in the charge channel@xcite . the lower `` @xmath264 '' at @xmath260 in fig . [ fig : n2ddx ] is the rpa moment in the mbz renormalized by @xmath284 and restricting the integral only up to the highest energy measured in ref . , we see that the obtained value is roughly 50% higher than the experimental weight . in order to see at what energy the disagreement arises we show in fig . [ fig : seffexptheo ] the experimental @xmath286 compared with the theoretical one using 1/2 as polarization factors [ eq . ( [ eq : seffqz ] ) ] and neglecting the longitudinal weight . the lower panel shows a comparison of the partial sum rule weight with the two limiting values of the polarization factor . the theoretical computation shows a strong resonance at @xmath287mev which in the experiment appears with a reduced spectral weight and in part distributed on a broader energy range . ( the resonance appears at higher energy if one looks only at the @xmath288 response@xcite ) . the overall weight integrated up to energies immediately above the resonance is in agreement with the experimental one . as the energy increases there are strong deviations between theory and experiment . the simplest explanation for the smaller weight in the experiment is that some of the structures that contribute at high energy may be broad and assigned to the background ( see ref . ) . we should say that because we neglect life - time effects even the sharp structures of our computation will be much broader in reality and this will be more important as the energy increases since , quite generally , the phase space for decay processes increases . also high - energy excitations will be much more sensitive to disorder since they involve short wave - lengths and this will also tend to make them broader . the broad structures will be assigned to the background explaining the difference between theory and experiment . we should also take into account that @xmath167 can be more depressed due to the larger impact of quantum fluctuations in the metallic phase . for @xmath258 ( ref . ) , la@xmath2ba@xmath3cuo@xmath4 @xmath5 ( ref . ) and la@xmath2sr@xmath3cuo@xmath4 with @xmath289 ( ref . ) and @xmath290 ( ref . ) . we used the following conversions from the quantity reported in the quoted reference : @xmath291 ( ref . ) @xmath292(ref . ) @xmath293 ( ref . ) .,width=302 ] in order to have a broader view of the trends in the evolution of the spectral weight , we show in fig . [ fig : seffexp ] a compilation of measurements of @xmath286 from different groups and at different dopings . for @xmath294 and 0.14 the overall weight seem to be roughly conserved whereas for @xmath290 the detected weight seems to be significantly smaller . it will be interesting to see weather this is due to a transfer to high - energies where absolute measurements are not available . in this work we have analyzed the distribution of spectral weight in neutron scattering experiments in cuprates . in the insulator the spectral weight has been estimated in detail in the heisenberg model and then translated into a hubbard response with the use of an effective shielding factor . we find an effective shielding factor which essentially coincides with a proposal of hirsch and tang obtained by analyzing numerical data.@xcite in addition we have estimated , with the aid of the tdga , the weight at the energy of the charge transfer transitions . the inelastic longitudinal spectral weight at low energies is in broad features and thus hard to detect , although progress has recently been made.@xcite the transverse part has the well know quantum renormalized propagating spin wave modes which will be the features more accessible to experiment . theory and experiment agree well in the insulator . for the doped phase we computed the dynamical response within the tdga . we considered metallic stripes but we expect our results for overall weights to be largely insensitive to the specific texture . one finds extra broad features due to the particle - hole continua plus propagating collective modes . in analogy with the insulator we expect the latter to be reduced in intensity due to quantum fluctuation and to dominate the spectral weight reported in experiments . our motivation was to understand why such a small fraction of the naively expected spectral weight is actually detected . we see that after taking into account all the reduction factors we passed from a weight which was an order of magnitude smaller than expected to a weight which is @xmath295 of that expected . given the uncertainties involved we think this value is reasonable . by far the most important factor in reducing the detectable weight turns out to be the fact that a large fraction of the spectral weight is in broad features or at too high energies and this make the detection of the total spectral weight experimentally very challenging . in the doped phase a consistent fraction of the transverse spectral weight is predicted to lay at intermediate energies which may become accessible to experiment in a near future . we see no reason why the broad / high energy modes can not have a strong impact on the effective attraction between holes relevant for superconductivity and therefore their experimental characterization remains an important open problem . we acknowledges support from the uk epsrc under grant number gr / r76714/01 ( rc ) . jl thanks the university of oxford , isis and the cclrc s rutherford appleton laboratory for financial support and hospitality during part of this work and a. t. boothroyd , s. m. hayden , d.f . mcmorrow , t. g. perring and j. tranquanda for illuminating discussions . we neglect the small @xmath296 variation of @xmath297 along the antiferromagnetic zone boundary predicted by ref . and r.r.p . singh and m. gelfand , b * 52 * , r15695 ( 1995 ) ; o.f . syljusen and h.m . rnnow , j. phys . : condens . matter * 12 * , l405 ( 2000 ) .

we present a model of optically thin , two - temperature , accretion flows using an exact monte carlo treatment of global comptonization , with seed photons from synchrotron and bremsstrahlung emission , as well as with a fully general relativistic description of both the radiative and hydrodynamic processes . we consider accretion rates for which the luminosities of the flows are between and of the eddington luminosity . the black hole spin parameter strongly affects the flow structure within the innermost gravitational radii . the resulting large difference between the coulomb heating in models with a non - rotating and a rapidly rotating black hole is , however , outweighed by a strong contribution of compression work , much less dependent on spin . the consequent reduction of effects related to the value of the black spin is more significant at smaller accretion rates . for a non - rotating black hole , the compressive heating of electrons dominates over their coulomb heating , and results in an approximately constant radiative efficiency of per cent in the considered range of luminosities . for a rapidly rotating black hole , the coulomb heating dominates , the radiative efficiency is per cent and it slightly increases ( but less significantly than estimated in some previous works ) with increasing accretion rate . our study neglects the direct heating of electrons , which effect can lead to larger differences between the radiative properties of models with a non - rotating and a rapidly rotating black hole than estimated here . flows with the considered parameters produce rather hard spectra , with the photon spectral index , and with high energy cut - offs at several hundred kev . we find an agreement between our model , in which the synchrotron emission is the main source of seed photons , and observations of black - hole binaries in their hard states and agns at low luminosities . in particular , our model predicts a hardening of the x - ray spectrum with increasing luminosity , as indeed observed below or so in both black - hole binaries and agns . also , our model approximately reproduces the luminosity and the slope of the x - ray emission in cen a. [ firstpage ] accretion , accretion discs black hole physics x - rays : binaries x - rays : general .

1107.0860

optically thin , two - temperature accretion flows have been considered as an explanation of a variety of black hole systems and a substantial work has been done for investigation of their dynamical and spectral properties ( see , e.g. , narayan & mcclintock 2008 for a review ) . however , the developed models still involve several approximations which significantly reduce their accuracy . these involve the use of a pseudo - newtonian potential of paczyski & wiita ( 1980 ) , which fails in the innermost region ( particularly if rotation of the black hole is considered ) , where most of the gravitational energy is dissipated , as well as local approximations of comptonization , which appear to be particularly incorrect in optically thin flows ( see xie et al . 2010 ; hereafter x10 ) . previous attempts to improve these two weaknesses are discussed in section [ previous ] . in this paper , we extend our previous treatment of global comptonization from x10 by using the hydrodynamical model from manmoto ( 2000 ; hereafter m00 ) , and develop a self - consistent model involving a fully general - relativistic ( gr ) description of both the hydrodynamical and the radiative processes . we consider moderate values of accretion rate , at which coulomb coupling between ions and electrons is relatively weak compared to the viscous heating of ions and , therefore , the ion temperature ( and hence the total pressure ) is not affected significantly by the details of the description of radiative processes . this allows for a slightly simplified treatment of the flow structure in our computations leading to a self - consistent solution . already the foundational papers proposed that tenuous , two - temperature flows may be responsible for the hard spectral states of black - hole binaries ( ichimaru 1977 ; see also narayan & yi 1995 ) and for the low nuclear luminosities in radio galaxies with large radio lobes ( rees et al . 1982 ) . we consider two ranges of the key parameters ( black hole mass and accretion rate ) which should be relevant for these two major application areas . for each of these two cases , we illustrate the impact of the black hole spin by considering a non - rotating and a maximally - rotating black hole . on the other hand , we neglect here the dependence on some other parameters of hot - flow models . in particular , we consider only weakly magnetized flows , with the magnetic pressure of 1/10th of the total pressure , and we neglect the direct viscous heating of electrons ; see section [ delta ] for a discussion of these two assumptions . furthermore , we neglect outflows , which may play an important role in some systems ( e.g. yuan , quataert & narayan 2003 ) . then , our solutions are strongly dominated by the advection of energy by ions , as in the original formulation of the advection - dominated accretion flow ( adaf ) model by , e.g. , narayan & yi ( 1995 ) . in this paper , we focus on modelling an innermost part of an accretion flow , namely inside @xmath7 , where @xmath8 is the gravitational radius and @xmath9 is the black - hole mass , where the bulk of the observed radiation is produced . global comptonization has a net effect of compton heating outside @xmath10 ( e.g. park & ostriker 2001 , yuan , xie & ostriker 2009 , yuan & li 2011 ) , which effect is not considered here . we consider a black hole , characterised by its mass , @xmath9 , and angular momentum , @xmath11 , surrounded by a geometrically thick accretion flow with an accretion rate , @xmath12 . we define the following dimensionless parameters : @xmath13 , @xmath14 , @xmath15 , where @xmath16 and @xmath17 is the eddington luminosity . the inclination angle of the line of sight to the symmetry axis is given by @xmath18 . we refine here our previous study presented in x10 by including a fully gr hydrodynamical description of the flow . our gr hydrodynamical model follows strictly m00 , except for a small difference in the advective term , as noted below equation ( [ eq : qint ] ) . we have thoroughly tested our hydrodynamical model by comparing it with the numerical code and results of li et al . ( 2009 ) , which work also follows m00 . other differences with respect to x10 involve : \(i ) the vertical structure of the flow : we assume here that the temperature and velocity components are vertically uniform , while the density distribution is given by @xmath19 , where @xmath20 is the scale height at @xmath21 ; in x10 we assumed a tangentially uniform structure of the flow ; \(ii ) the direct viscous heating of electrons , which is neglected here and included in x10 ; \(iii ) an outflow , neglected here and included ( with a large magnitude ) in x10 . we consider the pairs of the main parameters of ( @xmath22 , @xmath23 ) and ( @xmath24 , @xmath25 ) , which correspond to low - luminosity agns and hard states of black - hole binaries , respectively . for each pair of ( @xmath9 , @xmath26 ) we consider two values of the spin parameter , @xmath27 and @xmath28 . we assume the viscosity parameter of @xmath29 and the ratio of the gas pressure ( electron and ion ) to the total pressure of @xmath30 . the latter parameter determines the strength of the magnetic field in the accretion flow , and the synchrotron emission can then be determined . for it , we follow the method of narayan & yi ( 1995 ) , which was applied also by m00 and x10 . we define the compressive heating rate of electrons per unit area ( in the local reference frame ) of the flow , @xmath31 and the rate ( per unit area ) of advection of the internal energy of electrons , @xmath32 where @xmath33 is the mass density and @xmath34 is the electron pressure in the midplane , @xmath35 is the effective adiabatic index [ see , e.g. , eq . ( 62 ) in m00 for the definition ] ; for simplicity of notation we hereafter skip the e subscripts in symbols denoting electron heating / cooling rates . note that in our definition @xmath36 . then , @xmath37 is always positive ( therefore , we define it as a heating term ) as the density gradient is always negative in our models . in all models , @xmath38 is negative ( representing the release of internal energy ) in the innermost region , at larger distances it is positive ( the internal energy is stored ) or vanishes . following the usual notation , we define also the electron advection rate , @xmath39 , which is typically negative ( i.e. the advective effects result in effective heating ) except for an outer ( beyond @xmath40 , see below ) region in the local models . the above forms of @xmath37 and @xmath38 follow directly from eqs . ( 54 ) and ( 60 ) in m00 . we do not use the simplifying approximation , @xmath41 , adopted in m00 ; we checked that it yields deviations of the advective term from the actual values by a factor of a few . with the above definitions , the electron energy equation is @xmath42 where @xmath43 is the radiative cooling rate per unit area ( including synchrotron , @xmath44 , and bremsstrahlung , @xmath45 , and their comptonization , @xmath46 ) and @xmath47 is the electron heating rate per unit area by ions via coulomb collisions . the values of @xmath47 , @xmath37 and @xmath43 integrated over the whole body of the flow are denoted by @xmath48 , @xmath49 and @xmath50 , respectively . due to the effect of capturing by the black hole and the gravitational redshift ( see section [ collimation ] for details ) , the total power radiated by the flow , @xmath50 , is higher than the luminosity , @xmath51 , detected far away from the flow . for completeness we give also the ion energy equation : @xmath52 where the compressive heating and the advection of the internal energy of ions , respectively , are given by @xmath53 and the viscous dissipation rate , per unit area , is given by @xmath54 where @xmath55 , @xmath56 is the ion pressure , @xmath57 is the ion adiabatic index , @xmath58 is the angular velocity of the flow , @xmath59 is the lorentz factor of the azimuthal motion and @xmath60 . the above form of @xmath61 results from the usual assumption that the viscous stress is proportional to the total pressure , with the proportionality coefficient @xmath62 ; cf . eqs . ( 48 ) and ( 54 ) in m00 . this is almost certainly an oversimplified approach ; see gammie & popham ( 1998 ) for a discussion of various issues related to implementing a proper description of the viscous stress in gr models . the magneto - rotational instability ( mri ) is now widely believed to be an origin of the viscous stress and numerical simulations of this process ( e.g. krolik , hawley & hirose 2005 ) show radial profiles of the stress which are different from analytic prescriptions considered in the accretion theory . the uncertainty regarding the form of the stress is the major shortcoming in attempts to develop an analytical description of accretion flows . the form of the viscous stress , @xmath63 , is assumed here for computational simplicity . it has an unphysical property of leading to a negative dissipation rate close to the event horizon in some cases . however , this particular weakness of our model appears to be not important for our results . namely , we find that @xmath64 occurs at @xmath65 only in our models with @xmath27 ; in these models the heating of ions is strongly dominated by the compression work at @xmath66 . therefore , details of the viscous heating of ions do not affect the flow structure . clearly , the accuracy of @xmath61 would be much more important in models taking into account direct viscous heating of electrons . our modelling of comptonization makes use of a monte carlo ( mc ) method taking into account all effects relevant for the kerr metric ( niedwiecki 2005 ; see also niedwiecki & zdziarski 2006 and x10 ) . the seed photons are generated from synchrotron and bremsstrahlung radial emissivities of the flow , taking into account the vertical profile of the density ; their transfer and energy gains in consecutive scatterings are affected by both the special relativistic and gravitational effects . the mc comptonization simulation and the solution of structure equations are treated separately in our model . therefore , we have to iterate between their results to find a mutually consistent solution . in principle , a full set of hydrodynamical equations should be solved after each mc simulation , as in x10 . however , in the current study we apply a slightly simplified procedure , motivated by the fact that the structure of the flow is fully determined by the ion energy equation ( this condition is , however , satisfied only within inner several hundred @xmath67 , see below ) , while the electron energy equation determines the electron temperature . under such conditions , our procedure of refining the description of compton cooling should affect only the electron temperature and not the other flow parameters ( @xmath68 , @xmath33 , @xmath20 , and the velocity field ) . then , we first solve the dynamical structure with a local approximation of compton cooling and then we iterate between the mc comptonization results and the solutions of the electron energy equation , keeping the flow parameters , except for @xmath69 , unchanged . after each mc simulation , we determine the radially - dependent compton cooling . we then use this rate to solve the energy equation ( [ eq : energy ] ) , taking into account the dependence of all relevant processes ( synchrotron and bremsstrahlung emissivities , coulomb exchange rate , advective terms ) on @xmath69 . we repeat the above procedure until it converges . for the initial solution , the outer boundary of the flow is set at @xmath70 ( see section 2.4 in m00 for details ) . the mc simulations are performed within @xmath71 , using the hydrodynamical solution at @xmath72 ; note that details of this inner part of the solution should not depend on the outer boundary condition ( see e.g. fig . 5 in narayan , kato & honma 1997 ) . for our final solution , we balance the electron energy equation up to @xmath73 . due to boundary effects in mc simulation , combined with strong sensitivity of the advective term on the temperature gradient , the solution is very unstable at @xmath21 close to @xmath74 . furthermore , we note that an assumption underlying our computational procedure fails in these outer regions . namely , all our final solutions are characterised by an electron temperature larger ( typically by a factor of @xmath75 at @xmath76 ) than the initial solutions and the electron pressure can give a non - negligible contribution beyond several hundred @xmath67 . in that region , the scale height would increase , which effect is not taken into account here ; however , we expect it to be a rather minor , @xmath77 per cent , effect such a change of @xmath78 was found , in response to a similar in magnitude ( but opposite ) change of @xmath69 , in x10 . for one set of parameters , namely @xmath79 , @xmath80 and @xmath27 , @xmath69 approaches @xmath68 at @xmath81 , so the flow loses a two - temperature structure . we skip here more careful investigation of the outer region because it gives a negligible contribution to the observed radiation . furthermore , a detailed study of the outer region should involve a proper description of additional , poorly understood effects ( like a mechanism of transition from an optically thick disc to a tenuous flow ; e.g. , in hard states such a transition is likely to occur at @xmath82 ) which are beyond the scope of this paper . figs [ fig : rates ] and [ fig : rates_stel ] show the change of the heating and cooling rates , and the electron temperature , between our initial and final solutions . figs [ fig : flow ] and [ fig : flow_stel ] show the remaining parameters of the flow these are assumed to be the same in the initial and final solutions . we can clearly see how the value of the spin parameter , @xmath83 , affects the flow properties . rotation of the black hole stabilizes the circular motion of the innermost part of the flow ( an effect directly related with properties of test particle motion , analogous to the well - known dependence of the innermost stable orbit on @xmath83 in keplerian discs ) which yields a higher density ( through the continuity equation ) . this , in turn , implies both a stronger coulomb coupling ( see figs [ fig : rates]f and [ fig : rates_stel]f ) and a larger optical depth ( see figs [ fig : flow]a and [ fig : flow_stel]a ) in the innermost region in models with @xmath28 . on the other hand , heating by the compression work is typically stronger than , or comparable to , heating by the coulomb energy transfer from ions ( cf . nakamura et al . 1997 , mahadevan & quataert 1997 ) , which effect partly reduces the differences between the high and low @xmath83 models . these effects , and their impact on the radiative properties of the flow , are discussed in detail below . figs [ fig : rates](f ) and [ fig : rates_stel](f ) show also the dissipative heating rate of ions , @xmath61 . the stabilized rotation results in a much stronger dissipative heating in the inner region for @xmath28 , however , this property is rather unimportant for effects investigated in this paper ( neglecting both the direct heating of electrons and hadronic processes ) . we discuss here in some details the initial solutions although they are not self - consistent , because most of models available in literature use the same local approximation for comptonization . moreover , we amend some conclusions regarding the accuracy of local approximations , derived previously in x10 . we find the initial solutions using the prescription for the compton cooling rate in a slab geometry with initial photon energies of 1 ev , given in dermer et al . ( 1991 ; hereafter d91 ) . in x10 , we found that it gives a reasonably good approximation for the innermost region . for the models considered here , the slab approximation appears to be less accurate , however , our choice of the slab case allows for a direct comparison with previous studies , a number of which used the slab approximation based on d91 , or its modification introduced in esin et al . ( 1996 ) . figs [ fig : rates](a , c ) and [ fig : rates_stel](a , c ) show the heating and cooling rates in the initial solution . comptonization of synchrotron photons is the most efficient radiative cooling process . however , the synchrotron emissivity decreases rapidly beyond its transition radius , @xmath40 ( @xmath84 for @xmath23 and @xmath85 for @xmath86 ) . then , in the local model , the radiative cooling at @xmath87 is dominated by the ( much weaker ) bremsstrahlung and its comptonization . then , two regions can be distinguished in terms of the _ electron _ energy equation ( ions are always advection dominated ) . ( i ) at @xmath87 , the flow is adiabatically compressed , which is reflected in the increase of @xmath69 with decreasing @xmath21 ( see figs [ fig : rates]e and [ fig : rates_stel]e ) ; both the radiative cooling and the coulomb heating are much weaker than the advective terms . ( ii ) at @xmath88 , the radiative cooling becomes efficient and @xmath69 decreases ; the coulomb heating is more efficient in this region and it exceeds the compression work in the innermost part ( except for the model with @xmath27 and @xmath89 ) . in this region , the advection of the internal energy contributes also to the heating of electrons , however , this effect is rather weak . the solid curves in figs [ fig : rates](a , c ) and [ fig : rates_stel](a , c ) show the gr global compton cooling rate obtained in our initial mc simulations , i.e. with the @xmath69 satisfying the energy equation ( [ eq : energy ] ) with the compton cooling rate , @xmath90 , given by the local slab ( d91 ) prescription . as we can see , there are significant deviations between the local ( dotted curves ) and global cooling rates , resulting from several effects . two of these effects should be generic to models of black - hole flows regardless of specific values of parameters of the flow . first , the input of seed photons from the inner region strongly enhances the compton cooling beyond @xmath40 . the second one involves the presence of the event horizon ( neglected in the local approximation ) . the presence of an inner boundary of the flow at the event horizon results in an obvious difference with respect to the semi - infinite slab ; namely , a large fraction of photons generated in the innermost region is captured , reducing the input of seed photons . another effect affecting the structure of the flow is the dependence of the optical depth from a given point on direction . this strongly depends on the flow parameters and assumptions about heating and outflow . in particular , the flows for our parameters have the optical depth from a point in the equatorial plane along the outward radial direction , @xmath91 , similar to the vertical one , @xmath92 , at any @xmath21 , whereas @xmath93 in the model of x10 . furthermore , for @xmath28 , both @xmath91 and @xmath92 are much smaller than the optical depth in the inward radial direction , @xmath94 , except for the very innermost region ; in these models @xmath95 ( @xmath96 for @xmath97 and @xmath4 for @xmath80 ) at @xmath98 . for @xmath27 , @xmath95 ( @xmath99 for @xmath97 and @xmath100 for @xmath80 ) at @xmath101 ; in these models the difference between @xmath94 and @xmath91 typically does not exceed a factor of 2 . then , for a non - rotating black hole , the local radiative properties would be more accurately described by the spherical geometry , while for a rapidly rotating black hole the local properties are intermediate between the sphere and the slab . then , the assumption of the slab geometry overestimates the initial cooling rate in all models considered here . finally , as we note in x10 , the d91 formula is accurate for only two values of seed photon energies , 1 ev and 1 kev , and we use here the lower one . with this value , the formula gives a reasonable approximation for flows around a stellar black hole , where seed photons have energies @xmath102 ev . for flows around a supermassive black hole , however , with typical seed photon energies of @xmath1 ev ( for @xmath103 ) , it underestimates the cooling rate . specifically , for this @xmath9 we found deviations by a factor of @xmath104 between the formula of d91 for 1 ev and the results of our _ local _ slab mc simulation using the exact energy distribution of seed photons ( given by the synchrotron and bremsstrahlung emission distributions ; see x10 for the description of our local mc model ) . the initial solutions are qualitatively similar at both values of @xmath26 considered here , the most significant dependence on @xmath105 involves the increase of the value of @xmath40 with increasing @xmath26 . in contrary , detailed properties of the self - consistent , final solutions are more sensitive to @xmath26 . figs [ fig : rates](b , d ) show the heating and cooling rates of the final solutions for @xmath23 . at @xmath106 , a significant fraction of the power heating electrons is stored as their internal energy rather than radiated away . this is qualitatively similar to the initial solution , although @xmath43 is larger at @xmath106 in the global cooling model and hence @xmath38 is smaller than in the local cooling model ; therefore , the temperature gradient is smaller in the outer region in our final solutions ( see discussion below ) . we note that the strong contribution of the @xmath38 term significantly complicates the procedure for finding the self - consistent solution at this lower @xmath26 , as both the global compton cooling and the electron advection are strongly sensitive to , and affect , the electron temperature profile . [ fig : rates](e ) shows the electron temperature profiles for the initial and the final solutions . we see that , in the final solution , @xmath69 increases compared to the initial solution , by a factor of @xmath107 in the inner region , and up to a factor of 3 at several hundred @xmath67 . the increase of @xmath69 at @xmath108 is an obvious effect , related with the overestimation of the cooling rate by the slab approximation , combined with - crucial in that region - capturing of seed photons by the black hole , as discussed in section [ initial ] . somewhat surprisingly , the global solution requires higher @xmath69 also at @xmath109 , where the global radiative cooling is significantly enhanced . the temperature must remain high in this outer region because the energy balance could not be achieved for @xmath69 decreasing too rapidly . namely , we found that such a ( too rapid ) decrease of temperature at @xmath109 would yield a large positive @xmath110 term in the electron energy equation ( [ eq : energy ] ) , which , added to the ( increased in global cooling ) @xmath43 term , could not be balanced by the heating terms . we emphasise that this property is related to strong contribution of advective terms and it should not necessarily occur in the whole range of parameters , in particular at higher accretion rates . for @xmath25 ( fig . [ fig : rates_stel ] ) , the self - consistent @xmath69 is higher in the innermost part than @xmath69 in the initial solution , due to the same reasons as for @xmath23 . however , the compton cooling is relatively more important in the outer region , which results in an approximately isothermal structure at @xmath111 ; in particular , we do not find a local maximum of @xmath69 , marking the transition ( at @xmath112 ) from an adiabatically compressed flow to an efficiently cooled flow in global solutions with @xmath23 and in all our local solutions . the temperature at which the energy balance is achieved is twice lower for @xmath28 , due to a much larger input of seed photons from the central region . the strong contribution of the compressive heating has crucial consequences for the radiative efficiency , @xmath113 , where @xmath51 is the total observed luminosity of the flow , and for its scaling with @xmath26 . we compare here our models with the same @xmath83 for different @xmath26 , although we note that they have also different @xmath9 , which may slightly affect such a comparison ( the main dependence on the value of @xmath9 concerns the energy of synchrotron photons , influencing the cooling rate ) . comparing our results for @xmath23 and 0.5 , we find that the fraction of an accretion power transferred to electrons via coulomb collisions varies approximately as @xmath114 for @xmath28 and for @xmath27 we note even stronger dependence on @xmath26 . then , heating dominated by the coulomb coupling would lead to an approximately linear scaling of the radiative efficiency with @xmath26 , as derived in previous works on this subject ( see , e.g. , eq . 12 and references in narayan & mcclintock 2008 ; see also table 3 in rajesh & mukhopadhyay 2010 ) . however , @xmath115 is much less dependent , moreover negatively , on @xmath26 ; specifically , it is higher by a factor of @xmath104 for @xmath23 than for @xmath25 . as a result , the dependence of @xmath116 on @xmath26 is significantly reduced . for @xmath27 , @xmath117 at @xmath23 and @xmath118 at @xmath25 . the decrease of @xmath115 with increasing @xmath26 is partially balanced by the increased contribution of coulomb heating . two further effects decrease the radiative efficiency at lower @xmath26 : ( 1 ) the increase of the fraction of the energy given to electrons which is not radiated away ( and increases the internal energy advected with the flow ) , and ( 2 ) the increase of the fraction of photons captured by the black hole ( section [ collimation ] ) . as a result , the radiative efficiency for @xmath27 is approximately the same for both considered values of @xmath26 . more specifically , @xmath119 ( with @xmath120 erg s@xmath121 and @xmath122 ) for @xmath23 and @xmath123 ( @xmath124 erg s@xmath121 and @xmath125 ) for @xmath25 . for @xmath28 , @xmath48 exceeds @xmath49 at both @xmath23 and 0.5 . radiative efficiencies at these @xmath26 differ only by a factor of 2 , due to relatively stronger contribution of compressive heating at @xmath23 and very strong reduction of the observed @xmath51 by gr effects ( section [ collimation ] ) at @xmath25 . specifically , for @xmath23 , @xmath126 ( @xmath127 erg s@xmath121 and @xmath128 ) and for @xmath25 , @xmath129 ( @xmath130 erg s@xmath121 and @xmath131 ) . we note that for @xmath28 the apparent luminosity depends on @xmath132 , see section [ anisotropy ] , and the above values of @xmath51 are averaged over all viewing angles . and @xmath23 . the solid ( black ) and dashed ( red ) curves are for @xmath28 and @xmath27 , respectively . ( b ) the angular dependence of spectra for the final solution with @xmath28 ; the spectra from bottom to top are for @xmath1331 , 0.50.6 and 00.1 . , width=302 ] and @xmath23 . the solid ( black ) and dashed ( red ) curves are for @xmath28 and @xmath27 , respectively . ( b ) the angular dependence of spectra for the final solution with @xmath28 ; the spectra from bottom to top are for @xmath1331 , 0.50.6 and 00.1 . , width=302 ] figs [ fig : spectra](a ) and [ fig : spectra_stel](a ) show the angle - averaged spectra for the initial and final solution . in all models , the normalization of the spectra of the final solution is much higher than that of the initial solution , due to the increase of the synchrotron emissivity resulting from the increase of @xmath69 . the x - ray spectra are relatively hard , with the photon index @xmath134 , for both ( @xmath23 , @xmath28 ) and ( @xmath25 , @xmath27 ) and @xmath135 for ( @xmath25 , @xmath28 ) ; the last value is determined in the 210 kev range , as at higher energies a rather pronounced wien peak leads to further hardening of the spectrum . for ( @xmath136 , @xmath27 ) , the relativistic @xmath69 and small optical depth result in pronounced scattering bumps seen in this model spectrum ; the average slope in the 2200 kev range is @xmath137 . thus , we see that the spectra harden with the increasing luminosity , for @xmath26 increasing from 0.1 to 0.5 , from @xmath138 to @xmath139 at @xmath27 and from @xmath140 to @xmath141 at @xmath28 . in all models , the @xmath142 spectra have the maxima at several hundred kev . the electron temperature reaches the maximum values of @xmath143 mev for @xmath23 and @xmath144 kev for @xmath25 . note , however , that although these flows produce photons with energies @xmath145 , we do not expect an efficient e@xmath146 pair production in photon - photon collisions . the probability that a @xmath147-ray produces a pair is approximately given by the compactness parameter for this process , @xmath148 , where @xmath149 is the radiative power in photons with @xmath150 kev produced within @xmath151 . it is @xmath152 for all @xmath151 in all models except for @xmath25 and @xmath28 , where @xmath153 in the innermost few @xmath67 . the same conclusion was reached in previous studies of pair equilibria in advection - dominated flows ( e.g. esin 1999 ) where additional pair creation processes , namely electron - electron and photon - particle collisions , were also taken into account . but for @xmath154 and @xmath25 . , width=302 ] but for @xmath154 and @xmath25 . , width=302 ] the nature of the space - time metric ( specified by the value of @xmath83 ) affects the radiative properties of hot flows through various effects . as noted above , the stabilized circular motion results in a much larger density in the innermost region in models with large @xmath83 . this affects the radiative efficiency in a manner discussed in section [ final ] ; here we restate some of these results with the emphasis on the dependence on @xmath83 . two further effects , discussed in sections [ collimation][anisotropy ] , concern the properties of photon motion in curved space - time . higher density implies a stronger coulomb coupling for large @xmath83 . for @xmath27 , @xmath48 is smaller than for @xmath28 by a factor of 200 and 35 for @xmath23 and 0.5 , respectively . the difference between the heating of electrons for high and low @xmath83 is , however , strongly reduced by the high efficiency of compression , which shows very weak dependence on the value of @xmath83 . then , the total heating , @xmath155 , is higher in models with @xmath28 only by a factor of 3.5 ( @xmath156 ) and 16 ( @xmath25 ) . extrapolating our results we can expect that at @xmath157 the difference between the heating efficiencies in high and low @xmath83 models vanishes . on the other hand , a much larger difference between the heating rates for large and small values of @xmath83 could be expected in models involving the direct transfer of the dissipated energy to electrons ; compare @xmath47 and @xmath37 with @xmath61 shown in figs [ fig : rates](f ) and [ fig : rates_stel](f ) . higher density implies also a larger optical depth for large @xmath83 . then , despite a stronger heating in these models , the energy balance in the innermost region is achieved at smaller electron temperatures than for small @xmath83 . on the hand , the highly efficient comptonization occurs at very small @xmath21 , therefore , a much more centrally concentrated emissivity of comptonized radiation is achieved in models with @xmath28 and the _ observed _ luminosity is subject to stronger reduction by gr effects , discussed below . different velocity fields result in a different strength of collimation toward the black hole horizon . in principle , larger values of the radial velocity , @xmath158 , imply a stronger collimation for @xmath27 , which could give rise to a larger fraction of photons being captured by the black hole . however , the radiation is produced , on average , at a smaller radial distance for @xmath28 and , therefore , the reduction of the observed luminosity due to gr effects , including the photon capture and the gravitational redshift , turns out to be larger in that case . specifically , in models with @xmath28 , @xmath159 for @xmath25 and 0.3 for @xmath156 ; the large reduction for @xmath25 results from @xmath160 ( and trapping of photons ) in the central region . for @xmath27 , @xmath161 ( for @xmath23 ) and 0.9 ( for @xmath86 ; in this case the emissivity profile is flatter ) . for high values of @xmath83 , effects unique for the kerr metric , i.e. bending of photon trajectories to the equatorial plane combined with a dependence of the gravitational shift of energy on the direction of photon escape , result in an intrinsic anisotropy of radiation produced within the innermost few @xmath67 ( see piran & shaham 1977 , niedwiecki 2005 ) . the comptonized component from that region is softer ( i.e. has a larger spectral index and a smaller cut - off energy ) and has a smaller normalization when observed at a smaller @xmath132 . the gravitational effects , underlying this property , are strong only within @xmath162 , then , the magnitude of the anisotropy in the total spectrum depends on the proportion between contributions from the innermost and the surrounding ( beyond @xmath163 ) regions . figs [ fig : spectra](b ) and [ fig : spectra_stel](b ) show the @xmath132-dependence of the observed spectra in our models with @xmath28 ( in models with @xmath27 the dependence on @xmath132 is negligible ) . as we can see , the spectra indeed show some anisotropy but the effect is relatively moderate . for @xmath23 , the total flux emitted face - on ( @xmath1331 ) and edge - on ( @xmath1640.1 ) corresponds to the isotropic luminosity of @xmath165 and @xmath166 erg s@xmath121 , respectively . in this model , the compressive heating results in a rather flat emissivity profile , then , mixing of relatively weak contribution from @xmath167 with radiation produced at more distant regions washes out the anisotropic properties . for @xmath25 , the face - on and edge - on fluxes correspond to @xmath168 erg s@xmath121 and @xmath169 erg s@xmath121 , respectively . in this model , the emissivity is strongly centrally concentrated ; however , the optical depth of the inner region is large and a large fraction of photons produced there get reprocessed at larger distances or captured . at both @xmath26 , the spectral indices differ by @xmath170 between a face - on and an edge - on observer . we expect that a much larger anisotropy would be produced in a model with moderate @xmath26 ( yielding @xmath171 ) and strong direct heating of electrons ( giving a very centrally concentrated emissivity ) . an optically thick disc may extend to relatively low radii in some objects observed at @xmath172 similar to these characterising our models . the best example concerns xte j1118 + 480 , in which the derived truncation radius of the outer cold disc is @xmath173 ( esin et al . we consider here the cooling of the inner hot flow due to comptonization of thermal photons emitted by such a surrounding , cold disc . for this we take the innermost part , within @xmath174 , of our self - consistent solutions described in section [ final ] . we use our mc method to compute the comptonization of the seed photons emitted by a cold , keplerian disc extending from @xmath175 to @xmath176 . the local black - body temperature of its thermal emission is found using page & thorne ( 1974 ) ; note that this approach gives the largest possible luminosity of the outer disc as it involves the assumption that the rate of the outward transport of mechanical energy , which is the dominant source of the radiated energy at @xmath177 , is the same as in an untruncated keplerian disc . we do not look for self - consistent solutions including this additional source of seed photons . instead , we compute the total rate of the compton cooling by the outer disc photons , @xmath178 , and compare it with the total compton cooling rate due to the synchrotron and bremsstrahlung emission , @xmath179 , of our self - consistent solutions . we find that in models with @xmath27 the effect is small , with @xmath180 , and for @xmath28 it is negligible , @xmath181 . we note that the weakness of this effects results from the following properties . first , the total luminosity of the disc truncated at @xmath182 is relatively small ( specifically , it is approximately equal to the luminosity of the hot flow ) . second , the solid angle subtended by the flow as seen by the outer disc is rather small . third , the optical thickness of the outer parts of the flow is very small at the values of @xmath26 considered here , so a small fraction of photons illuminating the flow get scattered . our results are consistent with the conclusion of esin et al . ( 2001 ) that the outer disc is not an important source for compton cooling in this range of parameters . obviously , the effect would be more important for a smaller truncation radius ( cf . esin 1997 ) , at which the luminosity of the disc would be larger , as well as at higher @xmath26 , at which a larger fraction of photons irradiating the flow would be scattered . most of the specific properties discussed in this paper have been studied in previous works . however , their quantitative importance could not be assessed properly as neither of previous studies considered self - consistently all relevant effects . the solution of a hot - flow structure based on a self - similar model ( e.g. , narayan & yi 1995 ) is commonly used due to its relative simplicity . however , this simple solution introduces inaccuracies of up to an order of magnitude even for the basic flow parameters ( e.g. , the density given by eq . 1 in mahadevan & quataert ( 1997 ) is @xmath183 times larger than the density in our solutions ) , which leads to further inaccuracies in papers based on this model ; e.g. , we find that compressive heating dominates over coulomb heating at much higher accretion rates than the critical value given by eq . ( 35 ) in mahadevan & quataert ( 1997 ) . gammie & popham ( 1998 ) and popham & gammie ( 1998 ) presented an extensive study of various aspects of gr hydrodynamical description of a flow in the kerr metric and they show that the black hole spin has a large effect on the velocity field , density and temperature of the flow ( in ways also discussed in the present paper ) ; however , their study neglected radiative processes , so a more detailed impact on the observed spectrum could not be established . manmoto , mineshige & kusunose ( 1997 ) included an explicit description of radiative processes with local approximation of compton cooling in a pseudo - relativistic model . the same model including outflows and direct electron heating was used by yuan et al . ( 2003 , 2005 ) . then , m00 improved such models by including a fully gr treatment of the hydrodynamical processes ( but still with local approximation of compton cooling ) . we have fully reproduced their results in our initial solutions ; the accuracy of these models is directly illustrated by comparison of our initial and final ( global ) solutions . in an alternative ( to mc ) method for the treatment of global comptonization in an optically thin flow , developed by narayan , barret & mcclintock ( 1997 ) , the flow is divided into a set of nested spherical shells and the iterative scattering method is used to describe their mutual interactions . the model was used , e.g. by esin , mcclintock & narayan ( 1997 ) , quataert & narayan ( 1999 ) and esin et al . neither of these papers presents any details which could allow us to compare our cooling rates and only the produced spectra can be compared . the results of esin et al . ( 2001 ) , who use the gr hydrodynamical model from gammie & popham ( 1998 ) and include the special relativistic effects and the gravitational redshift in the calculation of radiative processes , is of particular interest for such a comparison . it appears that the model of narayan et al . ( 1997 ) yields a higher cooling rate than our monte carlo model , as the spectrum computed in esin et al . ( 2001 ) for parameters similar to these in our models ( see the solid curve in their fig . 1 ) has a break energy at @xmath184 kev , significantly lower than our model spectra . we do not know the cause of this difference . among previous studies of hot flows , the fully gr adaf model of kurpiewski & jaroszyski ( 1999 , 2000 ) , with a mc computation of global comptonization , seems to be the closest to our present work . their spectra show a much stronger dependence on @xmath83 , see e.g. figure 4 in kurpiewski & jaroszyski ( 2000 ) , which may be due to their values of @xmath26 significantly smaller than in our models . yuan et al . ( 2009 ) used an iteration method , similar to ours but with a different approach to computing global comptonization , to find the radiative cooling rate mutually consistent between the dynamics and radiation of the flow in a non gr model including strong outflows and electron heating . taking into account significant differences between the results of our present work and x10 , pointed out below , we speculate that gr flows with a different strength of these two effects may be characterised by different critical accretion rates , and different maximum luminosities , than those found in the model of yuan et al . ( 2009 ) . finally , we compare our results with the pseudo - relativistic model of x10 . first , we note that models based on the pseudo - newtonian potential fail to properly describe the innermost region even for the @xmath27 case . specifically , the pseudo - relativistic model predicts a decrease of the optical depth at @xmath185 ( see the dot - dashed curve in fig . 2b in x10 ) while the fully gr model with the same parameters ( see our fig . [ fig : flow]a ) yields the optical depth monotonically increasing with decreasing @xmath21 . the reason for this discrepancy can be traced to the ( unphysically ) large radial velocity in the pseudo - relativistic model , formally exceeding @xmath186 at small @xmath21 , which implies a much smaller density than that in the gr model . second , in the present work we use the equation for vertical equilibrium in the form derived for the kerr metric by abramowicz , lanza & percival ( 1997 ) . in x10 we used its simplified version , @xmath187 , where @xmath188 is the sound speed and @xmath189 is the keplerian angular velocity , and we note significant differences in scale heights derived in these two approaches . specifically , the equation from abramowicz et al . ( 1997 ) yields a larger scale height , up to a factor of @xmath190 , for the flow parameters describing our models . third , additional effects included in x10 , namely the strong outflow and direct viscous heating of electrons affect the flow structure in a complex manner and lead to a qualitatively different effect of taking into account the global nature of the comptonization process . obviously , the outflow leads to a reduction of @xmath191 at small @xmath21 . therefore , @xmath191 is much smaller , by a factor of @xmath183 at @xmath66 , in the model of x10 than in our present model with @xmath27 and @xmath25 ( which parameters are the same as those in x10 ) . then , the outflow reduces the compressive heating of ions via reducing the density gradient , and the direct heating of the electrons reduces the viscous heating of ions . also , the outflow reduces the viscous heating rate per unit volume , flattening its radial profile , but the viscous heating rate per ion remains unchanged . we note that the compressive heating of ions exceeds their viscous heating in our models , and it is much larger , by a factor of @xmath192 , than the compressive heating of electrons . together , both reductions of the ion heating cause a reduction of @xmath68 , which , in turn , leads to a decrease of @xmath78 by @xmath193 per cent ( with respect to the pseudo - relativistic model without both an outflow and electron heating ) within the innermost several @xmath67 . the simplified hydrostatic equilibrium condition together with the outflow and electron heating considered in x10 yield a much smaller @xmath78 than our present model ( compare our fig . [ fig : flow_stel]b with fig . 3b in x10 ) . as a result , in spite of the strong mass loss to the outflow , the total _ radial _ optical depth in the model of x10 is still higher than that in our model with @xmath27 and @xmath25 . also , the radial optical depth of x10 is much larger than @xmath191 ; then , the geometry of the flow resembles that of a slab . fourth , comparing the radiative efficiencies of the x10 model and the present one with @xmath27 , @xmath25 illustrates the strong impact of the direct heating of electrons . namely , in x10 the value of @xmath116 is three times larger , although most of the potential energy of the flow is lost to the outflow in their model . we discuss here briefly our assumption on the strength of the magnetic field as well as our neglect of the viscous heating of electrons . our assumption of rather weak magnetic field , with the magnetic pressure of @xmath194 , is supported by results of the magnetohydrodynamic ( mhd ) simulations in which amplification of magnetic fields by the mri typically saturates at such a ratio of the magnetic to the total pressure ( e.g. machida , nakamura & matsumoto 2004 and references therein ) . it is rather easy to assess changes of the flow structure and the spectrum resulting from a change of this assumption . namely , a stronger magnetic field would result in a stronger synchrotron emission and this would cause a decrease of the electron temperature and a softening of the comptonization spectrum ( see , e.g. , esin et al . 1997 ) . an issue of the direct viscous heating of electrons is more uncertain . current applications of adaf models typically assume large values of electron heating to total heating , @xmath195 ( e.g. , yuan et al . 2003 ) , however , the model has a degeneracy between the value of @xmath196 and the strength of an outflow ( quataert & narayan 1999 ) . various attempts to assess this effect on theoretical grounds seem to favour small values of @xmath196 for weak magnetic fields and rather high @xmath26 , which are considered in our models . analytic investigation of particle heating by mhd turbulence in adafs indicates that the turbulence primarily heats protons for weak magnetic fields , while electrons are primarily heated for strong magnetic fields ( see quataert & gruzinov 1999 ) . however , large uncertainties in the division between electron and proton heating result from uncertainties in the description of the turbulence ; @xmath197 , assumed in our models , may correspond to the values of @xmath196 between @xmath198 and 0.6 ( see fig . 2 in quataert & gruzinov 1999 ) . details of another potential mechanism of electron heating discussed by bisnovatyi - kogan & lovelace ( 1997 ) and quataert & gruzinov ( 1999 ) , i.e. reconnection , remain even more uncertain . simulations of mri turbulence by sharma et al . ( 2007 ) indicate that pressure anisotropy , created in turbulent plasmas , may give an additional mechanism for particle heating ; they approximate the fraction of the viscous energy which heats electrons as @xmath199 . however , their model assumes a fully collisionless plasma , which approximation is valid only for @xmath200 . at higher accretion rates , coulomb collisions suppress the electron pressure anisotropy and electron heating is negligible . then , their prescription for @xmath196 seems to be not relevant for our models . we note that the direct viscous heating would exceed the coulomb and compressive heating of electrons for @xmath201 in models with @xmath27 , and for @xmath202 in models with @xmath28 . for these values of @xmath196 , flows should have larger luminosities than these assessed in our models . we can also expect that flows with such values of @xmath196 produce harder x - ray spectra , with higher cut - off energies , than obtained in our models . however , we note also that large values of @xmath196 , for which the heating of ions is significantly reduced , may have a strong impact on the flow structure , especially if a strong outflow is also present , which may lead to enhanced global compton cooling ( as discussed in section [ previous ] ) . then , the above simple estimation of spectral changes related to the increase of @xmath196 may be incorrect ; in particular , we note that the spectrum obtained in our model with @xmath203 in x10 has a smaller cut - off energy than our spectra computed here with @xmath204 . a substantial observational evidence indicates that accretion flows typically proceed through a radiatively - inefficient mode below a characteristic luminosity of @xmath205 in agns ( e.g. , ho 2008 ; see yuan 2007 for a review on the applications of hot flow models to low luminosity agns ) . fr i radio galaxies make an interesting class of objects observed at such luminosities . they do not show signs of the presence of an optically thick accretion disc ( e.g. , chiaberge , capetti & celotti 1999 ) . remarkably , an average radiative efficiency , @xmath206 , estimated for fr is by balmaverde , baldi & capetti ( 2008 ) is consistent with the values of @xmath116 obtained in most of our solutions . according to our results , hot flows with relevant accretion rates should produce rather hard spectra , extending to mev energies . then , mev observations are crucial for verification whether the high energy component is produced by the thermal comptonization in a hot flow ; however , the present data quality in this energy range is too poor for most objects to allow for such analysis . thanks to its proximity , centaurus a is the best studied fr i radio galaxy , suitable for detailed testing of various accretion models through modelling of its broadband spectrum . its x - ray luminosity ( @xmath207 , see below , where @xmath208 is the 210 kev luminosity ) satisfies the criterion , suggested in wu , yuan & cao ( 2007 ) , for a fr i galaxy to be dominated by a hot flow rather than a jet emission , of @xmath209 . indeed , many works attributed its spectral components to emission from accretion flow ( e.g. evans et al . 2004 ; whysong & antonucci 2004 ; meisenheimer et al . 2007 ) although they can also be explained in terms of jet models ( e.g. abdo et al . 2010 ) . below we briefly compare predictions of our model with the observed properties of the x - ray emission in cen a. \(i ) luminosity . over the last decade , cen a was typically observed with the unabsorbed x - ray luminosity of @xmath210 erg s@xmath121 ( e.g. evans et al . 2004 , markowitz et al . our model with @xmath23 and @xmath28 gives @xmath211 erg s@xmath121 ( assuming @xmath212 ) , which is approximately consistent with the average observed @xmath208 taking into account recent estimates of its black - hole mass of @xmath213 ( e.g. hring - neumayer et al . 2006 ) , which is @xmath75 times smaller than @xmath9 assumed in our model . also for @xmath23 and @xmath27 we get a roughly consistent @xmath214 erg s@xmath121 . \(ii ) spectral index . the time - averaged spectral index of the x - ray emission is @xmath2151.8 ( beckmann et al . 2011 , rothschild et al . the model spectrum for @xmath28 is slightly harder , @xmath216 . the average slope for @xmath27 is roughly consistent with the observed one , but this model spectrum has pronounced scattering bumps , which have not been observed in cen a. \(iii ) high energy cut - off . cen a is the only fr i galaxy with data at mev energies . steinle et al . ( 1998 ) have found a spectral break in the @xmath2151.8 spectrum at @xmath217150 kev . another power law spectral component was found above the break up to the mev range . grandi et al . ( 2003 ) and beckmann et al . ( 2011 ) fitted the hard x - ray spectra by an e - folded power law and found e - folding energies of @xmath217500 kev , which is roughly compatible with the break energy of @xmath217150 kev . rothschild et al . measured the power law spectrum up to 200 kev , within which energy range there was no indication for a break or cut - off . \(iv ) variability . the spectral index remains roughly unchanged while @xmath208 varies by a factor of 3 ( rothschild et al . 2011 ) . in our model , changes of @xmath208 can be attributed to changes of a few different parameters , namely , the strength of the magnetic field , the fraction of accretion power directly heating electrons or the accretion rate . the results of our current study allow us to assess only effects related to changing of @xmath26 ( which case is most commonly considered as the explanation of luminosity variations ) , and we note that the model predicts a small change of the photon index , by @xmath218 , corresponding to the change of @xmath172 by a factor 10 . this appears to be compatible with the reported spectral variability within @xmath1381.8 only . in summary , our model can reproduce most of the properties of the x - ray emission in cen a. however , it is inconsistent with the relatively low observed value of the break energy of @xmath219 kev . furthermore , the @xmath28 model predicts a slightly harder spectrum than that observed . however , an additional process , which would lead to a softening of the spectrum and a decrease of the cut - off energy , has to be included in the model with @xmath220 . namely , an efficient nonthermal synchrotron emission from relativistic e@xmath146 , coming from the decay of charged pions copiously produced through proton - proton collisions in a flow around a rapidly rotating black hole , where @xmath221 k , would provide a strong seed photon input in addition to the thermal synchrotron emission . however , we leave a more detailed comparison of the spectrum observed from cen a with the hot flow model spectra to a future work , in which we will take into account the relevant hadronic processes . black - hole binaries in the hard spectral state have typical x - ray photon spectra ( measured typically in the 310 kev range ) of @xmath2222 ( e.g. , zdziarski & gierliski 2004 ; mcclintock & remillard 2006 ) . in luminous spectral states , @xmath223 or so , the high energy cut - offs in those spectra occur at energies @xmath224200 kev ( e.g. , grove et al . 1998 ; wardziski et al . 2002 ; zdziarski & gierliski 2004 ) , as measured by the position of the spectral turnover in @xmath225 spectra . thus , the x - ray slope of our model spectra , @xmath2221.7 , is in the range of the observed values . on the other hand , the high - energy turnovers in our spectra are at @xmath226 kev , which is higher than the values observed . however , our models have @xmath227@xmath228 , which is lower than @xmath172 at which most of the spectral cut - offs have been measured . when measured , the cut - off energy has been found to increase with the decreasing luminosity ( wardziski et al . 2002 ; yamaoka et al . 2006 ; miyakawa et al . 2008 ) , and thus it is possible that the cut - off energy is at @xmath217500 kev in some low-@xmath51 hard states . the value of the x - ray photon index shows a negative correlation with the luminosity in luminous hard states of black - hole binaries , at @xmath223 or so , i.e. , the spectra soften with the increasing @xmath51 . however , the correlation changes its sign at low @xmath51 , where the spectra soften with the decreasing @xmath51 ( e.g. , corbel et al . 2004 ; zdziarski et al . 2004 ; yuan et al . 2007 ; wu & gu 2008 ; sobolewska et al . 2011 ) . as pointed out by sobolewska et al . ( 2011 ) , this indicates a change in the physical nature of the x - ray source at a luminosity of a fraction of a per cent . those authors discuss two scenarios explaining this change . one is that the emission is dominated by an accretion flow at high @xmath51 , and by a nonthermal jet radiation ( most likely synchrotron ) at low @xmath51 . this scenario for the black - hole binary xte j1550564 is discussed by russell et al . ( 2010 ) , who found the near - infrared / optical ( @xmath229 hz ) emission is tightly correlated with x - rays at low @xmath51 , but not at high @xmath51 . a potential problem for this model is presented by the fact that basically the same @xmath51-@xmath230 correlation , negative at low @xmath172 and positive at high @xmath172 is observed in agns . the transition is at @xmath231 or so , about the same as in black - hole binaries ( e.g. , gu & cao 2009 ; constantin et al . 2009 ; younes et al . 2011 ; veledina et al . 2011 ) . the magnetic field in jets in black - hole sources is expected to scale as @xmath232 ( e.g. , heinz & sunyaev 2003 ) , which ( since the cyclotron frequency is @xmath233 ) very strongly affects the jet emission , leading , in particular , to the strong mass dependence of the relative strength of the jet emission at a given frequency ( merloni , heinz & di matteo 2003 ) . thus , it appears highly unlikely that the x - ray jet emission would start to dominate the accretion - flow emission below about the same eddington ratio in both black - hole binaries and agns ( see also yuan & cui 2005 ; yuan , yu & ho 2009 ) . the second possibility explaining the form of the @xmath51-@xmath230 correlation discussed by sobolewska et al . ( 2011 ) , and independently by veledina et al . ( 2011 ) , is that it is due to a change of the character of the dominant seed photons for comptonization . the overall idea follows the model for spectral changes of black - hole binaries developed by esin et al . ( 1997 , 1998 ) , whose model did predict the changing @xmath51-@xmath230 correlation , see , e.g , fig . 1 in esin et al . ( 1998 ) . at high @xmath51 , above the transition @xmath172 , the seed photons are likely to be from an optically - thick disc overlapping with the hot flow . there is a lot of observational evidence for the presence of an outer cold disc in luminous hard states ( as summarized by done , gierliski & kubota 2007 ) , e.g. , a strong correlation of @xmath230 with the relative strength of compton reflection ( zdziarski , lubiski & smith 1999 ) . also , the inner disc radius decreases with the increasing @xmath51 ( e.g. , done et al . this naturally explains the positive @xmath51-@xmath230 correlation , since the resulting increasing overlap between the cold disc and the hot flow leads to both an increase of the strength of compton reflection and increased cooling of the comptonizing plasma , resulting in softening of the x - ray spectra . however , at low @xmath172 , the outer disc may be truncated far away , and the dominant source of seed photons for comptonization becomes synchrotron emission , as in the models studied in this work as well as in the adaf model in general ( narayan & yi 1995 ) . this model can also explain the correlation of the x - rays with the near - infrared / optical emission , which , in this scenario , would be from the adaf flow , see fig . [ fig : spectra_stel ] . as shown in section [ final ] , our models spectra do harden with the increasing @xmath51 , with @xmath234 for @xmath51 increasing by a factor of several . this strongly supports the suggestion of sobolewska et al . ( 2011 ) that the negative @xmath51-@xmath230 correlation seen at @xmath235 is due to the dominance of synchrotron seed photons in that regime . a hardening of the adaf spectra with increasing @xmath51 , below a few per cent of @xmath236 , was studied previously by esin et al . ( 1997 ) ; their model seems to predict a much stronger effect , with the increase of @xmath26 by a factor of 10 corresponding to @xmath237 . in this work , we have considered only two values of the accretion rate , and have not fully tested this model against observational data . we will consider this issue in detail in a paper in preparation . we note that oda et al . ( 2010 ) propose another possible explanation for the difference between the low and high luminosity hard states . it is a transition between two equilibrium states of an optically thin , two temperature flow with a weak and a strong magnetic field , respectively . such a change of magnetic field is outside the scope of our model . we have developed a fully gr model with global compton scattering of hot , tenuous accretion flows . the dominant source of seed photons in our model is the synchrotron process . we assume that an outer , optically - thick , accretion disc is truncated at large radii , so irradiation of the hot flow by it is negligible . we have also neglected direct viscous electron heating and outflows . we have considered the accretion rates of @xmath97 and 0.5 , for which our model gives the bolometric luminosities of @xmath227@xmath228 . in this range of @xmath172 , we find our models predict x - ray spectra hardening with increasing @xmath51 , which is in agreement with the correlation observed at @xmath235 in both black - hole binaries in the hard state and agns . this is also in agreement with the suggestion of sobolewska et al . ( 2011 ) that synchrotron seed photons dominate in this range of @xmath172 , whereas an optically - thick accretion disc overlapping with the hot flow provide dominant source of seed photons at higher @xmath51 . we have compared our models to the low-@xmath51 fr i radio galaxy cen a. we have found a good agreement with the observed x - ray spectral slope , but the position of the high - energy break , observed at @xmath217200 kev , is lower than the model cut - off energy . this may be possibly explained by the hadronic processes , which would be present if the black hole in this source has a high spin . we have studied the impact of the black hole spin and we find that despite a very strong influence of this parameter on the dynamical properties of the inner part of the flow , the difference between the observed radiation from a flow surrounding a maximally rotating and a non - rotating black hole , with other parameters unchanged , is rather moderate . in particular , their luminosities differ only by a factor of 23 ( i.e. smaller than a factor of @xmath238 in keplerian discs ) . we find this is due to a strong role of the compressive electron heating , which process only weakly depends on the black - hole spin . we note , however , that a much larger dependence on the spin value is expected in models involving direct heating of electrons , or taking into account hadronic processes . we stress that a proper treatment of global compton scattering is crucial for accurate modelling of the flow structure , most importantly , for the electron temperature profile , and for the spectral formation in optically thin flows . in models with @xmath80 , we find the entire inner flow is efficiently cooled by comptonization when the non - local nature of this process is taken into account . at @xmath97 , global comptonization is important , but also electron advection effects are strong at @xmath239 , giving rise to large temperature gradients . as compared to the results with the global comptonization , the local model assuming a slab geometry provides in general a very poor approximation , overestimating the cooling rate by an order of magnitude in the innermost region . we thank the referee and f. yuan for valuable comments . this research has been supported in part by the polish ncn grants n n203 582240 , n n203 581240 , n n203 404939 and 362/1/n - integral/2008/09/0 . fgx thanks dr . yan - rong li ( ihep ) for discussions and sharing his gr disc code for comparisons . fgx has been supported in part by nsfc ( grants 10973003 and 10843007 ) and nbrpc ( grants 2009cb824800 and 2009cb24901 ) .

this paper considers a class of generalized convex games where each player is associated with a convex objective function , a convex inequality constraint and a convex constraint set . the players aim to compute a nash equilibrium through communicating with neighboring players . the particular challenge we consider is that the component functions are unknown _ a priori _ to associated players . we study two distributed computation algorithms and analyze their convergence properties in the presence of data transmission delays and dynamic changes of network topologies . the algorithm performance is verified through demand response on the ieee 30-bus test system . our technical tools integrate convex analysis , variational inequalities and simultaneous perturbation stochastic approximation . ,

1507.01530

recent advances on information technologies facilitate real - time message exchanges and decision - making among geographically dispersed strategic entities . this has boosted the emergence of new generation of networked systems ; e.g. , the smart grid and intelligent transportation systems . these networked systems share some common features : on one hand , the entities do not belong to a single authority and may pursue different or even competitive interests ; on the other hand , each entity keeps private information which is unaccessible to others . it is of great interest to design practical mechanisms which allow for efficient coordination of self - interested entities and ensure network - wide performance . game theory along with its distributed computation algorithms represents a promising tool to achieve the goal . in many applications , distributed computation is executed in uncertain environments . for example , mobile robots are deployed in an operating environment where environmental distribution functions are unknown to robots in advance ; e.g. , @xcite . in traffic pricing , pricing policies of system operators may not be available to drivers . in optimal power flow control , the structural parameters of power systems are of national security interest and kept confidential from the public . the absence of such information makes game components ; e.g. , objective and constraint functions , inaccessible to players . very recently , the informational constraint has been stimulating the investigation of _ adaptive _ algorithms , including @xcite for continuous games and @xcite for discrete games . _ literature review . _ non - cooperative game theory has been widely used as a mathematical framework to reason about multiple selfish decision makers ; see for instance @xcite . these games have found a variety of applications in economics , communication and robotics ; see @xcite . in non - cooperative games , decision making of individuals is inherently distributed . very recently , this attractive feature has been utilized to synthesize cooperative control schemes , and a partial reference list for this regard includes @xcite . the set of papers more relevant to our work is concerned with generalized nash games where strategy spaces are continuous and the actions of players are coupled through utility and constraint functions . generalized nash games are first formulated in @xcite . since then , a great effort has been dedicated to studying the existence and structural properties of generalized nash equilibria in ; e.g. , @xcite and the recent survey paper @xcite . a number of algorithms have been proposed to compute generalized nash equilibria , including ode - based methods @xcite , nonlinear gauss - seidel - type approaches @xcite , iterative primal - dual tikhonov schemes @xcite and best - response dynamics @xcite . as mentioned , the set of papers @xcite investigates the _ adaptiveness _ of game theoretic learning algorithms . however , none of the papers mentioned in the last two paragraphs studies the _ robustness _ of the algorithms with respect to network unreliability ; e.g. , data transmission delays , quantization and dynamically changing topologies . in contrast , the robustness has been extensively studied for consensus and distributed optimization , including , to name a few , @xcite for time - varying topologies , @xcite for quantization and @xcite for time delays . yet the adaptiveness issue has not been addressed in this group of papers . _ contributions . _ in this paper , we aim to solve a class of generalized convex games over unreliable networks where the structures of component functions are unknown to the associated players . that is , we aim to simultaneously address the issues of adaptiveness and robustness for generalized convex games . in the games , each player is associated with a convex objective function and subject to a private convex inequality constraint and a private convex constraint set . the component functions are assumed to be smooth and are unknown to the associated players . we investigate distributed first - order gradient - based computation algorithms for the following two scenarios : [ scenario one ] the game map is pseudo - monotone and the maximum delay ( equivalently , the maximum number of packet dropouts or link breaks ) is bounded but unknown ; [ scenario two ] the inequality constraints are absent , the ( reduced ) game map is strongly monotone and the maximum delay is known . inspired by simultaneous perturbation stochastic approximation for optimization in @xcite , we utilize finite differences with diminishing approximation errors to estimate first - order gradients . we propose two distributed algorithms for the two scenarios and formally prove their asymptotic convergence . the comparison of the two proposed algorithms is given in section [ sec : comparison ] . the analysis integrates the tools from convex analysis , variational inequalities and simultaneous perturbation stochastic approximation . the algorithm performance is verified through demand response on the ieee 30-bus test system . a preliminary version of the current paper was published in @xcite where the adaptiveness issue was not investigated . in this section , we present the generalized convex game considered in the paper . it is followed by the notions and notations used throughout the paper . consider the set of players @xmath0 where the state of player @xmath1 is denoted as @xmath2}\in x_i\subseteq { \mathds{r}}^{n_i}$ ] . the players are selfish and pursue different interests . in particular , given the joint state @xmath3}\in x_{-i } \triangleq \prod_{j\neq i}x_j$ ] of its rivals throughout the paper . ] , each player @xmath1 aims to solve the following program parameterized by @xmath3}\in x_{-i}$ ] : @xmath4}\in x_i}f_i(x^{[i]},x^{[-i]}),\quad { \rm s.t . } \quad g^{[i]}(x^{[i]},x^{[-i]})\leq0,\label{e10}\end{aligned}\ ] ] where @xmath5 and @xmath6 } : { \mathds{r}}^n\rightarrow{\mathds{r}}^{m_i}$ ] with @xmath7 . in the remainder of the paper , we assume that the following properties about problem hold : the maps @xmath8 and @xmath6}$ ] are smooth , and the maps @xmath9})$ ] and @xmath6}(\cdot , x^{[-i]})$ ] are convex in @xmath2}$ ] . the set @xmath10 is convex and compact , and @xmath11 where @xmath12 and @xmath13 with @xmath14}(x ) \leq 0\}$].[asm7 ] we now proceed to provide an equivalent form of problem . to achieve this , we define the set - valued map @xmath15 as follows : @xmath16 } ) = \{x^{[i]}\in x_i \ ; | \ ; g^{[i]}(x^{[i]},x^{[-i]})\leq 0\}.\end{aligned}\ ] ] the set @xmath17})$ ] represents the collection of feasible actions for player @xmath1 when its opponents choose the joint state of @xmath3}\in x_{-i}$ ] . with the map @xmath18 , problem of player @xmath1 is equivalent to the following one : @xmath4}\in x_i^f(x^{[-i]})}f_i(x^{[i]},x^{[-i]}).\label{e11}\end{aligned}\ ] ] given @xmath3}\in x_{-i}$ ] , each player @xmath1 aims to solve problem . the collection of such coupled optimization problems consists of the _ generalized convex game _ ( for short , cvx ) . for the cvx game , we adopt the _ generalized nash equilibrium _ ( for short , gne ) as the solution notion which none of the players is willing to unilaterally deviate from : the joint state @xmath19 is a generalized nash equilibrium of the cvx game if the following holds : @xmath20},\tilde{x}^{[-i]}),\quad \forall x^{[i]}\in x_i^f(\tilde{x}^{[-i]}),\quad \forall i\in v.{\nonumber}\end{aligned}\]][def1 ] denote by @xmath21 the set of gnes of the cvx game . the following lemma verifies the non - emptiness of @xmath21 . the set of generalized nash equilibria of the cvx game is not empty , i.e. , @xmath22.[lem3 ] * proof : * recall that @xmath8 is convex and @xmath23 is compact . hence , @xmath22 is a direct result of @xcite . in the cvx game , the players desire to seek a gne . it is noted that @xmath8 , @xmath6}$ ] , and @xmath10 are private information of player @xmath1 and unaccessible to others . in order to compute a gne , it becomes necessary that the players are inter - connected and able to communicate with each other to exchange their partial estimates of gnes . the interconnection between players will be represented by a directed graph @xmath24 where @xmath25 is the set of edges . the neighbor relation is determined by the dependency of @xmath8 and/or @xmath6}$ ] on @xmath26}$ ] . in particular , @xmath27 if and only if @xmath8 and/or @xmath6}$ ] depend upon @xmath26}$ ] . denote by @xmath28 the set of in - neighbors of player @xmath1 . in this paper , we aim to develop distributed algorithms which allow for the computation of gnes in the presence of the following two challenges . 1 . data transmissions between the players in @xmath29 are unreliable , and subject to transmission delays , packet dropouts and/or link breaks . we let @xmath26}(k - \tau^{[i]}_j(k))$ ] with @xmath30}_j(k)\geq0 $ ] be either @xmath31 the outdated state of player @xmath32 received by player @xmath1 at time @xmath33 due to transmission delays ; or @xmath34 the latest state of player @xmath32 received by player @xmath1 by time @xmath33 due to packet dropouts and/or link breaks . for example , if the link from player @xmath32 to player @xmath1 is broken at time @xmath33 , then player @xmath1 can not receive the message @xmath26}(k)$ ] from player @xmath32 . for this case , player @xmath1 uses the more recent received message sent from player @xmath32 which is @xmath26}(k-\tau_j^{[i]}(k))$ ] . denote by @xmath35}(k-\tau^{[i]}_j(k))\}_{j\in{\mathcal{n}}^{\rm in}_i}$ ] the set of latest states of @xmath36 received by player @xmath1 at time @xmath33 . let @xmath37}_j(k)$ ] be the maximum delay or the maximum number of successive packet dropouts . each player @xmath1 is unaware of the structures of @xmath8 and @xmath6}$ ] but can observe their realizations . that is , if the players input the joint state @xmath38 into @xmath8 , then player @xmath1 can observe the realized value of @xmath39 . the disclosure of the value @xmath39 is case dependent . in @xcite , mobile sensors are unaware of environmental distribution functions but they can observe induced utilities via on - site measurements . this is an example of engineering systems . section [ sec : simulations ] will provide a concrete example of demand response of power networks where the system operator discloses realized values via communication given the decisions of end - users . this is an example of social systems . in the cvx game , each player is subject to an inequality constraint . in optimization literature , lagrangian relaxation is a widely used approach to handle inequality constraints . following this vein , we will perform lagrangian relaxation on the cvx game and obtain the unconstrained convex ( for short , uc ) game . in the uc game , there are two sets of players : the set of primal players @xmath29 and the set of dual players @xmath40 . define the following private lagrangian @xmath41 : @xmath42 } ) \triangleq f_i(x ) + \langle \mu^{[i ] } , g^{[i]}(x)\rangle$ ] , for primal player @xmath43 and dual player @xmath44 . given any @xmath3}\in x_{-i}$ ] and @xmath45}\in{\mathds{r}}^{m_i}_{\geq0}$ ] , each primal player @xmath43 aims to minimize @xmath46 over @xmath2}\in x_i$ ] ; i.e. , @xmath47}\in x_i}\mathcal{h}_i(x^{[i]},x^{[-i]},\mu^{[i]})$ ] , and , instead , the objective of the dual player @xmath44 is to maximize @xmath46 over @xmath45}\in m_i\subseteq{\mathds{r}}^{m_i}_{\geq0}$ ] ; i.e. , @xmath48}\in m_i}-\mathcal{h}_i(x,\mu^{[i]})$ ] where the set @xmath49 is convex , non - empty and will be introduced in the sequel . we let @xmath50 and the set @xmath51 with @xmath52 . the above game among the players in @xmath53 is referred to as the uc game parameterized by the set @xmath54 . the set of @xmath54 will play an important role in determining the properties of the uc game and we will discuss the choice of @xmath54 later . the solution concept for the uc game parameterized by @xmath54 is the standard notion of nash equilibrium given below : the joint state of @xmath55 is a nash equilibrium of the uc game parameterized by @xmath54 if the following holds for all @xmath43 : @xmath56})\leq \mathcal{h}_i(x^{[i]},\tilde{x}^{[-i]},\tilde{\mu}^{[i]}),\quad \forall x^{[i]}\in x_i$ ] , and the following holds for all @xmath44 : @xmath57})\leq \mathcal{h}_i(\tilde{x},\tilde{\mu}^{[i]}),\quad \forall \mu^{[i]}\in m_i$].[def2 ] denote by @xmath58 the set of nes of the uc game parameterized by @xmath54 . we now proceed to illustrate the relation between the uc game and the cvx game . before doing so , let us state the following boundedness assumption on the dual solutions : there is a vector @xmath59 such that for any @xmath60 , @xmath61}\| \leq \vartheta_i$ ] for @xmath43.[asm1 ] we would like to make a remark on assumption [ asm1 ] . for convex optimization , it is shown in @xcite that the lagrangian multipliers are uniformly bounded under the standard slater s condition . this boundedness property is used in @xcite and further in @xcite to solve convex and non - convex programs in centralized and distributed manners . in this paper , assumption [ asm1 ] , on one hand , ensures the existence of gnes , and on the other hand , guarantees the boundedness of estimates and gradients for the case of unknown @xmath62 . we will discuss the verification of assumption [ asm1 ] in section [ sec : discussion].[rem6 ] the following proposition characterizes the relations between the uc game and the cvx game . the following properties hold : 1 . consider any @xmath63 . we have @xmath64 if the following properties hold : * feasibility ; i.e. , @xmath6}(\tilde{x})\leq0 $ ] for all @xmath43 ; * slackness complementarity ; i.e. , @xmath65 } , g^{[i]}(\tilde{x})\rangle = 0 $ ] for all @xmath43 . 2 . suppose assumption [ asm1 ] holds . consider the uc game parameterized by @xmath54 with @xmath66}\in{\mathds{r}}^{m_i}_{\geq0}\;|\;\|\mu^{[i]}\|\leq \vartheta_i+r_i\}$ ] for some @xmath67 . then @xmath68 . in addition , for any @xmath63 , it holds that @xmath64 . [ pro5 ] * proof : * the proof of ( p1 ) is a slight extension of the results in @xcite to our game setup . for the sake of completeness , we summarize the analysis here . since @xmath65 } , g^{[i]}(\tilde{x})\rangle = 0 $ ] , we have the following relation : @xmath69 } ) - \langle\tilde{\mu}^{[i]},g^{[i]}(\tilde{x})\rangle = { \mathcal{h}}_i(\tilde{x},\tilde{\mu}^{[i]}).\label{e7}\end{aligned}\ ] ] since @xmath70 , then the following relation holds for all @xmath2}\in x_i$ ] : @xmath71 } ) -\mathcal{h}_i(x^{[i]},\tilde{x}^{[-i]},\tilde{\mu}^{[i ] } ) \leq 0.\label{e17}\end{aligned}\ ] ] substitute into , and it renders the following for any @xmath2}\in x_i$ ] : @xmath72},\tilde{x}^{[-i ] } ) + \langle\tilde{\mu}^{[i]},g^{[i]}(x^{[i]},\tilde{x}^{[-i]})\rangle,{\nonumber}\end{aligned}\ ] ] and thus @xmath73},\tilde{x}^{[-i]}),\quad \forall x^{[i]}\in x_i^f(\tilde{x}^{[-i]})$ ] , by noting that @xmath74},g^{[i]}(x^{[i]},\tilde{x}^{[-i]})\rangle \leq 0 $ ] for any @xmath75}\in x_i^f(\tilde{x}^{[-i]})$ ] . recall that @xmath6}(\tilde{x})\leq0 $ ] for all @xmath43 . the above arguments hold for all @xmath43 , and thus it establishes that @xmath76 . we now proceed to show ( p2 ) . since @xmath45}\geq0 $ ] , @xmath77 is a positive combination of convex functions of @xmath2}$ ] . thus @xmath77 is convex in @xmath2}$ ] . it is easy to see that @xmath78 is convex ( actually affine ) in @xmath2}$ ] . since @xmath79 and @xmath54 are convex and compact , it follows that @xmath80 by the results on generalized convex games in ; e.g. , @xcite . pick any @xmath70 and then @xmath61}\| \leq \vartheta_i$ ] by assumption [ asm1 ] . we then have the following relation : @xmath81 } ) -{{\mathcal{h}}}_i(\tilde{x},\mu^{[i ] } ) \geq 0,\quad \forall\mu^{[i]}\in m_i,{\nonumber}\end{aligned}\ ] ] and thus , @xmath82}-\tilde{\mu}^{[i ] } , g^{[i]}(\tilde{x})\rangle \leq 0,\quad \forall\mu^{[i]}\in m_i.\label{e6}\end{aligned}\ ] ] we now proceed to verify by contradiction the feasibility of @xmath6}(\tilde{x } ) \leq 0 $ ] and the complementary slackness of @xmath83 } , g^{[i]}(\tilde{x})\rangle = 0 $ ] . assume that @xmath6}(\tilde{x})_{\ell } > 0 $ ] . we choose @xmath45}$ ] such that @xmath45}_{\ell } = \tilde{\mu}^{[i]}_{\ell } + \pi r_i$ ] and @xmath45}_{\ell ' } = 0 $ ] for @xmath84 where @xmath85 is sufficiently small such that @xmath45}\in m_i$ ] . then it follows from that @xmath86}(\tilde{x})_{\ell } \leq 0 $ ] which is a contradiction . hence , we have the feasibility of @xmath6}(\tilde{x})\leq 0 $ ] . combine it with @xmath87}\geq0 $ ] , and it renders that @xmath65}\rangle , g^{[i]}(\tilde{x})\leq 0 $ ] . on the other hand , we let @xmath45 } = 0 $ ] in the relation and have @xmath65 } , g^{[i]}(\tilde{x})\rangle \geq 0 hence , it renders that @xmath65 } , g^{[i]}(\tilde{x})\rangle = 0 $ ] . we reach the desired result by using the property ( p1 ) . the vector @xmath88 represents the column vector with @xmath89 ones . the vector @xmath90}_{n}$ ] is the one in @xmath91 whose @xmath92-th coordinate is one and whose other coordinates are all zero . for any pair of vectors @xmath93 , the relation @xmath94 means @xmath95 for all @xmath96 . since the function @xmath8 is continuous and @xmath10 is compact , then the following quantities are well - defined : @xmath97 . given a non - negative scalar sequence @xmath98 , it is _ summable _ if @xmath99 and _ square summable _ if @xmath100 . in the remainder of the paper , we will use some notions about _ monotonicity _ , and the readers are referred to @xcite for detailed discussion . the mapping @xmath101 is _ strongly monotone _ with constant @xmath102 over @xmath103 if for each pair of @xmath104 , the following holds : @xmath105 the mapping @xmath101 is _ monotone _ over @xmath103 if for each pair of @xmath104 , the following holds : @xmath106 the mapping @xmath101 is _ pseudo - monotone _ over @xmath103 if for each pair of @xmath104 , it holds that @xmath107 implies @xmath108 . it is known that strong monotonicity implies monotonicity , and monotonicity implies pseudo - monotonicity @xcite . given the non - empty , convex , and closed set @xmath109 , the projection operator onto @xmath103 , @xmath110 , is defined as @xmath111 = { \rm argmin}_{x\in z}\|x - z\|$ ] . the following is the non - expansiveness of the projection operator ; e.g. , @xcite . let @xmath103 be a non - empty , closed and convex set in @xmath112 . for any @xmath113 , the following holds for any @xmath114 : @xmath115 - y \|^2 \leq \| z - y \|^2 - \| p_x[z ] - z\|^2 $ ] . [ lem6 ] we define @xmath116}}{\mathcal{h}}_i(x,\mu^{[i]})$ ] as the gradient of the convex function @xmath117},\mu^{[i]})$ ] at @xmath2}$ ] , and @xmath118}}{\mathcal{h}}_i(x,\mu^{[i ] } ) = g^{[i]}(x)$ ] as the gradient of the concave ( actually affine ) function @xmath119 at @xmath45}$ ] . define @xmath120 as the map of partial gradients of the players objective functions : @xmath121}}{\mathcal{h}}_1(x,\mu^{[1]})^t\cdots \nabla_{x^{[n]}}{\mathcal{h}}_n(x,\mu^{[n]})^t\\ & \nabla_{\mu^{[1]}}-{\mathcal{h}}_1(x,\mu^{[1]})\cdots \nabla_{\mu^{[n]}}-{\mathcal{h}}_n(x,\mu^{[n]})]^t.\end{aligned}\ ] ] the map @xmath120 is referred to as the _ game map_. it is well - known that @xmath122 is monotone ( resp . strongly monotone ) over @xmath123 if and only if its jacobian @xmath124 is positive semi - definite ( resp . positive definite ) for all @xmath125 . for @xmath6}(x)\leq0 $ ] being absent , we define @xmath116}}f_i(x)$ ] as the gradient of the convex function @xmath9})$ ] at @xmath2}$ ] . define @xmath126 as the map of partial gradients of the players objective functions : @xmath127}}f_1(x)^t\cdots \nabla_{x^{[n]}}f_n(x)^t]^t.\end{aligned}\ ] ] the map @xmath126 is referred to as the _ ( reduced ) game map _ for @xmath6}(x)\leq0 $ ] being absent . given any convex and closed set @xmath103 , let @xmath128 be the projection operator onto the set @xmath103 . the following lemma is a direct result of the smoothness of the component functions and the compactness of the constraint sets . the image set of the game map @xmath122 is uniformly bounded over @xmath123 . in addition , the game map @xmath122 ( resp . @xmath129 ) is lipschitz continuous over @xmath123 with constant @xmath130 ( resp . @xmath131 ) . [ lem4 ] in this section , we will study the scenarios mentioned in the introduction . proposition [ pro5 ] reveals that it suffices to compute a nash equilibrium in @xmath132 if assumption [ asm1 ] holds . in the remainder of this section , we will suppose assumption [ asm1 ] holds , and then each dual player @xmath1 can define the set of @xmath66}\in{\mathds{r}}^{m_i}_{\geq0}\;|\;\|\mu^{[i]}\|\leq \vartheta_i+r_i\}$ ] with @xmath67 . assumption [ asm1 ] will be justified in section [ sec : assumption ] . in this section , we synthesize a distributed first - order algorithm for the case where the quantity @xmath62 is unknown and the game map @xmath120 is merely pseudo - monotone . algorithm [ ta : algo1 ] is based on projected primal - dual gradient methods where each primal or dual player @xmath1 and dual player @xmath1 are both executed by real player @xmath1 . ] updates its own estimate by moving along its partial gradient with a certain step - size and projecting the estimate onto the local constraint set . recall that @xmath8 and @xmath6}$ ] are unknown to player @xmath1 . then player @xmath1 can not compute partial gradient @xmath116}}{\mathcal{h}}_i$ ] . inspired by simultaneous perturbation stochastic approximation in ; e.g. , @xcite , we use finite differences to approximate the partial gradient @xmath116}}{\mathcal{h}}_i$ ] ; i.e. , for @xmath133 , @xmath134}}{\mathcal{h}}_i(x(k),\mu^{[i]}(k))]_{\ell}{\nonumber}\\ & \approx \frac{1}{2c_i(k)}\big({\mathcal{h}}_i(x^{[i]}(k)+c_i(k)e^{[\ell]}_{n_i},x^{[-i]}(k),\mu^{[i]}(k)){\nonumber}\\ & -{\mathcal{h}}_i(x^{[i]}(k)-c_i(k)e^{[\ell]}_{n_i},x^{[-i]}(k),\mu^{[i]}(k))\big),\label{e51}\end{aligned}\ ] ] where @xmath135}_{n_i}$ ] and @xmath136}_{n_i}$ ] are two - way perturbations . in addition , @xmath118}}{\mathcal{h}}_i(x(k),\mu^{[i]}(k ) ) = g^{[i]}(x(k))$ ] . in algorithm [ ta : algo1 ] , @xmath137}_x(k)$ ] ( resp . @xmath137}_{\mu}(k)$ ] ) is the right - hand side of ( resp . @xmath118}}{\mathcal{h}}_i(x,\mu^{[i]})$ ] ) with delayed estimates . here we assume that player @xmath1 can observe the _ values _ associated with @xmath8 and @xmath6}$ ] and thus @xmath137}_x(k)$ ] and @xmath137}_{\mu}(k)$ ] . the scalar @xmath138 is the perturbation magnitude . when it is small , the finite - difference approximation is close to the partial gradient . in order to asymptotically eliminate the error induced by the finite - difference approximation , @xmath139 needs to be diminishing . the decreasing rate of @xmath139 should match that of computation step - sizes @xmath140 . otherwise , the convergence to nash equilibrium may be prevented . this is captured by ( a4 ) of theorem [ the1 ] . * require : * each primal player @xmath43 chooses the initial state @xmath2}(0)\in x_i$ ] . and each dual player @xmath44 defines the set @xmath141 and chooses the initial state @xmath45}(0)\in m_i$ ] . * ensure : * at each @xmath142 , each player in @xmath53 executes the following steps : \1 . each primal player @xmath43 updates its state according to the following rule:@xmath143}(k+1 ) = \mathbb{p}_{x_i}[x^{[i]}(k ) - \alpha(k){\mathcal{d}}^{[i]}_x(k)],\end{aligned}\ ] ] where the approximate gradient @xmath137}_x(k)$ ] is given by @xmath144}_x(k)]_{\ell } = \frac{1}{2c_i(k ) } \{f_i(x^{[i]}(k)+c_i(k)e^{[\ell]}_{n_i},\lambda_i(k)){\nonumber}\\ & + \langle\mu^{[i]}(k),g^{[i]}(x^{[i]}(k)+c_i(k)e^{[\ell]}_{n_i},\lambda_i(k))\rangle{\nonumber}\\ & -f_i(x^{[i]}(k)-c_i(k)e^{[\ell]}_{n_i},\lambda_i(k)){\nonumber}\\ & - \langle\mu^{[i]}(k),g^{[i]}(x^{[i]}(k ) -c_i(k)e^{[\ell]}_{n_i},\lambda_i(k))\rangle\},\label{e38}\end{aligned}\ ] ] for @xmath133 . each dual player @xmath44 updates its state according to the following rule:@xmath145}(k+1 ) = \mathbb{p}_{m_i}[\mu^{[i]}(k ) + \alpha(k){\mathcal{d}}^{[i]}_{\mu}(k)],\end{aligned}\ ] ] where the gradient @xmath137}_{\mu}(k)$ ] is given by @xmath137}_{\mu}(k ) = \nabla_{\mu^{[i]}}{\mathcal{h}}_i(x^{[i]}(k),\lambda_i(k),\mu^{[i]}(k ) ) = g^{[i]}(x^{[i]}(k),\lambda_i(k))$ ] . repeat for @xmath146 . the following theorem demonstrates that algorithm [ ta : algo1 ] is able to achieve a gne from any initial state in @xmath123 . suppose the following hold : 1 . the quantity @xmath62 is finite ; 2 . assumptions [ asm7 ] and [ asm1 ] hold ; 3 . the sequence of @xmath147 is positive , not summable but square summable ; 4 . the sequence of @xmath148 is summable ; 5 . the game map @xmath122 is pseudo - monotone over @xmath123 . for any initial state @xmath149 , the sequence of @xmath150 generated by algorithm [ ta : algo1 ] converges to some @xmath151 where @xmath76.[the1 ] * proof : * first of all , it is noted that @xmath123 is compact since each @xmath10 and @xmath10 is compact in assumption [ asm1 ] and ( p2 ) in proposition [ pro5 ] as well as @xmath122 is uniformly bounded over @xmath123 in lemma [ lem4 ] . then we write algorithm [ ta : algo1 ] in the following compact form : @xmath152,\label{e19}\end{aligned}\ ] ] where @xmath153}_x(k)^t)^t_{i\in v},(-{\mathcal{d}}^{[i]}_{\mu}(k)^t)^t_{i\in v_m})^t$ ] is subject to time delays and perturbations . then pick any @xmath125 . it follows from the non - expansiveness of the projection operator @xmath154 , lemma [ lem6 ] , that for any @xmath125 , the following relation holds : @xmath155 it follows from that @xmath156 for any @xmath157 and @xmath1 , we define the following : @xmath158}_x(k ) = \nabla_{x^{[i]}}{\mathcal{h}}_i(x^{[i]}(k),\{x^{[j]}(k)\}_{j\in{\mathcal{n}}^{\rm in}_i},\mu^{[i]}(k)),{\nonumber}\\ & \hat{{\mathcal{d}}}^{[i]}_{\mu}(k ) \triangleq \nabla_{\mu^{[i]}}{\mathcal{h}}_i(x^{[i]}(k),\{x^{[j]}(k)\}_{j\in{\mathcal{n}}^{\rm in}_i},\mu^{[i]}(k)),{\nonumber}\\ & \hat{{\mathcal{d}}}(k ) \triangleq ( ( \hat{{\mathcal{d}}}^{[i]}_x(k)^t)^t_{i\in v},(-\hat{{\mathcal{d}}}^{[i]}_{\mu}(k)^t)^t_{i\in v_m})^t,\label{e16}\end{aligned}\ ] ] which are the gradients evaluated at the delay - free states . so , the quantity @xmath159 is free of time delays and perturbations . similarly , for any @xmath157 and @xmath1 , we define the following : @xmath160}_x(k ) = \nabla_{x^{[i]}}{\mathcal{h}}_i(x^{[i]}(k),\lambda_i(k),\mu^{[i]}(k)),{\nonumber}\\ & \tilde{{\mathcal{d}}}^{[i]}_{\mu}(k ) \triangleq \nabla_{\mu^{[i]}}{\mathcal{h}}_i(x^{[i]}(k),\lambda_i(k),\mu^{[i]}(k)),{\nonumber}\\ & \tilde{{\mathcal{d}}}(k ) \triangleq ( ( \tilde{{\mathcal{d}}}^{[i]}_x(k)^t)^t_{i\in v},(-\tilde{{\mathcal{d}}}^{[i]}_{\mu}(k)^t)^t_{i\in v_m})^t,{\nonumber}\end{aligned}\ ] ] which are the gradients evaluated at the delayed states . then the quantity @xmath161 is free of perturbations but subject to time delays . with the above notations at hand , the relation implies the following : @xmath162 now let us examine the term with @xmath163 on the right - hand side of . since @xmath164 and @xmath79 is convex and closed , it then follows from the non - expansiveness of the projection operator @xmath165 , that @xmath166-\mathbb{p}_k[\eta(k)]\|{\nonumber}\\ & \leq\alpha(k)\|{\mathcal{d}}(k)\|.\label{e14}\end{aligned}\ ] ] consequently , it follows from the lipschitiz continuity of the game map @xmath122 and , that @xmath167}_x(k)-\tilde{{\mathcal{d}}}^{[i]}_x(k)\| + \|\hat{{\mathcal{d}}}^{[i]}_{\mu}(k)-\tilde{{\mathcal{d}}}^{[i]}_{\mu}(k)\|){\nonumber}\\ & \leq 2n l_{\theta}\sum_{\tau = k-\tau_{\max}}^{k-1}\|\eta(\tau+1)-\eta(\tau)\|{\nonumber}\\ & \leq 2n l_{\theta}\sum_{\tau = k-\tau_{\max}}^{k-1}\alpha(\tau)\|{\mathcal{d}}(\tau)\|.\label{e45}\end{aligned}\ ] ] since @xmath123 is compact and the image of @xmath122 is uniformly compact , implies that there is @xmath168 such that @xmath169 now consider the term with @xmath170 on the right - hand side in . recall that @xmath123 is compact . by the taylor expansion , we reach that @xmath171}_x(k ) = \nabla_{x^{[i]}}f_i(x^{[i]}(k ) , \lambda_i(k)){\nonumber}\\ & + \sum_{\ell=1}^{m_i}\mu^{[i]}_{\ell}(k)\nabla_{x^{[i ] } } g^{[i]}_{\ell}(x^{[i]}(k),\lambda_i(k))+o(c_i(k))\textbf{1}_{n_i}.\label{e21}\end{aligned}\ ] ] with the above relation , we have the following for @xmath172}(k)-{\mathcal{d}}^{[i]}(k)$ ] : @xmath173}_{\mu}(k)-{\mathcal{d}}^{[i]}_{\mu}(k ) = 0,\quad \tilde{{\mathcal{d}}}^{[i]}_x(k)-{\mathcal{d}}^{[i]}_x(k ) = o(c_i(k))\textbf{1}_{n_i}.\label{e12}\end{aligned}\ ] ] notice that @xmath174 is uniformly bounded . substitute and into , sum over @xmath175 $ ] , and it renders that there is some @xmath176 such that the following estimate holds : @xmath177 since @xmath178 and @xmath179 are summable , then the right - hand side of is finite when we let @xmath180 . we now show the following by contradiction : @xmath181 assume that there are @xmath182 and @xmath183 such that @xmath184 for all @xmath185 . since @xmath186 and @xmath147 is not summable , then we have the following : @xmath187 we then reach a contradiction , and thus holds . equivalently , the following relation holds : @xmath188 since @xmath122 is closed , the above relation implies that there is a limit point @xmath189 of the sequence @xmath150 such that the following holds : @xmath190 . since @xmath122 is pseudo - monotone , then we have the following relation : @xmath191 we now set out to show the following by contradiction : @xmath192 assume that there is @xmath193 such that the following holds : @xmath194 now choose @xmath195 , and define @xmath196 . since @xmath123 is convex , then we have @xmath197 . the following holds : @xmath198 where in the inequality we use . it follows from that the following relation holds for any @xmath195 : @xmath199 since @xmath122 is closed , letting @xmath200 in gives that : @xmath201 , which contradicts . as a result , the relation holds . by @xcite , the relation implies that @xmath202 . we replay @xmath203 by @xmath204 in , and establish the following : @xmath205 since @xmath122 is pseudo - monotone , it follows from that@xmath206 replace @xmath203 with @xmath207 in , apply , and it renders : @xmath208 sum up over @xmath209 $ ] and we have @xmath210 we take the limits on @xmath33 first and then @xmath211 on both sides of . since @xmath147 is square summable and @xmath212 is uniformly bounded , we have @xmath213 it implies that @xmath214 converges . since @xmath207 is a limit point of @xmath150 , @xmath150 converges to @xmath202 . furthermore , we have @xmath76 by ( p2 ) in proposition [ pro5 ] . in the last section , the convergence of algorithm [ ta : algo1 ] is ensured under the mild assumptions : @xmath62 is unknown and the game map @xmath122 is merely pseudo - monotone . this set of assumptions requires the usage of diminishing step - sizes . note that diminishing step - sizes may cause a slow convergence rate . the shortcoming can be partially addressed by choosing a constant step - size when the inequality constraints are absent and ( a5 ) in theorem [ the1 ] is strengthened to the following one : the map @xmath129 is strongly monotone over @xmath79 with constant @xmath215.[asm5 ] * require : * each player in @xmath29 chooses the initial state @xmath2}(0)\in x_i$ ] . * ensure : * at each @xmath142 , each player @xmath43 executes the following steps : \1 . each player @xmath43 updates its state according to the following rule:@xmath143}(k+1 ) = \mathbb{p}_{x_i}[x^{[i]}(k ) - \alpha d^{[i]}(k)],\end{aligned}\ ] ] where the approximate gradient @xmath216}(k)$ ] is given by : @xmath217}(k)+c_i(k)e^{[\ell]}_{n_i},\lambda_i(k)){\nonumber}\\ & -f_i(x^{[i]}(k)-c_i(k)e^{[\ell]}_{n_i},\lambda_i(k))\},\end{aligned}\ ] ] for @xmath133 . repeat for @xmath146 . algorithm [ ta : algo2 ] is proposed to address scenario two . in particular , algorithm [ ta : algo2 ] is similar to algorithm [ ta : algo1 ] , but has two distinctions . firstly , algorithm [ ta : algo2 ] excludes the inequality constraints . because the game map @xmath120 can not be strongly monotone due to @xmath77 being linear in @xmath45}$ ] . secondly , thanks to the strong monotonicity of @xmath120 , a constant step - size replaces diminishing step - sizes . the following theorem summarizes the convergence properties of algorithm [ ta : algo2 ] . suppose the following holds : 1 . the quantity @xmath218 is known ; 2 . assumptions [ asm7 ] and [ asm1 ] hold ; 3 . the step - size @xmath219 ; 4 . the sequence @xmath220 is summable . assumption [ asm5 ] holds . for any initial state @xmath221 , the sequence of @xmath222 generated by algorithm [ ta : algo2 ] converges to @xmath76.[the2 ] * proof : * first of all , for any @xmath157 and @xmath1 , we define @xmath223}(k ) \triangleq \nabla_{x^{[i]}}f_i(x^{[i]}(k),\{x^{[j]}(k)\}_{j\in{\mathcal{n}}^{\rm in}_i})$ ] , which is the gradient evaluated at the delay - free states . so , the quantity @xmath159 is free of time delays and perturbations . similarly , for any @xmath157 and @xmath1 , we define @xmath172}(k ) \triangleq \nabla_{x^{[i]}}f_i(x^{[i]}(k),\lambda_i(k))$ ] which is the gradient evaluated at the delayed states . then the quantity @xmath161 is free of perturbations but subject to time delays . we then write algorithm [ ta : algo2 ] in the following compact form : @xmath224.\label{e39}\end{aligned}\ ] ] it is noticed that @xmath225 is a fixed point of the operator @xmath226 $ ] for any @xmath227 ; i.e. , it holds that @xmath228 $ ] . by this , we have the following relations : @xmath229 - \mathbb{p}_x[\tilde{x}-\alpha \nabla\theta^r(\tilde{x})]\|^2{\nonumber}\\ & \leq \|(x(k ) - \tilde{x})-\alpha(d(k)-\nabla\theta^r(\tilde{x}))\|^2{\nonumber}\\ & = \|x(k ) - \tilde{x}\|^2 - 2\alpha\langle x(k ) - \tilde{x } , \hat{d}(k)-\nabla\theta^r(\tilde{x})\rangle{\nonumber}\\ & + \alpha^2\|(d(k)-\tilde{d}(k))+(\tilde{d}(k)-\hat{d}(k))+(\hat{d}(k)-u)\|^2{\nonumber}\\ & -2\alpha\langle x(k ) - \tilde{x } , d(k)-\tilde{d}(k)\rangle{\nonumber}\\ & -2\alpha\langle x(k ) - \tilde{x } , \tilde{d}(k)-\hat{d}(k)\rangle{\nonumber}\\ & \leq ( 1 - 2\rho\alpha+4l_{\theta^r}^2\alpha^2)\|x(k)-\tilde{x}\|^2{\nonumber}\\ & + 4\alpha^2(\|d(k)-\tilde{d}(k)\|^2+\|\tilde{d}(k)-\hat{d}(k)\|^2){\nonumber}\\ & + 2\alpha\|x(k ) - \tilde{x}\|(\|d(k)-\tilde{d}(k)\|+\|\tilde{d}(k)-\hat{d}(k)\|),\label{e42}\end{aligned}\ ] ] where we use the non - expansiveness property of the projection operator @xmath165 in the first inequality , and the strong monotonicity and the lipschitz continuity of @xmath129 in the last inequality . for the term of @xmath230 in , one can derive the following relations : @xmath231 similar to , we have @xmath232}(k)-d^{[i]}(k ) = o(c_i(k))\textbf{1}_{n_i}.\label{e13}\end{aligned}\ ] ] substituting and into yields : @xmath233 where we use the relation @xmath234 . to study the convergence of @xmath235 in , we define the following notation : @xmath236 then is rewritten as follows : @xmath237 where @xmath238 , @xmath239 and the sequence @xmath240 is diminishing . from the recursion of , we can derive the following relation for any pair of @xmath241 : @xmath242 recall that @xmath220 and thus @xmath243 are summable . take the limits on @xmath33 and @xmath211 at both sides of , and it gives the following relation : @xmath244 on the other hand , one can see that @xmath245 since @xmath219 , then @xmath246 . the combination of and renders that @xmath247 apparently , @xmath248 the combination of and establishes the convergence of @xmath222 to @xmath249 . it completes the proof . the two proposed algorithms are complementary . algorithm [ ta : algo1 ] can address inequality constraints , does not need to know @xmath62 and merely requires the game map to be pseudo monotone . it comes with the price of potentially slow convergence due to the utilization of diminishing step - sizes . in contrast , algorithm [ ta : algo2 ] can not deal with inequality constraints , needs to know @xmath62 and requires the game map to be strongly monotone . it comes with the benefit of potentially fast convergence due to the utilization of a constant step - size . the proposed algorithms rely upon assumption [ asm1 ] . in what follows , we will provide two sufficient conditions for assumption [ asm1 ] . recall that @xmath250 , @xmath251 , @xmath252 and @xmath253 are defined in section [ subsection : notations ] . there exist @xmath254 and @xmath255 such that @xmath256}(\bar{x})\|_{\infty } \geq \sigma'$ ] for all @xmath43.[asm3 ] the vector @xmath257 that satisfies assumption [ asm3 ] is referred to as the _ global _ slater vector . the existence of the global slater vector ensures the boundedness of @xmath58 as follows . if assumption [ asm3 ] holds and @xmath6}_{\ell}$ ] is convex in @xmath38 , then for any @xmath60 , it holds that @xmath61}\|_{\infty } \leq \frac{\sigma_{\max}-\sigma_{\min}}{\sigma'}$ ] for @xmath43.[lem2 ] * proof : * pick any @xmath258 . it holds that @xmath259 choose @xmath260 in , and we have @xmath261},g^{[i]}(\tilde{x})\rangle{\nonumber}\\ & + \sum_{i\in v}\sum_{\ell=1}^{m_i}\tilde{\mu}_{\ell}^{[i]}\langle \nabla_{x^{[i ] } } g^{[i]}_{\ell}(\tilde{x}),\tilde{x}^{[i]}-\bar{x}^{[i]}\rangle{\nonumber}\\ & \geq \langle \nabla\theta^r(\tilde{x}),\tilde{x}-\bar{x}\rangle - \sum_{i\in v}\langle \tilde{\mu}^{[i]},g^{[i]}(\bar{x})\rangle,\label{e56}\end{aligned}\ ] ] where the last inequality uses @xmath87}_{\ell}\geq0 $ ] and taylor theorem and the convexity of @xmath6}_{\ell}$ ] in @xmath38 . it follows from and assumption [ asm3 ] that @xmath262}\|_{\infty}\leq\frac{-\langle \nabla\theta^r(\tilde{x}),\tilde{x}-\bar{x}\rangle}{\sigma'}\leq \frac{\sigma_{\max}-\sigma_{\min}}{\sigma'}.{\nonumber}\end{aligned}\ ] ] in lemma [ lem2 ] , each player has to access the global information of @xmath257 , @xmath263 , @xmath252 and @xmath253 . in what follows , we will derive a sufficient condition which only requires private information of each player @xmath1 to determine an upper bound on @xmath45}$ ] . for each @xmath43 , there exists @xmath264 and @xmath265 such that for any @xmath266 , @xmath256}(\bar{x}^{[i]},\tilde{x}^{[-i]})\|_{\infty } \geq \sigma_i$ ] holds.[asm2 ] the vector @xmath267}$ ] that satisfies assumption [ asm3 ] is referred to as the _ private _ slater vector . one case where assumption [ asm2 ] holds is that @xmath6}$ ] only depends upon @xmath2}$ ] and , for each @xmath43 , there is @xmath267}\in x_i$ ] such that @xmath256}(\bar{x}^{[i]})\|_{\infty } \geq \sigma_i$ ] holds for some @xmath264 . if assumption [ asm2 ] holds , then for any @xmath60 , it holds that @xmath61}\|_{\infty } \leq \frac{\sigma_{i,\max}-\sigma_{i,\min}}{\sigma_i}$ ] for @xmath43.[lem5 ] * proof : * pick any @xmath268 . by assumption [ asm2 ] , there is @xmath267}\in x_i$ ] such that @xmath269}(\bar{x}^{[i]},\tilde{x}^{[-i]})\|_{\infty } \geq \sigma.\end{aligned}\ ] ] since @xmath266 , we then have the following : @xmath270},\tilde{x}^{[-i]},0)\leq { \mathcal{h}}_i(\tilde{x}^{[i]},\tilde{x}^{[-i]},\tilde{\mu}^{[i]}){\nonumber}\\ & \leq { \mathcal{h}}_i(\bar{x}^{[i]},\tilde{x}^{[-i]},\tilde{\mu}^{[i]}){\nonumber}\end{aligned}\ ] ] where in the first inequality we use @xmath271 and in the second inequality we use @xmath267}\in x_i$ ] . the above relations further render the following : @xmath272},\tilde{x}^{[-i ] } ) + \langle \tilde{\mu}^{[i ] } , g^{[i]}(\bar{x}^{[i]},\tilde{x}^{[-i]})\rangle.\label{e24}\end{aligned}\ ] ] recall @xmath256}(\bar{x}^{[i]},\tilde{x}^{[-i]})\|_{\infty } \geq \sigma_i$ ] . then the relation implies the relation of @xmath61}\|_{\infty } \leq \frac{\sigma_{i,\max}-\sigma_{i,\min}}{\sigma_i}$ ] . the current paper imposes several assumptions : ( 1 ) the compactness of @xmath10 ; ( 2 ) the smoothness of component functions ; ( 3 ) identical @xmath140 or @xmath273 for all the players ; ( 4 ) the convexity of component functions and constraints . in particular , the compactness of @xmath10 ensures the uniform boundedness of partial gradients . the convexity guarantees that local optimum are also globally optimal in the sense of nash equilibrium . it is of interest to relax these assumptions . in this section , we will provide a set of numerical simulations to verify the performance of our proposed algorithms . the procedure @xmath274 returns a uniform sample from the set @xmath275 . consider a power network which can be modeled as an interconnected graph @xmath276 where each node @xmath277 represents a bus and each link in @xmath278 represents a branch . the buses in @xmath279 are indexed by @xmath280 and bus @xmath281 denotes the feeder which has a fixed voltage and flexible power injection . a lossless dc model is used to characterize the relation between power generations and loads at different buses and power flows across various branches . each bus is connected to either a power load or supply and each load is associated with an end - user . the set of load buses is denoted by @xmath282 and the set of supply buses is denoted by @xmath283 . the maximum available power supply at bus @xmath284 is denoted by @xmath285 , and the intended power load at bus @xmath286 is denoted by @xmath287 . if the total power supply exceeds the total intended power load , then all the intended loads can be satisfied . otherwise , some loads need to reduce . the reduced load of end - user @xmath1 is denoted by @xmath288 $ ] . the objective function of each end - user @xmath1 is given by @xmath289 . the quantity @xmath290 represents the disutility induced by load reduction @xmath291 with @xmath292 . the scalar @xmath293 is the charged price given the total actual load @xmath294 . the value @xmath295 stands for the benefit produced by load @xmath296 . each end - user @xmath1 aims to minimize its own objective function @xmath297 by adjusting @xmath291 . such decision making is subject to the physical constraints of the power grid . the first constraint is that the total actual load can not exceed the maximum available supply ; i.e. , @xmath298 another set of constraints are induced by the limitations of power flows on branches . let @xmath299^{|{\mathcal{e}}|\times|\mathcal{v}_g|}$ ] ( resp . @xmath300^{|{\mathcal{e}}|\times|\mathcal{v}_l|}$ ] ) be the generation ( resp . load ) shift factor matrix . for @xmath301 , the @xmath302 entry represents the power that is distributed on line @xmath92 when @xmath303 is injected into bus @xmath1 and withdrawn at the reference bus . denote by @xmath304 the maximum capacity of branch @xmath305 . the power flow constraint for branch @xmath305 can be expressed as : @xmath306 where @xmath307_{e\in { \mathcal{e}}_p}$ ] , @xmath308_{i\in\mathcal{v}_s}$ ] , @xmath309_{i\in\mathcal{v}_l}$ ] and @xmath310_{i\in\mathcal{v}_l}$ ] . the above description defines a non - cooperative game among end - users in @xmath311 as follows : @xmath312}c_i r_i + p(\textbf{1}_{|\mathcal{v}_l|}^t(l - r))(l_i - r_i ) - u_i(l_i - r_i){\nonumber}\\ & { \rm s.t.}\quad \textbf{1}_{|\mathcal{v}_l|}^t(l - r ) \leq \textbf{1}_{|\mathcal{v}_s|}^ts{\nonumber}\\ & \quad\quad h_g s - h_l ( l - r ) - f^{\max}\leq 0{\nonumber}\\ & \quad\quad - h_g s + h_l ( l - r ) -f^{\max}\leq 0.\label{e54}\end{aligned}\ ] ] in game , @xmath313 is the pricing policy of load serving entity ( lse ) . this is confidential information of lse and should not be disclosed to end - users . so the objective function @xmath8 is partially unknown to end - user @xmath1 . in addition , the numerical values of generation and load shift factor matrices are of national security interest and are kept confidential from the public . therefore , the power flow constraints are unknown to end - users . assume that all the end - users can communicate with lse . given @xmath314 , lse broadcasts the price value @xmath315 , the power imbalance @xmath316 , and power limit violations @xmath317 and @xmath318 . in this way , end - users can access the values of @xmath8 and @xmath6}$ ] in algorithm [ ta : algo1 ] without knowing their structures . we choose @xmath319 with @xmath320 . for the pricing mechanism , we use ( 4 ) in @xcite ; i.e. , @xmath321 with @xmath322 , and @xmath323 . so @xmath324 . one can compute the following : @xmath325 choose @xmath326 such that @xmath327 . then the monotonicity property holds . assume that there is a global slater vector . figure [ fig1 ] shows the simulation results for the ieee @xmath328-bus test system @xcite and the system parameters are adopted from matpower @xcite . the delays at each iteration are randomly chosen from @xmath329 to @xmath330 . consider @xmath330 players and their components functions are defined as follows : @xmath331})^tx)^2 + ( q^{[i]})^tx,\end{aligned}\ ] ] where * @xmath332 $ ] with @xmath333)$ ] and @xmath334)$ ] ; * @xmath335}\in{\mathds{r}}^{10}$ ] with @xmath335}_i = 1 $ ] and @xmath335}_j = \frac{1}{10}$ ] for @xmath336 ; * @xmath337}\in{\mathds{r}}^{10}$ ] with @xmath337}_j = { \rm sample}([-50\;\;50])$ ] for @xmath338 . one can verify that @xmath339 and @xmath340 in theorem [ the2 ] . by ( b4 ) , we have @xmath341 . in the simulation , we choose @xmath342 . figure [ fig4 ] presents the estimate evolution of the players . we have studied a set of distributed robust adaptive computation algorithms for a class of generalized convex games . we have shown their asymptotic convergence to nash equilibrium in the presence of network unreliability and the uncertainties of component functions and illustrated the algorithm performance via numerical simulations .

let be a class of finite combinatorial structures . the _ profile _ of is the function which counts , for every integer , the number of members of defined on elements , isomorphic structures been identified . the _ generating function of _ is . many results about the behavior of the function have been obtained . albert and atkinson have shown that the generating series of several classes of permutations are algebraic . in this paper , we show how their results extend to classes of ordered binary relational structures ; putting emphasis on the notion of hereditary well quasi order , we discuss some of their questions and answer one . * ams subject classification : * 05c30 , 06f99 , 05a05 , 03c13 . * keywords * : profile , well quasi - ordering , indecomposability , permutations .

1409.1108

the context of this paper is the enumeration of finite relational structures . a relational structure @xmath5 is _ embeddable _ in a relational structure @xmath6 , in notation @xmath7 , if @xmath5 is isomorphic to an induced substructure of @xmath6 . the embeddability relation is a quasi order . several significant properties of relational structures or classes of relational structures can be uniquely expressed in term of this quasi order . this is typically the case of hereditary classes : a class @xmath0 of structures is _ hereditary _ if it contains every relational structure which can be embedded in some member of @xmath0 . interesting hereditary classes abound . in the late forties , frass , following the work of cantor , hausdorff and sierpinski , pointed out the role of the quasi - ordering of embeddability and hereditary classes in the theory of relations ( see his book @xcite for an illustration ) . recent years have seen a renewed interest for the study of these classes , particularly those made of finite structures . many results have been obtained . some are about obstructions allowing to define these classes , others on the behavior of the function @xmath1 , the _ profile _ of @xmath0 which counts , for every integer @xmath2 , the number @xmath3 of members of @xmath0 defined on @xmath2 elements , isomorphic structures being identified . general counting results have been obtained , as well as precise results , for graphs , tournaments and ordered graphs ( see the survey @xcite ) . enumeration results on permutations , motivated by the _ stanley - wilf conjecture _ , solved by marcus and tards ( 2004 ) , fall also under this frame , an important fact due to cameron @xcite . indeed , to each permutation @xmath8 of @xmath9:=\{1 , \dots , n\}$ ] we may associate the relational structure @xmath10 , \leq , \leq_{\sigma})$ ] , that we call _ bichain _ , made of two linear orders on @xmath9 $ ] ( @xmath11 being the natural order on @xmath9 $ ] and @xmath12 the linear order defined by @xmath13 if and only if @xmath14 ) . as it turns out , the order defined on permutations and the embeddability between bichains coincide ( see subsection [ permutation - bichain ] for details and examples ) . in this paper , we show how some results obtained by albert and atkinson @xcite for classes of permutations extend to classes of ordered binary relational structures . we prove notably theorem [ theo : algebraic ] . for this purpose , we recall in section [ sect : basic ] some basic definitions of the theory of relations , we survey in section [ sect : bichains ] some results concerning classes of permutations and show how permutations are related to relational structures . then , we illustrate the role of indecomposable structures ( see section [ sect : indecomposable ] ) and of well quasi order ( see section [ sect : well - quasi - order ] ) in enumeration results . finally , in section [ sect : conjecture ] , we present a conjecture and a partial solution , a special case answering a question of albert and atkinson @xcite . our results have been presented at the international conference on discrete mathematics and computer science ( dimacos11 ) held in mohammedia , morocco , may-5 - 8 , 2011 , and at the international symposium on operational research ( isor11 ) , held in algiers , algeria , may 30-june 2 , 2011 @xcite . we are pleased to thanks the organizers of these meetings for their help . our terminology agree with @xcite . let @xmath2 be a positive integer . a @xmath2-ary relation with _ @xmath15 is a subset @xmath16 of the @xmath2-th power @xmath17 of @xmath15 ; for @xmath18 and @xmath19 we use the words _ unary relation _ and _ binary relation _ , in this later case we rather set @xmath20 instead of @xmath21 . a _ relational structure _ with _ domain _ @xmath15 is a pair @xmath22 made of a set @xmath15 and a family @xmath23 of @xmath24-ary relations @xmath25 on @xmath15 , each @xmath25 being a subset of @xmath26 . the family @xmath27 is the _ signature _ of @xmath5 . we will denote by @xmath28 the domain of @xmath29 . we denote by @xmath30 the class of these structures and by @xmath31 the subclass of the finite ones . a relational structure @xmath5 is _ ordered _ if it can be expressed as @xmath32 where @xmath33 is a linear order on @xmath15 and the @xmath34 s are @xmath35-ary relations ; the ( truncated ) signature in this case is @xmath36 . a relational structure @xmath37 is a _ binary relational structure _ , _ binary structure _ for short , if each @xmath25 is a binary relation ; the class of those finite binary structures will be denoted by @xmath38 instead of @xmath39 basic examples of ordered binary structures are chains @xmath40 , bichains ( @xmath41 and @xmath42 is a linear order ) and multichains ( @xmath34 is a linear order for all @xmath43 ) . we denote by @xmath44 the collection of finite ordered binary structures made of a linear order and @xmath45 binary relations . let @xmath22 be a relational structure ; the _ substructure induced by @xmath5 on a subset @xmath46 of @xmath15 _ , simply called the _ restriction of @xmath5 to @xmath46 _ , is the relational structure @xmath47 where @xmath48 let @xmath22 and @xmath49 be two relational structures of the same signature @xmath50 a map @xmath51 is an _ isomorphism from @xmath5 onto @xmath6 _ if @xmath52 is bijective and @xmath53 if and only if @xmath54 for every @xmath55 , @xmath56 the relational structure @xmath5 is _ isomorphic _ to @xmath6 if there is some isomorphism from @xmath5 onto @xmath6 , it is _ embeddable _ into @xmath6 , and we set @xmath7 , if @xmath5 is isomorphic to some restriction of @xmath57 the embeddability relation ( called `` abritement '' by _ frass _ in french ) is a quasi - order . a class @xmath0 of relational structures is _ hereditary _ if @xmath58 and @xmath59 imply @xmath60 relational structures which are not in @xmath0 are _ obstructions _ to @xmath0 . the _ age _ of a relational structure @xmath29 is the class @xmath61 of finite @xmath62 which are embeddable into @xmath29 ( equivalently , this is the set of finite restrictions of @xmath29 augmented of their isomorphic copies ) . an age is non - empty , hereditary and up - directed ( that is for every @xmath63 there is some @xmath64 which embeds @xmath62 and @xmath65 ) . in the terminology of posets , this is an _ ideal _ of @xmath31 . if the signature is finite , every ideal of @xmath31 is the age of some relational structure ( _ frass _ 1954 ) . if @xmath66 is a subset of @xmath31 then @xmath67 denotes the subclass of members of @xmath31 which embed no member of @xmath66 . clearly , @xmath67 is an hereditary class . moreover , every hereditary subclass @xmath0 of @xmath31 has this form . this fact , due to _ frass _ , is based on the notion of bound : a _ bound _ of an hereditary subclass @xmath0 of @xmath31 is every finite @xmath5 not in @xmath0 such that every @xmath6 which strictly embeds into @xmath5 belongs to @xmath0 . clearly , every finite obstruction to @xmath0 contains a bound . hence , if @xmath68 denotes the collection of bounds of @xmath0 considered up to isomorphism then @xmath69 . + the _ profile _ of an hereditary class @xmath0 is the function @xmath70 which counts , for every @xmath2 , the number of members of @xmath0 defined on @xmath2 elements , isomorphic structures been identified . the _ generating function for _ @xmath0 is @xmath4 . these two notions are the specialization to hereditary classes of basic notions in enumeration . many results on the enumeration of classes of permutations are about the enumeration of relational structures . indeed , as mentioned in the introduction , permutations can be considered as special cases of binary structures , and more specifically ordered binary structures , in fact bichains . we introduce these notions below and point out the relationship between permutations and bichains in the next section . let @xmath2 be a non negative integer . let @xmath71 be the set of permutations on @xmath9:=\{1,\dots , n\}$ ] and @xmath72 an order relation on @xmath73 is defined as follows : the permutation @xmath74 of @xmath9 $ ] _ contains _ the permutation @xmath8 of @xmath75 $ ] and we write @xmath76 if some subsequence of @xmath74 of length @xmath77 is order isomorphic to @xmath8 . more precisely , @xmath76 if there exist integers @xmath78 such that for @xmath79 + @xmath80 for example , @xmath81 contains @xmath82 as it can be seen by considering the subsequence @xmath83 ( @xmath84 ) . a subset @xmath85 of @xmath73 is _ hereditary _ if @xmath86 implies @xmath87 . its counting function , that we call the _ profile _ of @xmath85 , is @xmath88 . how much does @xmath3 drop from @xmath89 if @xmath90 the _ stanley - wilf _ conjecture asserted that it drops to exponential growth . the conjecture was proved in @xmath91 by _ marcus and tards _ @xcite : if @xmath85 is a proper hereditary set of permutations , then , for some constant @xmath92 for every @xmath2 . _ kaiser and klazar _ @xcite proved that if @xmath85 is hereditary , then , either @xmath93 is bounded by a polynomial and in this case is a polynomial , or is bounded below by an exponential , in fact the _ generalized fibonacci function _ @xmath94 . + we recall that the generalized fibonacci number is given by the recurrence @xmath95 for @xmath96 , @xmath97 and @xmath98 for @xmath99 . @xmath94 is the coefficient of @xmath100 in the power series expansion of the expression @xmath101 the kaiser and klazar theorem reads as follow : if @xmath85 is an hereditary set of permutations , then exactly one of the four cases occurs . 1 . for large @xmath2 , @xmath102 is eventually constant . there are integers @xmath103 @xmath104 and @xmath105 , such that @xmath106 for large @xmath2 . moreover , @xmath107 for every @xmath108 3 . there are constants @xmath109 in @xmath110 , @xmath111 , such that @xmath112 for every @xmath2 . 4 . one has @xmath113 for every @xmath2 . in the cases ( 1 ) to ( 3 ) the generating function is rational . _ albert and atkinson _ gave in 2005 examples of hereditary classes whose generating function is algebraic @xcite . in order to state their result , we recall first that a power series @xmath114 with @xmath115 in @xmath116 is _ algebraic _ if there exists a nonzero polynomial @xmath117 in @xmath118 $ ] such that @xmath119 . the series @xmath120 is _ rational _ if @xmath121 has degree 1 in @xmath122 , that is , @xmath123 for two polynomials in @xmath124 $ ] where @xmath125 . recall next that a permutation @xmath126 of @xmath9 $ ] is _ simple _ if no proper interval of @xmath9 $ ] ( @xmath127 , \varnothing$ ] or @xmath128 ) is transformed into an interval . in other words , @xmath129 is not an interval in @xmath9 $ ] for every @xmath130 , and either @xmath131 or @xmath132 . if @xmath133 all permutations are simple and called trivial . albert and atkinson s theorem is the following : if @xmath0 is an hereditary class of permutations containing only finitely many simple permutations , then the generating series of @xmath0 , namely @xmath134 is algebraic . as an illustration of this result , let us mention that the class of permutations not above @xmath135 and @xmath136 contain no non trivial simple permutation ( these permutations are called _ separable permutations _ ) . the generating series of this class is @xmath137 ( see @xcite ) . the simple permutations of small degree are @xmath138 let @xmath139 be the number of simple permutations of @xmath9 $ ] . the values of @xmath139 for @xmath18 to @xmath140 are : @xmath141 @xcite . asymptotically , @xmath139 goes to @xmath142 , a result obtained independently in @xcite . let @xmath8 be a permutation of @xmath9 $ ] . to @xmath8 we associate the _ bichain _ @xmath143 , \leq , \leq_{\sigma})$ ] where @xmath11 is the natural order on @xmath9 $ ] and @xmath12 the linear order defined by @xmath13 if and only if @xmath14 . for example , let @xmath8 be the permutation of @xmath144 given by the sequence of its values : @xmath145 . the sequence of elements of @xmath144 ordered according to @xmath12 is : @xmath146 . hence , this is the sequence of values of @xmath147 , the inverse of @xmath8 . let us represent @xmath8 by its graph in the product @xmath148 , that is the set @xmath149 and order this set componentwise , that is set @xmath150 if @xmath151 and @xmath152 . since @xmath8 is bijective , the poset @xmath153 is the intersection of two linear orders , given respectively by the natural order on the first and on the second coordinate . if we identify each @xmath154 to @xmath155 , the order induced on @xmath156 is the intersection of @xmath11 and @xmath12 . see figure [ critique ] . 1 . if @xmath157 is a finite bichain then @xmath158 is isomorphic to a bichain @xmath159 for a unique permutation @xmath8 on @xmath160.$ ] 2 . if @xmath8 and @xmath74 are two permutations then @xmath76 if and only if @xmath161 the correspondence between permutations and bichains was noted by cameron @xcite ( who rather associated to @xmath8 the pair @xmath162 . it allows to study classes of permutations by means of the theory of relations . in particular , via this correspondence , hereditary classes of permutations correspond to hereditary classes of bichains and , as we will see below , simple permutations correspond to indecomposable bichains . let @xmath37 be a binary structure . a subset @xmath46 of @xmath15 is an _ interval _ of @xmath5 if for each @xmath163 : @xmath164 the empty set , the singletons and the whole set @xmath15 are intervals and said _ trivial_. if @xmath5 has no non trivial interval it is _ indecomposable_. for example , if @xmath165 is a chain , its intervals are the ordinary intervals . if @xmath166 is a bichain then @xmath46 is an interval of @xmath5 if and only if @xmath46 is an interval of @xmath167 and @xmath168 . hence : a permutation @xmath8 is simple if and only if the bichain @xmath159 is indecomposable . the notion of indecomposability is rather old . the notion of interval goes back to frass @xcite , see also @xcite . a fundamental decomposition result of a binary structures into intervals was obtained by gallai @xcite ( see @xcite for further extensions ) . hence , it is not surprising that several results on simple permutations were already known ( for example their asymptotic evaluation ) . albert and atkinson result recasted in terms of relational structures asserts that if @xmath85 _ is an hereditary class of finite bichains containing only finitely many indecomposable bichains then the generating series of @xmath85 is algebraic_. the paper @xcite contains several examples of infinite bichains @xmath169 whose infinitely many members of @xmath170 are indecomposable . + we will establish an extension to ordered binary structures in theorem [ theo : algebraic ] . in the sequel , we recall the facts we need on lexicographic sums and the links with the indecomposability notion . some are old ( the notion of lexicographic sum goes back to cantor ) . let @xmath37 be a binary structure and @xmath171 be a family of binary structures @xmath172 , indexed by the elements of @xmath15 . we suppose that @xmath15 and the @xmath173 are non - empty . the _ lexicographic sum _ of @xmath174 over @xmath5 , denoted by @xmath175 , is the binary structure @xmath176 obtained by replacing each element @xmath177 by the structure @xmath178 . more precisely , @xmath179 where @xmath180 and for each @xmath163 , @xmath181 if either @xmath182 and @xmath183 or @xmath184 and @xmath185 . trivially , if we replace each @xmath178 by an isomorphic binary structure @xmath186 , then @xmath187 is isomorphic to @xmath188 hence , we may suppose that the domains of the @xmath178 s are pairwise disjoint . in this case we may slightly modify the definition above , setting @xmath189 and for two elements @xmath190 and @xmath191 , @xmath192 with this definition , each set @xmath173 is an interval of the sum @xmath193 furthermore , let @xmath194 be the quotient of @xmath195 made of blocks of this partition into intervals , let @xmath196 be the natural projection and let @xmath197 be the image of @xmath198 ( that is @xmath199 where @xmath200 . if we identify each block @xmath173 to the element @xmath201 then @xmath29 and @xmath197 coincide on pairs of distinct elements . they coincide if we consider only reflexive relations . conversely , if @xmath202 is a binary structure and @xmath203 is a partition of @xmath195 into non empty intervals of @xmath204 then @xmath198 is the lexicographic sum of @xmath205 over the quotient @xmath206 in simpler words : the decompositions of a binary structure into lexicographic sums are in correspondence with the partitions of its domain into intervals . an important property of these decompositions is the following : the set of partitions of @xmath15 into intervals of @xmath29 , once ordered by refinement , is a sublattice of the set of partitions of @xmath15 . let us illustrate . let us say that a lexicographic sum @xmath175 is _ trivial _ if @xmath207 or @xmath208 for all @xmath177 , otherwise it is _ non trivial _ ; also a binary structure is _ sum - indecomposable _ if it can not be isomorphic to a non trivial lexicographic sum . we have immediately : a binary structure is sum - indecomposable if and only if it is indecomposable . [ pro : sumlex ] let @xmath5 be a finite binary structure with at least two elements . then @xmath5 is isomorphic to a lexicographic sum @xmath209 where @xmath210 is indecomposable with at least two elements . moreover , when @xmath62 has at least three elements , the partition of @xmath29 into intervals is unique . if the set @xmath62 in proposition [ pro : sumlex ] has two elements , then the decomposition is not necessary unique , a fact which leads to the notion of strong interval . we recall that an interval @xmath46 of a binary structure @xmath5 is _ strong _ if it is non - empty and overlaps no other interval , meaning that if @xmath158 is an interval such that @xmath211 then either @xmath212 or @xmath213 . we say that @xmath46 is _ maximal _ if it is maximal for inclusion among strong intervals distinct from the domain @xmath15 of @xmath29 . the maximal strong interval form a partition of @xmath15 , provided that some maximal exists ; in this case @xmath15 is _ non - limit _ @xcite or , equivalently , _ robust _ evidently , this partition exists whenever @xmath15 is finite . the reader will easily check that when this partition exists and the quotient is indecomposable then every other non - trivial partition into intervals is finer . hence , in proposition [ pro : sumlex ] above , if @xmath62 has at least three elements , the intervals in the decomposition are strong and thus the decomposition is unique . let us say that @xmath214 is _ chainable _ if there is a linear order , @xmath11 , on @xmath15 such that , for each @xmath154 , @xmath215 for every @xmath216 such that @xmath217 . if @xmath29 is reflexive , this amounts to the fact that each @xmath25 is either the equality relation @xmath218 , the complete relation @xmath219 or a linear order ; moreover if @xmath25 and @xmath34 are two linear orders , they coincide or are opposite . note that if furthermore @xmath29 is ordered then the @xmath25 which are linearly ordered are equal or opposite to the given order . we arrive to the fundamental decomposition theorem of gallai @xcite ( see for example @xcite , @xcite , @xcite for extensions to infinite structures ) [ theo : decomposition ] let @xmath29 be a finite binary structure with at least two elements , then @xmath29 is a lexicographic sum @xmath220 where @xmath62 is either indecomposable with at least three elements or a chainable binary structure with at least two elements and the @xmath221 s are strong maximal intervals of @xmath29 . for our purpose , we need to introduce the following notion . let @xmath222 be an ordered structure with two elements . an ordered structure @xmath62 is said @xmath222-_indecomposable _ if it can not be decomposed into a lexicographic sum indexed by @xmath223 if @xmath224 is a class of structures , we denote by @xmath225 the set of members of @xmath224 which are @xmath222-indecomposable . [ lem : decomposition ] let @xmath226 , with @xmath227 , be an ordered structure with two elements . if @xmath29 is a lexicographic sum @xmath228 and @xmath229 is @xmath62-indecomposable , then the partition @xmath230 is unique . we extend the notion of lexicographic sum to collections of non - empty binary structures . given a non - empty binary structure @xmath5 and classes @xmath231 of non - empty binary structures for each @xmath232 , let us denote by @xmath233 the class of all binary structures of the form @xmath175 with @xmath234 . if @xmath235 for every @xmath236 we denote this class by @xmath237 . if @xmath238 , with @xmath227 , @xmath239 and @xmath240 , we set @xmath241 . also if @xmath224 and @xmath242 are two classes of binary structures , we set @xmath243 . we say that a collection @xmath85 of binary structures is _ sum - closed _ if @xmath244 . the _ sum - closure _ @xmath245 of @xmath0 is the smallest sum - closed set that contains @xmath0 . if we define @xmath246 and @xmath247 , then @xmath248 . if @xmath0 is a class of bichains , @xmath245 is also a class of bichains and the class of corresponding permutations is said _ wreath - closed _ @xcite . if @xmath85 is made of reflexive structures and contains a one element structure , say @xmath249 , then @xmath250 . if in addition @xmath85 contains the empty structure then @xmath251 is hereditary . we denote by @xmath252 the collection of finite indecomposable members of @xmath253 if @xmath5 is a binary structure , we denote by @xmath254 the collection of its finite induced substructures which are indecomposable . for example , if @xmath5 is a cograph or a serie - parallel poset then the members of @xmath254 have at most two elements ( a graph ( undirected ) is a _ cograph _ if no induced subgraph is isomorphic to @xmath255 , the path on @xmath256 vertices , and a poset is _ serie - parallel _ if its comparability graph is a cograph ) . + let @xmath257 be a hereditary class of @xmath252 . set @xmath258 . [ theo : closed ] if all members of @xmath259 are reflexive , then @xmath260 inclusion @xmath261 holds under assumption that all members of @xmath257 are reflexive . conversely , if @xmath262 then either @xmath5 is indecomposable in which case @xmath263 or @xmath5 can not be expressed as a lexicographic sum of structures of @xmath257 hence @xmath264 in the sequel , we consider only ordered structures made of reflexive binary relations . let @xmath265 be the subclass of reflexive members of @xmath44 . let @xmath224 be a subclass of @xmath265 ; for @xmath266 let @xmath267 , resp . @xmath268 , be the subclass made of its members which have @xmath154 elements , resp . at least @xmath154 elements . [ lem : union ] let @xmath259 be a class made of non - empty indecomposable members of @xmath265 such that @xmath269 is reduced to the one - element structure @xmath270 . let @xmath224 be the sum - closure of @xmath259 and for each @xmath271 , let @xmath272 be the subclass of @xmath62-indecomposable members of @xmath224 . set @xmath273 if @xmath274 and otherwise set @xmath275 if @xmath271 and @xmath226 with @xmath227 . then : @xmath276 and @xmath277 for every @xmath278 . furthermore , all sets in equation ( [ eq : classe ] ) are pairwise disjoint . let s denote by @xmath279 ( respectively by @xmath280 ) the left - hand side ( respectively the right - hand side ) of equation [ eq : classe ] . inclusion @xmath281 is obvious because @xmath224 is sum - closed according to theorem [ theo : closed ] . to prove inclusion @xmath282 , let @xmath29 be in @xmath279 , if @xmath29 has one element then it is in @xmath280 , otherwise , according to theorem [ theo : decomposition ] , @xmath29 is a lexicographic sum @xmath283 where @xmath62 is either indecomposable with at least three elements or a chainable binary structure with at least two elements and each @xmath284 is a strong interval of @xmath29 for each @xmath285 . in the first case , @xmath29 is in @xmath286 , hence in @xmath287 in the second case , @xmath62 is chainable with @xmath2 elements , @xmath288 , and we may set @xmath289 with @xmath290 . set @xmath291 , @xmath292 , @xmath293 and @xmath294 . we have obviously @xmath295 . since @xmath229 is a strong interval of @xmath29 , @xmath296 is @xmath65-indecomposable , hence @xmath29 belongs to @xmath297 which is a subset of @xmath287 the fact that these sets are pairwise disjoint follows from proposition [ pro : sumlex ] and lemma [ lem : decomposition ] . equality [ eq : classeindec ] is obvious : the @xmath62-indecomposable members of @xmath224 are those which can not be writen as @xmath62-sums . in the sequel , we count . our structures being ordered we may choose a unique representative of an @xmath2-element structure on the set @xmath298 , the ordering being the natural order . let @xmath299 and @xmath300 be the generating series of @xmath301 and @xmath302 and let @xmath303 be the series obtained by substituting the indeterminate @xmath304 by @xmath299 . let @xmath305 and @xmath306 be the generating series of @xmath307 and @xmath308 for @xmath309 . and let @xmath310 be the cardinality of @xmath311 . [ lem : hereditary - series ] @xmath312 @xmath313 let us prove that equation [ eq2 ] holds . let @xmath314 . since by definition in lemma [ lem : union ] , @xmath315 , we have @xmath316 . from equation ( [ eq : classeindec ] ) , we deduce @xmath317 . since the coefficients of @xmath299 are non - negative , the series @xmath318 is invertible , hence @xmath319 as claimed in equation ( [ eq2 ] ) . let us prove that equation [ eq1 ] holds . let @xmath320 , with @xmath321 . since by definition in lemma [ lem : union ] , @xmath322 , we have @xmath323 . from this , we deduce that the generating series of @xmath324 is equal to @xmath325 . from equation ( [ eq2 ] ) , we deduce that the generating series of @xmath326 is @xmath327 . hence the generating series of @xmath328 is equal to @xmath329 . substituting these values in equation ( [ eq : classe ] ) , we obtain @xmath330 a straightforward computation yields equation ( [ eq1 ] ) . let us say that a class of finite structures is _ algebraic _ if its generating series is algebraic . [ cor : wqoalgebraic ] let @xmath259 be a class made of non - empty indecomposable members of @xmath265 such that @xmath269 is reduced to the one - element structure @xmath270 . if @xmath259 is algebraic then its sum - closure and the subclass @xmath308 consisting of the @xmath62-indecomposable members of the sum - closure @xmath224 are algebraic for each @xmath331 , . let @xmath0 be a subclass of @xmath332 and @xmath333 be a poset . set @xmath334 and @xmath335 if there is an embedding @xmath336 such that @xmath337 for all @xmath338 . + we recall that @xmath333 is _ well - quasi ordered ( wqo ) _ if @xmath333 contains no infinite antichain and no infinite descending chain . we say that @xmath0 is _ hereditary wqo _ if @xmath339 is _ wqo _ for every wqo @xmath333 . it is clear that every class which is hereditary wqo is wqo . if @xmath0 is reduced to a single structure @xmath5 , it is hereditary wqo provided that @xmath29 is finite ( this follows from the fact that if @xmath340 is wqo then its power @xmath341 ordered coordinatewise is wqo for each integer @xmath2 ) . if @xmath29 is infinite , this does not hold . also , a finite union of hereditary wqo classes is hereditary wqo ; hence every finite subclass @xmath85 of @xmath342 is hereditary wqo . a longstanding open question ask whether @xmath0 is hereditary wqo whenever the class @xmath343 of the elements of @xmath0 labelled by @xmath344 , the @xmath345-element antichain , is wqo . if @xmath346 is the class of finite chains , @xmath347 identifies to the set @xmath348 of finite words over the alphabet @xmath340 equipped with the higman ordering . the fact that @xmath346 is hereditary wqo is a famous result due to higman @xcite . we also note the following fact : if a subclass @xmath85 of @xmath31 ( with @xmath349 finite ) is hereditary wqo , then @xmath350 , the least hereditary subclass of @xmath31 containing @xmath85 , is hereditary wqo . we recall the following result of @xcite . [ thm : bounds ] if the signature is finite , a subclass of @xmath31 which is hereditary and hereditary wqo has finitely many bounds . behavior of the profile of special hereditary classes , the _ ages _ of frass , and the link with _ wqo _ classes were considered by the second author in the early seventies ( see @xcite and @xcite for a survey ) . the case of graphs , tournaments and other combinatorial structures was elucidated more recently ( see the survey of @xcite ) . + [ prop : wqo ] if a hereditary class @xmath257 of @xmath252 is hereditary _ wqo _ then @xmath351 is hereditary _ wqo _ and @xmath351 has finitely many bounds . the second part of the proposition follows from theorem [ thm : bounds ] above . in our case of binary structures , we may note that the proof is straightforward . the first part uses properties of wqo posets , and follows from higman theorem on algebras preordered by divisibility ( 1952 ) @xcite . instead of recalling the result we give a direct proof . let @xmath340 a poset which is wqo and consider @xmath352 if @xmath353 is not wqo , then according to one of preliminary result of higman , it contains some non finitely generated final segment ( @xmath354 is a final segment if @xmath355 and @xmath356 imply @xmath357 ) . according to zorn lemma , there is a maximal one , say @xmath354 , with respect to inclusion among final segments having this property . let @xmath358 be the complement of @xmath354 in @xmath352 the set @xmath359 is then wqo . let @xmath360 be an infinite antichain of minimal elements of @xmath361 as @xmath362 is wqo because @xmath259 is hereditary wqo , we can suppose that no element of this antichain is in @xmath363 then , according to proposition [ pro : sumlex ] and theorem [ theo : decomposition ] , for every @xmath364 there exists an indecomposable structure @xmath365 and non - empty structures @xmath366 such that @xmath367 . since @xmath368 strictly embeds into @xmath369 we have @xmath370 for every @xmath364 and @xmath371 . since @xmath359 is wqo , and @xmath259 is hereditary wqo , @xmath372 is wqo , thus the infinite sequence @xmath373 of @xmath372 , where @xmath374 , contains an increasing pair @xmath375 for some @xmath376 . which means that there is an embedding @xmath377 such that @xmath378 for all @xmath379 , that is @xmath380 for all @xmath381 . it follows that @xmath382 which contradicts that @xmath383 is an antichain . thus @xmath353 is wqo and hence @xmath384 is hereditary wqo . proposition [ prop : wqo ] particulary holds if @xmath257 is finite . if @xmath257 is the class @xmath385 of indecomposable structures of size at most @xmath77 then according to a result of schmerl and trotter , 1993 ( @xcite ) , the bounds of @xmath386 have size at most @xmath387 . when @xmath257 is made of bichains , proposition [ prop : wqo ] was obtained by albert and atkinson @xcite . + an immediate corollary is : [ cor : wqo ] if a hereditary class of @xmath38 contains only finitely many indecomposable members then it is wqo and has finitely many bounds . we say that a class @xmath0 of relational structures is _ hereditary rational _ , resp . _ hereditary algebraic _ if the generating function of every hereditary subclass of @xmath0 is rational , resp . albert , atkinson and vatter @xcite proved that hereditary rational classes of permutations are wqo . this fact can be extended to hereditary algebraic classes . [ lem : wqo ] a hereditary class @xmath0 which is hereditary algebraic is wqo . if @xmath0 contains an infinite antichain , there are uncountably many hereditary subclasses of @xmath0 and in fact an uncountable chain of subclasses ; these classes provides uncountably many generating series . some of these series can not be algebraic . indeed , according to c. retenauer @xcite , a generating series with rational coefficients which is algebraic over @xmath388 is algebraic over @xmath389 . since the generating series we consider have integer coefficients , there are algebraic over @xmath389 , hence there are only countably many such series . if @xmath85 and @xmath259 are two hereditary classes , then the generating series satisfy the identity @xmath390 . from this simple equality we have : [ lem : algebraic ] the union of two hereditary rational ( resp . algebraic ) classes is hereditary rational ( resp . algebraic ) . a minimal non - hereditary rational or a minimal non - hereditary algebraic class @xmath85 is the age of some relational structure . according to lemma [ lem : algebraic ] , @xmath85 can not be the union of two proper hereditary subclasses , hence this is an ideal , thus an age . [ theo : algebraic ] let @xmath45 be an integer . if an hereditary class @xmath85 of @xmath265 contains only finitely many indecomposable members then it is algebraic . we follows essentially the lines of albert - atkinson proof . we do an inductive proof over the hereditary subclasses of @xmath85 . but for that , we need to prove more , namely that @xmath85 and each @xmath391 for @xmath392 , are algebraic ( this is the only difference with albert - atkinson proof ) . to avoid unessential complications , we take out the empty relational structure of @xmath393 , that is we suppose that @xmath85 is made of non - empty structures . let @xmath394 . if @xmath395 then by corollary [ cor : wqoalgebraic ] , @xmath85 and each @xmath391 for @xmath392 , are algebraic . thus the result is proved . if @xmath396 , we may suppose that for each proper hereditary subclass @xmath397 of @xmath85 , both @xmath397 and @xmath398 for each @xmath399 , are algebraic . indeed , otherwise , since by corollary [ cor : wqo ] , @xmath85 is wqo , it contains a minimal hereditary subclass not satisfying this property and we may replace @xmath85 by this subclass . let @xmath392 . let @xmath391 be the subclass of @xmath85 made of @xmath62-indecomposable members of @xmath85 . let @xmath400 and @xmath401 , with @xmath227 , be the two elements of @xmath402 , we set @xmath403 . let @xmath404 we set @xmath405 . as in lemma [ lem : union ] we have @xmath406 and @xmath407 let @xmath299 and @xmath300 be the generating series of @xmath0 and @xmath408 respectively . let @xmath409 and @xmath410 be the generating series of @xmath411 and of @xmath412 . we have : @xmath413 and @xmath414 we deduce that @xmath415 and @xmath416 are algebraic for every @xmath417 , from the following claims that we will prove afterwards [ claim : claim1]the generating series of @xmath410 is a polynomial in the generating series @xmath418 whose coefficients are algebraic series . [ claim : claim2 ] for each @xmath417 , the generating series @xmath416 of @xmath419 is either a linear polynomial in the generating series @xmath418 of the form @xmath420 whose coefficients are algebraic series or is a rational fraction of the form @xmath421 substituting in formula [ eq : global ] the values of @xmath410 and @xmath422 given by claim [ claim : claim1 ] and claim [ claim : claim2 ] we obtain a polynomial in @xmath423 whose coefficients are algebraic series . this polynomial is not identical to zero . indeed , it is the sum of a polynomial @xmath424 and @xmath425 whose coefficients are algebraic series ( in fact , @xmath426 ) . the valuation of @xmath46 and @xmath158 as series in @xmath304 are distinct . indeed , the valuation of @xmath46 is @xmath401 ( notice that @xmath427 where @xmath428 is either zero or an algebraic series of valuation at least @xmath345 ) . hence , if @xmath429 ( when @xmath408 is non empty ) its valuation is at least @xmath430 . since @xmath46 and @xmath158 do nt have the same valuation , then @xmath431 is not identical to zero . being a solution of a non zero polynomial , @xmath418 is algebraic . with this result and claim [ claim : claim2 ] , @xmath416 is algebraic . with this , the proof of theorem [ theo : algebraic ] is complete . in order to prove our claims , we need the following lemmas ( respectively lemma 15 and lemma 18 in @xcite ) . [ lem : intersection ] let @xmath62 be an indecomposable ordered structure and @xmath432 , @xmath433 be two sequences of subclasses of ordered binary structures indexed by the elements of @xmath62 . if @xmath62 has at least three elements then @xmath434 if @xmath226 with @xmath227 then @xmath435 the first equality follows from proposition [ pro : sumlex ] and the second one follows from lemma [ lem : decomposition ] . let @xmath85 be a class of finite structures and @xmath436 be a sequence of finite structures , we will set @xmath437 if @xmath85 is hereditary , a proper hereditary subclass @xmath397 of @xmath85 is _ strong _ if every bound of @xmath397 in @xmath85 is embeddable in some bound of @xmath438 note that the intersection of strong subclasses is strong . let @xmath432 , where @xmath62 is indecomposable with at least three elements . decomposition of a binary structure _ @xmath169 _ over _ @xmath224 is a map @xmath439 such that @xmath440 and @xmath441 for all @xmath442 hence , each @xmath443 is an interval of @xmath444 let @xmath445 be the set of all such decompositions of @xmath169 . [ lem : egality]let @xmath62 be an indecomposable ordered structure , @xmath432 be a sequences of subclasses of ordered binary structures indexed by the elements of @xmath62 , @xmath446 be a sequence of finite structures , and @xmath447 . if @xmath62 has at least three elements then @xmath85 is a union of sets of the form @xmath448 where each @xmath449 is either @xmath450 or one of its strong subclasses . we prove the result for @xmath451 and we set @xmath452 . for that we prove that : @xmath453 where @xmath454 and @xmath455 for @xmath456 . let s call by @xmath279 ( respectively by @xmath280 ) the left - hand side ( respectively the right - hand side ) of equation [ eq : eqstrong ] . inclusion @xmath282 holds without any assumption . indeed , let @xmath198 in @xmath279 . we prove that @xmath198 is in @xmath280 . if @xmath457 is a decomposition of @xmath169 , we want to find @xmath458 such that @xmath459 since @xmath198 is in @xmath279 , it has a decomposition over @xmath460 let @xmath461 , since @xmath462 , there exist @xmath463 such that @xmath464 hence @xmath465 inclusion @xmath281 holds under the assumption that a structure in @xmath280 has a unique decomposition over @xmath62 and that it is ordered , ( what means that @xmath62 is rigid , that is @xmath62 has no automorphism distinct from the identity ) . let @xmath198 in @xmath280 , then for every @xmath461 there exist @xmath466 such that @xmath467 , thus , @xmath468 we have @xmath469 because , @xmath470 for every @xmath471 hence , @xmath472 we claim that @xmath473 . suppose @xmath474 let @xmath52 be an embedding of @xmath169 into @xmath198 and @xmath475 , where @xmath310 is the projection map from @xmath198 into @xmath476 we must have @xmath477 for @xmath478 which is a contradiction with the fact that @xmath479 using distributivity of intersection over union , we may write @xmath280 as a union of terms , each of which is an intersection of terms like @xmath480 , where @xmath481 is an interval of @xmath169 such that , there exist a decomposition @xmath457 of @xmath169 and @xmath482 these intersections , by lemma [ lem : intersection ] and the fact that among all decompositions of @xmath169 are all ones which send @xmath169 into a single element @xmath304 of @xmath62 have the form @xmath483 where each @xmath449 is of the form @xmath484 where the structures occurring after @xmath169 ( if any ) are intervals of @xmath444 hence , @xmath449 is either @xmath485 or one of its strong subclasses . the case @xmath486 follows by induction . * proof of claim [ claim : claim1 ] . * since @xmath487 is wqo and @xmath85 is a proper hereditary subclass , we have @xmath488 for some finite family @xmath489 of elements of @xmath224 . let @xmath404 , lemma [ lem : egality ] asserts that @xmath490 is an union of classes , not necessarily disjoint , of the form @xmath491 where each @xmath492 is either @xmath85 or one of its strong subclasses . the generating series of@xmath491 is a monomial in the generating series @xmath418 of @xmath85 whose coefficient is a product of generating series of proper strong subclasses of @xmath85 . from the induction hypothesis , the generating series of of these strong subclasses are algebraic series , hence this coefficient is an algebraic series . using the principle of inclusion - exclusion , we get that the generating series @xmath493 of @xmath490 is a polynomial in the generating series @xmath418 whose coefficients are algebraic series . since the @xmath490 s are pairwise disjoint , the generating series @xmath410 is also a polynomial in the generating series @xmath418 whose coefficients are algebraic series . @xmath494 [ lem : twoelements ] if @xmath62 has two elements @xmath400 and @xmath401 , @xmath226 with @xmath227 , then @xmath495 is an union of classes of the form @xmath496 , where @xmath497 and @xmath498 are either equal to @xmath499 or to some strong subclasses of @xmath499 . as above we suppose first @xmath451 . equation [ eq : eqstrong ] yields @xmath500\ ] ] an induction take care of the case @xmath486 . * proof of claim [ claim : claim2 ] . * let @xmath404 . lemma [ lem : twoelements ] asserts that @xmath490 is an union of classes , not necessarily disjoint , of the form @xmath501 , where @xmath397 and @xmath502 are either equal to @xmath85 or to some strong subclasses of @xmath85 . the generating series of these classes are of the form @xmath503 or @xmath504 or @xmath505 , where @xmath506 and @xmath507 are algebraic series . using the principle of inclusion - exclusion , we get that the generating series @xmath493 is either of the form @xmath503 or of the form @xmath508 , where @xmath509 and @xmath428 are algebraic series . in particular @xmath493 is of the form @xmath510 where @xmath511 and @xmath512 are polynomials in @xmath418 of degree at most @xmath401 with algebraic series as coefficients . using equation [ eq : classeindecbis ] we obtain @xmath513 when all bounds @xmath514 of @xmath85 in @xmath224 are @xmath62-indecomposable or @xmath420 if at least one bound @xmath515 is not @xmath62-indecomposable . @xmath494 the conclusion of theorem [ theo : algebraic ] above does not hold with structures which are not necessarily ordered . let @xmath516 be the direct sum of infinitely many copies of the complete graph on an infinite set . as it is easy to see the generating function of @xmath517 is the integer partition function . this generating series is not algebraic . however , @xmath517 contains no indecomposable member with more than two elements . more generally , note that the class @xmath518 of finite cographs contains no indecomposable cograph with more than two vertices and that this class is not hereditary algebraic . finite cographs are comparability graphs of serie - parallel posets which in turn are intersection orders of separable bichains . by albert - atkinson s theorem , the class of these bichains is hereditary algebraic . this tells us that algebraicity is not necessarily preserved by the transformation of a class into an other via a process as above ( processes of this type are the free - operators of frass @xcite ) . in their paper @xcite , albert and atkinson indicate that there are infinite sets of simple permutations whose sum closure is algebraic but , as it turns out , some hereditary subclasses are not necessarily algebraic . an example is the collection of decreasing oscillations ( see the end of the section ) . in order to extend their proof to some other classes , they ask whether there exists _ an infinite set of simple permutations whose sum - closure is well quasi ordered_. as we indicate in proposition [ criticalwqo ] below , the set of exceptional permutations has this property . in fact , it is hereditary wqo . we guess that this notion of hereditary wqo is the right concept for extending albert - atkinson theorem . exceptional permutations correspond to bichains which are _ critical _ in the sense of schmerl and trotter . let us recall that a binary structure @xmath29 with domain @xmath15 is _ critical _ if @xmath29 is indecomposable but @xmath519 is not indecomposable for every @xmath177 . schmerl and trotter @xcite gave a description of critical posets . they fall into two infinite classes : @xmath520 and @xmath521 where @xmath522 , @xmath523 , @xmath524 if @xmath525 ; @xmath526 and @xmath527 if @xmath528 . as it is well known , an indecomposable two - dimensional poset @xmath535 has a unique realizer ( that is there is a unique pair @xmath536 of linear orders whose intersection is the order @xmath537 of @xmath538 ) . hence , there are at most two bichains , namely @xmath539 and @xmath540 such that @xmath541 . the critical posets described above yield four kind of bichains , namely @xmath542 , @xmath543 , @xmath544 and @xmath545 . these bichains are critical . indeed , a bichain is indecomposable if and only if the intersection order is indecomposable ( @xcite for finite bichains and @xcite for infinite bichains ) . the isomorphic types of these bichains are described in albert and atkinson s paper in terms of permutations of @xmath546 for @xmath547 : + @xmath548 @xmath549 + @xmath550 @xmath551 + @xmath552 @xmath553 + @xmath554 @xmath555 + for example , the type of the bichain @xmath556 is the permutation given in @xmath554 , whereas the type of @xmath557 is its inverse , given in @xmath550 ( enumerate the elements of @xmath558 into the sequence @xmath546 , this according to the order @xmath559 , then reorder this sequence according to the order @xmath560 ; this yields the sequence @xmath561 ; according to our definition the type of @xmath556 is the permutation @xmath8 , this is the one given in @xmath554 ) . for @xmath562 , the permutations given in @xmath548 and @xmath554 coincide with @xmath135 whereas those given in @xmath550 and @xmath552 coincide with @xmath136 ; for larger values of @xmath563 , they are all different . the four classes of indecomposable bichains are obtained from @xmath564 by exchanging the two orders in each bichain or by reversing the order of the first one , or by reversing the second one . hence the order structure w.r.t . embedabbility of these classes is the same , and it remains the same if we label the elements of these bichains . this class is the union of four classes hence , in order to prove that it is hereditary wqo , it suffices to prove that each one of these classes is hereditary wqo . according to the observation above , it suffices to prove that one , for example @xmath242 , is hereditary wqo . let @xmath340 be a wqo poset . we have to prove that @xmath565 is wqo . for that , set @xmath566 , where @xmath567 , and order @xmath169 componentwise . let @xmath568 be the set of all words over the ordered alphabet @xmath169 . we define an order preserving map @xmath120 from @xmath568 onto @xmath565 . this will suffice . indeed , @xmath169 is wqo as a product of two wqo sets ; hence , according to higman theorem on words over ordered alphabets @xcite , @xmath568 is wqo . since @xmath565 is the image of a wqo by an order preserving map , it is wqo . we define the map @xmath120 as follows . let @xmath569 . set @xmath570 where @xmath571 and @xmath572 for @xmath573 . we observe first that @xmath574 in @xmath568 implies @xmath575 in @xmath565 . indeed , if @xmath574 there is an embedding @xmath457 of the chain @xmath576 into the chain @xmath577 such that @xmath578 for all @xmath579 . let @xmath580 defined by setting @xmath581 . as it is easy to check , @xmath582 is an embedding of @xmath583 into @xmath584 . next , we note that @xmath120 is surjective . indeed , if @xmath585 with @xmath571 , then the word @xmath586 with @xmath587 yields @xmath588 . in @xcite it is mentioned that this class has finitely many bounds . the generating series of the class of critical bichains is rational ( the class is covered by four chains ) . according to corollary 13 of @xcite their sum - closure is algebraic . indeed , let @xmath589 be the doubly infinite path whose vertex set is @xmath590 and edge set @xmath591 . the edge set @xmath15 has two transitive orientations , e.g. @xmath592 and its dual @xmath593 . as an order , @xmath594 is the intersection of the linear orders @xmath595 and @xmath596 let @xmath597 and @xmath598 . members of @xmath259 of size @xmath2 are obtained by restricting @xmath599 to intervals of size @xmath2 , @xmath600 , of the chain @xmath601 ( observe first that the graph @xmath602 is indecomposable as all its restrictions to intervals of size different from @xmath430 of the chain @xmath601 and furthermore there are no others indecomposable restrictions ; next , use the fact that the indecomposability of a comparability graph amounts to the indecomposability of its orientations @xcite , and that the indecomposability of a two - dimensional poset amounts to the indecomposability of the bichains associated with the order @xcite ) . up to isomorphy , there are two indecomposable bichains of size @xmath2 , @xmath603 , namely @xmath604 and @xmath605 where @xmath606 . these two bichains embed all members of @xmath259 having size less than @xmath2 . being covered by two chains , @xmath259 is wqo . to see that @xmath259 is not hereditary wqo , we may associate to each indecomposable member of @xmath259 the comparability graph of the intersection of the two orders and observe that this association preserves the embeddability relation , even though label are added . the class of graphs obtained from this association consists of paths of size distinct from @xmath430 . it is not hereditary wqo . in fact , as it is immediate to see , if a class @xmath607 of graphs contains infinitely many paths of distinct sizes , then @xmath608 is not wqo . indeed , if we label the end vertices of each path by @xmath401 and label the other vertices by @xmath400 , we obtain an infinite antichain . thus @xmath609 is not wqo . the generating series of @xmath259 is rational ( its generating function is @xmath610 ) . in fact , @xmath259 is hereditary algebraic ( every hereditary subclass of @xmath259 is finite ) . by corollary 13 of @xcite , the sum - closure @xmath611 is algebraic . ( in fact , if @xmath612 is the generating function of @xmath613 then @xmath614 ) . but @xmath611 is not hereditary algebraic . for that , it suffices to observe that it is not wqo and to apply lemma [ lem : wqo ] . the fact that @xmath611 is not wqo is because we may embed the poset @xmath615 into @xmath611 via an order preserving map . a simpler argument consist to observe first that the family @xmath616 , where @xmath617 is the graph obtained from the @xmath2-vertex path @xmath618 by replacing its end - vertices by a two - vertex independent set , is an antichain , next that these graphs are comparability graphs associated to members of @xmath259 . the downward closure @xmath619 is @xmath620 , the age of @xmath599 ; this age has four obstructions , it is rational : the generating series is @xmath621 , the generating function being the sequence a05298 of @xcite , starting by 1 , 1 , 2 , 5 , 11 , 24 . for all of this see @xcite . 1 . a hereditary class of indecomposable ordered binary structures @xmath259 is hereditary wqo whenever its sum closure is hereditary algebraic ? the generating series of a hereditary class of relational structures is rational whenever the profile of this class is bounded by a polynomial ? this is true for graphs @xcite and tournaments @xcite . the profile of a wqo hereditary class of relational structures is bounded above by some exponential ? t. gallai , . acta math . . hungar . * 18 * ( 1967 ) 25 - 66 ( english translation by f. maffray and m. preissmann in j.j . ramirez - alfonsin and b. reed ( eds ) , perfect graphs , wiley 2001 , pp.25 - 66 . g. higman . ordering by divisibility in abstract algebras . ( 1952 ) 326336 , . d. kelly . _ comparability graphs . _ graphs and orders , i. rival ( ed ) , nato asi series , vol.147 , d. reidel , dordrecht , 1985 , pp . t. kaizer , m. klazar , * 9 * ( 2002/2003 ) , no . 2 , research paper 10 , 20 pp . m. klazar , overview of general results in combinatorial enumeration , in _ permutation patterns _ , london math . lecture note ser . 376 , ( 2010 ) , 340 , cambridge univ . press , cambridge c. laflamme , m. pouzet , n. sauer , i. zaguia , pairs of orthogonal countable ordinals , 15pp . 2014 , to appear in j. discrete math . c. retenauer , personnal communication , nov.2011 . i. rival , n. zaguia . _ perpendicular orders_. discrete math . * 137 * ( 1995 ) , no . 1 - 3 , 303313 . schmerl , w.t . trotter , , _ discrete math . _ , * 113 * ( 1 - 3 ) ( 1993 ) 191205 . n. j. a. sloane , the on - line encyclopedia of integer sequences , sequence a111111 . i. zaguia . _ prime two - dimensional orders and perpendicular total orders . _ europ . j. of combinatorics * 19 * ( 1998 ) , 639649 .

we studied the semiconductor response with respect to high intensity resonant excitation on short time scale when the contribution of the fermi statistics of the electrons and holes prevails . we studied both the single and double pulse excitations . for the latter case we considered the time evolution of the multi - wave mixing exciton polarization . the main difference between the excitation by a single pulse or by two non - collinear pulses is that the rabi oscillations of the multi - wave mixing response are characterized by two harmonics . analyzing the operator dynamics governed by the external excitation we found that there are three invariant spin classes , which do not mix with the evolution of the system . two classes correspond to the bright exciton states and one contains all dark states . we found that the dynamics of the classes is described by six frequencies and the rabi frequencies are only two of them ( one for each bright class ) . we discuss the effect of the dispersion of the electrons and holes and the coulomb interaction describing the semiconductor by the semiconductor bloch equation ( sbe ) . we show that if initially the system is in the ground state then the sbe preserves the invariant spin classes thus proving absence of the dark excitons in the framework of this description . we found that due to the mass difference between holes of different kind additional rabi frequencies , two of those present in the operator dynamics , should appear in the evolution of the exciton polarization .

0712.4022

one of the main tools of probing the complex character of the many - body correlations and interactions in semiconductors is the multi - wave mixing response . the multi - wave mixing polarization in optically excited semiconductors produces the signal in directions that are prohibited in the linear regime because of momentum conservation , thus giving access to the semiconductor many - body excitations . the typical example is the four - wave mixing spectroscopy.@xcite recently also wave mixing of higher orders started to draw attention.@xcite a great success in understanding the mechanism of the formation of the nonlinear response in general was achieved in relatively low - field limit using the perturbation theory with respect to the external field.@xcite perturbational description , however , is not suitable for investigating the coherent reconstruction of the spectrum , such as in the case of rabi oscillations . for a qualitative analysis of such phenomena few - level quantum models have been used.@xcite this approach , however , misses the important property of the excitations in semiconductors . these excitations constitute a quantum field rather than a canonical quantum mechanical system . the system supports infinitely many states unless , of course , the spectrum has truly discrete component , that is when the excitations are localized , such as in quantum dots . thus a description non - perturbative with respect to the external field must deal with many - body aspects of the dynamics of the system.@xcite in order to treat this problem a variety of methods based on derivation of respective closed equations of motion was developed.@xcite unfortunately , the dynamical equations turn out to be very complex owing to the coulomb interaction , which is shown to be crucially important at relatively long time scales.@xcite therefore , one has to resort to numerical calculations investigating the dynamics of the exciton polarization . the numerical simulations proved to describe successfully the dynamics @xcite but they are difficult to apply for studying the detailed effect of different contributions to the dynamics and its dependence on parameters of the system . as a result , the theory of the semiconductor response suffers from a lack of exact results obtained in controllable approximations which would guide the respective numerical , theoretical , and experimental studies . at short time scales , however , one can rely on significant simplification of the dynamics of the system due to negligible phase change during the optical excitation . effectively , the system follows the external field , which is illustrated best by a simple dynamical model @xmath0 p(t ) + e(t)\ ] ] with @xmath1 being a real positive function . if initially the system is at rest , @xmath2 , we can present the time dependence @xmath3 in the form @xmath4\right ] e(t').\ ] ] this representation is convenient for comparing the responses of the system with respect to the @xmath5-functional excitation @xmath6 and to the excitation with the piece - wise constant amplitude @xmath7 , where @xmath8 is the duration of the forced regime . as follows from eq . , in the case of the @xmath5-pulse immediately after the excitation is switched off one has @xmath9 . for the excitation with constant amplitude eq . yields an estimate maximally different from the @xmath5-pulse case in the form @xmath10 , where @xmath11 . thus if the excitation pulse is shorter than the time scale determined by @xmath12 , i.e. if @xmath13 , then the response with respect to the constant pulse of finite duration differs up to quadratic terms from the @xmath5-pulse response only by the phase factor @xmath14 . an immediate application of this consideration to the evolution of the exciton polarization in optically excited semiconductors is prevented by the more complex character of the polarization dynamics , namely , by the field induced coupling between the polarization and the charge densities . in order to see the principal differences introduced by this coupling we consider a model @xmath15 straightforward substitution of the @xmath5-shaped pulse into this equation obviously fails since now the amplitudes in front of the @xmath5-function in the coupling terms are given by the quantities , @xmath16 and @xmath3 , which experience discontinuity at exactly the point of singularity of the @xmath5-function . therefore , we need to consider the case of the external pulse of finite duration . for the pulse with piece - wise constant amplitude @xmath17 , as in the previous example , eq . with the initial conditions @xmath2 , @xmath18 is solved for @xmath19 by @xmath20 , \\ % n(t ) = & - \frac{\epsilon^2}{\omega^2 t^2}\left[1- \cos ( \omega t)\right ] , \end{split}\ ] ] where @xmath21 . it is seen that in the limit of the excitation pulses short compared to the typical dynamical time scale , @xmath22 , one has to distinguish the cases of weak and strong excitations , @xmath23 and @xmath24 respectively . in the first case up to the quadratic terms , @xmath25 , the response @xmath26 looks like the one of the system where @xmath27 and @xmath28 are decoupled with respect to @xmath5-pulse excitation . this illustrates the relation with the @xmath5-functional approximation popular in studying the nonlinear optical response by using perturbational approach , which corresponds to weak scattering . in the limit of strong external field when @xmath22 while @xmath29 the evolution of the system is significantly different . neglecting the terms quadratic in @xmath30 we obtain that the evolution of the system is similar to the one we have in the case @xmath31 ( i.e. rabi flopping ) with the only difference that @xmath3 acquires a ( small ) imaginary part . applying now the similar idea for estimating a phase shift as for the analysis of eq . we can see that the conclusion about small perturbations vanishing with @xmath32 holds also in the case of @xmath27-dependent frequency and renormalized coupling @xmath33 we consider the limiting case when the effect of the internal dynamics can be completely neglected , which corresponds to the cases @xmath34 in the examples considered above . we obtain classes of state related to the spin selection rules and related to these classes the spectrum of the rabi frequencies . in section [ sec : internal ] we discuss the effect of the internal dynamics and the coulomb interaction . we consider the dynamics of excitation of a semiconductor quantum well during the action of an ultrashort pulse of high intensity so that the period of the rabi oscillations is smaller than the typical time scales determined by the internal dynamics . we consider the question of validity of this approximation in section [ sec : internal ] . we assume that the pulse is tuned to resonance with the exciton levels lying below the band edge and that the envelope of the pulse is constant while the field is on . by means of a canonical transformation the harmonic time dependence of the excitation field can be excluded , ( see e.g. ref . ) so that in the rotating wave approximation the transformed hamiltonian becomes time - independent . for the description of the semiconductor dynamics one can use the sbe accounting for degenerate valence bands.@xcite having the dynamical equations at hand one can work out the short time approximation similarly to the one shown in the introduction . however , in the short time approximation , it is more convenient to study directly the exciton polarization . in the coherent regime it is defined as @xmath35 , where @xmath36 is the state of the semiconductor , @xmath37 denotes the whole set of relevant quantum numbers describing the specific exciton ( bound or unbound ) state and @xmath38 is the exciton annihilation operator defined as @xmath39 . in terms of the electron and hole operators the exciton operator @xmath38 is presented as @xmath40 where @xmath41 is the exciton wave functions , @xmath42 destroys a hole in the valence band with the spin state @xmath43 at the point with the coordinate @xmath44 and @xmath45 destroys an electron in the conduction band with the spin state @xmath46 at point @xmath47 . neglecting the coulomb interaction and the effect of slow internal dynamics determined by small detuning @xmath48 means the semiconductor hamiltonian is approximated by the hamiltonian of light - matter interaction @xmath49 . in the rotating wave and dipole approximations the interaction hamiltonian has the form @xcite@xmath50,\ ] ] where @xmath51 is a piece - wise constant function of time , it takes the values @xmath52 when the field is off and @xmath53 with @xmath54 being the envelope of the external field when the excitation is on . here we have introduced @xmath55 the respective matrix elements of the dipole moment , which are assumed to be independent on both the electromagnetic field momentum and its angle of incidence.@xcite thus , in this approximation one has @xmath56 ( throughout the paper we write formulas using units with @xmath57 ) and accordingly @xmath58\ldots]]},\ ] ] where @xmath59 . we calculate this series using the observation that commuting @xmath60 $ ] with @xmath49 we arrive at an operator which can be presented in a form similar to eq . ( [ eq : exciton_operator_elementary ] ) . we employ this observation introducing operators @xmath61,\ ] ] which are related to each other through repetitive commutation with @xmath49 , namely , @xmath62 $ ] . with the initial condition " @xmath63 the operators @xmath64 represent the _ even _ terms in series ( [ eq : p_t_series_presentation ] ) . these terms themselves do not contribute into @xmath65 since @xmath66 , which in turn follows from @xmath67 and @xmath68 . after commuting with @xmath69 , however , they produce @xmath70-number terms @xmath71}$ ] resulting in @xmath72 in order to simplify the expressions we assume below that the quantum well can be approximated by a 2d plane . in this case the exciton states are characterized by the spin states of the electron and hole constituting the exciton , the center of mass momentum in the plane of the well , @xmath73 , and other quantum numbers , @xmath74 , so that @xmath75 . using the relation between @xmath76 and @xmath64 we find @xmath77 \\ % d^{(n+1)}_{\sigma , s}(\mathbf{x},\mathbf{y } ) = \sum_{\sigma ' , s ' } & \left [ d^{(n)}_{\sigma ' , s}(\mathbf{x},\mathbf{y})\mathcal{e}^*_{\sigma , s'}(\mathbf{x } ) \mathcal{e}_{\sigma ' , s'}(\mathbf{x})\right . \\ % & \left . \qquad + d^{(n)}_{\sigma , s'}(\mathbf{x},\mathbf{y})\mathcal{e}^*_{\sigma ' , s}(\mathbf{y } ) \mathcal{e}_{\sigma ' , s'}(\mathbf{y } ) % \right . \\ % % % & \left . \qquad -2 u^{(n)}_{\sigma ' , s'}(\mathbf{y},\mathbf{x})\mathcal{e}^*_{\sigma , s'}(\mathbf{x } ) \mathcal{e}^*_{\sigma ' , s}(\mathbf{y } ) \right ] , \end{split}\ ] ] where @xmath78 . in terms of the kernels of the operators @xmath64 the coefficients @xmath79 are expressed as @xmath80 with @xmath81.\ ] ] some general results can be obtained directly from eqs . ( [ eq : kernels_relation ] ) and ( [ eq : a_kernels ] ) . first , the polarization of dark excitons ( states with helicity @xmath52 and @xmath82 ) is zero . second , if the excitation pulse is circularly polarized then summation over the spin indices in eqs . ( [ eq : kernels_relation ] ) and ( [ eq : a_kernels ] ) reduces to the single terms with the electron and hole spins determined by @xmath46 and @xmath43 , respectively . if the excitation pulse is linearly polarized then only the electron spin is fixed to @xmath46 and as a result different electron - hole states get coupled.@xcite the summation over the hole spins , as will be explicitly demonstrated below , to be limited to such values that meet the condition @xmath83 . for example , if @xmath84 ( heavy holes ) then @xmath85 can take values @xmath86 and @xmath87 ( light hole ) . it follows from eqs . ( [ eq : kernels_relation ] ) and ( [ eq : a_kernels ] ) that @xmath88 , where @xmath89 with @xmath90 being the second value satisfying the spin selection rule discussed above . applying these results to eq . ( [ eq : p_mu_sum_a ] ) we find the general representation , which is valid for an excitation pulse with an arbitrary spatial profile in the plane of the quantum well , @xmath91.\ ] ] this result can be generalized for the more general case when the envelope amplitude of the external field is not a piece - wise constant and can be presented as @xmath92 . we note that the hamiltonians @xmath93 taken at different instants @xmath94 and @xmath95 commute with each other . therefore , the line of arguments used above can be repeated with the substitution @xmath96 it can be easily seen that the expression for the exciton polarization @xmath97 remains essentially the same @xmath98,\ ] ] where @xmath99 and @xmath100 are defined by the same expressions but with the time - dependent envelope function of the external field , i.e. they differ by the factor @xmath101 . the phases @xmath102 have the meaning of the pulse areas . since such generalization does not bring new physics but complicates the discussion of the time dependence in what follows we will consider only the case of piece - wise constant amplitude of the external excitation . first we briefly discuss the case when the excitation is a single plane wave @xmath103 . in this case @xmath104 does not depend on the coordinate and the integration in eq . can be easily performed taking into account that @xmath105 , where @xmath106 is the coordinate of the center of mass of the exciton and @xmath107 is its relative wave function . for example , for @xmath108 exciton polarization we obtain @xmath109 where @xmath110 is the value of the exciton wave function at the origin , which in the 2d approximation is @xmath111 with @xmath112 being the respective exciton bohr radius , and @xmath113 equation ( [ eq : plane_wave_excitation_h ] ) presents the rabi oscillations of the exciton polarization . it is interesting to note that this result explicitly shows the statistical origin of the rabi oscillations . following the same line of arguments one can show that the oscillations would absent and the exciton polarization would increase monotonously with time if the exciton operators had obeyed bosonic commutation relations or if the electrons and holes were bosons . indeed , let , for example , the exciton operators be bosonic then all terms in eq . with @xmath114 turn to @xmath52 leaving @xmath115 . because of the coupling between different exciton states the evolution of the exciton polarization is characterized by two rabi frequencies , @xmath116 and @xmath117 . the ratio between these frequencies is shown in fig . [ fig : rabi_ratio ] as a function of @xmath118 , where for calculations we have taken@xcite @xmath119 . the effect of sharing the rabi frequency leads to the same time dependence of the heavy- and light - hole contributions into the signal with fixed helicity if the excitation is linearly polarized . if , however , the external field has elliptic polarization these contributions oscillate with different frequencies leading to beatings and to a nontrivial time dependence of the polarization state of the signal . as a function of ellipticity of the excitation pulse @xmath118 . the ratio varies from @xmath120 to @xmath121 . for calculations the ratio of heavy- and light - hole dipole moments is taken @xmath119 . , width=288 ] before we proceed we would like to discuss in details the spin selection rules noticed above from the perspective of the operator equations of motion . for this we consider an operator of the form @xmath122 where the sum is taken over all spin indices . the amplitudes in this expression are chosen in such a way that @xmath123 is the heisenberg representation of the operator @xmath124 defined by the light - matter hamiltonian @xmath125 the last term in eq . determines the average value @xmath126 and for the case of the exciton polarization was considered above . for the purpose of our discussion it suffices to consider the dynamical equations for the time - dependent amplitudes @xmath127 and so on for the case of @xmath128 and spatially independent external field @xmath129 . from the heisenberg equation of motion @xmath130 $ ] we find ( see appendix [ app : external_field_algebra ] ) @xmath131 where the summation over the dashed spin variables is implied . with the respective initial conditions these equations define the heisenberg representation of the operator @xmath124 . for example , if all amplitudes but @xmath132 are initially zero , then this would be the representation of the _ interband _ polarization creation operator and so on . it should be noted that no restrictions on the amplitudes being bright " or dark " are imposed in eq . . thus this is a system with respect to @xmath133 unknowns and , counting degeneracy , has as many characteristic frequencies . the actual number of different frequencies , however , turns out to be much smaller . taking the derivative of the last pair of equations with respect to time we obtain the closed system @xmath134 where as as well the summation over dashed spin indices is performed . taking the complex conjugation of these equations we obtain the same system but with respect to @xmath135 . because of this symmetry the solutions of eq . are divided into two classes @xmath136 and @xmath137 , which correspond to @xmath138 and @xmath139 irreducible representations of the group @xmath140 , respectively . additional simplifications come from the structure of the coupling induced by the external field . having in mind successive application for the dynamics determined by the semiconductor bloch equation we consider the additional symmetries on a general ground . due to the electric dipole selection rules only four elements of @xmath129 are non - zero , @xmath141 where the columns are enumerated by the electron spin projections @xmath142 , @xmath87 and rows denote different hole spins @xmath143 . we denote by @xmath144 the vector space containing the elements of form but without any relation between the nonzero elements , that is the @xmath145-dimensional subspace of the interband polarizations . for two elements @xmath146 , @xmath147 we consider @xmath148 , where the composition is the usual matrix multiplication of two ( rectangular ) matrices @xmath149 the space spanned by such elements will be called @xmath150 . this is a @xmath151-dimensional subspace of the @xmath145-dimensional space of the electron densities . using eq . in one can see that in the basis of the electron - hole spins @xmath150 consists of diagonal @xmath152 matrices . finally , for the elements @xmath146 and @xmath153 we consider @xmath154 , which are the matrices with the matrix elements @xmath155 we will call the respective linear vector space @xmath156 , which , as can be easily checked , is @xmath157-dimensional subspace of the @xmath158-dimensional space of the hole - hole correlation functions . the elements of @xmath150 and @xmath156 can be considered as operators acting on the electron and the hole spin variables , respectively . the important properties of these operators is that their action is reducible , namely , @xmath150 leaves the spin state intact and @xmath156 mixes only states within the groups @xmath159 and @xmath160 but not from the different groups . using these results we obtain the modes of system . these modes are naturally classified by the invariant spin classes , which do not mix during the evolution , and the irreducible representations of the group @xmath140 . inspecting eqs . one can see that the spin classes in the spin space of the electron - hole polarizations are @xmath161 or @xmath162 , @xmath163 and @xmath164 . the first two classes correspond to the bright exciton operators , the last class contains the dark excitons . in each class both states have the same projection of the electron spin , @xmath166 for @xmath167 and @xmath168 for @xmath169 , while the hole spin can take two values . if we need to distinguish these values we will denote them @xmath85 and @xmath170 . in order to find the modes one can present @xmath171 in eq . . studying the resulting system of homogeneous equations one finds the frequencies and ( non - normalized ) eigen - vectors of the modes @xmath172 where latexmath:[$\beta_{s}^2 = |\mathcal{e}_{\sigma , s}|^2 + with the frequency defined in eq . ) and @xmath138 and @xmath139 denote the even and odd irreducible representations , respectively . using these solutions in the first two equations of eqs . we find the electron and hole densities coupled to the respective interband polarization operators . one can immediately see that these densities belong to @xmath150 and @xmath156 , respectively . as immediately follows from eqs . and only @xmath139-solutions can contribute into the average value of the exciton polarization , since for the symmetric solutions both terms under the integral in eq . cancel each other . moreover , for the @xmath139-solutions with the frequencies @xmath174 the sum @xmath175 vanishes , which leaves only solution with @xmath176 in accordance with eq . . the dark exciton operators are specified by the amplitudes in the complement to @xmath144 . the states are naturally enumerated by the hole spin state @xmath85 and for each amplitude @xmath178 there is a correspondent component of the excitation field obtained by inversion of the electron spin @xmath179 with @xmath180 . the frequencies and non - zero components of different modes are found to be @xmath181 since for each @xmath178 in this class the respective components of the external field are zero , i.e. @xmath182 , the dark exciton amplitudes , according to eq . , do not contribute to @xmath183 . thus the average value of the dark exciton polarization equals to its initial value @xmath184 , where @xmath185 with @xmath85 and @xmath186 from @xmath187 . if the initial state of the system is vacuum then @xmath188 confirming the result about absent polarization of dark excitons . the frequencies found @xmath189 define all the frequencies introduced into the dynamics due to the interaction with the external field as defined by eq . . the remaining @xmath157 frequencies can be shown to be @xmath52 . four solutions ( two for each class @xmath167 and @xmath169 ) corresponding to the zero frequency are presented in eqs . and the remaining @xmath145 can be found from the first two equations of eqs . . since eqs . have the form of a schrdinger equation with a hermitian hamiltonian the solutions corresponding to @xmath31 are not secular solutions but rather invariants , that is they correspond to operators , which are invariant with respect to action of the external field . for example , as follows from eqs . , for the exciton polarization such operators are @xmath190 , where the spin summation is restricted to a particular class @xmath191 or @xmath192 . in the short - time limit only frequencies @xmath193 contribute to the rabi oscillations of the exciton polarization . however , as will be demonstrated later the slow dynamics leads also to admixture of other frequencies . a more complex situation arises when the excitation pulse consists of two plane waves with non - collinear wave - vectors , @xmath194 . this corresponds to the standard problem of excitation by two pulses with zero delay between them . factoring out the exponential factor corresponding to the first pulse we can rewrite eq . ( [ eq : p_mu ] ) as @xmath195 where @xmath196 is a periodic function @xmath197 with @xmath198 . thus , integration over @xmath47 yields the exciton polarization in the form @xmath199 where @xmath200 are the amplitudes of multi - wave mixing ( mwm ) polarizations . the amplitudes @xmath201 and @xmath202 correspond to the polarization along the directions of linear response , @xmath203 and @xmath204 , respectively . for @xmath205 one has four - wave mixing polarization , @xmath206 correspond to six - wave mixing and so on . performing the fourier transform of the periodic factor @xmath196 we derive the integral representation for the amplitudes of mwm polarization @xmath207 where @xmath208 and the frequencies @xmath209 are defined as the amplitudes of @xmath104 corresponding to different pulses , @xmath210 . ) and the four - wave mixing ( @xmath211 ) directions for ( a ) @xmath212 , ( b ) @xmath213 . the effective time is defined as @xmath214 . , width=576 ] in fig . [ fig : two_pulse_rabi ] we show the numerical evaluations of eq . ( [ eq : n_amplitudes_mwm ] ) for different ratios between the amplitudes of the plane waves constituting the excitation along one of the forward and four - wave mixing directions . the initial stage of the dynamics for @xmath215 can be obtained directly from series ( [ eq : p_mu_sum_a ] ) in the form ( for @xmath216 ) @xmath217 where @xmath218 . the evolution of the exciton polarization in the opposite limit @xmath219 is found from integral representation ( [ eq : n_amplitudes_mwm ] ) using the stationary phase method @xmath220 \right.\\ % & \left . + \frac{\mathcal{e}_\mu^{(1 ) } + \mathcal{e}_\mu^{(2)}}{\sqrt{\omega_\mu^{(1 ) } + \omega_\mu^{(2 ) } } } \sin\left[\left(\omega_\mu^{(1 ) } + \omega_\mu^{(2)}\right)t - \frac{\pi}{4}\right ] \right\}. \end{split}\ ] ] it shows that in this limit the @xmath221 demonstrates power decay @xmath222 and is essentially a superposition of two harmonics with frequencies @xmath223 . of course , the limit @xmath224 should not be taken too literally , since in this limit the main assumption @xmath225 is no longer valid and the effect of desynchronization of the mwm polarization of high orders ( see below ) becomes important . however , if @xmath226 and @xmath227 are not too close then after just a few oscillations the dependence @xmath221 is satisfactory approximated by its asymptotic form . for example , the complex behavior of the exciton polarization shown in fig . [ fig : two_pulse_rabi]b is a result of existence of two harmonics with the frequencies @xmath228 , which exist if @xmath229 . if @xmath230 then the slow harmonic vanishes . indeed , in this case , as follows from eq . ( [ eq : n_amplitudes_mwm ] ) , mwm polarizations can be presented in the form @xmath231 , \end{split}\ ] ] where @xmath232 are the bessel functions of the first kind . this solution asymptotically oscillates with the frequency @xmath233 , which is consistent with the simple picture of the rabi frequency renormalized due to doubling the external field . the interesting difference between the rabi oscillations in the case of a single pulse excitation and the oscillations of the mwm polarization is that the latter asymptotically decays in time @xmath222 . it can be understood taking into account the dynamical origin of formation the mwm polarization . as follows from eq . ( [ eq : n_amplitudes_mwm ] ) the mwm polarizations @xmath200 satisfy the equation of motion of a classical tight - binding model @xmath234 where @xmath235 . equations ( [ eq : tbm_polarizations ] ) are supplemented by the initial conditions @xmath236 for all @xmath237 and @xmath238 for all @xmath237 except @xmath239 , for which one has @xmath240 and @xmath241 . this representation allows one to estimate the effect of the dependence of the exciton energy on the in - plane momentum . it leads to modification of the eigenfrequency " , @xmath242 , by the term @xmath243 . even if @xmath244 this contribution becomes essential for mwm exciton polarizations of significantly high orders . however , for the most important case of mwm polarizations of low orders on the time scale of few rabi flops the effect of the correction due to the exciton dispersion can be neglected . it follows from the representation ( [ eq : tbm_polarizations ] ) that initially the excitation is localized at the sites @xmath239 ( the directions of linear response ) . with time the energy spreads along the chain thanks to the coupling @xmath245 between different sites . in terms of the mwm polarization the dynamics of this spreading corresponds to excitation mwm polarizations of higher and higher orders . the dynamical model described by eq . ( [ eq : tbm_polarizations ] ) has the first integral @xmath246 \right\rbrace = |\phi_\mu(0)|^2 \left(\left|\mathcal{e}_\mu^{(1)}\right|^2 + \left|\mathcal{e}_\mu^{(2)}\right|^2\right).\ ] ] from this perspective it is clear that the decay of the rabi oscillations is the consequence of spreading the excitation among mwm polarizations of different orders . in the previous section we have studied in details the dynamics of the exciton polarization in the short time limit , when the main contribution into dynamics comes from the interaction with the external field . here we consider the effect of the internal semiconductor dynamics and study how the results of the previous section appear in the framework of a more general description . we approach this problem describing the dynamics of the interband polarizations @xmath247 , the electron - electron @xmath248 and the hole - hole @xmath249 correlation functions by the semiconductor bloch equation ( see appendix [ app : sbe_coordinate_derivation ] ) . first we notice that the most significant contribution of the full dynamics would be the production of the response , which was absent in the approximation used in the previous section . this would be generating the interband polarization corresponding to dark states ( class @xmath187 ) . it can be proven , however , that the dynamics described by eqs . does not support excitation of the dark states . indeed , integrating both sides of the dynamical equations with respect to time in the interval @xmath250 with @xmath251 one can see that initially the spin state of the interband polarization belongs to @xmath144 . next , considering the integral over the interval @xmath252 one can see that if at instant @xmath253 the state of the system , i.e. @xmath254 , @xmath255 and @xmath256 , belongs to @xmath144 , @xmath156 and @xmath150 , respectively , then so does the state at @xmath257 . it should be noted , thereby , that in quantum wells the valence band mixing does not lead to a violation of invariance of the spin classes since the hamiltonian describing the mixing is an element of @xmath156 . thus , if initially the system is in the ground state then the interband polarization corresponding only to the bright states will be produced at least as long as the sbe and the approximation of angular independence of the dipole moment hold . as the next step we consider the limit of very short time response such that following the analysis of simple model the contribution of the kinetic energy terms can be neglected , while other terms including non - linear are kept . the solution found in the previous section , eq . , suggests the interband polarization in the form @xmath258 . substituting this ansatz into eqs . with neglected kinetic energy terms and introducing a cut - off of the coulomb potential such that @xmath259 one arrives at the system of equations independent of @xmath260 , which is structurally similar to eqs . and which yields the same result for the exciton polarization as eq . . the cancelation of the terms proportional to the coulomb potential is particularly evident from the representation of the sbe in form . this consideration shows that the most significant correction to the approximate solution results from the kinetic energy terms rather than from the coulomb interaction as one might expect considering that the coulomb interaction is responsible for the non - linear terms in the sbe . we would like to emphasize at this point that the quantity of our main interest is the exciton polarization @xmath65 since these are the excitons that constitute true single - particle semiconductor states . as follows eq . , @xmath65 is related to the interband polarization @xmath261 through @xmath262 for bound exciton states this naturally introduces a characteristic spatial scale , the exciton bohr radius @xmath263 . this circumstance together with the assumption that the typical spatial variation of the external excitation in the plane of the quantum well is small comparing to the bohr radius @xmath264 allows one to simplify the sbe using the parametric approximation . for this we introduce new spatial variable @xmath265 and @xmath266 . next we notice that in the case of single pulse or two pulse excitations the solutions of the sbe are invariant with respect to either infinitesimal or finite ( by @xmath267 ) translations in @xmath268 plane . having in mind the consecutive convolution of the solutions with the exciton wave function we can neglect the terms @xmath269 and leave only the dependence on @xmath268 as a parameter.@xcite thus , assuming for simplicity that @xmath270 we approximate eqs . by @xmath271 + \mathcal{e}_{\sigma , s}(\mathbf{r})\delta(\mathbf{r } ) - \int d\mathbf{r } ' \left [ \widetilde{\mathcal{e}}_{\sigma',s}(\mathbf{r}',\mathbf{r})h_{\sigma',\sigma}(\mathbf{r } ' - \mathbf{r},\mathbf{r } ) - \widetilde{\mathcal{e}}_{\sigma , s'}(\mathbf{r}',\mathbf{r})e_{s',s}(\mathbf{r } ' + \mathbf{r},\mathbf{r } ) \right ] , \\ % -i \dot{e}_{s_1,s_2}(\mathbf{r } , \mathbf{r } ) = & \ , \widehat{k}_{s_1,s_2}[e ] + \int d\mathbf{r } ' \left [ \widetilde{\mathcal{e}}^*_{\sigma',s_1}(\mathbf{r}',\mathbf{r})p_{\sigma',s_2}(\mathbf{r } ' + \mathbf{r},\mathbf{r } ) - \widetilde{\mathcal{e}}_{\sigma',s_2}(\mathbf{r}',\mathbf{r})p^*_{\sigma',s_1}(\mathbf{r } ' - \mathbf{r},\mathbf{r } ) \right ] , \\ % -i \dot{h}_{\sigma_1,\sigma_2}(\mathbf{r } , \mathbf{r } ) = & \ , \widehat{k}_{\sigma_1,\sigma_2}[h ] + \int d\mathbf{r } ' \left [ \widetilde{\mathcal{e}}^*_{\sigma_1,s'}(\mathbf{r}',\mathbf{r})p_{\sigma_2,s'}(-\mathbf{r } - \mathbf{r}',\mathbf{r } ) - \widetilde{\mathcal{e}}_{\sigma_2,s'}(\mathbf{r}',\mathbf{r})p^*_{\sigma_1,s'}(\mathbf{r } + \mathbf{r}',\mathbf{r } ) \right ] . \end{split}\ ] ] here @xmath272 and the integro - differential operators in eqs . are substituted by @xmath273f_{s_1}(\mathbf{r},\mathbf{r } ) -\int d\mathbf{r}'\ , v(\mathbf{r}')e_{s_1,s'}(\mathbf{r}',\mathbf{r } ) f_{s'}(\mathbf{r } - \mathbf{r}',\mathbf{r } ) , \\ % \widehat{h}_{\sigma_1,\sigma'}f_{\sigma'}= & -\frac{1}{2m_{\sigma_1,\sigma ' } } \frac{\partial^2}{\partial \mathbf{r}^2 } f_{\sigma'}(\mathbf{r } , \mathbf{r } ) + h_{\sigma_1}f_{\sigma_1}(\mathbf{r},\mathbf{r } ) -\int d\mathbf{r}'\ , v(\mathbf{r}')h_{\sigma_1,\sigma'}(\mathbf{r}',\mathbf{r } ) f_{\sigma'}(\mathbf{r } - \mathbf{r}',\mathbf{r } ) . \end{split}\ ] ] system in the short time limit when all terms in the right - hand side but those depending on the external field are neglected reproduces the results obtained in the previous section from eq . . in particular the multi - wave mixing response is obtained by the fourier transform over the parameter @xmath268 similarly to eq . . in what follows we will use eqs . in even simpler form assuming the single - pulse excitation only and thus omitting the dependence on @xmath268 . it suffices for our purpose of discussion of the effect of the internal dynamics and the coulomb interaction . we also omit the off - diagonal elements of the holes hamiltonian @xmath274 describing the mixing of the valence bands . thus we characterize these bands by @xmath275 and @xmath276 that is by the masses of the heavy- and light - holes , respectively . considering the terms besides the kinetic energy in the right - hand side of these equations as sources the responses are governed by the free - particle propagators for the correlation functions and the coulomb propagator for the interband polarization . the fundamental property of these propagators is that in the short time limit @xmath277 they turn to spatial @xmath5-functions owing to fast spatial oscillations of the kernel @xmath278.\ ] ] in order to evaluate the parameter @xmath237 we take into account that the evolution of the semiconductor initially being in the ground state is spanned by the spin classes @xmath165 . thus for the electron correlation function @xmath279 so that the propagator remains local . for the hole correlation functions this parameter is finite for the off - diagonal elements @xmath280 . for the interband polarization in the short time limit the coulomb propagator can be approximated by the free - particle propagator@xcite with @xmath281 and @xmath282 , where @xmath283 are the excitons reduced masses and @xmath284 is the electron mass . the effect of the dispersion of the electrons and holes on the _ exciton _ polarization is expressed as a distortion of the exciton wave function under the action of the free - particle propagators . the time scale , at which the distortion becomes essential , can be estimated as the time required for the initial @xmath5-shape of the kernel to acquire the width of the order of the exciton bohr radius . this leads to the estimate @xmath285 the numerical value of the typical time scale for gaas is determined by @xmath286 , @xmath287 , @xmath288 and @xmath289 , where @xmath290 is the electron mass in empty space and @xmath291 is the background dielectric function . using these values in eq . we find @xmath292 fs . thus for pulses with duration shorter than @xmath293 fs and the intensity sufficiently high to produce several rabi flops , the approximation used in the previous section is satisfactory . however , as will be shown below the exciton polarizations corresponding to the states in the discrete spectrum are less sensitive to the deviation of the dynamics from its short time limit . first we discuss the general effect of the electron - hole dispersion on the rabi oscillations . for this we consider the case of excitation by a single circularly polarized pulse of high intensity . for such excitation we neglect the contribution of the non - linear coulomb terms ( see below ) and additionally simplify the internal dynamics assuming the effects of the detuning and the energy offsets to be small at the time scales under consideration . since in this case the translational invariance in the plane of the quantum well is restored it is more convenient to solve eqs . rewriting them in the momentum representation.@xcite the laplace transformation @xmath294 of the solution is found to be @xmath295}\ ] ] where @xmath296 and latexmath:[$\omega^2_{\sigma , s } = 4 ( |\mathcal{e}_{\sigma , s}|^2 + being the hole spin state complimentary to @xmath85 in the respective spin class . in the case under consideration because of circular polarization of the excitation pulse only interband polarization corresponding to the same helicity is created , thus , if @xmath298 then @xmath299 and vice versa . we , however , write down the full " expression for @xmath300 in order to show the relation with eq . . the poles of the right - hand side of eq . determine the characteristic frequencies yielding the time dependence in the form @xmath301 - 2i \frac { \omega_{\sigma , s}(\mathbf{k } ) } { \widetilde{\omega}_{\sigma , s}(\mathbf{k } ) } \sin^2 [ \widetilde{\omega}_{\sigma , s}(\mathbf{k } ) t/2])\right\ } , \\ % e_{s , s}(t , \mathbf{k } ) & = h_{\sigma,\sigma}(t , \mathbf{k } ) = \frac{4|\mathcal{e}|^2_{\sigma , s}}{(2\pi)^2\widetilde{\omega}^2_{\sigma , s}(\mathbf{k } ) } \sin^2 [ \widetilde{\omega}_{\sigma , s}(\mathbf{k } ) t/2 ] \end{split}\ ] ] where @xmath302 . it is interesting to note the similarity to the simple dynamical models discussed in introduction [ see eq . ] . the _ exciton _ polarization is found by convoluting eq . with the exciton wave function @xmath303 according to eq . . in fig . [ fig : exciton_dispersion ] we plot the time dependence of the exciton polarization for different intensities of the external excitation . since the exciton wave function mostly localized inside the region @xmath304 and by assumption @xmath305 this convolution does not lead to appearance of new frequencies , thus the oscillations of the exciton polarization are characterized by a slightly modified rabi frequency @xmath300 . ( solid line ) and @xmath306 ( dashed line ) . , title="fig:",width=576 ] + there are two significant corrections to the picture of the oscillations obtained in the previous section . first , there is slower than exponential decay due to dephasing of the components corresponding to different @xmath307 . the decay can be estimated considering the long time asymptotic @xmath224 . since the phase in eq . reaches its stationary points at the ends of the @xmath307 interval , i.e. at @xmath308 , it implies the asymptotic power decay of the amplitude as @xmath309 independent on the intensity of the excitation . the second modification is related to the second term in the brackets in eq . [ fig : exciton_dispersion]b ] . this term has the form of oscillations with the rabi frequency @xmath300 near a constant level , which position strongly depends on the intensity of the excitation for @xmath310 and slowly decays towards @xmath52 for higher values of the rabi frequency as shown in fig . [ fig : level_absolute ] . . , title="fig:",width=288 ] + solutions allow us to estimate the effect of the coulomb interaction and to show that the contribution of the non - linear terms into the right - hand side of equations of motion vanishes with increasing the intensity of the excitation . physically this is supported by , first , observation that the interband polarization and the single particle correlation functions as functions of the external excitation are limited from above and , hence , the nonlinear terms are scaled as the exciton binding energy@xcite , while the other terms increase with increasing the intensity of the external field . second , one can also apply the result obtained from eq . about vanishing the non - linear terms in the limit of @xmath5-functional profile of the interband polarization and the single - particle correlation functions . as follows from eqs . with increasing the intensity of the excitation the spatial distribution of @xmath261 , @xmath311 and @xmath312 narrows with the characteristic width @xmath313 , where @xmath314 . there are different ways to estimate the contribution of the non - linear coulomb terms into the dynamics . in order to illustrate the application of the physical ideas mentioned above we consider the contribution into the right - hand side of the equation of motion for the interband polarization coming from the term written in the second line in eq . . in the simplest case of a single pulse excitation this term yields an addition to the equation with respect to @xmath315 in the form @xmath316,\ ] ] where @xmath317 and we have taken into account that the dynamics is restricted to a particular spin class . we illustrate the typical line of calculations leaving only the first term in the expression for @xmath318 and approximating the oscillating terms by @xmath319 thus presenting @xmath320.\ ] ] rescaling the wave vectors and using for the single - pulse excitation @xmath321 we obtain @xmath322 where @xmath323 is the exciton binding energy and @xmath324 the dependence of this function on @xmath325 is shown in fig . [ fig : integral_kappa ] . it varies slowly over the interval @xmath326 , where its magnitude is close to @xmath327 . thus , if the rabi frequency exceeds the exciton binding energy or , equivalently , if the typical spatial scale of the interband polarization @xmath328 is smaller than the exciton bohr radius the effect of the non - linear terms on the time dependence of the exciton polarization reduces to a relatively small modification of the amplitude of the external field . of the integral in eq . describing the effective modification of the source due to the nonlinear terms in the limit of high intensity of the external field.,title="fig:",width=288 ] + this argument allows one to reformulate the criterion of validity of the short time approximation for the exciton polarization in terms of relation between the exciton bohr radius @xmath263 and the typical spatial width of the _ interband _ polarization due to the interaction with the external field @xmath329 . if @xmath330 then the result of convolution of the exciton wave function with the interband polarization @xmath261 does not depend on the details of the dependence of @xmath318 on @xmath331 . indeed , since in this case the main contribution into the integral @xmath332 comes from the small area near @xmath333 we can approximate the integral by @xmath334 which turns eq . into eq . exactly ( including the cancelation of the factor @xmath335 ) . this consideration shows that the exciton polarization corresponding to the states in the discrete spectrum is less sensitive to the deviations from the short - time dynamics of the system than the interband polarization . the latter contains information about all exciton states including those from the continuum spectrum whose spatial variation is no longer characterized by the exciton bohr radius . the condition @xmath336 imposes the limitation on the intensity of the external field that the period of the rabi oscillations should be smaller than @xmath337 fs ( for gaas ) . an interesting effect of the dispersion is revealed when the excitation pulse is not circularly polarized . in this case both states constituting a spin class are involved in the dynamics resulting in the appearance of a new characteristic frequency . in order to illustrate this effect we consider the case when the intensity of the excitation is significantly high so that the effect dispersion of the hole - hole correlation function , which is quantified by the parameter @xmath338 , can be treated perturbatively . in order to simplify the expression we assume additionally that both heavy- and light - hole excitons are described by the same dispersion law @xmath339 . under these assumptions the laplace transform of the interband polarization can be written as @xmath340 [ ( \lambda + i \omega_{\sigma , s}(\mathbf{k})/2)^2 + \widetilde{\omega}^2_{\sigma , s}(\mathbf{k})/4]},\ ] ] where @xmath341 is given by eq . and @xmath342 with @xmath343 , i.e. @xmath344 . the perturbation term in the right - hand side of eq . has poles at the same rabi frequency @xmath345 as in the case of the circularly polarized excitation and at the frequency @xmath346 yielding a component of the exciton polarization oscillating with doubled period we would like to emphasize that the new frequency was initially present among the eigenfrequencies of the bright exciton classes [ see eq . ] . its appearance , however , in the time evolution of the exciton polarization was prohibited in the dispersionless limit by the symmetry @xmath347 of dynamical equations . the exciton finite masses break the symmetry allowing for the frequency @xmath348 to contribute to the evolution . with this regard it could be speculated that if the dark excitons would be involved in the dynamics their presence could be traced by admixture of the frequencies specific for the dark exciton spin class @xmath187 [ see eq . ] . it should be emphasized that neither effect of decay of the rabi oscillations nor the admixture of extra " frequencies are present in atomic analogues of the semiconductor rabi oscillations . the decay is essentially caused by the spatial spread of the true ( single - particle ) eigenstates of the system . the appearance of additional frequencies is related to the existence of different kinds of holes with continuous spectra characterized by different masses . we have studied the short time semiconductor response with respect to resonant high intensity excitations . the main results are obtained neglecting the slow dynamics , that is on the time scale provided by detuning . we have calculated exactly the excitation polarization @xmath349 in the coherent limit . we have found that if the semiconductor is excited by a field consisting of one or two plane waves the polarization of bright excitons demonstrates rabi oscillations , while the polarization of dark excitons identically zero . the main difference between the single - wave and two - wave excitations is that in the latter case the rabi oscillations of the multi - wave mixing ( mwm ) polarizations are generally a superposition of two harmonics ( @xmath350 ) unless the amplitudes of the waves satisfy a special relation . for example , if both waves are circularly polarized with the same helicity then @xmath351 , where @xmath352 is amplitude of the weaker wave , and the slow harmonic @xmath353 collapses only if the amplitudes of the both waves are the same . also , in the case of the two pulse excitation oscillations of the exciton polarization decay @xmath222 owing to redistribution of the excitation over the multi - wave mixing polarizations . in order to establish the dynamical origin of the vanishing polarization of the dark states we considered the operator dynamics governed by the external field . we found that there exist three invariant spin classes , which do not mix with the evolution of the system . two of these classes correspond to bright excitons and the third one contains all the dark states . it has turned out that the operator dynamics is described by six frequencies . only two of those frequencies show up as the rabi frequencies in the time evolution of the exciton polarizations ( one for each bright classes ) , while others are prohibited by a hidden symmetry . the short time approximation used in the first part of the paper appears naturally in the more general context of the semiconductor bloch equation ( sbe ) . we use the sbe in order to discuss the effect of the coulomb interaction and the effect of the continuous spectra of the electrons and holes on the results obtained in the limit of immediate response with respect to intensive excitation . we found that if the system is initially in the ground state then the sbe preserves the invariant spin classes . thus , it proves that as long as the sbe holds and the approximation of angular independent dipole moment is justified the dark excitons do not appear in the dynamics . an interesting result of studying the short time approximation within the context of the sbe is that the coulomb interaction does not lead to essential changes if the intensity of excitation is sufficiently high compared to the exciton binding energy . moreover , including the nonlinear exchange terms directly to the equations of the short time limit ( i.e. without the kinetic energy ) adds nothing new since they cancel each other . in turn , the effects of the continuous spectrum of the electrons and holes are much more significant . we show that the electron - hole dispersion while does not change significantly the frequency of the oscillations of the exciton polarization leads to the power law decay of the oscillations . an interesting effect is related to the mass difference between heavy- and light - holes . it breaks the symmetry that selects only two rabi frequencies out of six that are introduced into the system by the external field and leads to admixture of the second nontrivial frequency for a specific bright class . we would like to thank lu sham and cristiano ciuti for useful discussions . we acknowledge partial support through the nsf under grant no . eccs 0725514 and through the darpa / mto young faculty award under grant no . hr0011 - 08 - 1 - 0059 . in order to derive eqs . one can either consider a regularization of the @xmath5-function , which appears in the anticommutation relations and in the right - hand - side of eq . or consider a more general form of eq . with the amplitudes being scalar functions , which are convoluted with the electron and hole operators , @xmath354 furthermore the heisenberg equation of motion produces the dynamical equations with respect to the amplitudes @xmath355 , \end{split}\ ] ] where the summation over dashed spin variables is implied . the last equation gives the time dependence of @xmath356 ( compare with eq . ) . considering the dynamics of the diagonal , @xmath128 , values of the amplitudes in the case of spatially homogeneous excitation @xmath357 one obtains eq . . it should be emphasized that eqs . are not a semiconductor bloch equations . these are equations with respect to amplitudes of different operators entering the heisenberg representation of the operator @xmath358 . the semiconductor bloch equation , in turn , is written for average values of the specific operators [ see e.g. eq . ] . while these equations have similar structure ( in the linear approximation for the sbe ) they are different as can be see comparing closely eqs . and . in order to find an average value of a particular operator using eqs . one needs to solve these equations for specific initial conditions and then use this solution in the last equation of eqs . . in the main text we consider the semiconductor response under the action of an external field , which is spatially inhomogeneous in the plane of the quantum well . because the system does not posses the translational invariance it is convenient to use the semiconductor bloch equation in the coordinate representation . the hamiltonian of an excited semiconductor in the rotating wave approximation has the form @xmath359 where @xmath360 and @xmath361 are the standard electron and hole single - particle hamiltonians@xcite , @xmath362 is the hamiltonian of the coulomb interaction and @xmath49 describes the light - matter interaction @xmath363 where @xmath364 , operator @xmath365 creates a hole with the spin state @xmath85 at point @xmath47 and @xmath366 is the respective electron creating operator . the equations of motion for the interband polarizations @xmath247 , the electron - electron @xmath248 and the hole - hole @xmath249 are derived from the equation of motion @xmath367}$ ] using the hartree - fock approximation for the terms describing the coulomb interaction . straightforward calculations give immediately quite cumbersome system of equations . we provide explicitly only one equation , which shows the structure of the non - linear coulomb terms @xmath368 + \left[u_\delta(\mathbf{x}_1 ) - u_\delta(\mathbf{x}_2)\right]{p}_{\sigma , s}(\mathbf{x}_1,\mathbf{x}_2 ) \\ % & -\int d\mathbf{x}'\,\left[v(\mathbf{x}_1 -\mathbf{x } ' ) - v(\mathbf{x}_2 -\mathbf{x}')\right ] \left[h_{\sigma',\sigma}(\mathbf{x}',\mathbf{x}_1)p_{\sigma',s}(\mathbf{x}',\mathbf{x}_2 ) - p_{\sigma , s'}(\mathbf{x}_1,\mathbf{x}')e_{s',s}(\mathbf{x}',\mathbf{x}_2)\right ] \\ % & + \mathcal{e}_{\sigma , s}(\mathbf{x}_1)\delta(\mathbf{x}_1 - \mathbf{x}_2 ) - h_{\sigma',\sigma}(\mathbf{x}_2,\mathbf{x}_1)\mathcal{e}_{\sigma',s}(\mathbf{x}_2 ) - \mathcal{e}_{\sigma , s'}(\mathbf{x}_1 ) e_{s',s}(\mathbf{x}_1,\mathbf{x}_2 ) . \end{split}\ ] ] here and below a summation over dashed spin indices is implied . this equation clearly demonstrates that at the diagonal @xmath369 the contributions of the coulomb terms cancel each other . in eq . @xmath370 describes an effective background potential created by the local imbalance between the electrons and holes @xmath371.\ ] ] for a semiconductor excited by a single plane wave this potential vanishes in an overall neutral system due to the translational invariance . in the two wave excitation setup the effect of @xmath372 is small for bound exciton states if the order of the multi - wave mixing response is not too high . the dynamical equations significantly simplify after the terms being rearranged to form renormalizations of the single - particle energies and the coupling between the charge densities and the interband polarizations @xmath373 + \mathcal{e}_{\sigma , s}(\mathbf{x}_1)\delta(\mathbf{x}_1 - \mathbf{x}_2 ) - \int d\mathbf{x } ' \left [ \widetilde{\mathcal{e}}_{\sigma',s}(\mathbf{x}',\mathbf{x}_2)h_{\sigma',\sigma}(\mathbf{x}',\mathbf{x}_1 ) - \widetilde{\mathcal{e}}_{\sigma , s'}(\mathbf{x}',\mathbf{x}_1)e_{s',s}(\mathbf{x}',\mathbf{x}_2 ) \right ] , \\ % -i \dot{e}_{s_1,s_2}(\mathbf{x}_1,\mathbf{x}_2 ) = & \ , \widehat{k}_{s_1,s_2}[e ] + \int d\mathbf{x } ' \left [ \widetilde{\mathcal{e}}^*_{\sigma',s_1}(\mathbf{x}',\mathbf{x}_1)p_{\sigma',s_2}(\mathbf{x}',\mathbf{x}_2 ) - \widetilde{\mathcal{e}}_{\sigma',s_2}(\mathbf{x}',\mathbf{x}_2)p^*_{\sigma',s_1}(\mathbf{x}',\mathbf{x}_1 ) \right ] , \\ % -i \dot{h}_{\sigma_1,\sigma_2}(\mathbf{x}_1,\mathbf{x}_2 ) = & \ , \widehat{k}_{\sigma_1,\sigma_2}[h ] + \int d\mathbf{x } ' \left [ \widetilde{\mathcal{e}}^*_{\sigma_1,s'}(\mathbf{x}_1,\mathbf{x}')p_{\sigma_2,s'}(\mathbf{x}_2,\mathbf{x } ' ) - \widetilde{\mathcal{e}}_{\sigma_2,s'}(\mathbf{x}_2,\mathbf{x}')p^*_{\sigma_1,s'}(\mathbf{x}_1,\mathbf{x } ' ) \right ] , \end{split}\ ] ] where we have introduced the modified coupling between the interband polarization and the charge densities @xmath374 we represent the time dependence of the external field in the form @xmath375 , where the amplitude @xmath376 is assumed to be constant during the excitation . in the rotating frame the operators @xmath377 = & \ , \widehat{h}_{\sigma,\sigma'}p_{\sigma',s } + p_{\sigma , s'}\widehat{h}_{s',s } - v(\mathbf{x}_1-\mathbf{x}_2)p_{\sigma , s}(\mathbf{x}_1,\mathbf{x}_2 ) - \omega p_{\sigma , s}(\mathbf{x}_1,\mathbf{x}_2 ) , \\ % \widehat{k}_{s_1,s_2}[e ] = & \ , \widehat{h}_{s_1,s'}e_{s',s_2 } - e_{s_1,s'}\widehat{h}_{s',s_2 } , \\ % \widehat{k}_{\sigma_1,\sigma_2}[h ] = & \ , \widehat{h}_{\sigma_1,\sigma'}h_{\sigma',\sigma_2 } - h_{\sigma_1,\sigma'}\widehat{h}_{\sigma',\sigma_2 } \end{split}\ ] ] are expressed in terms of the integro - differential operators @xmath378 . the action of these operators is defined by @xmath379f_{s_1}(\mathbf{x}_1,\mathbf{x}_2 ) -\int d\mathbf{x}'\ , v(\mathbf{x}_1 - \mathbf{x}')e_{s_1,s'}(\mathbf{x}_1,\mathbf{x } ' ) f_{s'}(\mathbf{x}',\mathbf{x}_2 ) , \\ % \widehat{h}_{\sigma_1,\sigma'}f_{\sigma'}= & -\frac{1}{2m_{\sigma_1,\sigma ' } } \frac{\partial^2}{\partial \mathbf{x}_1 ^ 2 } f_{\sigma'}(\mathbf{x}_1 , \mathbf{x}_2 ) + \left[v_h(\mathbf{x}_1 ) - u_\delta(\mathbf{x}_1 ) + h_{\sigma_1}\right]f_{\sigma_1}(\mathbf{x}_1,\mathbf{x}_2 ) \\ % & \qquad -\int d\mathbf{x}'\ , v(\mathbf{x}_1 - \mathbf{x}')h_{\sigma_1,\sigma'}(\mathbf{x}_1,\mathbf{x } ' ) f_{\sigma'}(\mathbf{x}',\mathbf{x}_2 ) \end{split}\ ] ] with @xmath380 and @xmath381 being the electron and hole confinement potentials , which effect on the spatial motion of the electrons and holes is estimated by the @xmath5-functional approximation for the quantum well . the parameters @xmath382 and @xmath383 denote the energy offsets of the respective bands . the diagonal elements of @xmath384 are the masses of the light- and heavy - holes , while the off - diagonal elements account for the valence bands mixing . thus , @xmath385 has the meaning of a two - particle hamiltonian with the coulomb interaction between them , while the properties of free particles are described by the hamiltonians @xmath274 and @xmath386 .

we present near - infrared photometry and optical spectroscopy of very low - mass stars and brown dwarf candidates in the pleiades open cluster . the membership status of these objects is assessed using color - magnitude diagrams , lithium and spectral types . eight objects out of 45 appear to be non - members . a search for companions among 34 very low - mass pleiades members ( m.09 m ) in high - spatial resolution images obtained with the hubble space telescope and the adaptive optics system of the canada - france - hawaii telescope produced no resolved binaries with separations larger than 0.2 arcsec ( a au ; p years ) . nevertheless , we find evidence for a binary sequence in the color - magnitude diagrams , in agreement with the results of steele & jameson ( 1995 ) for higher mass stars . we apply the lithium test to two objects : cfht - pl-16 , which lies in the cluster binary sequence but is unresolved in images obtained with the hubble space telescope ; and cfht - pl-18 , which is binary with 0".33 separation ( martn et al . 1998 ) . the first object passes the test , but the second object does not . we conclude that cfht - pl-16 is an pleiades brown dwarf binary with separation au , and that cfht - pl-18 is a foreground system . we compare the multiplicity statistics of the pleiades very low - mass stars and brown dwarfs with that of g and k - type main sequence stars in the solar neighborhood ( duquennoy & mayor 1991 ) . we find that there is some evidence for a deficiency of wide binary systems ( separation au ) among the pleiades very low - mass members . we briefly discuss how this result can fit with current scenarios of brown dwarf formation . we correct the pleiades substellar mass function for the contamination of cluster non - members found in this work . we find a contamination level of 33% among the brown dwarf candidates identified by bouvier et al . ( 1998 ) . assuming a power law imf across the substellar boundary , we find a slope dn / dmm , implying that the number of objects per mass bin is still rising but the contribution to the total mass of the cluster is declining in the brown dwarf regime .

astro-ph0004174

brown dwarfs ( bd ) provide the opportunity of extending classical studies of stellar properties into the hitherto unexplored realm of objects with masses lower than the h - burning limit ( 0.075 m@xmath1 ; e.g. baraffe et al . 1998 ) . stars are frequently associated with other stars in multiple systems . 57% of nearby g - type dwarfs have h - burning companions ( abt & levy 1976 ; duquennoy & mayor 1991 , hereafter dm91 ) . the multiplicity frequency seems to be lower among nearby m - type dwarfs ( 38% , henry & mccarthy 1990 ; fisher & marcy 1992 ) . this difference could be due to the smaller mass range available for the stellar secondaries of m dwarfs . on the other hand , the distribution of binary separations is similar for g- and m - type stars . it has a broad maximum from 3 to 30 astronomical units ( au ) . there is no indication that stellar binary properties depend on primary mass . stellar clusters are important for studying the multiplicity properties as a function of age , metallicity and mass . one of the best studied open clusters is the pleiades . it is nearby ( d=125 pc ) , young ( 120 myr ) , the metallicity is close to solar , and there are more than 800 known members ( see hambly 1998 for a review ) . the pleiades mass function has been studied over a broad mass range , from 4 m@xmath1 to 0.04 m@xmath1 ( hambly , hawkins & jameson 1993 ; meusinger , schilbach & souchay 1996 ; bouvier et al . 1998 , hereafter b98 ) . cluster bds have been confirmed via the lithium test ( basri , marcy & graham 1996 ; rebolo et al . several imaging surveys have recently identified a numerous population of pleiades bd candidates ( b98 ; festin 1998 ; zapatero osorio et al . 1999 ; hambly et al . follow - up observations of these objects is necessary for assessing their membership . the search for bd binaries is interesting because of the following reasons : 1 ) any fainter and cooler secondary of a bd should have even lower mass than the primary . 2 ) the binary frequency among bds and the distribution of orbital periods and eccentricities is an important clue for understanding the formation of these objects . 3 ) bd binaries provide the opportunity to measure dynamical masses , which are necessary for calibrating evolutionary models . 4 ) the pleiades substellar mass function needs to be corrected for binarity . the paper is organized as follows : in section 2 we present the observations and describe the data analysis . in section 3 we search for companions and we construct color magnitude diagrams . in section 4 we discuss the cluster membership of our objects , we derive parameters for the likely members , we estimate the binary frequency , and we discuss the implications of our results for the substellar mass function . our list was selected among the following pleiades vlm candidate members ( [email protected] m@xmath1 ) : ` calar ' and ` teide ' objects from zapatero osorio et al . ( 1997b ) and martn et al . ( 1998b ) ; ` cfht - pl ' objects from b98 ; ` hhj ' objects from hambly et al . ( 1993 ) ; ` mho ' objects from stauffer et al . ( 1998b ) ; ` ppl ' objects from stauffer et al . ( 1989 , 1994 ) and ` roque ' objects from zapatero osorio et al . ( 1997a , 1997c , 1999 ) . these surveys include most of the known pleiades bd candidates . for the hst / nicmos program , we gave preference to objects with known near - infrared ( nir ) magnitudes and/or spectral types , and we included all the pleiades bds with lithium detections ( rebolo et al . 1996 ; martn et al . 1998 ; stauffer et al . 1998a ) , with the exception of ppl 15 ( stauffer et al . 1994 ; basri et al . 1996 ) because archive hst observations were already available . we observed a total of 30 objects , but one of them was a repetition ( ppl1=roque15 ) . for the ground - based near - ir and spectroscopic observations , we selected all the cfht objects that had not been observed in previous campaigns . we obtained near - ir photometry for 22 objects and spectroscopy for 17 . the targets were centered in the field of view of the nicmos camera 1 ( nic1 ) . exposures of 447.95 s , 383.95 s and 383.95 s were obtained in multiple - accumulate mode with filters f110 m , f145 m , and f165 m ( thompson et al . 1998 ) , respectively . each target was observed during one orbit ( average visibility 52 minutes ) . we used the iraf / digiphot package for data reduction and analysis . magnitudes for all the targets were computed based on the header keyword photfnu and are given in table 1 . these values should be used with caution because the values for photfnu are valid for sources with a constant flux per unit wavelength across the band pass , which might not be the case for our sources . other limiting factors in the accuracy of the magnitudes are that we used model point spread functions ( psf ) for fitting the data that did not always provide a perfect match to the observed psf , and uncertainties in the nicmos darks and resulting spatial variations in the background . we used on - orbits darks ( as opposed to the model darks used in the standard nicmos pipeline ) to improve the photometric accuracy , but the correction was still not perfect . the following pleiades vlm objects have been observed with wfpc2 in other hst programs : hhj 3 , hhj 5 , hhj 6 , hhj 10 , hhj 11 , hhj 14 , hhj 19 , hhj 36 and ppl 15 . we retrieved the data from the hst archive . none of them shows any companion in the f785lp filter up to 4 magnitudes fainter than the primary at separations between 0``.15 and 4.''0 . near - ir broad - band data was collected in the following observing runs : 1 ) on 21 - 23 september 1997 at the 1 m nickel telescope of lick observatory using the lirc ii camera . the wide field of view ( fov ) was selected ( 7.29 arc mins@xmath6 , 1.71 arc secs pix@xmath7 ) . 2 ) on 14 - 16 november 1997 using the 1.2 m telescope and the stelircam ir camera . stelircam obtains j and k band data simultaneously using two 256x256 insb detector arrays and a dichroic filter . the pixel scale used for these observations was 0.3 arcseconds per pixel . the typical total integration time per object was about 20 minutes , and the seeing was about 1.5 arcseconds on average . 3 ) on 13 - 15 january 1998 at the 3.6 m canada - france - hawaii telescope ( cfht ) using the adaptive optics ( ao ) system pueo ( rigaut et al . 1998 ) and the kir ( doyon et al . 1998 ) ir camera with a 30 arcsec fov . integration times ranged from 100 to 150 seconds in jhk broad - band filters . images were dark and flat - field corrected and aperture photometry performed with iraf / apphot . ukirt faint standards were observed every night for photometric calibration . for 10 of the 12 cfht candidates observed during this run ( see table 2 ) , the adaptive optics system provided an angular resolution of 0.2 to 0.3 arcsec fwhm . the 2 remaining objects , cfht-17 and 18 , had no proper ao guiding star and have 0.4 - 0.5 arcsec fwhm on the final images . we obtained spectra using the keck ii telescope with the low resolution imaging spectrograph ( lris ; oke et al . 1995 ) , the 4 m telescope at kitt peak national observatory ( kpno ) with the cryogenic camera ( cryocam ) spectrograph , and the 4.2 m william herschel telescope ( wht ) with isis . the observing log is given in table 3 . we rotated the slit in order to be at parallactic angle at the beginning of the exposure in order to minimize refraction losses . all the ccd frames were bias subtracted , flat fielded , sky subtracted , and variance extracted using routines within iraf . we calibrated in wavelength using the emission spectrum of hgnear lamps . correction for instrumental response was made using spectra obtained in the same night of flux standards with fluxes available in the iraf environment . we did not use order - blocking filters for the keck observations of the pleiades bd candidates . these objects are so red that second - order contamination is negligible . we observed vb10 without filter and with the og570 order - blocking filter . we did not find any significant contamination due to second - order light . we calibrated the lris spectra with the flux standard star bd+262606 , which was observed the same night as the pleiades bd candidates and with same instrumental configuration . the spectrum of the standard star obtained without filter was used blueward of 680 nm , and that obtained with the og570 filter was used redward of 680 nm . at kpno we used the og550 filter to block second - order light . the cryocam and wht spectra were flux calibrated with standard stars that have data available in the iraf database . fig . 1 shows a color - magnitude diagram ( cmd hereafter ) in the nicmos f110 m and f165 m filters . the f145 m filter is strongly affected by water vapour in the spectrum of very cool objects . the comparison with the theoretical models is less reliable because of uncertainties in the steam line lists and opacities . our objects are plotted in the cmd together with the theoretical isochrones ( age 120 myr ) of chabrier et al . the solid line denotes the dust - free models ( nextgen ; allard et al . 1997 ; hauschildt et al . the dashed line represents the new models that include dust effects ( dusty ; allard et al . it is expected that for temperatures cooler than 2800 k dust forms in the atmosphere of ultracool dwarfs ( jomes & tsuji 1997 ; marley et al . 1999 ) . in the nextgen models , a t@xmath8 of 2800 k corresponds to a mass of 0.075 m@xmath1 and m@xmath9=10.31 . the effect of dust formation in fig . 1 is to make the objects bluer than the nextgen isochrone because of `` backwarming '' ( e.g. leggett , allard & hauschildt 1998 ) . we start seeing clear evidence for a turnover to bluer colors in the pleiades sequence at [email protected] , which corresponds to a t@xmath8=2320 k. hence , we find that for t@xmath8 lower than about 2300 k , dust becomes an important opacity source in the atmospheres of pleiades bds . the dusty isochrone gives a good fit to the position of our faintest object ( roque 25 , spt = l0 ) . intermediate models between nextgen and dusty are not available . they seem to be required to fit the location of pleiades bds between [email protected] and m@xmath1012 . all of our targets with previously known lithium detections are located close ( f110m - [email protected] mag ) to the nextgen isochrone in fig . 1 . these lithium bds are bona - fide pleiades members and help to define the cluste sequence . they are : cfht - pl 15 , mho 3 , ppl 1 , roque 13 , teide 1 and teide 2 . calar 3 has not been plotted because the photometry is very uncertain due to guiding problems . ppl 1 was observed with nic1 at two different epochs ( 1 sep 1998 and 9 sep 1998 ) because we had not realized that it is actually the same object as roque 15 ( zapatero osorio et al . 1997c did not notice it neither ) . we obtained different magnitudes for the two sets of observations ( table 1 ) . in the second epoch the object was brighter and bluer . we have marked the variation in the location of ppl 1 in the cmd of fig . 1 . the difference is much larger in the f145 m filter ( 8 @xmath12 ) than in the other two filters , suggesting variability in the steam absorption . this could be a hint of a `` weather '' change in the brown dwarf . periodic photometric variability has been observed in i - band ccd observations of two cluster vlm stars , one in @xmath13per ( martn & zapatero osorio 1997 ) and the other in the pleiades ( terndrup et al . the light changes in these stars are probably due to surface thermal inhomogeneities ( dark spots ) modulated by the rotation of the object(s ) . on the other hand , bailer - jones & mundt ( 1999 ) have recently failed to detect photometric variability in calar 3 , roque 11 and teide 1 . further photometric monitoring of ppl 1 is necessary to clarify if the observed variability in nicmos filters is due to weather or magnetic spots . the example of ppl 1 indicates that some of the spread observed in the pleiades sequence of fig . 1 could be due to intrinsic variability of the objects . other sources of scatter in the cmd are unresolved binaries and non - members of the cluster . we note two examples of each kind of object . cfht pl 16 is brighter and redder than the rest of the sequence . it also stands out in other diagrams that are discussed in the next sections , and was noted as a possible binary by b98 . we consider it as a likely unresolved binary . cfht pl 22 is much bluer and fainter than the pleiades sequence . it is likely a non - member . the membership status of our sample is discussed in detail in section 4.1 . we have combined our new broad - band near - ir photometry with data available in the literature ( b98 ; festin 1998 ; zapatero osorio et al . 1997a , b ; martn et al . 1998c ) to produce the i vs i - k cmd shown in fig . 2 . we used the following extinction corrections for all the objects : a@xmath14=0.06 , a@xmath15=0.01 . we have compared the data with theoretical isochrones ( ages=100 and 120 myr ) . dusty and nextgen models are represented with dashed and solid lines , respectively . there is a well defined observational sequence , which is conveniently fitted by the the nextgen isochrones down to m@xmath16=14 . our faintest object , roque 25 , is bluer than the dust - free isochrone . its location in fig . 3 is in good agreement with the dusty isochrone . this is explained by the greenhouse effect of dust grains ( chabrier et al . 2000 ) . the objects located well above the nextgen isochrone could be binaries with nearly identical components . one of them is ppl 15 , which is known to be a short - period nearly equal - mass binary ( basri & martn 1999 ) . the other objects are : not 1 , hhj 6 , cfht - pl 6 , cfht - pl 12 and cfht - pl 16 . only the later one was included in our hst / nicmos program . hhj 6 was observed with hst / wfpc2 , and cfht - pl 12 was observed with the cfht ao system . if they are binaries , they must have angular separations smaller than 0".2 ( 25 au ) . the objects located well below the nextgen isochrone for m@xmath1714 are likely non - members . they are identified with five - pointed star symbols in fig . 2 . they are cfht - pl 19 , 20 and 22 . we display the spectra of our faintest bd objects in fig . we derived spectral types for our targets using the calibration of the pseudocontinuum index ( pc3 ) given by martn et al . the results are given in table 4 . our values are in good agreement with previous work for the objects in common . using these spectral types and those published by steele & jameson ( 1995 ) , martn , rebolo & zapatero osorio ( 1996 ) , cossburn et al . ( 1997 ) , zapatero osorio et al . ( 1997c ) and festin ( 1998 ) , we have made the diagram shown in fig . two objects lie well outside the pleiades sequence , one above it ( cfht - pl 16 ) and one below it ( cfht - pl 26 ) . the first one is likely an unresolved binary , and the second one is probably a non - member . cfht - pl 18 is an important object because it is the only binary that we have resolved in our hst images . it lies on the cluster sequence in the cmd diagrams discussed above . martn et al . ( 1998a ) obtained a radial velocity consistent with cluster membership . in order to confirm its membership , we obtained additional mid - resolution spectra of cfht - pl 18 around the lii resonance line at 670.8 nm for applying the lithium test for bds ( magazz et al . 1993 ) . if cfht - pl 18 is indeed a member , both components should have preserved their initial lithium content because they are fainter than the pleiades substellar boundary ( martn et al . 1998b ; stauffer et al . 1998a ) . in fig . 5 , we show the final spectrum where we do not detect the lithium feature . we put an upper limit to the lii equivalent width of 200 ma , which is a factor of 5 lower than the equivalent width ( ew ) measured in teide 1 ( rebolo et al . 1996 ) , and a factor 2.5 lower than the ew measured in ppl 15 ( basri et al . thus , we are confident that cfht - pl 18 does not pass the lithium test . this binary system appears to be located at precisely the right distance to be confused with a pleiades member . we do not know the distance to this system because it is not a cluster member , but we can estimate it from the spectral type ( m8 , martn et al . 1998a ) . if it has the same absolute j - band magnitude as lhs 2397a ( m8 , m@xmath18=11.13 , leggett et al . 1998 ) , we obtain a spectroscopic parallax of 105 pc . the projected separation of the binary would then be 34.5 au . cfht - pl 16 is located more than 0.5 mag above the pleiades sequence in all the cmds . it could be an unresolved binary with nearly identical components . alternatively , cfht - pl 16 could be a non - member . we obtained keck / lris mid - resolution spectroscopy to test its membership . the reduced spectrum is shown in fig . we clearly detect h@xmath19 in emission with ew=-6.1 a. the quality of our spectrum is barely sufficient for a detection of the lii resonance line . there is an absorption feature at the position of the lii resonance line with [email protected] a , which is similar to the ew of teide 1 ( rebolo et al . 1996 ) , but there are also noise features of similar strength elsewhere in the spectrum . we estimate that the probability of having a noise feature at the position of the lii line is @xmath310% , i.e. low but not negligible . we measured a heliocentric radial velocity of 13@xmath115 km / s using the h@xmath19 feature , which is consistent with the radial velocities of vlm pleiades members ( martn et al . 1998b ) . since the nicmos psf can vary significantly from one hst orbit to the next , we used all of the hst / nicmos pointings to build up a library of hst / nic1 psfs in all three filters ( f110 m , f145 m , anf f165 m ) . for each set of observation , we then identified the best matching psf in the library for psf subtraction . sub - pixel offsets between target and psf were computed by cross - correlating the individual frames . the psf was then fourier - shifted , scaled , and subtracted . 6 shows examples of nic1 images before and after point spread function ( psf ) subtraction . the dark line seen in each image is the boundary between the individual nicmos quadrants . in the original image we can clearly see the first and second airy rings . they are considerably reduced in the subtracted image . we have estimated limiting sensitivities to the presence of companions as a function of separation from the primary . they are given in table 5 for two different separations . a few radial averages of the limits to the presence of companions in nicmos images are shown in fig . they were derived under the assumption that we would detect a source if the counts in the central pixel of its psf are at least 3 times the sigma of the background . the four objects , namely cfht - pl 15 , cfht - pl 19 , hhj 2 and mho 3 , for which the residuals of the psf subtraction were high have been flagged in table 5 . the residuals could be due to the presence of faint companions within 5 pixels ( 0.215 " ) of the primary , or to unrepeatable structure in the psf . we have not found any distant ( sep.@xmath41 " ) companions in the near - ir images taken at cfht and mt hopkins . any companion up to 3 magnitudes fainter than the primary in the separation range 1 to 8 arcseconds , should have been detected . the membership criteria that we have used are the following : 1 ) is the object a dwarf or a giant ? all the objects for which we have low - resolution spectra have ki and nai lines similar to those of dwarfs . because of the low - resolution of most of our spectra , the nai doublet at 818.3 and 819.0 nm is blended with telluric absorption and we can not derive an accurate equivalent width . thus , we do not use the strength of the line as a membership criterion ( martn et al . 2 ) is the radial velocity similar to those of known members ? we have used radial velocity data from the literature ( basri et al . 1996 ; martn et al . 1998b ; stauffer et al . 3 ) is the proper motion similar to the cluster ? we have used proper motion data from the literature ( hambly et al . 1993 ; hambly et al . 4 ) does the object have lithium ? pleiades members fainter than [email protected] should have lithium ( martn et al . 1998b ; stauffer et al . brighter objects could have lithium if they are binaries ( ppl 1 ) . 5 ) does the object have h@xmath19 in emission ? the majority of pleiades members do show h@xmath19 in emission because of their young age . however , h@xmath19 is variable and sometimes not detected . for example , we did not detect it in our keck spectrum of teide 1 , but it was previously detected by rebolo et al . we do not reject any candidate member solely on basis of lack of h@xmath19 emission , but we note that many non - members do not have h@xmath19 emission , suggesting that they are relatively old field dwarfs . 6 ) we checked the position of each object in the spectral type versus apparent i - band magnitude ( fig . 4 ) . the lithium bds are considered as benchmark objects , and most of the bd candidates are located near them . objects that deviate from the sequence by more than 0.8 magnitudes are considered as likely non - members . 7 ) the location of the lithium bds in the i versus ( i - k ) and f110 m versus ( f110m - f165 m ) cmds is well fitted with the nextgen isochrone . on the other hand , the location of roque 25 in the same cmds is fitted by the dusty isochrone . objects that are 3@xmath12 away from the isochrones are considered as likely non - members . we summarize our assessment of the membership status of 45 objects in table 6 . we confirm the pleiades membership of 37 vlm stars and bds , and reject 8 objects as non - members . the later objects are excluded from our calculation of binary statistics in the pleiades cluster . in order to derive the physical parameters of the objects that we consider as likely pleiades members , we adopted a common age , distance and metallicity of 120 myr , 125 pc and solar , respectively . the only way to infer masses for our objects is to compare their location in the cmd with theoretical calculations . in fig . 8 , we show the theoretical mass - luminosity relationships for several near - ir filters that we have adopted . the behaviour of the nicmos filters f110 m and f165 m mimics that of the @xmath20 and @xmath21 broad - band filters , respectively . the sensitivity to dust effects changes with @xmath22 and mass ( temperature ) across these plots . we have minimized the dependence of our results on the role of dust in the atmosphere by using the filters that show less sensitivity to the choice of model ( nextgen or dusty ) for a given mass . for all the objects we used the absolute magnitudes in the j and f110 m filters ( we averaged the two of them if both magnitudes were available ) and the nextgen models , but for roque 25 we used the k - band absolute magnitude and the dusty model . our mass estimates for the cfht objects are consistent within @xmath110.005 m@xmath1 with those derived by b98 using three different sets of models and absolute i - band magnitudes . the masses of the primaries considered in this study range from 0.14 m@xmath1 ( cfht - pl-1 , i=16.1 , b98 ) to 0.035 m@xmath1 ( roque 25 , i=21.17 , martn et al . the masses of the primaries observed with high - resolution techniques ( ao and hst ) are all lower than 0.11 m@xmath1 ( hhj19 , i=16.7 , hambly et al . 1993 ) . we estimated upper limits to the masses of undetected companions using the dusty models and the sensitivity limits in the @xmath23 and/or f165 m filters . the average limits on the mass ratios ( q = m@xmath24/m@xmath25 ) for different separation ranges are given in table 7 . the limits on close ( @xmath31 " ) companions come from the hst and ao images , and the limits on distant companions were derived from the ground - based k - band images . since the only binary that we found , cfht - pl-18 , did not pass the lithium test , we have found zero resolved binaries in our sample . however , this null result does not necessarily imply that the binary frequency among pleiades vlm members is very low . basri & martn ( 1999 ) found that ppl 15 is a double - lined binary with a period of only 5.8 days . zapatero osorio et al . ( 1997a ) had noticed that this object lies on the binary sequence . five other pleiades vlm members also lie in the binary sequence in the cmd of fig . 2 . we consider them as likely binaries with nearly - identical components . two of them ( lying near hhj 6 ) have not been observed with hst or ao . three of them ( hhj 6 , cfht - pl-12 and 16 ) are not resolved with hst or cfht / ao . steele & jameson ( 1995 ) have suggested that 46% of pleiades vlm stars are multiple on the basis of a spectroscopic study of temperature indicators . however , none of their hhj candidate binaries observed with hst / wfpc were resolved . bouvier , rigaut & nadeau ( 1997 ) carried out an ao search for binaries among 144 g- and k - type pleiades members . they found that the binary frequency and period distribution of their sample is similar to that of dm91 . since the dm91 study has better statistics , we have used their orbital period and mass ratio distributions for predicting the number of binaries expected in our sample . the results are given in the third column of table 7 . we have converted the observed upper limits on components separations on the plane of the sky ( s ) to semi - major axes ( a ) using a cluster distance of 125 pc and s / a=0.92 ( heacox & gathright 1994 ) . considering our sample of pleiades bds observed with hst / nicmos , we find that there is a factor of @xmath23 excess of binaries with respect to the distribution of dm91 for separations smaller than 10.9 au , which corresponds to orbital periods shorter than @xmath2110 years . however , because of the small size of our sample we find that the difference between the pleiades and the field binaries is significant only at the 1 @xmath12 level . on the other hand , it is more significant that we do not find any binaries with separations larger than @xmath227 au . the lack of pleiades vlm systems with periods longer than 444 years ( 3.4 expected , 0 found ) is a 2 @xmath12 effect for poisson statistics . reid & gizis ( 1997 ) obtained hst observations of 53 hyades m - type members ( primary mass @xmath00.3 m@xmath1 ) . they found nine binaries with [email protected] in the separation range 14 to 825 au , consistent with observations of m - dwarfs in the solar neighborhood . if the binary properties of our pleiades vlm sample was similar to that of the hyades sample of reid & gizis , we should have found 2 binaries with separations larger than 30 au . when they grow old , the pleiades vlm stars and bds will become cooler and will look similar to l - type field dwarfs . the first three field binaries with l - dwarf components have been recently discovered ( martn , brandner & basri 1999 ; koerner et al . all of them have separations less than 10 au , and would not have been resolved in our survey if they were at the pleiades distance . there is no bias against finding wider binaries because the fov in these searches is @xmath26100 au . in fact , no wide doubles have been reported in the infrared images of @xmath2 100 l dwarfs and @xmath215 methane dwarfs identified so far in the denis , sloan and 2mass surveys . however , there has not been a survey for spectroscopic binaries among ultracool dwarfs . there could be many short - period vlm binaries that have not been identified yet . the null result of our survey for pleiades bd binaries is consistent with the properties of the field bd binaries found so far . the median orbital separation of the main - sequence stars studied by dm91 is 23 au ( heacox 1998 ) . two pleiades bd binaries ( ppl 15 and cfht pl 16 ) and three field bd binaries have been identified with separations @xmath323 au , but no bd binaries are known to have separations @xmath2623 au . the systematically small separations of the bd binaries with respect to the dm91 sample , suggest that there is a difference in population characteristics . the small numbers involved preclude any definitive conclusion in this regard , but it is worth of mention and brief discussion . the mass of the primary could be an important factor in the the process of binary formation . basri & martn ( 1999 ) suggested that the formation of substellar binaries is biased toward smaller separations than the formation of stellar binaries because the probability of finding a short - period binary like ppl 15 from a normal ( dm91-type ) binary distribution is only @xmath210% . another interesting proposition is that of a `` brown dwarf desert '' , i.e. a scarcity of bd secondaries to g and k dwarfs within 2 au of the primary ( marcy & butler 1998 ) . it seems that bds , as objects representative of the tail of the stellar mass function , could provide important hints about binary formation . we briefly consider three possible scenarios that could explain a difference in the characteristics of bd binaries with respect to the dm91 sample : ( 1 ) disruption of wide binaries due to dynamical interactions with the stellar members of the cluster . this process has been invoked by kroupa , petr & mccaughrean ( 1999 ) to explain the binary frequency of stars in the trapezium cluster . an important test of this scenario is the binary frequency of hard binaries , which should be independent of primary mass . ( 2 ) if vlm stars and bds usually come from the dynamical ejection of vlm fragments from protostellar aggregates ( laughlin & lin , private communication ) , the ones that are more likely to survive are close binaries with nearly - identical component . such an scenario has two constraints to overcome . the typical ejection velocity of the ejected vlm star or bd can not be much larger than the escape velocity from the pleiades cluster . the mechanism has to be extremely efficient because bds are quite numerous . ( 3 ) numerical simulations of molecular cloud fragmentation have shown that slower rotating cloud cores form closer binaries than when the clouds are rapidly rotating ( boss 1993 ) . small cloud cores that start collapse from high initial densities form even closer binaries . this scenario seems provide a natural way of explaining the absence of wide bd binaries and an excess of close systems . furthermore , the bd desert around solar - type stars can also be explained by fragmentation processes ( bate 2000 ) . we have presented nir photometry and low - resolution spectroscopy for all the bd candidates reported by b98 . we have also applied the lithium test to cfht - pl-16 and 18 . the membership status of all the cfht bd candidates is summarized in table 6 . we restrict ourselves to the cfht sample for addressing the issue of the substellar mass function , because the level of follow - up observations of the bd candidates coming from other surveys is still incomplete . the level of field star contamination among pleiades bd candidates was assessed by b98 using two statistical approaches . they compared the mass of pleiades members and field stars in the same volume , and they used a field luminosity function . both methods were in good agreement , and suggested a contamination of 25% . we confirm the membership of 12 pleiades cfht bd candidates out of the 18 included in b98 . the 6 non - members are all foreground cool m dwarfs . we find a contamination of 33% , which is somewhat higher than the estimate of b98 . these authors obtained a power - law index in a log - log plot ( dn / dm@xmath2m@xmath27 ) of @xmath13=-0.6 for the cluster imf across the substellar boundary . we revise this slope downwards to a value of @xmath13=-0.53 , implying that the number of bds per decreasing mass bin is moderately rising , but their relative contribution to the total mass of the pleiades cluster is diminishing . we can not make a detailed correction of the imf for the binary fraction of the cfht bds because we have not resolved any system . nevertheless , we believe that the binary corrections are probably not very important because there does not seem to be many bd companions to stars , and because our search for binaries suggest that bds do not have a higher multiplicity frequency than m - type stars . _ acknowledgments _ : this research makes use of data collected at the hubble space telescope , the canada - france - hawaii telescope , the lick nickel telescope and the mt hopkins telescope . some 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 the national aeronautics and space administration . the observatory was made possible by the generous financial support of the w.m . keck foundation . we are grateful to the hst staff ( particularly , beth perriello and al schultz ) for their assistance . we thank france allard , isabelle baraffe , gilles chabrier and peter hauschildt for making their evolutionary models available to us . em thanks alan boss , greg laughlin and doug lin for discussions about the formation of brown dwarf binaries . gb acknowledges the support of nsf through grant ast96 - 18439 . fa was supported by nasa through grants 110 - 96ltsa and nag5 - 3435 . funding for this publication was provided by nasa through proposal go-7899 submitted to the space telescope science institute , which is operated by the association of universities for research in astronomy , inc . , under nasa contract nas5 - 26555 . abt , h. a. & levitt , s. g. 1976 , apjs , 30 , 273 f allard , f. , hauschildt , p. h. , alexander , d. r. & starrfield , s. 1997 , araa , 35 , 137 allard , f. , alexander , d. r. & hauschildt , p. h. 1998 , asp conf . 154 , the tenth cambridge workshop on cool stars , stellar systems and the sun , edited by r. a. donahue and j. a. bookbinder , p.63 bailer - jones , c. a. l. & mundt , r. 1999 , a&a , 348 , 800 baraffe , i. , chabrier , g. , allard , f. , & hauschildt , p.h . 1998 , a&a , 337 , 403 basri , g. , marcy , g. , & graham , j. r. 1996 , , 458 , 600 basri , g. , & martn , e.l . 1999 , aj , 1727 , 2460 bate , m. 2000 , mnras , in press boss , a. 1993 , in `` the realm of interacting binary stars '' , ed . j. sahade , kluwer academic publishers , p. 355 bouvier , j. , rigaut , f. & nadeau , d. 1997 , a&a , 323 , 139 bouvier , j. , stauffer , j. r. , martn , e.l . , barrado y navascus , d. , wallace , b. & bjar , v.j.s . 1998 , a&a , 336 , 490 ( b98 ) chabrier , g. , allard , f. , baraffe , i. & hauschildt , p. 2000 , a&a , submitted cossburn , m. r. , hodgkin , s. t. , jameson , r. f. & pinfield , d. j. 1997 , mnras , 288 , l23 doyon , r. et al . 1998 , in : fowler a.m. ( ed . ) , spie proc . 3354 , infrared astronomical instrumentation duquennoy , a. & mayor , m. 1991 , a&a , 248 , 485 festin , l. 1998 , a&a , 333 , 497 fisher , d. a. & marcy , g. w. 1992 , apj , 396 , 178 reid , i. n. & gizis , j. e. 1997 , aj , 114 , 1992 hambly , n. c. 1998 , in `` brown dwarfs and extrasolar planets '' , eds r. rebolo , e. l. martn & m. r. zapatero osorio , asp conf . series , 134 , 11 hambly , n. c. , hawkins , m. r. s. , & jameson , r. f. 1993 , a&as , 100 , 607 hambly , n. c. , hodgkin , s. t. , cossburn , m. r. & jameson , r. f. 1999 , mnras , 303 , 835 hauschildt , p. h. , allard , f. & baron , e. 1999 , apj , 512 , 377 heacox , w. d. & gathright , j. 1994 , aj , 108 , 1101 heacox , w. d. 1998 , aj , 115 , 325 henry , t. j. & mccarthy , d. w. 1990 , apj , 350 , 334 jones , h. r. a. & tsuji , t. 1997 , apj , 480 , l39 koerner , d. w. , kirkpatrick , j. d. , mcelwain , m. w. & bonaventura , n. r. 1999 , apj , 526 , l25 kroupa , p. , petr , m. g. & mccaughrean , m. j. 1999 , newa , 4 , 495 leggett , s. k. , allard , f. & hauschildt , p. h. 1998 , apj , 509 , 836 magazz , a. , martn , e. l. & rebolo , r. 1993 , apj , 404 , l17 marcy , g. & butler , r. p. 1998 , ara&a , 36 , 57 martn , e. l. , rebolo , r. , & zapatero osorio , m. r. 1996 , , 469 , 706 martn , e. l. , basri , g. , brandner , w. , bouvier , j. , zapatero osorio , m. r. , rebolo , r. , stauffer , j. , allard , f. , baraffe , i. & hodgkin , s. 1998a , , 509 , l113 martn , e. l. , basri , g. , gallegos , j. e. , rebolo , r. , zapatero osorio , m. r. & bjar , v. j. s. 1998b , , 499 , l61 martn , e. l. , basri , g. , zapatero osorio , m. r. , rebolo , r. , & garca lpez , r. j. 1998c , , 507 , l41 martn , e. l. & zapatero osorio , m. r. 1997 , mnras , 286 , l17 martn , e. l. , brandner , w. & basri , g. 1999 , science , 283 , 1718 marley , m. s. , gelino , c. , stephens , d. , lunine , j. i. & friedman , r. 1999 , apj , 513 , 879 oke , j. b. , cohen , j. g. , carr , m. et al . 1995 , pasp , 107 , 375 rebolo , r. , martn , e. l. , basri , g. , marcy , g. w. , & zapatero osorio , m. r. 1996 , apj , 469 , l53 rigaut , f. et al . 1998 , pasp 110 , 152 stauffer , j. r. , hamilton , d. , & probst , r. 1994 , aj , 108 , 155 stauffer , j. r. , hamilton , d. , probst , r. , rieke , g. & mateo , m. 1989 , apj , 344 , l21 stauffer , j. r. , schultz , g. & kirkpatrick , j.d . 1998a , apj , 499 , l199 stauffer , j. r. , schild , r. , barrado y navascus , d. , backman , d. e. , angelova , a. m. , kirkpatrick , j. d. , hambly , n. & vanzi , l. 1998b , apj , 504 , 805 steele , i. a. & jameson , r. f. 1995 , mnras , 272 , 630 terndrup , d. m. , krishnamurthi , a. , pinsonneault , m. h. & stauffer , j. r. aj , in press thompson , r.i . , rieke , m.j . , schneider , g. , hines , d. & corbin , m.r . 1998 , apj , 492 , l95 zapatero osorio , m. r. , martn , e. l. , & rebolo , r. 1997a , a&a , 323 , 105 zapatero osorio , m. r. , rebolo , r. , & martn , e. l. 1997b , a&a , 317 , 164 zapatero osorio , m. r. , rebolo , r. , martn , e. l. , basri , g. , magazz , a. , hodgkin , s. t. , jameson , r.f . & cossburn , m. r. 1997c , , 491 , l81 zapatero osorio , m. r. , rebolo , r. , martn , e. l. , hodgkin , s. t. , cossburn , m. r. , magazz , a. , steele , i. a. & jameson , r.f . 1999 , a&as , 134 , 537 lcccc calar 3 & cfht 21 & [email protected] & [email protected] & [email protected] + cfht - pl 15 & & [email protected] & [email protected] & [email protected] + cfht - pl 16 & & [email protected] & [email protected] & [email protected] + cfht - pl 17 & & [email protected] & [email protected] & [email protected] + cfht - pl 18 a & & [email protected] & [email protected] & [email protected] + cfht - pl 18 b & & [email protected] & [email protected] & [email protected] + cfht - pl 19 & & [email protected] & [email protected] & [email protected] + cfht - pl 20 & & [email protected] & [email protected] & [email protected] + cfht - pl 22 & & [email protected] & [email protected] & [email protected] + cfht - pl 23 & & [email protected] & [email protected] & [email protected] + cfht - pl 25 & & [email protected] & [email protected] & [email protected] + hhj 2 & & [email protected] & [email protected] & [email protected] + hhj 8 & & [email protected] & [email protected] & [email protected] + mho 1 & & [email protected] & [email protected] & [email protected] + mho 3 & & [email protected] & [email protected] & [email protected] + mho 6 & & [email protected] & [email protected] & [email protected] + piz 1 & & [email protected] & [email protected] & [email protected] + ppl 1 & roque 15 & [email protected] & [email protected] & [email protected] + & & [email protected] & [email protected] & [email protected] + ppl 14 & & [email protected] & [email protected] & [email protected] + roque 4 & & [email protected] & [email protected] & [email protected] + roque 7 & cfht - pl 24 & [email protected] & [email protected] & [email protected] + roque 11 & & [email protected] & [email protected] & [email protected] + roque 12 & & [email protected] & [email protected] & [email protected] + roque 13 & & [email protected] & [email protected] & [email protected] + roque 14 & & [email protected] & [email protected] & [email protected] + roque 16 & cfht - pl 11 & [email protected] & [email protected] & [email protected] + roque 17 & & [email protected] & [email protected] & [email protected] + roque 25 & & [email protected] & [email protected] & [email protected] + teide 1 & & [email protected] & [email protected] & [email protected] + teide 2 & cfht - pl 13 & [email protected] & [email protected] & [email protected] + lcccc calar 3 & [email protected] & [email protected] & [email protected] & cfht + cfht - pl 1 & [email protected] & & [email protected] & mho + cfht - pl 2 & [email protected] & & [email protected] & mho + cfht - pl 3 & [email protected] & & [email protected] & mho + cfht - pl 4 & [email protected] & & [email protected] & mho + cfht - pl 5 & [email protected] & & [email protected] & mho + cfht - pl 6 & [email protected] & & [email protected] & mho + cfht - pl 7 & [email protected] & & [email protected] & mho + cfht - pl 8 & [email protected] & & [email protected] & mho + cfht - pl 9 & [email protected] & & & lick + & [email protected] & & [email protected] & mho + cfht - pl 10 & [email protected] & & & lick + & [email protected] & & [email protected] & mho + cfht - pl 12 & [email protected] & [email protected] & [email protected] & cfht + & [email protected] & & & lick + & [email protected] & & [email protected] & mho + cfht - pl 15 & [email protected] & [email protected] & [email protected] & cfht + & [email protected] & & [email protected] & mho + cfht - pl 16 & [email protected] & [email protected] & [email protected] & cfht + & [email protected] & & & lick + cfht - pl 17 & [email protected] & [email protected] & [email protected] & cfht + & [email protected] & & & lick + cfht - pl 18 & [email protected] & [email protected] & [email protected] & cfht + cfht - pl 19 & [email protected] & [email protected] & [email protected] & cfht + & [email protected] & & & lick + cfht - pl 20 & [email protected] & [email protected] & [email protected] & cfht + & [email protected] & & & lick + cfht - pl 22 & [email protected] & [email protected] & [email protected] & cfht + cfht - pl 23 & [email protected] & [email protected] & [email protected] & cfht + cfht - pl 25 & [email protected] & [email protected] & [email protected] & cfht + roque 16 & [email protected] & [email protected] & [email protected] & cfht + & [email protected] & & [email protected] & mho + lcccccc cfht - pl 1 & 1998 dec 22 & kpno & 730 & 4.20 & 533.9959.7 & 900 + cfht - pl 2 & 1998 dec 22 & kpno & 730 & 4.20 & 533.9959.7 & 900 + cfht - pl 5 & 1998 dec 23 & kpno & 730 & 4.20 & 533.9959.7 & 900 + cfht - pl 6 & 1998 dec 23 & kpno & 730 & 4.20 & 533.9959.7 & 900 + cfht - pl 7 & 1998 dec 22 & kpno & 730 & 4.20 & 533.9959.7 & 1800 + cfht - pl 8 & 1998 dec 22 & kpno & 730 & 4.20 & 533.9959.7 & 1200 + cfht - pl 16 & 1998 dec 22 & kpno & 730 & 4.20 & 533.9959.7 & 1800 + cfht - pl 16 & 2000 jan 5 & keck ii & 900 & 0.85 & 631.2802.9 & 3600 + cfht - pl 17 & 1998 dec 22 & kpno & 730 & 4.20 & 533.9959.7 & 1800 + cfht - pl 18 & 1998 dec 21 & keck ii & 900 & 0.85 & 631.2802.9 & 5400 + cfht - pl 19 & 1998 dec 22 & kpno & 730 & 4.20 & 533.9959.7 & 1800 + cfht - pl 20 & 1998 dec 23 & kpno & 730 & 4.20 & 533.9959.7 & 1800 + cfht - pl 25 & 1998 dec 23 & keck ii & 150 & 4.82 & 355.51125.1 & 1200 + cfht - pl 26 & 1998 dec 23 & keck ii & 150 & 4.82 & 355.51125.1 & 1800 + roque 7 & 1997 dec 28 & wht & 158 & 2.91 & 629.1926.5 & 2000 + roque 25 & 1998 dec 23 & keck ii & 150 & 4.82 & 355.51125.1 & 1800 + roque 33 & 1998 dec 23 & keck ii & 150 & 4.82 & 355.51125.1 & 2000 + teide 1 & 1998 dec 23 & keck ii & 150 & 4.82 & 355.51125.1 & 1200 + lccccc & & & & & ( ) + cfht - pl 1 & 1.26 & dm4.9 & 2.56 & 2.41 & [email protected] + cfht - pl 2 & 1.26 & dm4.9 & 3.09 & 2.43 & [email protected] + cfht - pl 5 & 1.36 & dm5.5 & 3.26 & 2.52 & [email protected] + cfht - pl 6 & 1.61 & dm6.9 & 3.52 & 2.70 & @xmath26 - 2 + cfht - pl 7 & 1.37 & dm5.6 & 3.08 & 2.43 & @xmath26 - 2 + cfht - pl 8 & 1.37 & dm5.6 & 3.09 & 2.51 & [email protected] + cfht - pl 16 & 2.20 & dm9.3 & 4.41 & 2.74 & @xmath26 - 15 + cfht - pl 16 & & & & & [email protected] + cfht - pl 17 & 1.82 & dm7.9 & 4.48 & 2.86 & -7 : + cfht - pl 18 & & & & & [email protected] + cfht - pl 19 & 1.68 & dm7.5 & 4.07 & 2.50 & @xmath26 - 10 + cfht - pl 20 & 1.71 & dm7.5 & 3.55 & 2.51 & @xmath26 - 3 + cfht - pl 25 & 2.10 & dm9.0 & 4.75 & 3.06 & @xmath26 - 6 + cfht - pl 26 & 1.65 & dm7.1 & 3.76 & 3.20 & @xmath26 - 5 + roque 7 & 1.90 & dm8.3 & 4.60 & 2.89 & @xmath26 - 15 + roque 25 & 2.58 & dl0.1 & 2.61 & 2.51 & -9@xmath112 + roque 33 & 2.44 & dm9.8 & 6.35 & 3.20 & -55@xmath1110 + teide 1 & 1.81 & dm7.9 & 4.22 & 2.80 & -2 : + lcccc & ( 0``.1 ) & ( 0''.1 ) & ( 0``.4 ) & ( 0''.4 ) + calar 3 & & & 1.0 & 1.0 + cfht - pl 15 & 2.6 & 3.4 & 5.2 & 4.5 + cfht - pl 16 & 4.3 & 4.5 & 4.8 & 5.0 + cfht - pl 17 & 4.0 & 4.6 & 4.9 & 5.6 + cfht - pl 19 & 2.1 & 3.7 & 4.6 & 4.4 + cfht - pl 22 & 3.5 & 4.0 & 4.2 & 4.2 + cfht - pl 23 & 3.9 & 4.1 & 4.7 & 5.2 + cfht - pl 25 & 4.2 & 5.1 & 5.2 & 5.5 + hhj 2 & 3.3 & 4.5 & 5.6 & 5.5 + hhj 8 & 3.0 & 3.0 & 4.5 & 4.5 + mho 1 & 3.7 & 4.7 & 5.0 & 5.3 + mho 3 & 2.9 & 3.8 & 5.1 & 5.5 + mho 6 & 3.0 & 3.0 & 4.5 & 4.5 + piz 1 & 3.1 & 4.1 & 3.5 & 4.4 + ppl 1 & 4.6 & 4.8 & 5.8 & 5.8 + ppl 14 & 4.6 & 4.7 & 5.2 & 5.4 + roque 4 & 4.2 & 4.4 & 4.7 & 5.2 + roque 7 & 3.9 & 4.1 & 4.7 & 5.5 + roque 11 & 3.0 & 3.0 & 4.5 & 4.5 + roque 12 & 3.0 & 3.0 & 4.5 & 4.5 + roque 13 & 4.2 & 4.6 & 5.4 & 5.5 + roque 14 & 4.1 & 4.5 & 5.3 & 5.7 + roque 16 & 4.6 & 4.4 & 5.7 & 5.6 + roque 17 & 3.9 & 4.2 & 5.5 & 5.2 + roque 25 & 3.0 & 3.8 & 3.3 & 3.9 + teide 1 & 4.1 & 4.1 & 5.4 & 4.9 + teide 2 & 4.0 & 4.3 & 4.7 & 5.2 + lccccccccc calar 3 & yes & yes & & yes & yes & yes & yes & & yes + cfht - pl 1 & yes & & & & yes & yes & yes & & yes + cfht - pl 2 & yes & & & & yes & yes & yes & & yes + cfht - pl 3 & yes & & yes & & & & yes & & yes + cfht - pl 4 & & & & & & & yes & & yes ? + cfht - pl 5 & yes & & & & yes & yes & yes & & yes + cfht - pl 6 & yes & & & & no & yes & yes & & yes ? + cfht - pl 7 & yes & & & & no & yes & yes & & yes ? + cfht - pl 8 & yes & & & & yes & yes & yes & & yes + cfht - pl 9 & yes & & & no & yes & & yes & & yes + cfht - pl 10 & yes & & & no & yes & & yes & & yes + cfht - pl 12 & yes & & & yes & yes & & yes & yes & yes + cfht - pl 14 & yes & & no & no & no & & & & no + cfht - pl 15 & yes & & & yes & yes & & yes & yes & yes + cfht - pl 16 & yes & yes & & yes & yes & yes & yes & yes & yes + cfht - pl 17 & yes & & & & yes & yes & yes & yes & yes + cfht - pl 18 & yes & & & no & yes & yes & yes & no & no + cfht - pl 19 & yes & & & & & yes & no & yes & no ? + cfht - pl 20 & yes & & & & no & yes & no & no & no + cfht - pl 22 & yes & & & & & yes & no & no & no + cfht - pl 23 & yes & & & & & & yes & yes & yes + cfht - pl 25 & yes & & & & no & yes & yes & yes & yes + cfht - pl 26 & yes & & & & no & no & & & no + hhj 2 & yes & & yes & & yes & yes & yes & yes & yes + hhj 6 & yes & yes & yes & & yes & yes & yes & yes & yes + hhj 8 & yes & & yes & & yes & yes & yes & yes & yes + mho 1 & & & & & & & & no & no + mho 3 & yes & yes & & yes & yes & & & yes & yes + mho 6 & & & & & & & & yes & yes ? + piz 1 & yes & & & & & yes & yes & yes & yes + ppl 1 & yes & yes & & yes & yes & yes & yes & yes & yes + ppl 14 & yes & & & & yes & yes & & no & no ? + ppl 15 & yes & yes & yes & yes & yes & yes & yes & & yes + roque 4 & yes & yes & & & yes & yes & yes & yes & yes + roque 7 & yes & & & & & yes & & yes & yes + roque 11 & yes & yes & & & yes & yes & yes & yes & yes + roque 12 & yes & & & & yes & yes & yes & yes & yes + roque 13 & yes & & & yes & yes & & yes & yes & yes + roque 14 & yes & & & yes & yes & yes & & yes & yes + roque 16 & yes & & & yes & yes & & yes & yes & yes + roque 17 & yes & & & yes & yes & yes & & yes & yes + roque 25 & yes & & & & yes & yes & yes & yes & yes + roque 33 & yes & & & yes & yes & yes & & & yes + teide 1 & yes & yes & yes & yes & yes & yes & yes & yes & yes + teide 2 & yes & yes & & yes & yes & yes & yes & yes & yes + lcccccc 24 & 2 & 0.6 & @xmath30.08 & @xmath310.9 & 0.8 & @xmath34.60 + 34 & 4 & 3.0 & @xmath30.2 & @xmath327.2 & 0.6 & @xmath35.21 + 34 & 0 & 1.8 & 0.2 - 1.0 & 27.2 - 135.9 & 0.4 & 5.21 - 6.26 + 44 & 0 & 1.6 & 1.0 - 8.0 & 135.9 - 1087.0 & 0.5 & 6.26 - 7.62 +

contrary to a broadly accepted assumption we show that random ( ga , in)p is not an exception with respect to the crude classification of the phonon mode behavior of random mixed crystals in terms of 1-bond-mode systems or 2-bond-mode systems , as established from the simple criterion derived by elliott _ et al . _ [ r.j . elliott _ et al . _ , rev . mod . phys . * 46 * , 465 ( 1974 ) ] . consistent understanding of the puzzling raman / infrared behavior of ( ga , in)p , that has been a subject of controversy , is achieved via a basic version of our 1-bond-mode model originally developed for ( zn , be)-chalcogenides , that exhibit a large contrast in the bond properties , and recently extended under a simplified form to the usual ( ga , in)as alloy . the raman / infrared features from ( ga , in)p are accordingly re - assigned , with considerable change with respect to the previous approaches . in particular the in impurity mode , previously assigned within ( ) the optical band of the host gap compound ( 368403 ) , is re - assigned below it ( ) . accordingly the ga p and in p transverse optical branches do not overlap , which reconciles ( ga , in)p with the elliott s criterion . besides , we show that the idea of two bond lengths per species in alloys , supported by our 1-bond-phonon picture , opens an attractive area for the discussion of spontaneous ordering in gainp , and in mixed crystals in general . essentially this is because it allows to play with the related competition effects regarding the minimization of the local strain energy due to the bond length mismatch between the parent compounds . in particular the unsuspected issue of intrinsic limit to spontaneous ordering comes out (.5 in gainp ) . the whole discussion is supported by detailed re - examination of the ( ga , in)p raman / infrared data in the literature , full contour modeling of the transverse and longitudinal optical raman lineshapes via our phenomenological 1-bond-mode model , and first - principles bond length calculations concerned with the minority bond species close to the impurity limits ( ga 0 , 1 ) and to the ga p ( ga.19 ) and in p ( ga.81 ) bond percolation thresholds . in the latter case we discriminate between connected and isolated bonds , not in the usual terms of next - nearest neighbors .

cond-mat0602442

the long wave vibrational properties of ( a , b)c semiconductor mixed crystals , where c denotes indifferently the anionic or the cationic species , are well - documented both experimentally and theoretically . this has lead to a crude classification of the phonon mode behavior of _ random _ mixed crystals in the raman or infrared ( ir ) spectra in two categories.@xmath5 most random mixed crystals exhibit the so - called 1-bond@xmath01-mode behavior . this corresponds to well - separated a c and b c transverse - longitudinal optical ( to lo ) bands over the whole composition range . these degenerate into distinct ac : b ( b@xmath10 ) and bc : a ( a@xmath10 ) impurity modes , and have strengths that scale as the corresponding fractions of bonds in the crystal . the remaining random alloys exhibit the so - called 2-bond@xmath01-mode behavior , also referred to as the mixed - mode behavior . this corresponds to a single to lo band with ( a c , b c)-mixed character . the frequency range that the band covers shifts continuously from one end member to the other when the alloy composition changes , and the strength remains approximately constant throughout the whole composition range . in mixed - mode mixed crystals the impurity modes are expected to fall within the optical bands of the host lattices , and thereby a common believe is that they should not be observed as distinct and separate modes.@xmath6 a rather crude criterion is that for 1-bond@xmath01-mode behavior the to - lo bands of the parent materials must not overlap . as a matter of fact the condition is too strong , but if overlap is large , the 2-bond@xmath01-mode behavior is always found . _ et al.__@xmath7 derived a more accurate criterion , even though simple , based on the coherent potential approximation . this is now the standard to decide about the 1-bond@xmath01-mode or mixed - mode behavior of the random mixed crystals . basically for 1-bond@xmath01-mode behavior the relative change in the reduced mass of the bond induced by the impurity must be greater than the difference between the limit dielectric constants of the host lattice , normalized to the sum . it is worth to notice that this criterion neglects disorder in the force constants , only mass disorder is taken into account . there would be only one fascinating exception that does not fit in the above classification of phonon mode behavior , i.e. ( ga , in)p.@xmath8 according to the elliott s criterion , random ( ga , in)p should exhibit a typical 1-bond@xmath01-mode behavior in the raman / ir spectra . _ emphasize that the larger the difference between the average mass of the substituting species on one side , and the mass on the unperturbed site on the other side , the more reliable their criterion . precisely this is true for ( ga , in)p as both in ( 4@xmath9 row in the periodic table ) and ga ( 3@xmath10 row ) are much heavier than p ( 2@xmath11 row ) . as a matter of fact , the to lo bands from pure inp ( 303345 @xmath2 ) and pure gap ( 368403 @xmath2 ) do not overlap . however , the 1-bond@xmath01-mode behavior does not show up in the raman / ir spectra . in the following we propose a brief survey of the extended vibrational information available for random ( ga , in)p in the literature , and of the different models , for clear insight upon the puzzling phonon behavior of this alloy . we proceed chronologically , for more clarity . lucovsky _ _ et al.__@xmath12 have realized the first vibrational study of ( ga , in)p alloys , by using a combination of far - ir reflectance and raman scattering on polycrystalline samples . they observed a main to lo band that behaved as expected in case of a mixed - mode behavior ( refer above ) . in addition a minor to - lo band was evidenced within the main to - lo band . surprisingly the minor lo mode shows up at lower frequency than the to counterpart , which discouraged an assignment of the minor to - lo band in terms of a zone - center signal . this was rather attributed to zone - edge modes . the authors concluded that ( ga , in)p was the first example of a mixed - mode behavior among * iii**v * s . the impurity modes were tentatively located at @xmath1330 @xmath2 ( gap : in ) and @xmath1390 @xmath2 ( inp : ga ) , i.e. within the optical bands of the host lattices . beserman _ _ et al.__@xmath13 have reported an exhaustive raman study of a large amount of polycrystalline ( ga , in)p alloys made of small pieces of single crystals . the whole composition range could be analyzed . both the to and lo modes were allowed in their scattering geometries . first , beserman _ _ showed that the minor to lo band persists all the way from the intermediate composition range down to both the ga- and in - dilute limits . in addition the authors could perform proper lo symmetry analysis by using a convenient piece of single crystal with small in content ( @xmath10.04 ) . somewhat surprisingly the minor lo mode , localized at @xmath1390 @xmath2 , was found to be highly polarized , just as the main zone - center lo mode . the apparent zone - center 2-modes lo behavior at small in content was attributed to a splitting of the nominal zone - center lo mode due to coupling with two - phonon combinations from the zone - edge . besides , beserman _ et al_. derived decisive information in the to symmetry , which we detail below [ refer to point ( i ) ] . jahne _ _ et al.__@xmath14 discussed the minor to - lo band as an individual zone - center response , which made it difficult to cover the mixed - mode behavior . for modeling of their ir spectra obtained with polycrystalline ( ga , in)p ingots , jahne _ et al . _ used a version of the cluster model originally developed by verleur and barker to account for the complex raman / ir spectra of ga(as , p)@xmath15 and cd(s , se)@xmath16 , that were attributed to local segregation effects . in this model the entire mixed crystal is built up from five basic units corresponding to the possible first - neighbor arrangements around the unperturbed site . basically the minor to lo band would have its origin in the change of the ga p force - constant from one type of basic unit to another . however , a model with potentially four oscillators per bond species in the alloy seems oversized to account for a single additional mode in the raman / ir spectra . at this stage further discussion of the phonon mode behavior of ( ga , in)p clearly required deeper insight upon the impurity modes . careful raman investigation of ( ga , in)p samples in the ga- and in - dilute limits was achieved by jusserand and slempkes@xmath17 ( js ) and kato _ et al_.,@xmath18 respectively . for this purpose a new generation of samples with well - defined symmetry was used , i.e. ( ga , in)p samples grown as epitaxial layers or single crystals . js did perform raman measurements with mostly in - rich ( ga , in)p epitaxial layers . they have used a standard backscattering geometry along the [ 001]-growth axis . with this geometry only the lo modes are allowed , the to modes are forbidden . js did observe the progressive emergence of the gap - like lo line on the high - frequency side of the dominant inp - like lo line from very low ga incorporation ( @xmath10.1% ) . the inp : ga impurity mode was accordingly identified as a distinct mode at @xmath12 @xmath2 above the inp lo mode , contrary to first expectations . interestingly , js noted that the in p and ga p lines exhibit similar strengths for as small ga incorporation as 1.5% . this was attributed to some coupling between the two lo modes , due to their vicinity . however , the discussion remained qualitative only , and the physical mechanism behind was not identified . with the inp : ga impurity mode being out of the inp optical band , as for a typical 1-bond@xmath01-mode system , and the gap : in impurity mode being within the gap optical band , as for a mixed - mode system , a strong overlapping of the ga - p and in - p bands is expected in the alloy , resulting in a complicated phonon behavior . js@xmath17 proposed that ( ga , in)p is an exceptional alloy in the classification of phonon mode behavior . it would obey the so - called modified 2-mode behavior , with a dominant to mode at low - frequency that joins the inp ( 303 @xmath2 ) and gap ( 368 @xmath2 ) parent to modes , and a minor to mode at high - frequency that connects the inp : ga ( 347 @xmath2 ) and gap : in ( 390 @xmath2 ) impurity modes . kato _ _ et al.__@xmath18 performed a thorough raman study of ( ga , in)p single crystals and epitaxial layers covering the whole composition range , with special emphasis upon the ga - rich side . they could confirm the existence of a quasi - degenerate to - lo mode at @xmath1390 @xmath2 in the in - dilute limit , which at first sight supported the original assignment in terms of the gap : in impurity mode , and thereby the phonon picture proposed by js . interestingly , we note from their exhaustive data that the to mode at high - frequency , earlier referred to as the minor to mode , does not seem so ` minor ' . in particular close to the stoichiometry ( [email protected] ) the low- and high - frequency to modes have similar strengths ( refer to fig . 5 in ref . this has attracted little attention so far . in the past decade most of the attention was focused on ga@xmath19in@xmath20p , abbreviated gainp@xmath3 for simplicity , which has become the leader material for the study of spontaneous ordering in semiconductor mixed crystals.@xmath8 precisely to close this brief overview we mention the reference first - principles calculations of the transverse phonon density of states ( dos ) in disordered gainp@xmath21 , recently performed by ozolin and zunger ( oz).@xmath22 a theoretical dielectric function @xmath23 was derived for pre - insight upon the to raman lineshapes , via @xmath24 . what emerged is that the dominant to mode around 330 @xmath2 involves both ga p vibrations , as expected . our view is that such behavior is consistent with the mixed - mode description as well as with the modified 2-mode alternative , as proposed by lucovsky _ et al_. and js , respectively . also , oz derived decisive insight upon the minor to mode , which is summarized below [ refer to point ( ii ) ] . following js the consensus so far is that random ( ga , in)p exhibits the modified 2-mode behavior in the raman / ir spectra , which contradicts the elliot s criterion . in addition the modified 2-mode picture fails to explain two key to features that have attracted little attention so far : \(i ) the dominant to mode exhibits a marked antagonist asymmetry on each side of the stoichiometry ( ga @xmath10.5 ) , as detected by beserman _ et al . _ by using raman scattering ( see fig . 2 in ref . this was attributed to a fano interference with the disorder - induced combination of transverse ( ta ) and longitudinal ( la ) acoustical modes at the @xmath25 zone - edge , that was assumed to be at higher frequency than the zone - center to mode in pure inp , as opposed to gap . however , later measurements of the phonon dispersion in inp invalidate this mechanism . indeed the ta(@xmath25)+la(@xmath25 ) acoustical combination was found below the zone - center to mode in inp ( ref . 14 ) , as in gap ( ref . 15 ) . \(ii ) the minor to mode in disordered gainp@xmath3 has a gap - like character , as shown by oz in their first - principles calculations ( see fig . 3-a in ref . this , in particular , is challenging for the modified 2-mode picture , as the latter implies a ( ga p , in p)-mixed character for both the dominant and the minor to modes , at any alloy composition . in this work we investigate whether our 1-bond@xmath02-mode picture , earlier referred to as the ` percolation ' picture , may provide consistent understanding of the puzzling phonon behavior of ( ga , in)p , which is still lacking . this picture was originally developed for the long - wave phonons of ( zn , be)-chalcogenides,@xmath26 that opened the class of mixed crystals with contrast in the bond force constant , and has been extended recently to ( ga , in)as , with much success.@xmath27 the key issue when considering the physical properties of random ( a , b)c mixed crystals is how to handle the problem of alloy disorder . certainly the most convenient way is to describe the system in terms of the virtual crystal approximation ( vca ) . accordingly each atom c from the unperturbed site is ideally surrounded by four virtual nearest neighbors , each of these consisting of a statistical average of the a and b substituting species depending on the alloy composition . this way perfect order is artificially re - built in the crystal , where it does not exist in reality , so that the approaches finalized at the microscopic scale for the basic understanding of the physical properties of the perfectly ordered parent compounds can be directly extended to the alloys . regarding vibrational properties , such an approach leads to the idea that each bond in the alloy should bring a single feature in the raman / ir spectra , as in the corresponding parent compound , but with characteristics ( strength , frequency ) depending on the alloy composition . this corresponds to a typical 1-bond@xmath01-mode behavior in the raman / ir spectra , as accounted for by the well - admitted modified - random - element - isodisplacement ( mrei ) model developed by chang and mitra,@xmath5 based on a vca description of the mixed crystals . provided some adjustment is made , this model also accounts for the mixed - mode behavior.@xmath6 while the vca seems actually relevant for the integral physical properties of semiconductors , that operate a natural average on alloy disorder , such as the band gap or the lattice constant , our view is that it should not apply to vibrational properties because these address directly the _ bond force constant _ , which is a local physical property . we claim that their basic understanding requires detailed insight upon the local neighborhood of the substituting species , which falls into the scope of the percolation site theory.@xmath28 essentially this is concerned with the statistical properties ( population , internal structure , ) of clusters formed by sites occupied at random on a regular lattice . while there is an obvious analogy between the topology of such systems and the topology of the random mixed crystals , we note that the concept of percolation remains basically outside the schemes used for the very basic understanding of the routine physical properties of mixed crystals . in our 1-bond@xmath02-mode picture for the raman / ir spectra of the random ( a , b)c mixed crystals we describe these as true composite media made of two coexisting a - rich and b - rich regions , resulting from natural fluctuations in the alloy composition at the local scale . this way the alloy disorder is explicitly recognized , while it was totally eclipsed with the mrei - vca description . separate resolvable phonon modes are envisioned for each bond species , corresponding to the different force constants experienced in each of the a - rich and b - rich regions . in particular our phenomenological model envisages singularities in the bond force constant , and thereby in the phonon behavior , at the bond percolation thresholds . we remind that the bond percolation thresholds are the critical compositions corresponding to the first formations of pseudo - infinite chains of the b c and a c bonds in the a@xmath29b@xmath30c crystal . by using computer simulation based on random substitution on the ( a , b ) cfc sub - lattice , these were estimated by stauffer as @xmath310.19 and @xmath320.81 , respectively.@xmath28 by crossing the percolation threshold the host region undergoes a dispersion@xmath33pseudocontinuum topological transition . as a matter of fact a singularity at the bond percolation threshold was clearly observed in the raman spectra of the reference ( zn , be)se ( see fig . 1 in ref . 16 ) and ( zn , be)te@xmath34 systems . the nature of the singularity is briefly discussed in sec . it was tentatively discussed in terms of the different internal structures of the dispersion and the pseudocontinuum , as predicted by the percolation site theory . detail is given in ref . it is worth to notice that a similar singularity at the bond percolation thresholds was also observed by bellaiche _ _ et al.__@xmath35 in their first - principles calculations dedicated to another local physical property of random mixed crystals , i.e. the _ bond length_. there again the singularity was discussed in terms of a percolation behavior . the microscopic mechanism for the 1-bond@xmath02-mode behavior in the raman / ir spectra was previously identified as the difference in bond length due to the different local bond distortions according to whether the bonds belong to the randomly formed a - rich or b - rich regions.@xmath36 on this basis the larger the contrast between the bond lengths , the more clearly the 1-bond@xmath02-mode behavior is expected to show up in the raman / ir spectra . in fact the contrast is large for ( ga , in)p , as detailed below . at this stage , let us emphasize that our 1-bond@xmath02-phonon picture can not be derived from the elliott s criterion because the latter considers mass disorder only , not disorder in the force constant.@xmath7 basically we expect two well - separated phonon branches for the ga - based bond and two tight ones for the in - based bond in ( ga , in)p , as for ( ga , in)as.@xmath27 actually ( ga , in)as can be used as a reference for the study of alloying effects on the phonon properties of ( ga , in)p , because the substituting species are the same , hence a similar contrast in the bond properties and thereby similar local bond distortions in the mixed crystals , with concomitant impact upon the phonon frequencies . while p has a smaller covalent radius ( 1.06 ) than as ( 1.20 ) , so that the p - based bonds are shorter / stiffer than the as - based ones , the contrasts between the bond lengths ( @xmath37 ) and the bond stiffness , best described by the ratio @xmath38 between the bond stretching ( @xmath39 ) and the bond bending ( @xmath40 ) force constants , remain similar in ( ga , in)as ( @xmath416.3% , @xmath4228.1% ) and ( ga , in)p ( @xmath417.1% , @xmath4234.3%),@xmath43 the ga - based bond being shorter / stiffer . besides , we have checked that ( ga , in)as is classified as a 1-bond@xmath01-mode system according to the elliott s criterion , as ( ga , in)p . we discuss the to modes mainly , because these consist of purely mechanical vibrations , i.e. quasi - independent oscillators , and thereby carry reliable strength / frequency information on each oscillator . we have shown in earlier work that proper investigation of the 1-bond@xmath02-mode behavior via the raw lo raman data is basically hopeless due to strong coupling between neighbor individual lo modes via their common long range polarization field @xmath44 .@xmath26 accordingly our lo study here entirely proceeds from the safe to modes . the paper is organized as follows . in sec . ii we remind briefly the basis of our phenomenological 1-bond@xmath02-mode picture for contour modeling of the ( ga , in)p raman lineshapes , and we outline the ab initio calculations that we implement to support our corresponding re - assignment of the raman / ir features . in sec . iii our attention is focused on the phonon behavior of random ( ga , in)p . in sec . iii - a we re - assign the gap : in impurity mode below the gap to lo optical band from careful re - examination of the raman / ir data in the literature , and propose on this basis a simple version of our 1-bond@xmath02-phonon picture that applies to random ( ga , in)p . in sec . b we perform ab initio calculations of the bond length distribution of the minority bond species in large ( ga , in)p supercells corresponding to alloy compositions close to the dilute limits and to the in p ( in @xmath10.19 ) and ga p ( in @xmath10.81 ) bond percolation thresholds . we discriminate between connected and isolated bonds , not in the usual terms of next - nearest neighbors . the ab initio calculations are used to validate our novel assignment of the gap : in impurity phonon mode , and to provide qualitative insight upon the magnitude of the phonon splitting @xmath45 within each of the ga p double - branches . in sec . iii - c we show that our model provides consistent understanding of the whole phonon behavior of random ( ga , in)p , as summarized in sec . i. in sec . iv we tackle the key issue of spontaneous ordering in gainp@xmath3 on this novel basis . we propose a possible mechanism that accounts for the disconcerting evolution of the raman / ir lineshapes with increasing ordering . conclusions are outlined in sec . first , we outline briefly , for more clarity , the technical aspects of our phenomenological 1-bond@xmath02-mode model for long wave phonons in random zincblende a@xmath29b@xmath30c mixed crystals . extensive detail is given elsewhere.@xmath26 we start with the frequency aspect . schematically , our view is that for each bond species the to response over the whole composition range consists of two quasi - parallel branches tied up to the corresponding parent and impurity modes at the two ends of the composition range , and separated by a characteristic finite frequency gap @xmath45 in the dilute limits . somewhat surprisingly , for a given bond species , the bonds are longer ( shorter ) within the host region that refers to the parent material with the smaller ( larger ) lattice constant . this was explained in detail elsewhere , based on first - principles bond length calculations in the two host regions.@xmath36 now , the shorter the bond length , the larger the bond force constant , and thereby the phonon frequency . accordingly within each double - branch the low ( high ) frequency branch refers to the host region with the parent material corresponding to the smaller ( larger ) lattice constant . now we come to the singularity in the bond force constant at the bond percolation thresholds , as mentioned in sec . i. basically , for each individual to branch the model envisages two different regimes on each side of the bond percolation threshold to which the host region refers : one in which the optical mode of the most dilute substitutional species vibrates with a frequency that is basically independent on its concentration ( regime 1 ) , and one in which the frequency of the same mode depends smoothly on the alloy composition according to a traditional but ` rescaled ' modified - random - element - isodisplacement ( mrei ) description ( regime 2 ) . we turn next to the strength aspect . a very general trend is that the oscillator strengths and faust - henry coefficients from the overall a c and b c signals in the raman / ir spectra do scale as the fraction of the related oscillators in the a@xmath29b@xmath30c alloy , i.e. as @xmath46 and @xmath47 , respectively.@xmath5 these are referred to as the global weighting factors . now , simple symmetry considerations guarantee that in case of random substitution the scattering volumes from the a - rich and b - rich regions scale as @xmath46 and @xmath47 , respectively . these are referred to as the individual weighting factors . they fix the sharing of the available oscillator strength and faust - henry coefficient within each double - branch . eventually the multiplication of the global and individual weighting factors determines the relative strength of the four to modes ( 2 ac - like and 2 bc - like ) in the raman / ir spectra . once the two double - branches are properly set , full contour modeling of the multi - phonon raman lineshapes is eventually achieved on the above frequency / strength basis , while using no adjustable parameter , by using the generic equation @xmath48 ^{2 } + \sum\limits_{p}\!c_{p}^{2}\,\frac{k_{p}^{2}\,l'_{p } } { 4\pi z_{p}^{2 } } \right\ } .\ ] ] the to modes are obtained from the imaginary part of the second term , while the full expression provides the lo modes . here , the summation runs over the relevant number of oscillators , that may be smaller than four , depending on the alloy ( see sec . @xmath49 , @xmath50 and @xmath51 are defined on a per - oscillator basis and are , respectively , the faust - henry coefficient of the @xmath52-mode , its to frequency squared , and its related lorentzian responses . @xmath53 is defined according to the standard mrei notations;@xmath5 it relates to the oscillator strength @xmath54 , and thereby to the ( to - lo)@xmath55-splitting . @xmath54 and @xmath49 are normalized to the fraction of @xmath52-oscillators in the crystal with respect to the corresponding parent values . @xmath56 is the relative dielectric function of the mixed crystal in a form generalized to multi - oscillators , as established according to the standard mrei scheme . in our model , once the to phonon double - branches of the ga p bonds are properly set , the raman / ir lineshapes are entirely determined . each double - branch is fixed by two parameters only on top of the frequency of the parent to mode , i.e. , the magnitude of the phonon splitting @xmath45 , and the frequency of the impurity mode . for clear insight upon these two key parameters in ( ga , in)p we determine a representative distribution of bond lengths , that we have done in a first - principles supercell calculation . we applied the calculation method and the computer code siesta,@xmath57 and allowed full relaxation ( of lattice parameters along with internal coordinates ) in a prototype ` percolation - threshold ' supercell similar to that used in ref . 21 , along with ` dilute limit ' supercells . we used the local density approximation throughout . if comparing with previous calculations for the ( ga , in)p system , by oz@xmath22 , we used the same type of norm conserving pseudopotentials ( constructed along the troullier - martins scheme , see ref . 25 ) , as oz , specifically using the following configurations and pseudoization radii ( in brackets , in bohr ) : ga 4@xmath58(2.00 ) 4@xmath59(2.00 ) 3@xmath60(1.78 ) , in 5@xmath58(2.19 ) 5@xmath61(2.48 ) 4@xmath60(1.68 ) , p 3@xmath62(1.83 ) 3@xmath63(1.83 ) 3@xmath64(1.83 ) . however , we used a different basis set ( localized atom - centered functions , vs. planewaves as by oz ) , and larger supercells ( of 64 atoms , vs. a maximum of 16 by oz ) . generally , care must be taken that our bond length calculations are strictly valid only at zero temperature , which might generate slight discrepancy with respect to room temperature , of present interest here . according to the general criterion of localization derived by anderson,@xmath65 the condition for clear observation of the 1-bond@xmath02-phonon behavior in the raman / ir spectra of random ( ga , in)p is that the ga - rich and in - rich environments generate fluctuations in the to frequency that are typically larger than the reference to dispersion in the parent material . precisely the to mode in gap is nearly dispersionless , with a difference between the frequencies of the zone - center and zone - edge ( @xmath25 ) to modes of @xmath11.5 @xmath2 ( ref . this brings a much favorable context for the observation of a 1-bond@xmath02-phonon behavior in the ga - p spectral range . in contrast the same frequency difference is rather large in inp , of @xmath120 @xmath2 ( ref . actual phonon localization in the in p spectral range would require extremely large fluctuations in the to frequency , and thereby in the in - p bond length . however , by analogy with ( ga , in)as@xmath27 this is rather unlikely . on the above basis we are lead to describe random ( ga , in)p as a three - oscillator [ 1(in p ) , 2(ga p ) ] phonon system , in a first approximation . we propose the 1-bond@xmath02-mode to picture outlined in fig . 1 for the random ga@xmath30in@xmath29p alloy ( thick lines ) , from careful re - examination of the available raman / ir data in the literature ; details are given below . it consists of three quasi - parallel branches : two well - separated ga p branches that converge in the dilute limits , above two in p branches so tight ( dashed lines ) that they merge into an overall in p branch . the high and low - frequency ga - p branches refer to ga p vibrations within the in - rich and ga - rich regions , respectively ( refer to sec . the to modes are accordingly labeled as to@xmath66 , to@xmath67 and to@xmath68 with increasing frequency , where the superscript refers to the host region . for each individual ga p branch the so - called regime 2 ( refer to sec . ii ) is ideally modeled as a straight line , in the first approximation . we have checked that the proper re - scaled mrei curves exhibit a bowing of merely @xmath11 @xmath2 . regarding the strength aspect , the to@xmath66 , to@xmath67 and to@xmath68 modes scale as the corresponding fractions of bonds in the alloy , i.e. as @xmath47 , @xmath69 and @xmath70 , respectively ( refer to sec . the relative strengths of the different to modes are explicitly indicated in fig.1 , for more clarity . p to double - branches do not overlap in our description , which reconciles ( ga , in)p with the elliott s criterion . thick lines refer to to modes . the corresponding fractions of bonds are indicated on the left . thin and dotted lines refer to coupled and uncoupled lo modes , respectively . the circles mark strong @xmath44-coupling regimes . calculated raman lineshapes in the ga - dilute limit ( lo , top inset ) and at large ga - content ( to , bottom inset ) are shown . in the latter case the experimental data from kato _ ( reproduced from fig . 11 ) are added ( squares ) , for comparison . , scaledwidth=90.0% ] from fig . 1 we re - assign the s - called minor and dominant to modes in the raman / ir spectra as the to@xmath68 mode and the sum of the nearby to@xmath71 and to@xmath67 modes , respectively . accordingly the minor to mode is gap - like , as predicted by the first - principles calculations of oz@xmath22 [ refer to point ( ii ) above ] . also , the strengths of the nearby to@xmath66 and to@xmath67 modes vary in opposite sense versus the ga - content @xmath47 , which accounts for the antagonist asymmetry of the sum at large and small @xmath47 values as detected by beserman _ et al . _ by using raman spectroscopy [ refer to point ( i ) above ] . more quantitative insight is given in sec . iii - c . on top of the parent to lo bands three empirical parameters were used to build up the entire 1-bond@xmath02-mode picture : the frequencies of the gap : in and inp : ga impurity modes , as for the more simple 1-bond@xmath01-mode mrei picture,@xmath5 plus the ga p splitting , noted @xmath45 . we have taken @xmath7222 @xmath2 corresponding to the frequency gap between the dominant ( 368 @xmath2 ) and the minor ( 390 @xmath2 ) to modes close to the in - dilute limit , as accurately measured by kato _ et al . _ by using raman spectroscopy.@xmath18 this fixes directly the profile of the whole to@xmath68 branch once the singularity at the in p bond percolation threshold is taken into account ( refer to sec . the to@xmath67 branch is derived by symmetry , as expected in case of a random ga substitution to in over the whole composition range.@xmath26 on this basis a typical phonon splitting of @xmath133 @xmath2 should be observed between the two ga p to modes at the stoichiometry , in remarkable agreement with the value of @xmath135 @xmath2 found by oz via first - principles calculations ( refer to figs . 3 from refs . 12 - 13 ) . the location of the inp : ga impurity mode at @xmath1347 @xmath2 , as detected by js@xmath17 by using raman spectroscopy , is not questioned at this stage , owing to the strong experimental support . on the other hand we re - assign the gap : in mode below the gap optical band , i.e. at @xmath1350 @xmath2 . this is estimated from linear extrapolation of the to@xmath66 frequency from the bulk ( 303 @xmath2 ) to the in - dilute limit , passing through an intermediary value ( dot in fig . 1 ) measured close to the in - dilute limit for more accuracy . this is adjusted so as to achieve full contour modeling of the experimental ( to@xmath66 + to@xmath67 ) raman signal as obtained by kato _ _ at the in content of @xmath10.14 ( refer to the bottom spectrum from fig . 11 ) , the other parameters being otherwise fixed by the 1-bond@xmath02-mode model . technical detail is given in sec . the same phonon damping was taken for the two modes . the best fit , obtained for @xmath73(in p)=344 @xmath2 , is superimposed to the experimental curve in the bottom inset of fig . 1 , for comparison . clear theoretical insight upon the phonon behavior of random ( ga , in)p was derived by oz@xmath22 at the representative alloy composition corresponding to the stoichiometry , which was much helpful to build up our to phonon picture displayed in fig . 1 [ refer to point ( ii ) above ] . additional phonon calculations at other alloy compositions are of little interest for our purpose because the relative positions of the three to modes is not expected to be much dependent on the alloy composition , as can be inferred from fig . 1 . what we rather need to fully validate fig . 1 is complementary bond length information close to other critical alloy compositions , i.e. the bond percolation thresholds and the dilute limits , as detailed below . we address two issues . first , we need direct evidence that the gap : in impurity mode stays below the gap optical band , not within ( 390 @xmath2 ) as is currently admitted . for direct insight we achieve full relaxation of a ( ga , in)p supercell containing one in atom only out of 32 cations ( in @xmath13 at.% ) . the single - site in substitution to ga shortens the in p bond length from @xmath742.540 in the pure inp crystal to @xmath752.485 because the in impurity has to fit into the gap - like host media characterized by a smaller bond length . the key point is that the local in p compression is hydrostatic here . accordingly , while explicit ab initio calculations of the phonons either by the linear response , as was done by oz,@xmath22 or by the frozen phonon scheme , as in ref . 21 , are the obliged way to get theoretical phonon insight out of the dilute limits , where the individual bonds undergo complex bond distortions , they are not required here in a first approximation . instead we simply estimate the expected shift @xmath76 in the square to frequency due to the local compression via the grneisen parameter @[email protected] , as measured from the frequency - dependence of the zone - center to mode of pure inp under hydrostatic pressure.@xmath79 we use the relation@xmath80 @xmath81 where @xmath82 is the relative change in the bond length . this yields the estimate of @[email protected] @xmath2 for the displaced frequency of the gap : in impurity mode , i.e. even lower than the measured value ( @xmath1350 @xmath2 ) . we have performed similar bond length calculations by using the same supercell but with ga and in interchanged . @[email protected] for pure gap@xmath79 so that the observed lengthening of the ga p bonds from @xmath742.360 in pure gap to @xmath752.393 , due to outward hydrostatic relaxation of the four p neighbors around the ga impurity , yields the expected frequency @[email protected] @xmath2 for the inp : ga impurity mode , in very good agreement with the actually measured value ( @xmath1347 @xmath2 , see ref . this validates our procedure to estimate the impurity - related phonon frequencies . as the second issue we need evidence of a 1-bond@xmath02-mode bond length distribution , that would mirror the 1-bond@xmath02-mode phonon behavior . basically the bond length distribution must be narrow for the in - p bond , and well - resolved for the ga - p one . also , the short ( long ) bonds must refer to the in - rich ( ga - rich ) region , for each species . to address these aspects we consider the minority bond species close to the ga p ( [email protected] ) and in p ( [email protected] ) bond percolation thresholds , where it is easy to figure out the ga - rich and in - rich regions . we use our prototype ( zn , be)se supercell of ref . 21 , with 4 impurity atoms connected in a straight wall - to - wall chain plus 2 isolated impurities . the ga and in atoms are interchanged for the in - poor and ga - poor configurations . in the ga - poor supercell the connected and isolated ga - p bonds refer to the ga - rich and in - rich regions , respectively . the situation is reversed for the in - p bonds in the in - poor supercell . first , we consider the ga - poor supercell . we obtain an overall bi - modal bond length distribution for the stiff - short ga - p bond ( not shown ) that mimics the be - se reference in ( zn , be)se ( refer to fig . 4 top - left in ref . 21 ) , as expected . only , the distributions of the isolated and connected bonds are re - centered close to the bond length of pure gap , i.e. at @xmath12.383 and @xmath12.402 , respectively . the bond length difference is @xmath18 , which has to be related to @xmath7222 @xmath2 on the phonon side , as detailed above . remarkably , @xmath720 for the ga as bond in ( ga , in)as , while the difference in the ga as bond length reaches @xmath11%.@xmath27 this is consistent with our present view that for a given local bond distortion @xmath45 increases with the bond stiffness . this statement is based on our earlier observation in the reference ( zn , be)vi systems that similar changes in the be*vi * ( @xmath12% ) and zn*vi * ( @xmath11% ) bond lengths according to whether the bonds belong to the be - rich region or the in - rich one generate a much larger phonon shift , i.e. a much larger change in the bond force constant , for the stiff be*vi * bond ( @xmath18.5% of the to frequency in pure bevi ) than for the soft zn*vi * one ( @xmath11% of the to frequency in pure zn*vi*).@xmath27 on this basis we anticipate a large @xmath84 value for the ga - n bond in zincblende ( ga , in)n , with larger contrast in the bond properties as ( ga , in)as and ( ga , in)p , i.e. ( @xmath859.5% , @xmath8641.7% ) . this is an unexplored issue . in the in - poor supercell the connected in p bonds are shorter ( 2.487 ) than the isolated ( 2.493 ) ones , as expected . also , the difference in bond length is small , i.e. @xmath12 , as for the reference in as bond in ( ga , in)as.@xmath27 we expect that the corresponding small phonon splitting is screened by the large to dispersion in inp ( @xmath120 @xmath2 ) , and/or by the phonon damping . this justifies a posteriori our approximation of an apparent 1-bond@xmath01-mode behavior in the raman response of the in - p bond in fig . 1 . full contour modeling of the raman lineshapes of random ( ga , in)p is achieved by using the reduced ( to ) and extended ( lo ) forms of eq . ( 1 ) from the frequency / strength information displayed in fig . 1 . the other input parameters are the faust - henry coefficients of gap ( @xmath870.53 , ref . 29 ) and inp ( @xmath870.46 , ref . 30 ) and the parent oscillator strengths , as derived according to the traditional mrei scheme from the parent to - lo optical bands and the @xmath88 values of gap ( 8.45 , ref . 2 ) and inp ( 9.53 , ref . 30 ) . now we compare the raw theoretical lineshapes with the raman data and first - principles phonons calculations available in the literature . we emphasize that no adjustable parameter is used in our phenomenological calculations . the theoretical to raman lineshapes are shown in fig . a small phonon damping of 1 @xmath2 is taken for a clear overview of the whole collection of individual modes . in particular fig . 2 provides straightforward insight upon the puzzling antagonist asymmetry of the dominant ( to@xmath66 + to@xmath67 ) mode on each side of the stoichiometry , as detected by beserman _ [ refer to point ( i ) above ] . remarkably the nearby to@xmath66 and to@xmath67 modes have similar strengths at the stoichiometry , as ideally expected . the difference is less than 10% to the advantage of the gap - like mode , in close agreement with the zone - center to density of states derived by oz in their first - principles calculations ( see fig . 3-a in ref . we note that the relative strength of their to mode at @xmath1370 @xmath2 is small with respect to our prediction , apparently due to overdamping . otherwise the variation of our theoretical strength ratio between the to@xmath68 mode ( minor ) and the ( to@xmath66 + to@xmath67 ) mode ( dominant ) versus the ga - content @xmath47 fairly reproduces the experimental curve obtained by kato _ et al . _ from their exhaustive raman data ( see fig . 5 in ref . 11 ) , as shown in the inset of fig . in particular the different slopes observed in the dilute limits are rather well - reproduced by our model . to the best of our knowledge this is the first attempt to put a model on these data . in@xmath29p calculated from fig . the topologies of the ga - rich ( black ) and in - rich ( white ) regions are schematically represented at different @xmath47-values . the modes are accordingly labeled for clarity . the evolution of the theoretical strength ratio @xmath38 between the to@xmath68 mode and the ( to@xmath66 + to@xmath67 ) mode versus @xmath47 is shown in the inset . the experimental data from kato _ et al . _ ( taken from fig . 11 ) are added ( crosses ) , for comparison . , scaledwidth=90.0% ] now we discuss briefly the lo modes . @xmath44-coupling is explicitly allowed in our phenomenological model by taking a single dielectric function for the three - oscillator system . this tends to generate a single ` giant ' oscillation that receives most of the available oscillator strength , thereby blue - shifted from the rest of the series.@xmath26 three lo features are obtained but these have ( in p , ga p)-mixed character , which results in a strong distortion of the overall lo signal with respect to the reference uncoupled lo lines . examples are given in sec . the mixed lo features are simply labeled as lo@xmath89 , lo@xmath90 and lo@xmath91 with increasing frequency , where superscript ` int . ' stands for ` intermediary ' . the uncoupled lo lines are calculated via eq . ( 1 ) also , but by taking separate dielectric functions for each oscillator , i.e. by treating each oscillator independently . they are referred to as lo@xmath66 , lo@xmath67 and lo@xmath68 hereafter , by analogy with the to counterparts . the variations of the frequencies of the coupled and uncoupled lo modes versus the ga - content x are displayed as thin and dotted lines in fig . 1 , respectively . in particular the lo@xmath89 mode is systematically weak , and remains confined between the to@xmath66 and to@xmath67 phonon branches when the alloy composition changes . in analyzing the lo - allowed raman spectra obtained with the traditional backscattering geometry along the [ 001]-growth axis of the now available ( ga , in)p epitaxial layers , we suspect that this mode was previously mistaken as the parasitical activation of the dominant to mode ( theoretically forbidden ) , the result of breaking in the raman selection rules induced by the alloy disorder . also , we identify the minor optical band detected by lucovsky _ _ in their pioneer ir measurements ( refer to sec . i ) , as the lo@xmath90 to@xmath67 band . the key point here is that the to and lo features do not refer to the same vibration , which suppresses the enigma of the apparent lo - to inversion . three regimes corresponding to strong @xmath44-coupling , i.e. quasi - resonance for some of the uncoupled lo modes , are identified by circles in fig . 1 . in each case typical coupling - induced anti - crossing behaviors are observed ( compare the thin lines that cross each other , and the dotted lines that do not cross in fig . 1 ) . in particular , strong @xmath44-coupling occurs close to the stoichiometry , which prevents the discussion of any of the lo features in the raman / ir spectra in this very sensitive composition range ( refer to sec . iv ) as due to any specific bond vibration . strong @xmath44-coupling occurs also in the ga - dilute limit , which we identify as the mechanism behind the spectacular emergence of the inp : ga mode in the lo raman spectra , as predicted by js.@xmath17 a priori this strong distortion with respect to the uncoupled lo features ( refer to the thin lines in the top inset of fig . 1 ) might have generated misleading phonon shifts , with concomitant impact on the reliability of the frequency of the inp : ga impurity mode as determined from the raw lo raman spectra . as a matter of fact the exact location of the inp : ga mode was debated at a certain time . to be quite sure we have calculated the lo raman lineshape for a typical ga content of @xmath11% for the two proposed locations of the inp : ga impurity mode , i.e. @xmath12 @xmath2 above@xmath17 and below@xmath14 the lo mode of pure inp ( @xmath1345 @xmath2 ) . the two theoretical lo raman lineshapes exhibit antagonist asymmetries ( not shown ) , which indicates that the overall lo signal is extremely sensitive to the location of the inp : ga impurity mode . as a matter of fact only the assignment proposed by js fairly reproduces the low - temperature lo raman spectrum that these authors obtained , after the latter spectrum was slightly red - shifted to simulate ambient conditions ( see spectrum c in fig . 2 of ref . 10 ) . since the pioneer extended x - ray absorption fine structure ( exafs ) obtained by mikkelsen and boyce with the representative ( ga , in)as system,@xmath92 it is rather well - admitted that while the lattice constant exhibits a quasi - linear variation from one end member to the other when the alloy composition varies , each bond species tends more or less to keep its natural bond length in the alloy , as determined in the pure crystal . transferred to lattice dynamics this simple picture would support the standard 1-bond@xmath01-mode behavior , as envisaged by the mrei model.@xmath5 however , such a simple bond length description offers little flexibility for the discussion of spontaneous ordering in mixed crystals . as a matter of fact the mechanism behind spontaneous ordering still remains a debated issue.@xmath8 generally our 1-bond@xmath02-mode picture for the phonon behavior in the raman / ir spectra of mixed crystals opens an attractive area for the discussion of spontaneous ordering , because it brings the idea of two bond lengths per species in the alloy . this allows to play with the related competition effects regarding the minimization of the local strain energy due to the bond length mismatch between the end compounds . as a matter of fact our recent atomistic calculations of the bond length distribution in large ( ga , in)as supercells that mimic real random alloys clearly confirm the bi - modal character of the bond length distribution related to the short ga - based bond , corresponding to a clear bi - modal behavior in the raman / ir spectra.@xmath27 the configurations were analyzed to distinguish between bonds from the same species that are either inter - connected or isolated , not in the usual terms of next - nearest neighbors the analogy is straightforward with ( ga , in)p . there again the short ga p bond exhibits a clear bi - modal behavior in the raman / ir spectra ( refer to fig . 1 ) , corresponding to a bi - modal bond length distribution that discriminates between the ` short ' and ` long ' ga p bonds from the in - rich and ga - rich regions , respectively ( refer to sec . iii - b ) . as the phonon splitting between the two ga p to modes remains basically unchanged throughout the whole composition range ( @xmath133 @xmath2 , refer to fig . 1 ) , we expect the same for the corresponding difference in the ga p bond lengths . this should remain close to the value found at the ga p bond percolation threshold , i.e. @xmath18 ( refer to sec . iii - b ) . now we focus our attention on gainp@xmath3 . at this critical composition the ga p bond species are in similar proportion in the alloy , and also the two series of ga p bonds ( corresponding to the two ga p to modes in the raman / ir spectra , refer to sec . ii ) , which enhances the local strain energy . now , the two series of ga p bonds undergo a local tensile strain , due to the longer in p bonds . our view is that spontaneous ordering occurs so as to favor a single ga p bond length in the alloy , i.e. the larger one so as to minimize the local strain energy . basically with increasing order we expect that the topology of the ( ga , in ) substituting species becomes more like that in the ga - rich region , and thus that the to@xmath67 mode reinforces at the cost of the to@xmath68 mode . as a matter of fact we note that the minor to resonance ( to@xmath68 ) is deep in the ir transmission spectra obtained by alsina _ _ et al.__@xmath93 with nominally random gainp@xmath3 , while it is shallow for strongly ordered gainp@xmath3 ( compare figs . 1 and 2 in ref . 32 ) , suggesting that the minor to mode weakens with increasing order . oz@xmath94 arrive at the same conclusion via first - principles calculations . besides mestres _ _ et al.__@xmath95 observe by using raman scattering that the dominant to mode ( to@xmath66 + to@xmath67 ) strengthens and sharpens with increasing order , as can be expected from reinforcement of the gap - like character . we denote as @xmath96 ( 0@xmath971 ) the fraction of ` short ' ga - p bonds that has turned ` long ' due to spontaneous ordering . now we calculate the to and lo raman lineshapes while increasing progressively @xmath96 until we obtain fair agreement with the extended data in the literature related to the now available gainp@xmath3 films . care must be taken that all of these exhibit spontaneous ordering to some extent , as characterized by values of the order parameter @xmath98 in the range @xmath10.10.5.@xmath99 we remind that @xmath98 measures the average deviation with respect to equal representation of the substituting species in the ( 111 ) cationic planes , corresponding to the formation of a ( ga@xmath100in@xmath101p@xmath3)/(ga@xmath101in@xmath100p@xmath3 ) [ 111 ] superlattice . @xmath98=0 corresponds to the perfectly random situation , while @xmath98=1 corresponds to the perfectly ordered cupt - type gainp@xmath3 , i.e. a succession of full - ga and full - in cationic planes along the [ 111 ] direction . for direct comparison with the data we take a realistic phonon damping of 10 @xmath2 , and a reduced splitting of @xmath14 @xmath2 between the to@xmath66 and to@xmath67 modes as given by the ab initio calculations of oz ( see fig . 3 in ref . 12 ) , i.e. roughly half our prediction . the central frequency remains the same . incidentally this discrepancy of a few @xmath2 at the stoichiometry between the ab initio calculations and our phenomenological model is rather small when considering that the latter model is entirely built up from a reduced set of three input parameters taken in the dilute limits ( the @xmath45 value for the ga - p phonon splitting , and the frequencies of the gap : in and inp : ga impurity modes ) . the ( to , lo ) situation for random gainp@xmath3 ( @xmath102 ) is shown in fig . 3-a , for reference purpose . the best agreement between the model and the data in the literature is obtained for @xmath1030.7 , typically . this corresponds to a dominant to mode at @xmath1332 @xmath2 , a weak to mode at @xmath1372 @xmath2 , and three clear lo signals at @xmath1333 , @xmath1362 and @xmath1381 @xmath2 . the curves are shown in fig . we note that as @xmath96 increases the lo@xmath90 and lo@xmath91 modes converge while the lo@xmath90/lo@xmath91 strength ratio enlarges . basically this is enough to explain the puzzling reduction of the so - called valley - to - depth ratio b / a , as schematically indicated in fig . 3-b , with increasing order.@xmath8 = 0 ) and spontaneously ordered ( b , @xmath1030.7 ) gainp@xmath3 . the individual ( to , lo ) modes ( thin lines ) are added , for reference purpose . in part ( b ) arrows indicate the strength / frequency variations of the individual gap - like lo modes when @xmath96 increases . the lo curves are translated along the vertical axis , for more clarity . , scaledwidth=90.0% ] at last , we expect an intrinsic limit to spontaneous ordering in mixed crystals . in gainp@xmath3 this should be reached when all the ga - p bonds are eventually ` long ' in the crystal ( @xmath1031 ) . at this limit the to@xmath68 mode , that represents the ` short ' ga - p bonds in the crystal , should be hardly detectable in the raman / ir spectra . experimentally , this corresponds to @xmath40.5 ( see fig . 2 in ref . in fact we are not aware that spontaneously ordered gainp@xmath3 films could be grown with an order parameter @xmath98 greater than @xmath10.5 . the consensus so far was that random ( ga , in)p is the only alloy that exhibits the so - called modified 2-mode behavior in the raman / ir spectra , with a dominant to mode at low frequency that joins the parent to modes and a minor to mode at high frequency that connects the impurity modes . here we show that ( ga , in)p is not an exception in the crude classification of phonon mode behavior as established by elliott _ et al . _ , to distinguish between 1-bond@xmath01-mode and 2-bond@xmath01-mode systems . consistent understanding of the phonon mode behavior of ( ga , in)p is achieved via a basic version of our 1-bond@xmath02-mode phenomenological model ( earlier referred to as the percolation model ) , supported by detailed re - examination of the raman / ir data available in the literature , phenomenological full contour modeling of the to and lo raman lineshapes while using no adjustable parameter , and first - principles bond length calculations in the impurity limits and close to the bond percolation thresholds . in the latter case it is essential that we discriminate between isolated and connected bonds , not in the usual terms of next - nearest neighbors . the to and lo modes in the raman / ir spectra are re - assigned , with notable difference from previous attributions . in particular the gap : in impurity mode , earlier identified within the gap optical band , i.e. at @xmath1390 @xmath2 , is re - assigned at a frequency significantly below the to mode ( 368 @xmath2 ) , i.e. at @xmath1350 @xmath2 . accordingly the two impurity modes stay out of the to lo bands of the host compounds , so that the in p and ga p to phonon branches do not overlap . this is enough to reconcile ( ga , in)p with the elliott s criterion . in the to symmetry the final picture consists of two well - separated ga p phonon branches just above two in p branches so tight that they merge into a single overall in p branch . on this basis the dominant to mode in the raman / ir spectra is re - assigned as a ( in p , ga p)-mixed mode resulting from the sum of the overall in p mode and the near - by low - frequency ga p mode . besides , the minor to mode is re - assigned as the remaining high - frequency ga p mode . regarding the lo symmetry we show that strong coupling occurs between the individual lo modes , via their long range longitudinal polarization field . the resulting lo signal is strongly distorted with respect to the individual uncoupled lo lines , which makes it not relevant to attribute any lo feature in the raman / ir spectra to any specific bond vibration , at any alloy composition . basically our simple 1-bond@xmath02-phonon phenomenological model appears to provide consistent understanding of the long wave ( to , lo ) phonon properties of random ( ga , in)p , which was still lacking . moreover , we propose a mechanism for spontaneous ordering in gainp@xmath3 , based on our observation of a bi - modal phonon behavior for the short ga p bond in random ( ga , in)p . our view is that spontaneous ordering tends to favor those local atomic arrangements around the ga sites that eventually result in longer ga - p bond length . this way the local strain energy in the crystal due to the bond length mismatch between the parent compounds is minimized . in particular this simple mechanism accounts for two puzzling behaviors in the raman / ir spectra when ordering increases : the reinforcement of the dominant mode to the cost of the minor mode in the to symmetry , and the reduction of the so - called valley - to - depth ratio in the lo symmetry . besides , as an unprewied issue , we predict an intrinsic limit to spontaneous ordering in mixed crystals . in gainp@xmath3 this should be reached when the minor to mode completely disappears from the raman / ir spectra . experimentally this corresponds to @xmath40.5 . it is worth to mention that the attempts to understand why spontaneous ordering failed to generate @xmath98 values greater than 0.5 had attracted little attention so far , whereas higher degrees or spontaneous ordering were currently under expectation . generally this work illustrates that detailed understanding of the phonon mode behavior in mixed crystals requires to take into account the disorder in the force constant on top of the mass disorder . as a matter of fact generalization of the coherent potential approximation to include the disorder in the force constant as a full theory was recently achieved by ghosh _ _ et al.__@xmath104 and alam and mookerjee,@xmath105 for example , but we are not aware of any application to optical properties . this work has been supported by the indo - french center for the promotion of advanced research ( ifcpar project no 3204 - 1 ) . @xmath106 i.f . chang and s.s . mitra , adv . phys . * 20 * , 359 ( 1971 ) . + @xmath107 i.f . chang and s.s . mitra , phys . rev . * 172 * 924 . + @xmath108 r.j . elliott , j.a . krumhansl and p.l . leath , rev . phys . * 46 * , 465 ( 1974 ) . + @xmath109 a. mascarenhas , h.m . cheong , m.j . seong and f. alsina , in _ spontaneous ordering in semiconductor alloys _ , + edited by a. mascarenhas ( springer physics , 2002 ) , p. 391 . + @xmath110 g. lucovsky , m.h . brodsky , m.f . chen , r.j . chicotka and a.t . ward , phys . b * 4 * , 1945 ( 1971 ) . + @xmath111 r. beserman , c. hirlimann and m. balkanski , solid state commun . * 20 * , 485 ( 1976 ) . + @xmath112 e. jahne , w. pilz , m. giehler and l. hildisch , phys . stat . sol . ( b ) * 91 * , 155 ( 1979 ) . + @xmath113 h.w . verleur and a.s.jr . barker , phys . rev . * 149 * , 715 ( 1966 ) . + @xmath114 h.w . verleur and a.s.jr . barker , phys . rev . * 155 * , 750 ( 1967 ) . + @xmath17 b. jusserand and s. slempkes , solid state commun . * 49 * , 95 ( 1984 ) . + @xmath18 t. kato , t. matsumoto and t. ishida , jpn . * 27 * , 983 ( 1988 ) . + @xmath115 v. ozolin and a. zunger , phys . b * 57 * , r9404 ( 1998 ) . + @xmath94 v. ozolin and a. zunger , phys . rev . b * 63 * , 87202 ( 2001 ) . + @xmath116 p.h . borcherds , g.f . alfrey , d.h . saunderson , and a.d.b . woods , j. phys . . state phys . * 8 * , 2022 ( 1975 ) . + @xmath117 p.h . borcherds , k. kunc , g.f . alfrey and r.l . hall , j. phys . c : solid state phys . * 12 * , 4699 ( 1979 ) . + @xmath26 o. pags , m. ajjoun , t. tite , d. bormann , e. tourni and k.c . rustagi , phys . b * 70 * , 155319 ( 2004 ) , + and references therein . + @xmath27 o. pags , t. tite , k. kim , p.a . graf , o. maksimov and m.c . tamargo , j. phys . : matter * 18 * , 577 ( 2006 ) . + @xmath28 d. stauffer , _ introduction to percolation theory _ ( taylor and francis , london , 1985 ) . + @xmath34 o. pags , t. tite , a. chafi , d. bormann , o. maksimov and m.c . tamargo , j. appl . , to be published . + @xmath35 l. bellaiche , s .- h wei and a. zunger , phys . b * 54 * , 17568 ( 1996 ) . + @xmath36 a.v . postnikov , o. pags and j. hugel , phys . b * 71 * , 115206 ( 2005 ) . + @xmath43 r.m . martin , phys . b * 1 * , 4005 ( 1970 ) . + @xmath118 p. ordejn , e. artacho and j.m . soler , phys . b * 53 * , r10441 ( 1996 ) . + @xmath119 j.m . soler , e. artacho , j.d . gale , a. garca , j. junquera , p. ordejn and d. snchez - portal , + j. phys . : condens . matter * 14 * , 2745 ( 2002 ) . + @xmath120 n. troullier and j.l . martins , phys . b * 43 * , 1993 ( 1991 ) . + @xmath65 m. cardona , p. etchegoin , h.d . fuchs and p. molins - mata , j. phys . : condens . matter * 5 * , a61 ( 1993 ) . + @xmath79 r. trommer , h. mller and m. cardona , phys . b * 21 * , 4869 ( 1980 ) . + @xmath80 r. carles , g. landa and j.b . renucci , solid state commun . * 53 * , 179 ( 1985 ) . + @xmath121 w.l . faust , and c.h . henry , phys . * 17 * , 1265 ( 1966 ) . + @xmath122 j.f . young , and k. wan , phys . b * 35 * , 2544 ( 1987 ) . + @xmath92 j.c . mikkelsen and j.b . boyce , phys . b * 28 * , 7130 ( 1983 ) . + @xmath93 f. alsina , j.d . webb , a. mascarenhas , j.f . geisz , j.m . olson and a. duda , phys . b * 60 * , 1484 ( 1999 ) . + @xmath95 n. mestres , f. alsina , j. pascual , j.m . bluet , j. camassel , c. geng and f. scholz , phys . b * 54 * , 17754 ( 1996 ) . + @xmath99 a.m. mintairov , j.l . merz and a.s . vlasov , phys . b * 67 * , 2052 ( 2003 ) . + @xmath123 h.m . cheong , f. alsina , a. mascarenhas , j.f . geisz and j.m . olson , phys . b * 56 * , 1888 ( 1997 ) . + @xmath104 s. ghosh p.l . leath and m.h . cohen , phys . b * 66 * , 214206 ( 2002 ) . + @xmath105 a. alam and a. mookerjee , phys . b * 71 * , 94210 ( 2005 ) .

this paper describes the lifting of automorphic characters of to . it does so by matching the image of this lift with the lift of automorphic characters from to . our matching actually gives a matching of individual automorphic forms , and not just of representation spaces . let be a dimensional quadratic vector space and a certain dimensional quadratic space . to an automorphic form determined by the schwartz function in the lift of the character we match an automorphic form determined by the schwartz function in the lift of the character . our work shows that , the space is explicitly determined by the character . the character is explicitly determined by the space and the function is given by an orbital integral involving . theta lift , automorohic characters

math0509486

the siegel - weil formula identifies the integral of a certain theta function as an eisenstein series on a symplectic group . in modern language , this is an example of a global theta correspondence for the dual pair @xmath15 in which the trivial representation of @xmath16 corresponds to an automorphic representation of @xmath17 generated by the values of eisenstein series . the group @xmath16 has many other one dimensional automorphic representations ( automorphic characters ) and it is an open problem to describe where they all go in the correspondence . this paper deals with the dual pair , @xmath18 . the local and global versions of the theta correspondence , in this case , were studied by waldspurger @xcite @xcite and he obtained very explicit information about corresponding representations . however he did not consider global one dimensional representations of @xmath0 . our main results describes explicitly the lifts of automorphic characters @xmath19 of @xmath20 in terms of theta series coming from various @xmath21 s , these results provide a complement to the work of waldspurger and are analagues to the seigel - weil formula . let @xmath22 be a number field and let @xmath23 be a quaternion algebra over @xmath22 . for any place @xmath24 of @xmath22 , let @xmath25 be the completion of @xmath22 , and let @xmath26 . the local algebra @xmath27 can be either split , @xmath28 or non - split , so that @xmath27 is a division algebra . the invariant @xmath29 is defined as @xmath30 when @xmath27 is split and @xmath31 when @xmath27 is non - split . @xmath32 let @xmath33 by the product formula @xmath32 the cardinality of @xmath34 , is even . the isomorphism class of @xmath23 is determined uniquely by @xmath34 . let @xmath3 be the trace zero elements of @xmath23 , @xmath35 and equip @xmath3 with a quadratic form @xmath36 , where @xmath37 is the reduced norm . the space @xmath3 is thus a three dimensional quadratic space over @xmath22 and , up to scaling @xmath38 , all three dimensional spaces over @xmath22 are constructed this way . since the dimension of @xmath3 is odd , the orthogonal group factors as @xmath39 . moreover , the group @xmath40 acts on @xmath3 by conjugation and the action gives rise to the exact sequence @xmath41 where @xmath42 is the center of @xmath43 . we can thus identify @xmath43 with @xmath44 let @xmath45 be the adeles of @xmath22 and @xmath46 be the ideles of @xmath22 . let @xmath9 be a character of @xmath47 such that @xmath48 . such a character is given by @xmath49 for some @xmath50 where the coset @xmath51 in @xmath52 is uniquely determined and where the global hilbert symbol @xmath53 is defined by the local hilbert symbols @xmath54 . since @xmath55 is the map @xmath56 the composition @xmath57 is a quadratic character on @xmath58 which is trivial on @xmath59 . hence it is an automorphic character of @xmath58 which descends to @xmath60 . all automorphic characters of @xmath60 are obtained in this way . let @xmath1 , be the @xmath61-fold metaplectic cover of @xmath62 fix a nontrivial additive character @xmath63 . then @xmath1 acts on the schwartz space @xmath64 by the weil representation @xmath65 determined by @xmath66 . the action of @xmath1 commutes with the linear action of @xmath58 on @xmath64 given by @xmath67 . using the commuting actions , we define for @xmath68 and @xmath69 the theta function @xmath70 this function is left invariant by @xmath71 . we now assume that @xmath72 is a division algebra , i.e. @xmath73 . in this case , the orthogonal group @xmath74 is anisotropic and the quotient @xmath75 is compact . thus , for a quadratic character @xmath9 as above , the integral @xmath76 is absolutely convergent and defines an automorphic form on @xmath1 . the space @xmath77 is a subspace of @xmath78 where @xmath78 is the space of genuine automorphic forms on @xmath1 . furthermore , the space @xmath79 is right invariant under the action of @xmath1 by right translation of the functions in @xmath78 . our main result is a description of the space @xmath79 for all nontrivial automorphic characters @xmath9 . this description will be given in terms of another theta correspondence which we outline next . let @xmath5 be the one dimensional space with a quadratic form @xmath80 . the group @xmath81 is just @xmath82 and the adelic group @xmath83 is the unrestricted direct product . the group @xmath84 is compact in the usual product topology and @xmath85 is embedded diagonally in @xmath86 a continuous character @xmath12 of @xmath84 determines a set of local characters @xmath87 of @xmath88 by restriction to the subgroup @xmath89 and continuity of @xmath12 implies that at almost all places @xmath90 , the trivial representation of @xmath88 . at the places where @xmath91 it is the sign character @xmath92 . let @xmath93 the automorphy condition on @xmath12 i.e. @xmath94 implies that @xmath95 is even . conversely , for any finite set of places @xmath34 where @xmath95 is even there is a unique automorphic character @xmath96 the group @xmath1 acts on the schwartz space @xmath97 by the weil representation @xmath98 as in the case above , the linear action of @xmath84 given by , @xmath99 commutes with the action of @xmath1 on @xmath97 . using the commuting action , we define for @xmath100 the theta function @xmath101 as above , this function is left invariant by @xmath102 and for an automorphic character @xmath12 of @xmath84 , we define the automorphic form on @xmath1 @xmath103 note that in this case , @xmath84 is compact so the integral is absolutely convergent . the space @xmath104 is a right @xmath1 invariant subspace of @xmath105 . such spaces for varying @xmath5 and @xmath12 , are colloquially referred to as coming from @xmath21 . we can now state our main result . main theorem[main ] let @xmath23 be a quaternion algebra over a totally real field @xmath22 let @xmath9 be a quadratic character of @xmath47 given by @xmath106 , with @xmath50 . assume that @xmath19 and that @xmath73 so that @xmath23 is a division algebra . let @xmath79 be the corresponding automorphic representation of @xmath1 constructed above . then @xmath79 is cuspidal irreducible and @xmath107 where @xmath5 is the one dimensional space with quadratic form @xmath108 and @xmath109 is the automorphic representation of @xmath84 determined by @xmath34 . the case of @xmath110 is covered by the classical siegel - weil formula . in that case @xmath111 is given by eisenstein series whereas in our case @xmath79 is a space of cusp forms . one similarity between our result and the siegel weil formula is that in the classical case , the theta integral attached to @xmath14 is identified with an explicit eisenstein series constructed from @xmath14 . similarly , we are able to construct a function @xmath11 from @xmath14 and identify the theta lift attached to @xmath14 with a theta lift attached to @xmath13 . this is made precise in the following [ matching ] suppose that @xmath112 and corresponding @xmath113 are as in the main theorem . then there is a unique map @xmath114 such that @xmath115 for all @xmath116 . more explicitly , for factorizable functions @xmath117 the map @xmath118 is given by @xmath119 where @xmath120 is defined on @xmath121 by @xmath122 we will say that the pair of functions @xmath123 in theorem [ matching ] are matching . in fact theorem [ main ] follows from theorem [ matching ] and the later is proved via a matching of whittaker functions . it turns out that in our case , the automorphic form @xmath7 is distinguished @xcite , i.e. , involves only one square class of fourier coefficients @xmath124 . that square class is the one where @xmath125 , and @xmath126 note that although many authors ( e.g. gelbart and piatetski shapiro @xcite for @xmath127 and vignras , for classical automorphic forms , ) proved that `` distinguished automorphic forms come from @xmath21 '' , their results do not apply in our case . the automorphic forms in @xmath128 are also distinguished . for any @xmath11 the fourier expansion of @xmath10 is @xmath129 where , in fact , @xmath130 . to establish theorem [ matching ] for a given @xmath14 it suffices to construct @xmath13 such that @xmath131 for all @xmath116 . thus the problem of matching reduces to a problem of matching the whittaker functions @xmath132 then , we are reduced to the problem of local matching : given @xmath133 , find @xmath134 such that @xmath135 for all @xmath136 . the analysis of the local matching depends on the bruhat decomposition @xmath137 in the following way . the equality in equation for @xmath138 determines @xmath120 uniquely for @xmath139 by the formula @xmath140 there are then two issues . first we must show that @xmath13 defined on @xmath121 as in equation can be extended to a schwartz function on @xmath25 . then , with this schwartz function @xmath141 , we must show that the equality in holds on the big cell @xmath142 as well and hence on all of @xmath143 . the calculation that establishes the local matching on the big cell amounts to establishing an identity involving orbital integrals , gauss sums and weil indexes . for the global identity we must also check the fundamental lemma which says that , in the unramified situation , the standard unramified vector @xmath144 is matched by the standard unramified vector @xmath145 . one technical point which should be noted is that the usual theta correspondence is between @xmath16 and @xmath1 whereas we have only discussed automorphic characters of @xmath146 . however , in our situation , @xmath147 where @xmath148 and this product allows us to extend our results to @xmath16 . for an automorphic character @xmath149 of @xmath150 , let @xmath151 where @xmath152 the space @xmath153 is the usual theta lift of the automorphic character @xmath154 of @xmath16 and as a consequence of our calculation we can determine it . [ ue ] there is a unique character @xmath149 of @xmath150 such that @xmath155 and @xmath156 for @xmath157 . more precisely , @xmath158 with @xmath159 a final note about the type of representation we get is that the local representations coming from @xmath21 which we match are the irreducible even or odd components of the weil representation for a one dimensional quadratic space . in fact , what we show is that , when @xmath160 , then in the local matching , @xmath161 is matched by an even @xmath162 and , when @xmath163 , then @xmath161 is matched by an odd @xmath162 . this means that the local component @xmath164 of @xmath128 is the image of the irreducible representation of @xmath1 on either the space of even schwartz functions @xmath165 or the space of odd schwartz functions @xmath166 according to the invariant of @xmath23 . by theorem [ matching ] the local component @xmath167 of @xmath79 coincides with @xmath164 . therefore the local components @xmath167 and the global representation @xmath79 are exactly the ones coming from @xmath21 . in @xcite waldspurger considers the space @xmath168 which is the orthogonal complement in @xmath169 to the representation coming from @xmath21 which he calls the elementary weil representations . these elementary representations are exactly the ones that correspond to @xmath79 in our case and so our result complements the classification carried out by waldspurger . our result can be applied to the question of representations by spinor generas to obtain results similar to those of r. schulze - pillot in @xcite . details will be included in a future article . this paper is based on a thesis directed by stephen kudla who conceived of the idea and directed me in carrying it out . i would like to take this opportunity to thanks him for all his help . thanks are also due to jeff adams for advice on the archemedian places and to larry washington for advice on gauss and kloosterman sums . all errors or misstatements that remain are the responsibility of me alone . when @xmath25 is non archimedean we denote the ring of integers by @xmath170 and a generator of the maximal subring @xmath171 of @xmath170 by @xmath172 . let the modular degree of @xmath25 over @xmath173 be @xmath174 so that @xmath175 fix @xmath66-a non trivial additive character of @xmath45 which is trivial on @xmath22 . the local components of @xmath66 , will be denoted by @xmath176 except when it is clear from context that @xmath66 means its local component . the character @xmath66 is defined in terms of the following parameters . we can define a map @xmath177 as the composition of the maps @xmath178 . where the map from @xmath25 to @xmath173 is the trace map . we then define characters of @xmath25 @xmath179 the product of the local characters @xmath180 is a character of @xmath181 and our fixed character @xmath66 and its local factors are defined as @xmath182 for some @xmath183 . for @xmath184 , let @xmath185 . then @xmath186 is the space of all non trivial characters on @xmath187 . the conductor of @xmath188 is the smallest power of @xmath171 on which @xmath188 is trivial . it is equal to the inverse different @xmath189 . the differental exponent is @xmath190 , it is defined such that @xmath191 . note that the conductor of @xmath66 is @xmath192 . @xmath23 is a quaternion algebra over @xmath22 , and @xmath193 . the invariant @xmath194 if @xmath195 and @xmath163 if @xmath27 is the ( unique up to isomorphism ) division quaternion algebra over @xmath25 . the isomorphism class of @xmath23 is uniquely determined by the set of places @xmath196 . @xmath3 is the 3-dimensional @xmath22-vector space @xmath197 we also define @xmath198 and @xmath199 . we will often abuse notation and write @xmath3 instead of @xmath200 when the meaning of @xmath3 is clear from context . the group @xmath43 is the group of invertible quaternions @xmath201 the group @xmath202 is the centralizer of @xmath203 @xmath204 the standard definitions of schwartz spaces are [ scdef ] when @xmath3 is defined over a non archimedean field , the schwartz space @xmath205 is the set of locally constant compactly supported functions on @xmath3 . when @xmath3 is defined over an archimedean field , the schwartz space @xmath205 is the set of infinitely differentiable rapidly decreasing functions . when @xmath3 is split and archimedean , the `` restricted schwartz space '' is the subspace of functions of the form @xmath206 where @xmath207 is a polynomial . when @xmath3 is non split and archimedean , the `` restricted schwartz space '' is the space of functions @xmath208 where @xmath207 is a polynomial and @xmath209 . this section concludes with some basic geometric properties of @xmath23 . [ comp ] if @xmath210 then @xmath211 is not compact and if @xmath212 then @xmath211 is compact . the quotient @xmath213 is compact unless @xmath214 . this follows from corollary 3.2 in @xcite . the product of the reduced norms is isotropic if and only if the reduced norm is isotropic at every @xmath24 . [ ext ] for @xmath215 we have @xmath216 lemma [ ext ] also holds over all local fields @xmath25 and over the ring of adeles , @xmath217 . [ compactness ] if @xmath218 then @xmath219 if and only if @xmath220 is compact . the group @xmath1 is the unique non trivial 2-fold topological central extension of @xmath221 @xmath222 group elements are realized as pairs @xmath223 . multiplication is defined by : @xmath224 where @xmath225 is a 2-cocycle on @xmath221 . the generators of @xmath1 are : @xmath226 when calculating the action of group elements by the weil representation ( see below ) we will drop the @xmath227 from the notation . recall that we have fixed a division quaternion algebra @xmath23 with associated quadratic space @xmath3 and hence indentified @xmath229 . the actions of @xmath1 and @xmath58 on @xmath230 are the following : @xmath58 acts on elements of @xmath231 by conjugation and on elements of @xmath230 by acting on their arguments i.e. for @xmath232 , and @xmath233 , the action is @xmath234 the weil representation of @xmath1 as described in kudla s notes ( @xcite p. 32 - 34 ) is determined by an additive character @xmath66 of @xmath217 which we fixed in chapter [ dtl ] and by the quadratic space @xmath235 , determined by the algebra @xmath72 . the formulas for the actions of the generators of @xmath1 on schwartz functions are : @xmath236 here @xmath237 is the quotient of the weil indices @xmath238 . see the appendix in @xcite for the definitions and properties of the weil index . also , @xmath239 is the determinant of the matrix of the quadratic form on @xmath3 and @xmath240 where @xmath241 is the hasse invariant of @xmath3 , see @xcite . the measure @xmath242 in equation is the unique haar measure which makes the transform @xmath243 self dual . for more details see @xcite . for the 1-dimensional quadratic space @xmath244 with @xmath245 the weil representation @xmath246 of @xmath1 , commutes with the action of @xmath84.this representation is given by the formulas @xmath247 as above @xmath237 is the quotient of the weil indices @xmath238 and @xmath248 our matching of the spaces @xmath79 and @xmath128 will be reduced to matching their whittaker models so we first realize the two spaces as such . we will now calculate the fourier expansion of functions @xmath249 . this calculation will show that , like @xmath128 , the space @xmath79 is distinguished . we start by writing the automorphic forms as sums . @xmath250 we make the convention that @xmath251 should be read as `` a square '' and @xmath252 be read as a non zero square . as mentioned in sections [ quat ] , we exclude the case of the unramified global algebra @xmath253 and therefore by lemmas [ comp ] and [ compactness ] the quotient @xmath75 is compact and the integral in equation is absolutely convergent . also , @xmath3 is anisotropic so that the norms of @xmath215 are not equal to a negative of a non zero square . therefore @xmath254 is not equal to a non zero square . the term @xmath255 in the sum in equation is zero . the space @xmath3 is anisotropic so the only point of norm 0 is 0 and the the @xmath255-term in the sum is the integral of the character @xmath57 on the compact space @xmath75 so the term is zero . the integral in equation can be switched with the summation since the domain of integration is compact and the function has compact support . thus we have @xmath256 where @xmath257 is some point such that @xmath258 notice that the summation over the set @xmath259 is replaced by the sum over @xmath260 . since @xmath23 is a division algebra and @xmath3 is anisotropic @xmath261 acts transitively on @xmath262 and so the choice of @xmath263 does not matter . also , since @xmath9 is trivial on the rational points @xmath264 for @xmath265 so we can include @xmath266 in the argument of @xmath267 . @xmath268 we now combine the inner sum and integral by a change of variables @xmath269 . @xmath270 the next step is to factor @xmath271 with @xmath272 and , @xmath273 . this gives , @xmath274 to proceed , we need the following description of @xmath275 . [ ker ] let @xmath9 is a nontrivial quadratic character of @xmath276 given by the hilbert symbol @xmath277 and @xmath278 . let @xmath279 with @xmath257 @xmath280 if and only if @xmath281 2 . @xmath282 if and only if @xmath283 the first result is contained in proposition 4.8.2 in @xcite when we notice that since @xmath284 the kernel of @xmath285 is the set @xmath286 and since @xmath287 , @xmath288 where @xmath289 . the second statement follows from the fact that if @xmath290 and @xmath291 are quadratic extensions of @xmath22 then they determine characters @xmath292 and @xmath293 , where @xmath294 , and @xmath295 if and only if @xmath296 . [ ramtriv ] if @xmath297 for some place @xmath298 , the set of ramified places for @xmath23 then @xmath299 . if @xmath285 is trivial at a ramified place @xmath24 , the square class of @xmath300 is @xmath301 . but since @xmath302 and @xmath303 is not a norm of @xmath304 , @xmath303 is not in the square class of @xmath300 . therefore , by lemma [ ker ] @xmath305 and the integral over @xmath306 in equation will be zero for any @xmath307 and @xmath14 . using lemma [ ker ] we can see that the inner integral in is nonzero exactly when @xmath283 . if we pick @xmath308 such that @xmath309 then elements with norm @xmath310 are represented by @xmath311 , therefore @xmath312 where we normalized the volume such that @xmath313 when we replace @xmath307 in equation by @xmath314 then the @xmath124 term in the sum is multiplied by @xmath315 . therefore the sum in is actually the fourier expansion of @xmath7 . we make this explicit in the following definitions : [ wvarphi ] the whittaker function corresponding to the form @xmath7 is the function @xmath316 where @xmath317 such that @xmath309 [ ed ] the whittaker model of the automorphic representation @xmath79 is the space @xmath318 we collect the result of the above discussion in the proposition [ dist ] let @xmath319 then 1 . the function @xmath7 is a cuspidal automorphic form and therefore @xmath320 2 . the function @xmath7 is distinguished in the sense of gelbart and piatetski - shapiro @xcite with fourier coefficients in the square class of @xmath321 . the information we have so far , allows us to determine some information about the space @xmath5 . [ setm ] in order that @xmath322 it is necessary that the quadratic form on @xmath5 is defined by @xmath108 . in this section we compute the fourier expansion of the function @xmath323 . first , we write @xmath324 the group @xmath84 is compact , so the representation on @xmath64 is a direct sum of isotypic components and it is not hard to see that the inner integral is the volume of @xmath84 times the projection from @xmath97 onto the @xmath109 isotypic subspace . recall that at places @xmath298 the factor @xmath87 is the sign representation and at places @xmath325 the factor @xmath87 is the trivial character . therefore the @xmath109 isotypic subspace is the subspace @xmath326 where @xmath327 are the subspaces of odd and even functions respectively . this means that we can replace @xmath13 by its projection to @xmath328 which we also denote by @xmath13 and then @xmath329 the space @xmath128 is isomorphic to @xmath328 . to construct the whittaker model for @xmath128 we consider the fourier expansion of the forms @xmath10 . let @xmath330 be any element then for any @xmath331 @xmath332 the term @xmath333 can be removed from the sum since @xmath334 and @xmath335 , so @xmath336 . by equation @xmath337 so we can rewrite the fourier expansion as @xmath338 thus @xmath10 is cuspidal and the fourier expansion is completely determined by the function @xmath339 . notice that the character @xmath340 appears twice in the sum , at @xmath341 and that @xmath342 is even so the coefficient of the character is actually @xmath343 . the above calculation justifies the definitions : [ wvarphi0 ] the whittaker function corresponding to the automorphic form @xmath10 is @xmath344 [ whitt ] the global whittaker model of @xmath128 is the space @xmath345 recall that @xmath346 defined the quadratic form on @xmath347 . we note a result which was evident in the fourier expansion of @xmath10 . [ dist1 ] the automorphic forms @xmath10 are cuspidal and distinguished in the sense of gelbart and piatetski - shapiro @xcite . the characters in the fourier expansion of @xmath10 are all in the square class of @xmath346 . if @xmath8 is factorizable , we write @xmath348 and then @xmath349 where @xmath350 and the @xmath351-whittaker space of all such functions is @xmath352 returning to the case of one dimensional schwartz functions @xmath13 we have a similar factorization where we write an arbitarary function as a sum of factorizable functions . if @xmath11 is factorizable we write @xmath353 and then @xmath354 where @xmath355 and the @xmath351-whittaker space ( by lemma [ setm ] ) of all such functions is @xmath356 from here on , our calculation will consist of local matching . therefore we will drop the @xmath24 subscript from most notation . unless stated otherwise , all formulas should be understood to be over a completion @xmath25 . we will also write @xmath357 instead of @xmath358 . the various lemmas in this section will be used in the later sections . by lemma [ ramtriv ] we can assume that , whenever @xmath110 , then @xmath359 . [ conji ] with notation as above , for every place @xmath24 there is an element @xmath360 such that @xmath361 and @xmath362 . by equation ( 1 ) on page 1 and corollary 2.2 on page 6 . in @xcite we can realize the quaternion algebra as @xmath363 with @xmath364 such that if @xmath365 then @xmath366 . where @xmath367 is the conjugate of @xmath346 in @xmath368 . it follows that @xmath369 . also , by corollary 2.4 on page 6 in @xcite we have @xmath370 if and only if @xmath371 . this means that @xmath372 . [ halpha ] for @xmath373 , let @xmath374 be any element such that @xmath375 with @xmath376 . [ cosets ] let @xmath319 , then the coset space @xmath377 is equal to the disjoint union of right cosets @xmath378 over the set of @xmath379 such that @xmath380 if @xmath381 and @xmath382 are such that for some @xmath124 , @xmath383 and @xmath384 with @xmath203 and @xmath385 perpendicular to @xmath308 then @xmath386 and there exists an element of the orthogonal group of the orthogonal complement to the span of @xmath308 taking @xmath203 to @xmath385 . this means that @xmath387 so the decomposition into right cosets indexed by @xmath124 is proved . to verify the second part of the claim realize @xmath27 as @xmath363 as in the proof of [ conji ] to carry out the calculation we use the orthogonal basis @xmath388 for @xmath3 and write an arbitrary element @xmath203 in terms of this basis : @xmath389 . in terms of those coordinates , if @xmath203 is in the orbit of @xmath308 then @xmath391 this means that @xmath392 and then @xmath393 . now , since @xmath394 we are done . the second lemma describes the values of @xmath395 : [ chihalpha ] if @xmath396 is defined as in [ halpha ] then @xmath397 we extend the basis defined in the proof of lemma [ cosets ] for @xmath3 to all of @xmath27 and write @xmath398 then by multiplying out the conjugation action of @xmath396 : and collecting the coefficients of @xmath308 we get @xmath399 from that we get that @xmath400 and since @xmath401 we get @xmath402 we can now derive an explicit formula for the local component @xmath120 of @xmath13 and obtain some parity results from it . recall that @xmath24 has been fixed and we usually do nt include it in the notation . also , recall the bruhat decomposition @xmath403 [ bmatch ] in order for the functions @xmath404 and @xmath405 to agree on @xmath406 it is necessary and sufficient that @xmath407 since the quadratic form on @xmath408 is @xmath108 , ( see lemma [ setm ] ) @xmath404 and @xmath405 are in the same whittaker space . therefore , in order for them to agree on @xmath406 , it is necessary and sufficient that they agree on right @xmath409 cosets , i.e. on the set @xmath410 . so we compare @xmath411 with @xmath412 solving for @xmath413 gives a necessary and sufficient condition for matching . @xmath414 the quaternion algebra with structure constants @xmath415 has a reduced norm which can be diagonalized as @xmath416 . this means that the determinant of the bilinear form associated to the reduced norm is @xmath417 so @xmath418 in order to to define @xmath13 according to the formula in lemma [ bmatch ] we need to show that the formula can be extended to a schwartz function of @xmath25 . we start with the definition let @xmath14 be a schwartz function in @xmath205 and @xmath9 a quadratic character of @xmath121 . define a function on @xmath121 . @xmath419 [ ari ] if @xmath25 is non archimedean , and @xmath420 then @xmath421 defined by equation can be extended to a function @xmath162 . the next few lemmas will establish proposition [ ari ] . the fact that [ varphi0 ] defines a schwartz function at the archimedean place will follow from an indirect argument in lemmas [ archimatch ] and [ archimatch2 ] . [ lc ] when @xmath25 is non archimedean then @xmath422 is locally constant on @xmath121 and supported inside a compact set of @xmath22 . since @xmath14 has compact support there will be some @xmath423 such that @xmath424 . therefore for @xmath425 small enough , @xmath426 for any @xmath427 and so for @xmath425 small enough @xmath428 . since @xmath14 is locally constant and has compact support there is a neighborhood of @xmath429 such that for all @xmath430 , for all @xmath203 in the support of @xmath14 we have @xmath431 . thus , for @xmath432 and any @xmath430 , @xmath433 [ lcdiv0 ] when @xmath27 is the division algebra @xmath434 , @xmath25 is non archimedean , and @xmath9 is nontrivial , then @xmath13 vanishes on a neighborhood of @xmath435 . since in the division algebra case @xmath436 is compact , for @xmath437 small enough @xmath438 for all @xmath381 . for such @xmath437 @xmath439 since @xmath440 is compact . and the integral over @xmath436 is equivalent to an integral over the compact group @xmath441 . [ lcm20 ] when @xmath28 , @xmath25 is non archimedean , and @xmath9 is non trivial , then @xmath13 is constant on a neighborhood of @xmath435 . since @xmath9 is nontrivial , by lemma [ compactness ] the set @xmath440 is compact and we can define the function @xmath442 it is not hard to see that @xmath443 depends only on the norm of @xmath203 and on @xmath124 as defined in [ halpha ] . next , by lemma [ cosets ] we can see that the measure @xmath444 on @xmath445 is a positive scalar multiple of the product measure @xmath446 with @xmath447 and @xmath124 ranging over the set of @xmath124 such that @xmath448 . also , by lemma [ chihalpha ] @xmath449 . this means that for the purpose of showing that @xmath13 is constant near @xmath435 we can replace the integral over @xmath445 by the integral @xmath450 for some constant @xmath451 . with an abuse of notation we can write @xmath443 as a function of two variables @xmath452 where @xmath453 and @xmath454 . therefore we can write @xmath455 . furthermore , the characteristic function of the set @xmath456 is the function @xmath457 $ ] . when we multiply by the characteristic function we can extend the integral to an integral over all of @xmath25 . @xmath458 \chi(2(\alpha + 1))d\alpha \\ & = \frac{\chi(2)c}{2}\chi(r)|r| \int_{{\ensuremath{\textup{f}}}_{v}}\bar{\varphi}(\alpha r , r^{2}\kappa ) [ \chi(\alpha+1)+\chi(\alpha-1)]d\alpha\end{aligned}\ ] ] we change variables to @xmath459 and we get : @xmath460d\alpha'\ ] ] to continue , we will need to extend @xmath308 to a basis for @xmath3 with @xmath461 such that @xmath462 and @xmath463 and we also need the following lemma . [ eps ] for some @xmath464 , if @xmath465 and @xmath466 then @xmath467 the condition that @xmath466 means that @xmath468 so the square class of @xmath469 can be either @xmath300 or @xmath30 . if @xmath470 i.e. the square class of @xmath469 is @xmath30 , let @xmath471 , it is not hard to see that @xmath472 regardless of @xmath473 . let @xmath474 , then @xmath475 and the @xmath308 coordinate of @xmath203 is @xmath473 so @xmath476 since @xmath14 is locally constant and @xmath477 is compact , for @xmath437 small enough @xmath478 similarly when @xmath479 since @xmath9 is not the trivial character , @xmath480 so it is not possible that @xmath481 where @xmath307 is the index of the coset of squares in @xmath482 , also since in this case @xmath483 , it is not possible that @xmath484 . let @xmath485 and @xmath486 . as above , @xmath487 and if we let @xmath488 we get that for small enough @xmath437 @xmath489 to continue with the proof of lemma [ lcm20 ] notice that for @xmath473 in some neighborhood of @xmath435 we have @xmath490 since we normalized the volume of @xmath477 to be @xmath30 . to show that for @xmath437 small enough @xmath491 is independent of @xmath437 let @xmath437 be small enough that @xmath492 for any @xmath473 such that @xmath466 . this means that @xmath493d\alpha'.\ ] ] if @xmath494 then @xmath495 so we can split the integral into two parts @xmath496d\alpha'\end{aligned}\ ] ] for @xmath437 small enough the integral over @xmath497 becomes @xmath498d\alpha'\ ] ] which splits into the sum of the integrals @xmath499 and when we change variables in the two integrals we get @xmath500 so the sum of the two integrals is @xmath501 we have now established proposition [ ari ] . the following proposition is needed in order to define the global representation . [ fundi ] if @xmath25 is nonarchimedean with @xmath502 odd , @xmath9 is unramified , @xmath160 , @xmath503 and @xmath504 then @xmath505 . from lemma [ bmatch ] we have that when @xmath506 must be @xmath507 if we normalize the measure on @xmath508 such that @xmath509 then the volume of the set @xmath510 is one . notice that for @xmath511 the set @xmath512 is the same as the set @xmath513 and so by lemma [ volomegal ] its volume is @xmath514 . this means that for @xmath515 we have @xmath516 . also for @xmath517 so @xmath518 and @xmath519 . when @xmath9 is unramified non trivial we normalize the measure such that @xmath520 then by lemma [ unramfundi ] @xmath505 . [ volomegal ] when @xmath297 ( and hence @xmath521 ) , @xmath25 is non archimedean and @xmath522 is any lattice then if @xmath523 then @xmath524 . a few things are clear : it is enough to show the result in the case of @xmath525 , and that the volume is independent of the choice of @xmath308 of norm @xmath300 . it is also not hard to see that we can assume that @xmath526 . to carry out the calculation , pick an isomorphism of @xmath23 with @xmath527 such that @xmath528 is the trace zero matrices with integer coefficients and @xmath529 let @xmath530 . it turns out that the double coset @xmath531 is represented by the set @xmath532 with @xmath533 this means that the coset space @xmath534 is the disjoint union of the cosets @xmath535 . it is clear that @xmath536 and @xmath537 so @xmath538 if and only if @xmath539 and also @xmath540 if and only if @xmath541 . the only @xmath542 such that @xmath543 is @xmath544 , so that @xmath545 is equal to the volume of the set @xmath546 . similarly @xmath547 and @xmath544 are the only @xmath542 such that @xmath548 and so @xmath545 is the sum of the volumes of @xmath549 and @xmath550 . comparison of the volumes of the two lattices now follows from a simple group calculation . [ unramfundi ] if @xmath502 is odd , @xmath9 is unramified non trivial , @xmath551 and @xmath552 then @xmath13 as defined by equation [ varphi0 ] is equal to @xmath553 . pick @xmath554 with @xmath555 . by lemma [ ext ] we can see that @xmath556 notice that the function @xmath14 is invariant by the action of @xmath557 so to calculate @xmath558 the integral can be replaced by a sum over coset representative of @xmath559 to actually calculate the integral we need to calculate the volume of the cosets @xmath560 . since we set @xmath561 we obtain the volume of the other cosets by comparing them to the volume of @xmath549 . and in fact , for @xmath562 , @xmath563 . we can now calculate @xmath13 . first notice that if @xmath564 then @xmath565 and @xmath519 . when @xmath566 then @xmath567 now , @xmath568 so for @xmath569 is not in the support of @xmath14 so @xmath570 we can now obtain some parity properties for @xmath14 and @xmath13 . [ par ] the parity of @xmath13 is even when @xmath160 and odd if @xmath163 . be lemma [ conji ] there is an element @xmath571 such that @xmath361 and @xmath362 . with that we can calculate @xmath572 with @xmath573 @xmath574 [ par2 ] let @xmath575 with @xmath576 an even function and @xmath577 an odd function . if @xmath578 then @xmath579 and if @xmath580 then @xmath581 . consider @xmath576 . by corollary [ par ] @xmath582 so @xmath583 and we see that if @xmath584 then @xmath585 . the case of @xmath586 is similar . recall that the choice of parity at places of @xmath22 determines the character @xmath12 . we can restate corollary [ par ] in terms of @xmath12 and obtain the remaining part of the local data defining @xmath587 . in order that @xmath588 it is necessary that the character @xmath589 is defined by @xmath590 for the following , we will need to recall the bruhat decomposition . from lemma [ bmatch ] we know that if @xmath13 is defined according to equation then @xmath404 and @xmath405 agree on @xmath591 . it remains to show that with the same @xmath13 , the whittaker functions @xmath404 and @xmath405 will also agree on @xmath592 . the verification of this result is simplified by the following lemma . [ coin ] let the map @xmath593 be defined by @xmath594 where @xmath13 is given by equation , then for any @xmath595 @xmath596 since @xmath597 , it will suffice to show that for any @xmath598 , @xmath599 and for any @xmath600 @xmath601 to establish equation @xmath602 we compute @xmath603 @xmath604 on the other hand @xmath605 @xmath606 comparing the two sides of equation , we see that they are equal if @xmath607 recall from proof of lemma [ bmatch ] that @xmath608 is in the square class of @xmath61 for both division and matrix algebra so it is true that @xmath609 to establish equation we calculate @xmath610 @xmath611 [ coinvariant ] if for all @xmath612 for some constant @xmath451 , then for all @xmath420 and for all @xmath613 , @xmath614 . the subgroup @xmath615 can be factored as @xmath616 . since @xmath617 the cell @xmath592 can be written as @xmath618 . consider any @xmath619 , by assumption , for any @xmath620 we have @xmath621 so , using lemma [ coin ] @xmath622 so @xmath404 and @xmath623 coincide on all of @xmath624 . finally , since both @xmath625 and @xmath626 are in the same whittaker space by construction , they coincide on all of @xmath627 we will use two different arguments to show that @xmath404 matches @xmath405 . in some cases we will match a single @xmath14 and use general information about the weil representation to conclude that in fact , all other functions @xmath14 are also matched by @xmath13 . in other cases we will show directly that for any @xmath628 for some constant @xmath451 and then use lemma [ coinvariant ] to conclude that the @xmath629 on all of @xmath142 for some constant @xmath451 . finally , we will show that @xmath630 and deduce the actual equality on all of @xmath631 from that relation . from equation and the definition of @xmath405 , ( definition [ wvarphi0 ] . ) we have that @xmath632 notice that we set @xmath346 in equal to @xmath300 according to lemma [ setm ] . the measure @xmath242 is self dual with respect to the fourier transform @xmath633 . since the conductor of @xmath66 is @xmath634 it follows by a standard argument that @xmath635 where @xmath636 gives @xmath170 volume @xmath30 . since @xmath637 and @xmath638 we can simplify the expression for @xmath639 to @xmath640 . and substituting the explicit expression for @xmath13 in terms of @xmath14 , we have @xmath641 on the other hand , by the definition of the action of @xmath642 ( equation ) and the definition of @xmath404 we have @xmath643 as mentioned above , we need to verify the equality of and . notice that both equations involve integrals over @xmath304 and @xmath436 so verification of the equality is independent of the realization of the local algebra @xmath27 as long as the measures @xmath242 and @xmath636 are self dual with respect to the action of @xmath644 and the measure on @xmath436 is the same in both equations . we continue with the calculation of the right hand side of . since @xmath645 is nontrivial , and @xmath646 we know from lemma [ compactness ] that @xmath647 is compact . it is also clear that the integrand in is constant on left @xmath648-cosets so we will replace the integral over @xmath377 with an integral over @xmath649 . to do that , let @xmath650 and we have : @xmath651 we now change variables in the inner integral from @xmath652 to @xmath653 @xmath654 next we let @xmath655 be the characteristic function of a compact set @xmath656 and we note that for any increasing sequence of compact sets @xmath657 where in the last step we replaced the integral over @xmath304 by the integral over @xmath658 . this is justified by the fact that the set of elements with @xmath435 norm have measure @xmath435 in @xmath3 . now , from lemma [ ker ] we know that if @xmath659 is not in the square class of @xmath300 then there is an element @xmath660 in its stabilizer @xmath661 such that @xmath662 . in that case the inner integral is zero . so we can assume that the support of @xmath14 is contained in the set @xmath663 we state this as a lemma for reference . [ sup ] let @xmath664 be a nontrivial character and let @xmath665 be the restriction of @xmath14 to @xmath666 then @xmath667 . to continue , we will parameterize @xmath666 by using the two fold covering @xmath668 the measure on @xmath669 is the measure @xmath242 which makes the fourier transform @xmath670 self dual . this measure is invariant under the action of @xmath43 and transforms by @xmath671 under multiplication by a scalar @xmath672 . the measure @xmath673 on the cover @xmath674 is also invariant under the action of @xmath43 and also transforms by @xmath671 under multiplication by @xmath672 . it follows that the two measures are related by a positive constant @xmath451 . we defer the calculation of @xmath451 since we will soon change variables again and will normalize the measure that results from the two changes of variables . the relation between the two measures allows us to write @xmath625 in terms of integrals over @xmath675 @xmath676 notice that the integrand in the integral over @xmath441 is left invariant under @xmath647 and recall that the volume of @xmath647 is @xmath677 and so the inner integral can be changed to an integral over @xmath377 . @xmath678 comparing with equation we see that the third integral is @xmath679 so we get @xmath680 the integrand in @xmath625 depends only on cosets of the form @xmath681 so using lemmas [ cosets ] [ halpha ] we can change the integral in @xmath625 over @xmath377 to an integral over the product of the set of @xmath124 such that @xmath682 and @xmath683 . we have now changed measures from @xmath242 to the product of the measures @xmath684 , with @xmath124 as above , @xmath685 and @xmath686 . at this point in the calculation we note that the constant relating the two measures will depend on the type of algebra @xmath27 and on the norm of @xmath308 defining the torus @xmath648 . we denote the constant relating the measures @xmath242 and @xmath687 by @xmath688 . so far the formula for @xmath625 looks like : @xmath689 where @xmath124 ranges over elements such that @xmath690 . also recall that @xmath691 with @xmath692 so @xmath693 with the above observation and with lemmas [ cosets ] and [ chihalpha ] the two inner integrals become @xmath694 since we are taking the limit over all increasing sequences of lattices @xmath528 the limit would not change if we considered only lattices which are invariant under the action of the compact group @xmath647 . in that case , the integrand does not depend on @xmath695 . let @xmath696 and the inner integrals become @xmath697 the characteristic function of the set of @xmath698 such that @xmath690 is the function @xmath699 so we can extend the integral over @xmath124 to an integral over @xmath700 by multiplying the integrand by the characteristic function . the integral over @xmath25 is only defined as a limit and should be understood that when we write @xmath701 we will mean @xmath702 . @xmath703 we split the integral into two integrals and change variables to @xmath704 and to @xmath705 respectively . we also define @xmath706 and extend the integral to all of @xmath25 . @xmath707 we use another change of variable @xmath708 and we get @xmath709 we give the above integral a name [ gauss ] the gauss integral is the quantity @xmath710 so far the inner integrals in equation [ lookslike ] are @xmath711 putting equation [ sofar1 ] together with equation we get @xmath712 @xmath713 now by lemma [ par ] @xmath714 and so @xmath715 what we showed so far is that on the big cell , the two whittaker functions @xmath404 and @xmath405 are related by a constant which depends on @xmath716 but does not depend on the argument of the function . i.e. @xmath717 with @xmath718 if we showed that @xmath719 we would be done . [ nantm ] when @xmath25 is non archimedean and @xmath319 then for any @xmath720 @xmath721 furthermore , @xmath722 and for any @xmath420 we have @xmath723 . the discussion above shows that for any @xmath14 , and for @xmath13 defined by equation @xmath724 . by lemma [ coinvariant ] we have that @xmath725 for all @xmath726 . since the weil representation of @xmath631 is smooth , the stabilizers of @xmath14 and @xmath13 are both open subgroups and so the intersection of the stabilizers @xmath727 is an open neighborhood of @xmath728 . it is not hard to see that @xmath729 and @xmath730 we can find @xmath731 such that @xmath732 . if we pick @xmath14 such that @xmath733 then @xmath734 and @xmath735 since @xmath736 . therefore @xmath719 and we are done . the argument for matching in this case is different from the argument in the case of a nontrivial character . in this case , the equality of the spaces @xmath737 and @xmath738 follows from results of waldspurger and rallis . the explicit matching of the functions @xmath404 and @xmath405 will then follow from lemma [ bmatch ] . [ natm ] for @xmath110 and @xmath25 non archimedean @xmath739 furthermore , for every @xmath740 and @xmath13 defined by equation we have @xmath741 we define the map @xmath742 by @xmath743 . the map @xmath744 is @xmath143 equivariant and it is easy to see that it factors through the space @xmath745 of @xmath43 coinvariant . since we know from lemma [ par2 ] that @xmath746 , the map @xmath744 actually factors through the @xmath747 coinvariants @xmath748 . when @xmath110 the formula for @xmath404 is @xmath749 if we let @xmath14 be the characteristic function of some neighborhood containing elements of norm @xmath300 then it is easy to see that @xmath750 . therefore @xmath751 . on the other hand , let @xmath752 where @xmath149 is the character @xmath753 there is an @xmath143 equivariant map @xmath754 given by @xmath755 which also factors through @xmath748 . let @xmath756 be the image of @xmath757 . a result of rallis ( @xcite thm ii.1.1 ) says that the map @xmath757 induces an isomorphism @xmath758 . it is also known ( e.g. gelbart @xcite ) that since @xmath3 is a split space , then in fact @xmath759 . furthermore , the composition series of @xmath760 has length two , it has a unique irreducible submodule and a unique irreducible quotient . in @xcite , waldspurger analyzed the two factors in the composition series . his result is that the unique quotient is @xmath761 : the local theta correspondence of the trivial character from @xmath5 ( which waldspurger denotes by @xmath762 . ) this unique quotient has a @xmath351 whittaker model whereas the unique irreducible submodule of @xmath760 ( which waldspurger denotes by @xmath763 has whittaker models for all square classes of characters @xmath764 except for @xmath351 ( lemma 3 on p 227 in @xcite ) . since @xmath744 factors through @xmath765 there is a map @xmath766 and by considering the whittaker models we see that @xmath767 and so @xmath768 . i.e. @xmath769 is the unique irreducible quotient of @xmath760 which is isomorphic to @xmath761 , that is @xmath770 . the argument splits into three cases handled separately in the next subsections . [ [ chicircnu - is - trivial - on - ensuremathrm - oensuremathtextup - vensuremathmathbbr . ] ] @xmath57 is trivial on @xmath771 . ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ throughout this subsection , @xmath25 will be isomorphic to @xmath772 and we will pick a particular realizations of the matrix algebra and division algebra such that @xmath773 when @xmath9 is nontrivial and @xmath774 when @xmath9 is trivial . we will not mention these conditions in the statements of the lemmas , but they should be understood to be stated with respect to these choices of @xmath300 . since the equality we will establish is defined independently of a particular realization there is no loss of generality . the cases included in this subsection are the case of @xmath775 ( which by lemma [ ramtriv ] , can only exist when @xmath23 is the matrix algebra ) and the case @xmath19 and @xmath23 is the division algebra ( in which case @xmath776 so @xmath777 ) . note that in both cases , once we know that we can match all the forms coming from the representation of @xmath778 , by lemma [ par2 ] , the parity of all @xmath779 can be assumed to be even . hence only the trivial extension of @xmath267 to @xmath771 will have a nonzero lift . this is the reason for describing both cases in this subsection as the case of @xmath57 trivial on @xmath771 . the matching in this case will be established by the matching of the whittaker functions corresponding to the gaussians and using the fact that the the space @xmath780 is generated by the image of the gaussians . the matching of the gaussians will depend on calculating their @xmath549-types and using the iwasawa decomposition to conclude that the two whittaker functions agree on all of @xmath631 . the main argument is contained in the following theorem . [ archimatch ] let @xmath9 be a the trivial character when @xmath23 is a matrix algebra or the nontrivial character when @xmath23 is the division algebra . then @xmath781 and for any @xmath782 we have and @xmath783 . as explained above , in this case the characters extend to the trivial character on @xmath747 . therefore , as mentioned in @xcite , it follows from a result of howe @xcite that @xmath737 is generated by the image of the gaussian : @xmath404 . the calculation that we carry out in this proof will show that if @xmath14 is the gaussian then @xmath783 and therefore @xmath784 . therefore , for any @xmath14 , there is some @xmath785 such that @xmath786 and by lemma [ bmatch ] we know that @xmath787 must be @xmath13 . this would also show that the formula for @xmath13 defines a schwartz function in the restricted schwartz space . the second result is a consequence of lemma [ par2 ] , which implies that the space @xmath738 is the image of the irreducible representation on even functions @xmath788 so it is also irreducible and @xmath781 . by the above argument , the calculation that remains is to show that when @xmath14 is the gaussian we have @xmath783 . by lemma [ weights ] we know that for all elements of @xmath789 we have @xmath790 and @xmath791 for some @xmath346 . as we mentioned above we picked a realization of the algebras with @xmath796 when @xmath9 is nontrivial or trivial , respectively . by lemma [ setm ] , this means that we must choose the quadratic form on @xmath5 to be @xmath797 and @xmath798 respectively . the lie group @xmath549 is connected so its representation is determined by the action of the lie algebra . this lie algebra is generated by the element @xmath801 so it will be enough to calculate the actions of @xmath802 to show that the actions of the whole algebra are the same . we denote the lie algebra action of @xmath803 and @xmath246 by @xmath804 and @xmath805 , respectively . instead of calculating the action of @xmath802 directly , we write @xmath806 with @xmath807 and calculate the following actions @xmath808 and @xmath809 on @xmath13 and @xmath14 respectively . it is not hard to see that @xmath810 and @xmath811 the rest of the proof is a lengthy but standard calculus computation . just as an illustration of the type of calculation involved we calculate the action on @xmath14 when @xmath9 is trivial . in this case the gaussian is the function @xmath812 and we calculate the first term in the action of @xmath802 . @xmath813 we now calculate the second term in the action of @xmath802 . @xmath814 the last step uses a lemma ( which is also omitted ) which says that the gaussian is an eigenfunction of @xmath642 with eigenvalue @xmath815 . @xmath816 to continue the calculation we let @xmath817 and then @xmath818 using standard formulas from calculus and integrating with respect to the self dual measure @xmath819 so that @xmath820 adding up equation and we get the action of @xmath802 on the gaussian @xmath821 [ [ chicirc - nu - is - nontrivial - on - ensuremathtextup - hensuremathmathbbr . ] ] @xmath267 is nontrivial on @xmath822 . ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ the remaining archimedean case is the case of a nontrivial @xmath9 and @xmath23 is a matrix algebra . the argument that will show matching is similar to the one for the other real cases . we will use the fact that the representation is known to be irreducible and the matching of one function to conclude that the whittaker spaces of @xmath737 and @xmath738 coincide . since in this case the quadratic space @xmath3 is split with signature ( 2,1 ) we can use a theorem of zhu , ( 1.2 in @xcite ) which says that in this case , when @xmath826 , @xmath827 . by lemma [ archimatch ] we know that @xmath828 is irreducible and therefore so is @xmath737 . we will show that when @xmath14 is the gaussian then @xmath723 , this would mean that the whittaker spaces are equal and by lemma [ bmatch ] it follows that for any @xmath420 we have , @xmath723 . it will also follow that @xmath13 is actually a function in the restricted schwartz space of @xmath829 . as in the proof of theorem [ archimatch ] it will be enough to show that when @xmath14 is the gaussian then @xmath14 and @xmath13 are of the same k - type . this is done in lemma [ weights2 ] and completes the argument . the local data in the complex case is as simple as possible . since every complex number is a square , the quadratic character @xmath831 is trivial . for the same reason , the quaternion algebra is ramified i.e. @xmath160 . the correspondence in this case can be deduced from known facts without any additional calculation . in @xcite ( proposition 2.1 ) adams and barbasch show that the theta lift of the trivial character from @xmath830 to @xmath832 is isomorphic to the irreducible representation @xmath833 which is the weil representation of @xmath832 on @xmath834- the space of even schwartz functions . if @xmath835 is the automorphic form in the theta lift of the character @xmath57 extended trivially from @xmath836 to @xmath830 defined similarly to @xmath7 then @xmath837 . therefore , @xmath838 and to conclude the matching of functions we need to show that the spaces of functions @xmath79 and @xmath128 actually coincide . this follows from the multiplicity one of spherical representations of @xmath832 . the fact that the representations are spherical is verified by the existence of the gaussian functions in both representations . therefore , the matching for the actual function is given by the formula for @xmath13 .

we explore the viability of a bulk viscous matter - dominated universe to explain the present accelerated expansion of the universe . the model is composed by a pressureless fluid with bulk viscosity of the form where and are constants and is the hubble parameter . the pressureless fluid characterizes both the baryon and dark matter components . we study the behavior of the universe according to this model analyzing the scale factor as well as some curvature scalars and the matter density . on the other hand , we compute the best estimated values of and using the type ia supernovae ( sne ia ) probe . we find that from all the possible scenarios for the universe , the preferred one by the best estimated values of is that of an expanding universe beginning with a big - bang , followed by a decelerated expansion at early times , and with a smooth transition in recent times to an accelerated expansion epoch that is going to continue forever . the predicted age of the universe is a little smaller than the mean value of the observational constraint coming from the oldest globular clusters but it is still inside of the confidence interval of this constraint . a drawback of the model is the violation of the local second law of thermodynamics in redshifts . however , when we assume , the simple model evaluated at the best estimated value for satisfies the local second law of thermodynamics , the age of the universe is in perfect agreement with the constraint of globular clusters , and it also has a big - bang , followed by a decelerated expansion with the smooth transition to an accelerated expansion epoch in late times , that is going to continue forever .

1002.3605

in recent years the type ia supernovae ( sne ia ) observations have indicated a possibly late time _ accelerated expansion _ of the universe @xcite-@xcite . this discovery has been supported also by indirect observations of the expansion of the universe like the cosmic microwave background ( cmb ) @xcite and the large scale structure ( lss ) @xcite observations . it has been coined the name _ dark energy _ to refer to the unknown responsible of such acceleration that presumably is the @xmath8 of the total content of matter and energy in the universe @xcite-@xcite . since the time of this discovery in 1998 @xcite , a large amount of models have been proposed to explain this acceleration . the two most accepted dark energy models are that of a _ cosmological constant _ ( assumed possibly to be the quantum vacuum energy ) and a slowly varying rolling scalar field ( quintessence models ) @xcite-@xcite . despite the cosmological constant model is preferred by the observations , this has several strong problems , among them , the huge discrepancy between its predicted and observed value ( a difference of about 120 orders of magnitude ) @xcite-@xcite and the problem of that we are living _ precisely _ in a time when the matter density in the universe is of the same order than the dark energy density ( the `` _ _ cosmic coincidence problem _ _ '' ) @xcite , @xcite . on the other hand , in the context of _ inflation _ of the very early universe , it has been known since long time ago that an imperfect fluid with bulk viscosity in cosmology can produce an acceleration in the expansion of the universe without the need of a cosmological constant or some inflationary scalar field @xcite@xcite ( although some authors do not agree with this conclusion @xcite ) . so , extrapolating this theoretical idea used to induce an accelerating universe without the need of unknown components it is possible to postulate that one candidate to explain the present acceleration can be a bulk viscous pressure @xcite-@xcite . however , this idea faces some problems too , for example , the need to have a satisfactory mechanism for the origin and composition of the bulk viscosity , nevertheless , there are already some authors working in this aspect @xcite-@xcite . from a thermodynamical point of view the bulk viscosity in a physical system is due to its deviations from the _ local _ thermodynamic equilibrium @xcite ( for a review in the theory of relativistic dissipative fluids see @xcite ) . in a cosmological fluid , the bulk viscosity may arise when the fluid expands ( or contracts ) too fast so that the system does not have enough time to restore its local thermodynamic equilibrium and then it arises an _ effective pressure _ restoring the system to its thermal equilibrium . the bulk viscosity can be seen as a measurement of this effective pressure . when the fluid reaches again the thermal equilibrium then the bulk viscous pressure vanishes @xcite , @xcite@xcite . so , in an _ accelerated _ expanding universe , it may be natural to assume the possibility that the expansion process is actually a collection of states out of thermal equilibrium in a small fraction of time giving rise to the existence of a bulk viscosity . in the present work we study and test a matter - dominated cosmological model with bulk viscosity as an explanation for the accelerated expansion of the universe . the model is composed by a pressureless fluid ( dust ) with bulk viscosity of the form @xmath9 where @xmath1 and @xmath2 are constants to be determined by the observations and @xmath3 is the hubble parameter . the pressureless fluid characterizes both the baryon and dark matter components . the term `` @xmath1 '' takes into account the simplest parametrization for the bulk viscosity , a constant . and the term `` @xmath10 '' characterizes the possibility of a bulk viscosity proportional to the expansion ratio of the universe . the idea of this model is to drive the present acceleration using the trigger at recent times of the bulk viscous pressure of the dust fluid instead of any dark energy component . this fluid represents an unified description of the dark sector plus the baryon component in a similar way than the chaplygin gas model ( see for instance @xcite and references therein ) . one of the advantages of the model is that it solves automatically the cosmic coincidence problem because there is not any dark energy component . the explicit form of the bulk viscosity has to be assumed a priori or obtained from a known or proposed physical mechanism . in the present work we choose the first option to explore the possibilities , but with the idea to explore the second option in future works . different aspects of this model has been also studied in @xcite and references therein . there is already a large amount of works on viscous fluids in cosmology with different parametrizations and motivations . for instance , the particular case of a constant bulk viscosity ( the simplest parametrization ) it has been carefully studied in detail recently @xcite . some other works with the constant parametrization has been also analyzed in @xcite , @xcite-@xcite . other parametrizations or more general approaches for the viscous models have been also proposed and studied at @xcite-@xcite , @xcite-@xcite . another general approach to the bulk viscous cosmologies called `` fluids with inhomogeneous equation of state '' have been also proposed and studied by @xcite . in section [ sectionhubbleparameterwithoutradiation ] we present the main formalism of bulk viscous fluids in general relativity ( gr ) and we apply it to the bulk viscous matter - dominated model studied in this work . we write the expression of @xmath3 in terms of the density parameter of the viscous matter component in function of the redshift . with the expression of @xmath3 we find the explicit expression for the scale factor that will be used in section [ sectionscalefactor ] to analyze the possible scenarios for the universe according to the values of the _ dimensionless _ viscous coefficients @xmath11 and @xmath12 . in sections [ sectioncurvature ] and [ sectionmatterdensity ] we analyze the curvature scalars and the matter density to get a better understanding of the behavior of the universe according to the present model when it is evaluated at the best estimated values of @xmath11 and @xmath12 . in section [ sectionthermodynamics ] we do a brief review of the local second law of thermodynamics and its possible violation for these models . section [ sectionsnetest ] presents the sne ia probe to be used to constrain the model and to compute the best estimated values of @xmath13 . finally , in section [ sectionconclusions ] we give our conclusions . we analyze the cosmological model in a spatially flat universe framework . for that , we use the weinberg formalism @xcite of imperfect fluids . so , in the present work we consider a bulk viscous fluid as source of matter in the einstein fields equations @xmath14 , where @xmath15 is the newton gravitational constant . the energy - momentum tensor of the bulk viscous matter component is that of an _ imperfect fluid _ with a first - order deviation from the thermodynamic equilibrium . it can be expressed as @xcite : @xmath16 where @xmath17 @xmath18 is the four - velocity vector of an observer who measures the energy density @xmath19 and @xmath20 is the pressure of the fluid of matter , the term @xmath21 is the metric tensor , the subscript `` m '' stands for `` matter '' component . the term @xmath22 is a _ bulk viscous _ coefficient that arises in a fluid when it is out of the local thermodynamic equilibrium and that induces a _ viscous pressure _ equals to @xmath23 @xcite-@xcite . the term @xmath24 is an _ effective _ pressure composed by the pressure @xmath20 of fluid plus the bulk viscous pressure . the effective pressure ( [ pressure ] ) was proposed by eckart @xcite in 1940 for relativistic dissipative processes in thermodynamics systems out of local equilibrium . later , landau & lifshitz developed also an equivalent formulation @xcite . the eckart theory has some problems in its formulation , for example , all the equilibrium states in this theory are unstable @xcite , other issue is that signals can propagate through the fluids faster than the speed of light @xcite . to correct the problems of the eckart theory , israel - stewart @xcite developed in 1979 a more consistent and general theory that avoids these issues from which the eckart theory is the first - order limit when the relaxation time goes to zero . in this limit the eckart theory is a good approximation to the israel - stewart theory . in spite of the problems of the eckart theory , but taking advantage of the equivalence of both theories at this limit , it has been widely used by several authors because it is simpler to work with this than with the israel - stewart one . in particular , it has been used to model bulk viscous dark fluids as responsible of the observed acceleration of the universe assuming that the approximation of vanishing relaxation time is valid for this purpose ( see , for instance @xcite-@xcite , @xcite ) . it is important to point out that hiscock @xcite showed that a flat friedmann - robertson - walker cosmological model with a bulk viscous boltzmann gas expands faster when the eckart framework is used than with the israel - stewart one . he suggests that the inflationary acceleration due to the bulk viscosity could be just an effect of using the uncausal eckart theory and not a real acceleration . however , posterior studies seem to show that this conclusion could be wrong because consistent inflationary solutions have been found using the israel - stewart theory @xcite . the model could have a problem related with the cosmological density perturbations , where a rapid late time decay or irregular perturbations arise , leading to large modifications of the model lambda cold dark matter ( @xmath25cdm ) predictions for the matter power spectrum , cmb and weak lensing tests , as shown by b. li et al . however , hipolito et al . @xcite have shown that it is possible to have models with bulk viscosity proportional to @xmath26 that are compatible with the data from the 2dfgrs and the sdss surveys . due to the problem with the density perturbations , in the present work we assume that the viscosity appears only till recent times . it is convenient to mention that for irreversible processes , d. pavn et al . @xcite have developed a more general formulation than the israel - stewart theory , where the temperature could not be associated to the thermal equilibrium . however , in the cosmological scenario , this general formulation has not been totally explored . let s start assuming a spatially flat geometry for the friedmann - robertson - walker ( frw ) cosmology as suggested by wmap @xcite @xmath27 where @xmath28 is the _ scale factor_. moreover , the conservation equation for the viscous fluid is @xmath29 we write this conservation equation using the frw metric as @xmath30 where @xmath31 is the _ hubble parameter_. the dot over @xmath19 and @xmath32 means time derivative . for the matter component , we set @xmath33 , in concordance with a _ pressureless _ fluid . the bulk viscous pressure @xmath34 can be written as @xmath35 . so , the conservation equation ( [ conservationequationviscouscomponent ] ) for the pressureless viscous fluid becomes @xmath36 the friedmann equations for this model are @xmath37 @xmath38 using the ansatz @xmath0 and the first friedmann equation ( [ friedmannfirstequation ] ) , we can write the term @xmath39 of equation ( [ conservationequation-2 ] ) as @xmath40 where we have defined the dimensionless bulk viscous coefficients @xmath11 and @xmath12 as @xmath41 and @xmath42 is the hubble constant . using the expression ( [ conservationeq - termappart ] ) we can rewrite the conservation equation ( [ conservationequation-2 ] ) in terms of the scale factor as @xmath43 we divide the ordinary differential equation ( ode ) ( [ oderho - a ] ) by the _ critical density today _ @xmath44 yielding @xmath45 where we have defined the dimensionless density parameter @xmath46 . using the relation between the scale factor @xmath32 and the redshift @xmath47 we can write the ode ( [ ode_omega - a ] ) as @xmath48 where the function to be calculated in this ode is @xmath49 . this ode has the exact solution @xmath50 ^ 2 ; \quad { \mbox { for } } \ : \ ; \tilde{\zeta}_1 \neq 3\ ] ] where we have used the initial condition @xmath51 like many other authors ( see for instance @xcite and references therein ) . to be consistent with the dynamical observations measured in clusters of galaxies that suggest @xmath52 , it is assumed that this amount of matter is bounded in structures like galaxies and cluster of galaxies and that the remaining matter @xmath53 is distributed homogeneously in the flat space . ] in concordance with a flat universe . on the other hand , we divide the first friedmann equation ( [ friedmannfirstequation ] ) by the critical density today and then substitute there the expression ( [ parameterdensity-2 ] ) to obtain @xmath54 or in terms of the scale factor @xmath55 ; \qquad { \mbox{for } } \ ; \tilde{\zeta}_1 \neq 3\ ] ] we analyze the predicted behavior of the scale factor and the evolution of the universe according to the present model . so , from the definition of the hubble parameter @xmath56 we write the expression ( [ hubbleparameterintermsofa ] ) as @xmath57\ ] ] we define @xmath58 and @xmath59 , and integrate the equation ( [ scalefactorequation1 ] ) as follows @xmath60 } = \int^t_{t_0 } h_0 dt ' = h_0 ( t - t_0)\ ] ] where @xmath61 is the _ cosmic time_. solving this integral we obtain @xmath62 = h_0 ( t - t_0)\ ] ] solving for @xmath28 and substituting the values of @xmath63 and @xmath64 we arrive to @xmath65^{2/(3-\tilde{\zeta}_1 ) } ; \quad { \mbox { for } } \ : \tilde{\zeta}_0 \neq 0 , \ ; \tilde{\zeta}_1 \neq 3\ ] ] also , we compute the first and second derivatives of expression ( [ scalefactorexpression ] ) with respect to @xmath66 to study the accelerated or decelerated epochs of the universe and the transitions between them @xmath67^{(\tilde{\zeta}_1 - 1)/(3 - \tilde{\zeta}_1)},\ ] ] @xmath68^{2(\tilde{\zeta}_1 - 2)/(3-\tilde{\zeta}_1)}\ ] ] we have analyzed the behavior of the scale factor in all the possible combinations of values for @xmath69 . we have found that from all the possibilities , the cases where there is a big - bang as the origin of the universe , followed by a _ decelerated _ expansion epoch at early times and with a _ transition _ to an _ accelerated _ epoch in recent times , corresponds to the cases with values @xmath70 and @xmath71 . we have realized also that the best estimated values for @xmath69 , or @xmath11 alone ( when we set @xmath72 ) computed using the latest sne ia data set , lie precisely in these two ranges of values respectively [ see table [ tablesummary_z01 ] ] . we present below the details of the analysis of these two cases . it is important to highlight that in general for any other case different of the two cases @xmath73 ( with @xmath74 ) , and @xmath75 , the universe does not have the same behavior , i.e. , for some other cases there is not a big - bang or a late time accelerated epoch with a early time deceleration period , etc . first , we describe the case @xmath70 . there are two kinds of behaviors depending on the inequality : @xmath76 figures [ plotscalefactorz0gr0z1ls0][plotscalefactorz0gr0z1ls0gr3 ] show the behavior of the scale factor for these subcases using different values for @xmath13 . for different values of @xmath13 when @xmath77 and @xmath78 but with the condition @xmath79 [ see expression ( [ scalefactorexpression ] ) ] . the model predicts an _ eternally expanding _ universe that begins with a big - bang followed by a decreasing _ decelerated _ expansion until a time when the deceleration vanishes and then there is a _ transition _ to an _ accelerated _ expansion epoch that is going to continue _ forever _ , so that @xmath80 when @xmath81 . see section [ sectionscalefactor ] for details . the points indicate the time when the transition between the _ decelerated expansion _ epoch to an _ accelerated _ one happens [ see expression ( [ transitiondecacez0gr0z1ls0 ] ) ] . the dashed line corresponds to a _ de sitter universe _ that is obtained when @xmath82.,width=377 ] for the subcase * ( a ) * the model predicts an _ eternally expanding _ universe beginning with a big - bang in the past of the universe , i.e. , @xmath83 , where in addition , the curvature scalar @xmath84 and the matter density @xmath85 are singulars , i.e. , @xmath86 ( see sections [ sectioncurvature ] and [ sectionmatterdensity ] for details ) . the subscript `` bb '' stands for `` big - bang '' . ] in the past followed by a _ decelerated _ expansion where the deceleration is _ decreasing _ until a time when it vanishes and then there is a smooth _ transition _ to an _ accelerated _ expansion epoch that is going to continue _ forever _ , so that @xmath87 when @xmath88 ( see figures [ plotscalefactorz0gr0z1ls0]-[plotscalefactorbest12derive ] ) . the cosmic time of _ transition _ `` @xmath89 '' between the decelerated to the accelerated expansion epochs can be calculated equating to zero the expression ( [ scalefactor2derivez1neq3 ] ) yielding @xmath90 to find the expression of the transition between the decelerated to the accelerated expansion epochs in terms of the scale factor we derive with respect to @xmath32 the expression ( [ hubbleparameterintermsofa ] ) and using the fact of that @xmath31 we obtain @xmath91\ ] ] we equal to zero the equation ( [ procatransition ] ) yielding @xmath92^{2/(\tilde{\zeta}_1 - 3)}\ ] ] the transition happens in the _ past _ of the universe if @xmath93 , in the _ future _ if @xmath94 and _ today _ if @xmath95 . on the other hand , using @xmath96 , we find the cosmic time @xmath97 when the big - bang happens @xmath98 we define the _ age of the universe _ as the elapsed time between the time @xmath97 until the present time @xmath99 . so , using the expression ( [ bigbangtimez0ls0z1ls0 ] ) , the age of the universe is given by @xmath100 for different values of @xmath13 when @xmath77 and @xmath78 but with the condition @xmath79 [ see equations ( [ scalefactor1derivez1neq3 ] ) and ( [ scalefactor2derivez1neq3 ] ) ] . these plots show the decelerated and accelerated expanding epochs as well as the transition between them ( see section [ sectionscalefactor ] ) . for the left figure , the _ minimum _ value of each plot corresponds to the cosmic time when the transition happens , that for the right figure corresponds when a plot intercepts the @xmath101-axis.,width=604 ] the best estimated values for the bulk viscous dimensionless coefficients @xmath13 computed using the type ia sne data correspond to this subcase ( see table [ tablesummary_z01 ] and section [ sectionsnetest ] for details ) . the figures [ plotscalefactorbestestimate ] and [ plotscalefactorbest12derive ] show how the model , evaluated at the best estimates , predicts a big - bang as origin of the universe , followed by a _ decelerated _ expanding epoch at early times , with a _ transition _ to an accelerated epoch at late times that will continue forever . with respect to @xmath102 [ see expression ( [ scalefactorexpression ] ) ] when it is evaluated at the best estimated values of @xmath11 and @xmath12 ( see table [ tablesummary_z01 ] ) . the black line corresponds to the values @xmath103 coming from the use of a _ dirac delta prior _ distribution to marginalize over the hubble constant @xmath42 . the red line corresponds to the values @xmath104 coming from the use of a _ constant prior _ distribution to marginalize over @xmath42 ( see table [ tablesummary_z01 ] ) . the plot shows how the model , evaluated at the best estimates , predicts a big - bang as origin of the universe , followed by a _ decelerated _ expanding epoch at early times , with a _ transition _ to an accelerated epoch at late times that will continue forever . the points indicate the cosmic time when the transition between the decelerated epoch to an accelerated one happens [ see expression ( [ transitiondecacez0gr0z1ls0 ] ) and table [ tableagetransition ] ] . the grey and pink bands correspond to evaluate the expression ( [ scalefactorexpression ] ) at the errors in the estimations for bulk viscous coefficients , i.e. , for the dirac delta case ( pink band ) the expression ( [ scalefactorexpression ] ) was evaluated at @xmath105 to plot the upper limit of the band and @xmath106 for the lower one . in the same way , for the constant prior ( grey band ) @xmath107 for the upper and lower limits respectively.,width=377 ] table [ tableagetransition ] shows the values for the age of the universe , evaluated at the best estimates , as well as the value of the redshift of transition between deceleration - acceleration epochs . from table [ tableagetransition ] we can see that for the case of both @xmath13 as free parameters ( i.e. , when @xmath108 ) , and for both marginalizations over the hubble constant @xmath42 , the estimated values for the age of the universe ( 11.72 and 10.03 gyr ) are smaller than the mean value of @xmath109 gigayears coming from the oldest globular cluster observations but still inside of the confidence interval of these : @xmath110 gigayears ( see section [ sectionsnetest ] and table [ tableagetransition ] for details ) . with respect to @xmath111 [ see expressions ( [ scalefactor1derivez1neq3 ] ) and ( [ scalefactor2derivez1neq3 ] ) ] when they are evaluated at the best estimated values of @xmath11 and @xmath12 ( see table [ tablesummary_z01 ] ) . the points indicate the time when the transition between the _ decelerated expanding _ epoch to an _ accelerated _ one happens [ see expression ( [ transitiondecacez0gr0z1ls0 ] ) and table [ tableagetransition ] ] . for the left figure , the _ minimum _ value of each plot corresponds to the cosmic time when the transition happens , that for the right figure corresponds when a plot intercepts the @xmath101-axis . the legends `` _ _ delta prio__r '' , `` _ _ constant prior _ _ '' and `` _ _ constant prior _ _ , @xmath72 '' label the different marginalizations used to compute the best estimated values ( see table [ tablesummary_z01]).,width=566 ] for the subcase * ( b ) * the model predicts an eternally expanding universe and this expansion is always accelerated along the whole history of the universe ( see figure [ plotscalefactorz0gr0z1ls0gr3 ] ) . in this subcase _ there is not _ a big - bang as the origin of the universe , instead of that , the universe has a _ minimum _ value of the scale factor given by @xmath112 where the subscript `` min '' stands for `` minimum '' . from this minimum value @xmath113 the scale factor increases along the time until @xmath87 when @xmath81 . for this subcase the age of the universe is not defined [ see equation ( [ ageuniversez0ls0z1ls0 ] ) ] . for different values of @xmath13 when @xmath77 and @xmath78 but with the condition of @xmath114 [ see equation ( [ scalefactorexpression ] ) ] . the model predicts an eternally expanding universe and this expansion is always accelerated . _ there is not _ a big - bang as the origin of the universe , instead of that , the universe has a _ minimum _ value of the scale factor given by @xmath115^{2/(3 - \tilde{\zeta}_1)}$ ] when @xmath116 . from this minimum value @xmath113 the scale factor increases along the time until @xmath80 when @xmath81 . see section [ sectionscalefactor ] for details.,width=377 ] prior over @xmath42 & age universe & @xmath117 & @xmath118 + constant & 11.72 gyr & 0.61 & 0.63 + dirac delta & 10.03 gyr & 0.65 & 0.52 + constant ( @xmath72 ) & 14.82 gyr & 0.40 & 1.47 + when it is evaluated at the best estimated values for @xmath13 [ see expression ( [ ageuniversez0ls0z1ls0 ] ) and tables [ tableagetransition ] and [ tablesummary_z01 ] ] . @xmath42 is given in units of @xmath119 and the age in units of gigayears ( gyr ) . it was assumed a value of a year of 31556925.2 seconds ( a _ tropical _ year ) and a megaparsec @xmath120 km . the points indicate the computed value for the age of the universe when it is assumed a value of @xmath121 , as suggested by the observations of the hubble space telescope ( hst ) @xcite . the vertical lines correspond to the interval @xmath122\,({{\rm km}}/{{\rm s } } ) { { \rm mpc}}^{-1}$ ] , it is the permitted region according to values of @xmath42 consistent with the distance moduli used to derive ages for galactic globular clusters from the _ hipparcos _ parallaxes ( see ref . the horizontal lines correspond to the constraint for the age of the universe from the oldest globular clusters ( age@xmath123 gyr @xcite ) . so , the shaded area is the consistent region for the age of the universe . table [ tableagetransition ] summarizes the estimated values for the age of the universe.,width=377 ] now , for the second case @xmath124 , it has also the same behavior than the case @xmath125 ( with @xmath126 ) , it is , the model predicts a big - bang as origin of the universe , followed by an early time decelerated expanding epoch , with a transition to an _ accelerating _ epoch that is going to continue forever ( see figures [ plotscalefactorbest12derive ] and [ plotscalefactorbestu10constantz1eq0 ] ) . when we set @xmath72 , the best computed value for the other viscous coefficient is @xmath127 ( see table [ tablesummary_z01 ] ) . the transition between deceleration - acceleration happens when the scale factor has a value of @xmath128 [ see table [ tableagetransition ] and expression ( [ defscalefactortransition ] ) ] . in this case best estimated value for @xmath11 predicts an age for the universe of 14.82 gyr , that is in perfect agreement with the constraints of globular clusters . an important feature of this case is that it does not violate the second law of the thermodynamics ( see section [ sectionthermodynamics ] ) . this particular case has been also studied in @xcite . with respect to @xmath102 [ see expression ( [ scalefactorexpression ] ) ] evaluated at the best estimated value @xmath129 ( see table [ tablesummary_z01 ] ) . it corresponds to the case when @xmath12 is set to zero and it is computed the best estimate for @xmath11 alone , using a constant prior to marginalize over @xmath42 . the legend `` constant prior , @xmath72 '' indicates this . the plot shows how the scale factor , evaluated at the best estimated value for @xmath11 ( with @xmath72 ) , predicts a _ decelerated _ expanding epoch just after de big - bang followed by an _ accelerating _ epoch that is going to continue forever . the transition between deceleration - acceleration happens when the scale factor has a value of @xmath128 [ see table [ tableagetransition ] and expression ( [ defscalefactortransition ] ) ] , it is indicated with the black point in the plot . the thin pink band corresponds to evaluate the expression ( [ scalefactorexpression ] ) at the statistical error of the estimation of @xmath11.,width=302 ] to study the possible singularities of the model we analyze the behavior of the curvature scalar @xmath84 , as well as the contractions of the riemann tensor @xmath130 and the ricci tensor @xmath131 . we find that there is not any particular singularity other than that of the big - bang , that could indicate possible issues for the model coming from the curvature . about weyl tensor , it turns out that @xmath132 in general for the frw metric . the details of the analysis is shown below . the curvature scalar @xmath133 for a spatially flat universe is given as @xmath134\ ] ] the term @xmath135 corresponds to the second friedmann equation , i.e. , @xmath136 we substitute the expressions for : @xmath137 [ see eq . ( [ pressure ] ) ] , @xmath138 and @xmath139 from the first friedmann equation at the expression ( [ 2ndfriedmaneq ] ) to obtain @xmath140\ ] ] using the expressions of the dimensionless bulk viscous coefficients @xmath11 and @xmath12 [ see expressions ( [ defdimensionlesszetas ] ) ] yield @xmath141\ ] ] we substitute the equation ( [ secondfriedmanneq ] ) at ( [ defscalarcurvature ] ) to obtain @xmath142\ ] ] now , using the equation ( [ hubbleparameterintermsofa ] ) for the hubble parameter we arrive to @xmath143 \right\}\ ] ] for @xmath144 . for the case @xmath145 , when @xmath87 then @xmath146 ^ 2 $ ] . when in addition @xmath126 , then the curvature scalar diverges when the scale factor @xmath32 is equal to zero . this supports the existence of a big - bang in the past of the universe ( see figure [ plotrz0gr0z1ls0 ] and section [ sectionscalefactor ] ) . however , when @xmath147 ( with @xmath145 ) there is not big - bang in the past of the universe because the scale factor never becomes zero in the past , the _ minimum _ value that the scale factor can reach is @xmath148^{2/(3-\tilde{\zeta}_1)}$ ] when @xmath116 . so , the curvature scalar does not diverge but it is zero when the scale factor is equal to @xmath113 . then , there is not the conditions for a big - bang when @xmath147 ( see figure [ plotrz0gr0z1ls0 ] ) . when @xmath149 the curvature scalar has the _ constant value _ of @xmath150 along the whole evolution of the universe . when @xmath151 the curvature scalar is _ positive _ as @xmath152 and _ negative _ when @xmath153 . when @xmath154 then the curvature scalar is _ positive _ as @xmath152 if @xmath155 and _ negative _ if @xmath156 . the present central value of the curvature scalar is @xmath157 and @xmath158 when constant and dirac delta priors for @xmath42 are used , respectively . for the other important case @xmath159 , we find that the curvature scalar diverges when @xmath152 ( the big - bang singularity ) . from the infinite value of @xmath160 it decreases as the universe expands until to reach its minimum value of @xmath161 when @xmath87 ( see figure [ plotrz0gr0z1ls0 ] ) . with respect to the scale factor for different values of @xmath13 for the cases @xmath162 and @xmath163 [ see equation ( [ curvaturescalarz1neq3 ] ) ] . the short and long dashed lines correspond to evaluate @xmath84 at the best estimates for @xmath13 when the constant and dirac delta priors are used to marginalize over @xmath42 , respectively . the solid black line corresponds to evaluate @xmath84 at the best estimate of @xmath11 alone , when we set @xmath72 and marginalize @xmath42 using a constant prior . in general , the curvature scalar diverges when the scale factor is equal to zero for the subcase @xmath126 ( with @xmath164 ) . for the subcase @xmath147 ( with @xmath164 ) the _ minimum _ value that the scale factor reaches is @xmath165^{2/(3 - \tilde{\zeta}_1)}$ ] when @xmath116 . so , for this subcase the curvature scalar does not diverge but it is zero when the scale factor is equal to @xmath113 . then , there is not the conditions for a big - bang . when @xmath149 the curvature scalar has the _ constant value _ of @xmath150 along the whole evolution of the universe . when @xmath151 the curvature scalar is _ positive _ as @xmath152 and _ negative _ when @xmath153 . when @xmath154 then the curvature scalar is _ positive _ as @xmath152 if @xmath155 and _ negative _ if @xmath156 . when @xmath80 then @xmath166 ^ 2 $ ] . finally , for the case @xmath163 , the curvature scalar diverges when @xmath167 ( the big - bang singularity ) , and when @xmath87 then @xmath161.,width=377 ] we analyze the behavior of the contraction of the riemann tensor with itself , i.e. , @xmath168 . in general , for a frw metric in a curved space - time @xmath169 is found to be @xmath170\ ] ] where @xmath171 characterizes the constant _ spatial _ curvature of the frw metric ; for our model @xmath172 . so , using the fact of that @xmath173 and that @xmath135 corresponds to the second friedmann equation given by expression ( [ secondfriedmanneq ] ) we rewrite the equation ( [ defkriemanngeneraldef ] ) as @xmath174\ ] ] where @xmath3 is given by the expression ( [ hubbleparameterintermsofa ] ) , that we substitute to obtain @xmath175 ^ 2 \right\ } \nonumber\end{aligned}\ ] ] for the case @xmath176 , we find that @xmath169 diverges to infinity when @xmath152 and when @xmath87 then @xmath177 ^ 4 $ ] . for the best estimates , the present central value for the riemann contraction is @xmath178 and @xmath179 when constant and dirac delta priors are used , respectively . figure [ plotriemannricci ] shows the behavior of @xmath169 with respect to the scale factor , evaluated at the best estimates . for the other case of interest @xmath180 , we find that the scalar @xmath169 is infinite when @xmath152 ( the big - bang singularity ) , it decreases from @xmath181 as the universe expands until to reach its minimum value of @xmath182 when @xmath87 ( see figure [ plotriemannricci ] ) . ( left ) and @xmath183 ( right ) , coming from the contraction of the riemann tensor @xmath184 , and the ricci tensor @xmath185 , with respect to the scale factor [ see expressions ( [ kriemann ] ) and ( [ ariccicontraction ] ) respectively ] . both scalars are evaluated at the best estimates for @xmath13 together , and @xmath11 alone ( when we set @xmath72 ) . see table [ tablesummary_z01 ] . note that in both figures and for the six plots , @xmath169 and @xmath183 diverge when @xmath152 ( the big - bang singularity ) and when @xmath87 then @xmath169 and @xmath183 go to a constant minimum value.,width=604 ] the scalar derived of contracting the ricci tensor with itself , @xmath186 , for a curved frw metric has the form @xmath187\ ] ] for our model @xmath172 . we substitute @xmath188 by @xmath3 and @xmath135 by the expression ( [ secondfriedmanneq ] ) to obtain @xmath189\ ] ] we put the expression for the hubble constant ( [ hubbleparameterintermsofa ] ) in ( [ awithk0 ] ) yielding @xmath190 for the case @xmath191 , when @xmath152 then @xmath192 . and when @xmath87 then @xmath193 ^ 4 $ ] . for the best estimates , the present central value of the ricci contraction is @xmath194 and @xmath195 when constant and dirac delta priors are used , respectively . figure [ plotriemannricci ] shows the behavior of @xmath183 with respect to the scale factor . for the other case @xmath159 , the scalar @xmath183 is infinite when @xmath152 ( the big - bang singularity ) , it decreases from @xmath196 as the universe expands until to reach its minimum value of @xmath197 when @xmath87 ( see figure [ plotriemannricci ] ) . we analyze also the equation of the matter density to have a better understanding of the behavior of the universe according to the model . from the first friedmann equation ( [ friedmannfirstequation ] ) we have @xmath198 using expression for the hubble parameter ( [ hubbleparameterintermsofa ] ) at the expression above we arrive to @xmath199 ^ 2 ; \quad { \mbox{for } } \ ; \tilde{\zeta}_1 \neq 3\ ] ] when @xmath200 : for the subcase @xmath126 ( with @xmath201 ) the matter density diverges when the scale factor is equal to zero . this supports the assumption of a big - bang as the beginning of the universe ( see section [ sectionscalefactor ] and figure [ plotrhoz0gr0z1ls0 ] ) . however , for the subcase @xmath147 the minimum value that the scale factor reaches is @xmath202^{2/(3-\tilde{\zeta}_1)}$ ] when @xmath203 . the matter density is zero when the scale factor is equal to @xmath113 . then , there is not the conditions for a big - bang in this subcase . in both subcases , when @xmath87 then @xmath204 ^ 2 $ ] . for the subcase @xmath149 , the matter density has the _ constant _ value of @xmath205 . figure [ plotrhoz0gr0z1ls0 ] shows the behavior of the matter density with respect to the scale factor . for the other case of interest , @xmath163 , the matter density is infinite when @xmath152 ( the big - bang singularity ) , from this infinite value @xmath19 decreases as the universe expands until to reach its minimum value of @xmath206 when @xmath87 ( see figure [ plotrhoz0gr0z1ls0 ] ) . with respect to the scale factor for different values of @xmath207 where @xmath77 and @xmath78 ( left panel ) and also in the best estimates for @xmath69 ( right panel ) [ see equation ( [ densitymattergeneral ] ) and table [ tablesummary_z01 ] ] . for the subcase @xmath126 ( with @xmath201 ) the matter density diverges when the scale factor is equal to zero . this supports the assumption of a big - bang as the beginning of the universe ( see section [ sectionscalefactor ] ) . however , for the subcase @xmath147 the minimum value that the scale factor reaches is @xmath202^{2/(3-\tilde{\zeta}_1)}$ ] when @xmath203 . the matter density is zero when the scale factor is equal to @xmath113 . then , there is not the conditions for a big - bang in this subcase . in both subcases , when @xmath87 then @xmath204 ^ 2 $ ] . for the subcase @xmath149 , the matter density has the _ constant _ value of @xmath205 . for the right panel , the short and long dashed lines correspond to evaluate the matter density at the best estimates when it is used the constant and dirac delta priors to marginalize over @xmath42 , respectively . the solid black line corresponds to the best estimate for @xmath11 , setting @xmath72 and using a constant prior to marginalize over @xmath42.,width=604 ] with respect to the scale factor ( left panel ) and the redshift ( right panel ) [ see equation ( [ totaldimensionlessviscosity ] ) ] when it is evaluated at the best estimated values for @xmath69 ( see table [ tablesummary_z01 ] ) . at the early universe ( small values of @xmath32 , or greater values of @xmath208 ) the _ total _ bulk viscosity @xmath209 is negative for the case of `` delta prior '' and `` constant prior '' . negative values for the total bulk viscosity @xmath209 implies a violation of the local second law of thermodynamics at early times [ see equation ( [ entropy_condition ] ) ] . so , the model , @xmath210 , works well only for recent times of the universe . the plot `` constant prior , @xmath72 '' corresponds to the model @xmath211 evaluated at the best estimated value for @xmath11 ( we have set @xmath72 ) . this case does not violate the local second law of thermodynamics , the supernova data indicates a positive value for @xmath11 and it is constant along the whole history of the universe . the transition to a positive values in recent times of the model @xmath210 happens when the value of the scale factor @xmath212 is given by equation ( [ scalefactorviscositytransitionnegativepositive ] ) . and when @xmath87 , the total bulk viscosity for this case is @xmath213.,width=604 ] the law of generation of _ local _ entropy in a fluid on a frw space - time can be written as @xcite @xmath214 where @xmath215 is the temperature and @xmath216 is the rate of entropy production in a unit volume . with this , the second law of the thermodynamics can be written as @xmath217 so , from the expression , it implies that @xmath218 thus , for the present model the inequality ( [ 2ndlawthermodynamicszeta ] ) can be written as @xmath219 using the expression for the hubble parameter ( [ hubbleparameterintermsofa ] ) we find that the expression for the _ total _ bulk viscosity @xmath220 is given as @xmath221\ ] ] where we have defined the _ total dimensionless _ bulk viscous coefficient @xmath222 . figure [ plottotalviscosityu10sneonlyaz ] shows the behavior of @xmath209 with respect to the scale factor when @xmath13 are evaluated at the best estimated value ( see section [ sectionsnetest ] and table [ tablesummary_z01 ] ) . we find that for the best estimates of @xmath13 , the _ total _ bulk viscosity is _ negative _ in early times of the universe and _ positive _ at late times . for the best estimate of @xmath11 alone ( when we set @xmath72 ) , the total bulk viscosity is _ positive _ and constant along the cosmic time , because in this case @xmath211 . according to equation ( [ entropy_condition ] ) , negative values for the total bulk viscosity @xmath209 implies a violation of the local second law of thermodynamics . so , the model @xmath210 , evaluated at the best estimates , violates the entropy law at early times , but the model @xmath211 , evaluated at the best estimate for @xmath11 ( with @xmath72 ) , does not . the _ transition _ between negative to positive values of the total bulk viscosity for the case @xmath210 happens when the scale factor has the value @xmath223^{2/(\tilde{\zeta}_1 - 3)}\ ] ] the subscript `` np '' stands for `` negative to positive '' values . for the best estimates for @xmath69 using the sne ia probe we find that the value of the scale factor @xmath212 when transition happens are @xmath224 and @xmath225 when the _ constant _ and _ dirac delta _ priors are assumed to marginalize over @xmath42 , respectively . when @xmath87 then the total bulk viscosity is @xmath226 . prior over @xmath42 & @xmath11 & @xmath12 & @xmath227 & @xmath228 + constant & @xmath229 & @xmath230 & 542.38 & 0.977 + dirac delta & @xmath231 & @xmath232 & 561.15 & 1.011 + constant ( @xmath72 ) & @xmath233 & 0 & 544.587 & 0.979 + we test and constrain the viability of the model using the type ia supernovae ( sne ia ) observations . so , we calculate the _ best estimated values _ for the parameters @xmath11 and @xmath12 and the _ goodness - of - fit _ of the model to the data by a @xmath234-minimization and then compute the confidence intervals for @xmath13 to constrain their possible values with levels of statistical confidence . we use the `` union2 '' sne ia data set ( 2010 ) from `` the supernova cosmology project '' ( scp ) composed by 557 type ia supernovae @xcite . of a bulk viscous matter - dominated universe model with bulk viscosity of the form @xmath138 . we used the scp ( 2010 ) `` union2 '' sne data set composed of 557 type ia sne @xcite , to compute the best estimates @xmath235 and @xmath236 ; they are indicated with the blue point in the figure . the confidence intervals shown correspond to @xmath237 , @xmath238 and @xmath239 of probability . the grey region ( upper region ) corresponds to the values for @xmath69 to which the model @xmath138 predicts that there is not a big - bang in the past , it is , the scale factor never becomes zero in the past of the universe ( see figure [ plotscalefactorz0gr0z1ls0gr3 ] ) , as well as the curvature scalar @xmath84 did not diverge in the past ( see figure [ plotrz0gr0z1ls0 ] ) . the boundary line between the grey and white regions corresponds to a _ de sitter universe _ , it is obtained when @xmath240 . the white region ( central region ) indicates the values to which the universe is _ accelerating _ today , and the green region ( lower ) corresponds to deceleration today . note that the confidence intervals of the best estimates for @xmath69 lie precisely in the accelerating region today with @xmath241 of probability , at least . the dashed blue lines show two values for the age of the universe ( 10 and 20 gyr ) . the computed central value for the age using the best estimates for @xmath69 is 11.72 gyr ( see table [ tableagetransition ] ) . the hubble constant @xmath42 was marginalized assuming a _ constant prior _ distribution.,width=377 ] we use the definition of luminosity distance @xmath242 ( see @xcite , @xcite@xcite ) in a flat cosmology , @xmath243 where @xmath244 is the hubble parameter , i.e. , the expression ( [ hubbleparameter ] ) and ` @xmath245 ' is the speed of light given in units of km / sec . the _ theoretical distance moduli _ for the @xmath171-th supernova with redshift @xmath246 is defined as @xmath247 + 25\ ] ] where @xmath248 and @xmath249 are the apparent and absolute magnitudes of the sne ia respectively , and the superscript ` t ' stands for _ `` theoretical''_. we construct the statistical @xmath234 function as @xmath250 ^ 2}{\sigma_k^2}\ ] ] where @xmath251 is the _ observational _ distance moduli for the @xmath171-th supernova , @xmath252 is the variance of the measurement and @xmath253 is the amount of supernova in the data set , in this case @xmath254 . of a bulk viscous matter - dominated universe model with bulk viscosity of the form @xmath138 . it is like the figure [ plotallciz01sneu10fdm ] but now a _ dirac delta prior _ was used to marginalize over @xmath42 . the best estimates are @xmath255 and @xmath256 ( see table [ tablesummary_z01 ] ) . note how the confidence intervals of the best estimates lie precisely in the accelerating region today with @xmath241 of probability . the computed central value for the age using the best estimates is 10.03 gyr for this case ( see table [ tableagetransition]).,width=377 ] once constructed the @xmath234 function ( [ chisquaredefinition ] ) we numerically minimize it to compute the `` _ _ best estimates _ _ '' for the free parameters @xmath13 of the model . the minimum value of the @xmath234 function gives the best estimated values of @xmath13 and measures the goodness - of - fit of the model to data . the results are shown in table [ tablesummary_z01 ] and the confidence intervals for @xmath11 vs @xmath12 are shown in figures [ plotallciz01sneu10fdm ] and [ plotallciz01sneu10dirac ] . about the value of the hubble constant @xmath42 we marginalize it using two methods : constant prior : : : we assume that @xmath42 does not have any preferred value a priori , i.e , it has a _ constant _ prior probability distribution function . so , we solve _ analytically _ certain integrals using this assumption to obtain a @xmath234-function like the expression ( [ chisquaredefinition ] ) but that it is not going to depend on the variable @xmath42 anymore . this method is carefully described in the appendix a of @xcite . dirac delta prior : : : we assume that @xmath42 has a _ specific _ value ( suggested by some other independent observation ) , so , its probability distribution function has the form of a _ dirac delta _ , centered at the specific value . in particular we choose @xmath257 as suggested by the observations of the hubble space telescope ( hst ) @xcite . in practice , it simply means to set @xmath258 at the expression ( [ chisquaredefinition ] ) [ i.e. , at expression ( [ hubbleparameter ] ) ] and with this @xmath42 is not a free parameter of the model anymore . for the case when we set @xmath72 . the computed best estimated value ( the central value of the pdf ) is @xmath259 , the error is given at @xmath260 confidence level ( see table [ tablesummary_z01 ] ) . a constant prior was used to marginalize over @xmath42.,width=264 ] we have explored a bulk viscous matter - dominated cosmological model composed by a pressureless fluid ( dust ) with bulk viscosity of the form @xmath9 , representing both baryon and dark matter components . one of the advantage of this model is that it resolves automatically the `` coincidence problem '' because it is not introduced any dark energy component , i.e. , it is enough to think of the universe as composed by baryon and dark matter components as dominant ingredients of the cosmos . the expansion of the universe can be seen as a collection of states _ out _ of the thermal equilibrium during _ very short _ periods of time where it is produced local entropy . from all the possible scenarios predicted by the model according to different values of the dimensionless bulk viscous coefficients @xmath11 and @xmath12 , we find two where the universe begins with a big - bang , followed by a decelerated expansion epoch at early times , and with a transition to an accelerated expansion at recent times that is going to continue forever . these two scenarios are characterized for the ranges @xmath191 with @xmath261 , and @xmath262 . this behavior of the universe corresponds to what is actually observed and is close to the @xmath25cdm paradigm . on the other hand , we compute also the best estimated values for @xmath13 using the latest type ia sne data release . it turns out that from all the possible scenarios , the ones chosen by the best estimated values are _ precisely _ the two scenarios @xmath191 with @xmath261 , and @xmath262 . the predicted values of the age of the universe by the model when both @xmath13 are the free parameters , are a little smaller ( 11.72 and 10.03 gyr ) than the mean value of the observational constraint coming from the ages of the oldest galactic globular clusters but still inside of the confidence interval of this constraint ( @xmath110 gyr ) . however , when we set @xmath72 , the age of the universe evaluate at the best estimate of @xmath11 is 14.82 gyr . we find that the computed minimum values of the @xmath234-function by degrees of freedom ( i.e. , @xmath263 ) are very close to the value one ( or even smaller to one ) , indicating that the `` goodness - of - fit '' of the model to the sne ia data is excellent . one of the disadvantages of the model parameterized by @xmath210 , is that the best estimated _ total _ bulk viscosity function @xmath264 is positive for redshifts @xmath265 but _ negative _ for @xmath5 . according to the local entropy production , negative values of @xmath22 imply a violation to the local second law of the thermodynamics . so , this is one of the drawbacks of the model . we find that the cause of this violation is due the fact of having considered the existence of the term @xmath266 in the total bulk viscosity , this term characterizes a bulk viscosity proportional to the expansion ratio of the universe . for the case where we set @xmath72 , we find that the total bulk viscosity is now _ positive _ and constant along the whole history of the universe . this very simple model , in addition to the fact of that it does not violate the local entropy law , it has a good fit to the sne ia observations ( measured through the @xmath267 ) , the age of the universe is in perfect agreement with the observational constraints and it predicts a big - bang with a decelerated expansion followed with a transition to an accelerated epoch . this simple model of one parameter @xmath211 is a better candidate to explain the present accelerated expansion than the model with two parameters @xmath210 . finally , it is important to mention that to have _ viable _ bulk viscous models , a physical mechanism to explain the _ origin _ of this bulk viscosity must be proposed . there are already some proposals in that sense , for instance , a decay of dark matter particles into relativistic product is claimed as explanation of its origin . a preliminary study of this propose , using the sne ia test , have been done in @xcite showing that it could be a promising mechanism to explain this origin . 120 riess a g _ et al , _ 1998 _ astron . j. _ * 116 * 10091038 perlmutter s _ et al , _ 1999 _ astrophys . j. _ * 517 * 565 riess a g _ et al , _ 2004 _ astrophy . j. _ * 607 * , 665687 astier p _ 2006 _ astronomy and astrophysics _ * 447 * 3148 riess a g _ et al , _ 2007 _ astrophy . j. _ * 659 * 98121 davis et al . 2007 _ astrophys . j. _ * 666 * 716725 . wood - vasey _ et al , _ 2007 _ astrophys . j. _ * 666 * 694715 bennett c l _ et al , _ 2003 _ astrophys . j. _ * 148 * 1 tegmark m _ et al , _ 2004 _ phys . d. _ * 69 * 103501 caldwell r r , dave r and steinhardt p j 1998 _ appl . space sci . _ * 261 * 303 zlatev i , wang l and steinhardt p j 1999 _ phys . lett . _ * 82 * 896 wang l , caldwell r r , ostriker j p and steinhardt p j 2000 _ astrophys . j. _ * 530 * 17 steinhardt p j 2003 _ phil . r. soc . a _ * 361 * 2497 peebles p j e 2003 _ rev phys . _ * 75 * 559606 weinberg s 1989 _ rev . phys . _ * 61 * 1 carroll s m , press w h and turner e l 1992 _ annu . astrophys . _ * 30 * 499542 carroll s m 2001 _ living reviews in relativity _ * 4 * 1 padmanabhan t 2003 _ physics reports _ * 380 * , issues 56 , 235320 heller m and klimek z 1975 _ astrophys . and space science _ * 33 * l37-l39 ; diosi l , keszthelyi b , lukcs and pal g 1984 _ acta phys . b _ * 15 * 909 ; 1985 phys . lett . b * 157 * 23 ; morikawa m and sasaki m 1985 _ phys . b _ * 165 * 59 ; waga i , falcao r c and chanda r 1986 _ phys . * 33 * 1839 ; john d barrow 1986 _ phys . lett . b _ * 180 * 335339 ; padmanabhan t and chitre s m 1987 _ phys . lett . a _ * 120 * 443 ; john d barrow 1988 _ nucl . phys . b _ * 310 * 743763 ; john d barrow 1990 string - driven inflation . in _ the formation and evolution of cosmic strings _ eds . g gibbons & t vasasparti 449464 cup cambridge ; gron o 1990 _ astrophys . and space science _ * 173 * 191 g ; beesham a 1993 _ physical rev . d _ * 48 * 3539 ; maartens r 1995 _ class . quantum grav . _ * 12 * 14551465 ; coley a a , van den hoogen r j and maartens r 1996 _ phys . d _ * 54 * 13931397 . i. brevik and o. gorbunova , _ dark energy and viscous cosmology _ , _ gen . gravit . _ * 37*(12 ) ( 2005 ) 2039 - 2045 ; ming - guang hu and xin - he meng , _ bulk viscous cosmology : statefinder and entropy _ , _ phys . letters _ * b 635 * ( 2006 ) 186194 ` [ astro - ph/0511615 ] ` ; jie ren and xin - he meng , _ cosmological model with viscosity media ( dark fluid ) described by an effective equation of state _ , _ phys . lett . _ * b 633 * ( 2006 ) 18 ; jie ren and xin - he meng , _ modified equation of state , scalar field , and bulk viscosity in friedmann universe _ , _ phys * b 636 * ( 2006 ) 512 zimdahl w 2000 _ phys . d _ * 61 * 083511 zimdahl w , schwarz d , balakin a b and pavn d 2001 _ phys . rev . d _ * 64 * 063501 wilson j r , mathews g j and fuller g m 2007 _ physical rev . d. _ * 75 * 043521 mathews g j , lan n q and kolda c 2008 _ phys . d _ * 78 * , 043525 klimek z 1974 _ acta cosmologica _ * 2 * 49 heller m and suszycki l 1974 _ acta phys . * b5 * 345 klimek z 1975 _ acta cosmologica _ * 3 * 49 i. brevik , _ cardy - verlinde entropy formula in the presence of a general cosmological state equation _ , _ phys . * d 65 * ( 2002 ) 127302 chao - jun feng and xin - zhou li , _ viscous ricci dark energy _ , _ phys . * b 680 * ( 2009 ) 355 - 358 ` [ arxiv:0905.0527 ] ` ; chao - jun feng and xin - zhou li , _ cardassian universe constrained by latest observations _ , ` [ arxiv:0912.4793 ] ` ; hristu culetu , _ the membrane paradigm for a dynamic wormhole _ , ` [ arxiv:0906.0920 ] ` . a. tawfik , m. wahba , h. mansour and t. harko , _ viscous quark - gluon plasma in the early universe _ , ( 2010 ) , ` [ arxiv:1001.2814 ] ` ; a. tawfik , h. mansour , and m. wahba , ( 2009 ) , _ hubble parameter in bulk viscous cosmology _ , ` [ arxiv:0912.0115 ] ` ; a. tawfik , t. harko , h. mansour , m. wahba , _ dissipative processes in the early universe : bulk viscosity _ , ( 2009 ) , ` [ arxiv:0911.4105 ] ` . nojiri s and odintsov s d 2005 _ phys . d _ * 72*:023003 ; capozziello s _ et al , _ 2006 _ phys . d _ * 73 * : 043512 ; mubasher jamil and muneer ahmad rashid 2008 _ eur . j. _ * c 56 * 429434 ; hrvoje stefancic 2008 ` [ arxiv:0811.4548 ] ` ; hrvoje stefancic 2008 ` [ arxiv:0812.3796 ] ` ; hrvoje stefancic , _ the solution of the cosmological constant problem from the inhomogeneous equation of state : a hint from modified gravity ? _ , _ phys . _ * b 670 * ( 2009 ) 246 - 253 ` [ arxiv:0807.3692 ] ` . weinberg s 1972 _ gravitation and cosmology : principles and applications of the general theory of relativity _ ( john wiley & sons , inc . , new york ) ; 1971 _ astrophys . j. _ * 168 * , 175 ; misner c w , thorne k s and wheeler j a 1973 _ gravitation _ ( w. h. freeman and company ) p 567 ; hofmann s , schwarz d j and stoecker h 2001 _ phys . rev . d _ * 64 * 083507 . eckart c 1940 _ physical rev . _ * 58 * 919924 landau l d and lifshift e m 1958 _ fluid mechanics _ reading , ma : addison - wesley hiscock w a and lindblom l 1985 _ phys . d _ * 31 * 725 ; 1987 _ phys . d _ * 35 * 3723 muller i 1967 _ z. physik _ * 198 * 329 israel w 1976 _ ann . , ny _ * 100 * 310 wmap collaboration , g. hinshaw et al . , _ five - year wilkinson microwave anisotropy probe ( wmap ) observations : data processing , sky maps and basic results _ , _ astrophys . j. suppl . _ * 180 * ( 2009 ) ` [ arxiv:0803.0732 ] . ` carreta _ et al , _ 2000 _ astroph . j. _ * 533 * 215 - 235 turner m & riess a g 2002 _ astrophy . j. _ * 569 * , 18 sahni v and starobinsky a a 2000 _ int * d 9 * , 373 ; sahni v 2004 _ lect . notes phys . _ * 653 * , 141 . copeland e j , sami m and tsujikawa s 2006 _ int . _ * d 15 * , 1753

we consider the problem of slow activation dynamics in glassy systems undergoing a random first order phase transition . using an effective potential approach to supercooled liquids , we determine the spectrum of activation barriers for entropic droplets . we demonstrate that fluctuations of the configurational entropy and of the liquid glass surface tension are crucial to achieve an understanding of the barrier fluctuations in glassy systems and thus are ultimatively responsible for the broad spectrum of excitations and heterogeneous dynamics in glasses . in particular we derive a relation between the length scale for dynamic heterogeneity and the related barrier fluctuations . diluted entropic droplets are shown to have a gaussian distribution of barriers , strongly suggesting that non - gaussian behavior results from droplet - droplet interactions .

0809.3988

the glass transition and glassy behavior are dynamical phenomena characterized by slow relaxations and a broad spectrum of excitations . recently strong experimental evidence for dynamical heterogeneity in glassy systems , i.e. for spatially varying characteristic time scales for relaxation , has been obtained through nmr @xcite and nanometer - scale probing of dielectric fluctuations@xcite . the wide distribution of excitations , as seen in bulk dielectric measuments , is therefore believed to be caused largely by spatial variation of the characteristic time scales ; see also ref.@xcite . regions in a glassy system separated by only a few nanometers relax on time scales different by many orders of magnitude . whether these purely dynamical phenomena can be explained in terms of the underlying _ energy landscape _ is still debated . the question has been positively answered on the level of the mean field theory of glasses . the dynamical , ideal mode coupling theory@xcite and energy landscape based replica mean field theories were demonstrated to describe the same underlying physics , yet from rather different perspectives@xcite . the idealized version of the mode coupling theory@xcite falls out of equilibrium below a temperature @xmath0 , where @xmath1 is the laboratory glass temperature ( see below ) . this is consistent with static , energy landscape based mean field theories that find an emergence of exponentially many metastable states , @xmath2 , at the same temperature @xmath3 . here @xmath4 is the volume of the system . thus , the configurational entropy @xmath5 becomes extensive and only vanishes at the kauzmann tempeature @xmath6 . unfortunately , neither the emergence of slow activated dynamics , nor the possibility of spatial heterogeneity are correctly captured within mean field theory , making it impossible to use mean field theory alone to directly explain the heterogeneous dynamics observed in numerous glass forming liquids . on the other hand , mode coupling theory at higher temperatures ( @xmath7 ) has been shown to give a reasonably accurate account of the dynamics of liquids . at lower temperatures ( @xmath8 ) , bulk thermodynamic properties of glassforming liquids , most notably the temperature dependence of the configurational entropy and the related kauzmann paradox@xcite , agree with the results of the energy landscape based mean field theory ( see refs.@xcite ) . this motivated the development of a theory for glassforming liquids that has its foundation in the mean field theory , but that does take into account activated dynamics and spatial heterrogeneity which are beyond the strict mean field approachkw87,ktw89 . the resulting random first order ( rfot ) theory of glasses demonstrates that ergodicity above the kauzmann temperature @xmath9 can be restored via entropic droplets , in which a region of the liquid is replaced by any of an exponentially large number of alternatives . the entropy of the possible alternatives acts as a driving force for structure change . this driving force is off - set by the free energy cost of matching two alternative structures at their boundaries . this conflict gives a free energy barrier for activated motionskw87,biroli04 . various consequences of entropic droplets have been analyzed in refs.@xcite using for concrete calculation a density functional description of the liquid state . these quantitative results have been found to be in good agreement with experiment . within rfot , the number of metastable states determines the activation barrier of the glassy state . while this is similar to the adam - gibbs theory for glasses@xcite , there exist important distinctions between the two approaches , see ref.@xcite . in agreement with the rfot approach but not with the assumptions of adams - gibbs argument the complexity of a correlated region increases as the glass transition is approached @xcite . the need for instanton like events to understand the long time dynamics in glasses is also supported by recent findings extended mode - coupling theories for dense fluids@xcite . more recently , an effective landau theory for glasses , based on the replica method of ref.@xcite , has been developed . this framework naturally allows for activated dynamics beyond mean field theoryfranz05 - 1,dzero05 . this approach based on instantons offers a formal justification of the entropic droplet approach using the replica approach to glasses . it reproduces and thus confirms several results of the random first order transition ( rfot ) theory of structural glasses@xcite . in both approaches , a mean droplet activation energy @xmath10 occurs that is determined by the configurational entropy density @xmath11 ( for details see below ) : @xmath12@xmath13 vanishes linearly at the kauzmann temperature @xmath14 . the exponent @xmath15 relates the droplet size @xmath16 and the configurational entropy density : @xmath17 . @xmath18 is the spatial dimension . below we assume that @xmath19 . in its most elementary version the replica landau theory yields @xmath20@xcite . a renormalization of the droplet interface due to wetting of intermediate states on the droplet surface @xcite was shown to yield @xmath21 , leading to @xmath22 and correspondingly to a vogel - fulcher law @xmath23 for the mean relaxation time @xmath24a softening of the surface tension due to replica symmetry breaking of the instanton solution was also obtained in ref . @xcite and seems to be a first correction containing the effects that lead to a reduction of the exponent @xmath15 from the value @xmath20 . numerous experiments on supercooled liquids not only depend on the mean barrier , @xmath10 , but are sensitive to the entire , broad excitation spectrum in glasses@xcite . most notably , the broad peaks in the imaginary part of the dielectric function @xmath25 are most natually understood in terms of a distribution @xmath26 of relaxation times , such that @xmath27similarly dynamical heterogeneity with spatially fluctuating relaxation times yields nonexponential ( frequently streched exponential ) relaxation of the correlation function @xmath28and allows for an explanation of nonresonant hole burning experimentsdiezemann01 , even though interesting interpretations of the non - exponentiality based on `` dynamical homogeneity '' exist as well@xcite . other effects that are most likely caused by a distribution of relaxation rates include the breakdown of the stokes - einstein relation @xmath29 between the diffusion coefficient @xmath30 of a particle of size @xmath31 and the viscosity @xmath32@xcite . these experiments all call for a more detailed analysis of the fluctuations @xmath33of the activation barriers and , more generally , of the distribution function @xmath34 of barriers . the latter yields the distribution function of the relaxation times @xmath35through @xmath36 . for example , in case of a gaussian distribution of barriers one obtains a broad , log - normal distribution of relaxation rates:@xmath37with @xmath38and @xmath39 from eq.[meanr ] . while the distribution , eq.lognorm , does not yield precisely a stretched exponential form for the correlation function , it can often be approximated by @xmath40with @xmath41 . furthermore , the study of higher order moments of @xmath34 is important to determine whether the distribution is indeed gaussian or more complicated . in this paper we determine the fluctuations @xmath42 as well as higher moments of the distribution of the barrier distribution function @xmath34 . we start from the replica landau theory of ref.@xcite and use the replica instanton theory of refs.@xcite . we then generalize the replica formalism to determine barrier fluctuations in addition to the determination of the mean barrier @xmath10 . our approach is an important step to gain insight into the whole spectrum of activated processes in glassy systems in general . the role of barrier fluctuations within the rfot theory of structural glasses was first discussed by xia and wolynes in ref.@xcite . variations in the configurational entropy density were assumed to be the main cause of the fluctuations of the barrier , fig . [ fig1 ] . starting from @xmath43 , barrier and configurational entropy density fluctuations are related by @xmath44 , which yields @xmath45 . standard fluctuation theory then determines the entropy fluctuations in a region of size @xmath46 as @xmath47 where the configurational heat capacity remains finite as @xmath48 ( for a detailed discussion of @xmath49 see below ) . this reasoning finally yields@xmath50 in the approach of ref.@xcite , barrier fluctuations diverge with the volume of the entropic droplet . our results will demonstrate that the leading droplet size dependence of @xmath51 agrees with eq.[xw ] and ref.@xcite . we also show that there are additional terms originating in the fluctuations of the interface tension of entropic droplets that contribute to @xmath51 . in addition we analyze the temperature dependence of the static barrier fluctuations and of the exponent @xmath52 for stretched exponential relaxation as well as the variation of the dielectric response with frequency and temperature . this paper is organized as follows . in the next section we introduce the replica effective potential formalism to analyze barrier fluctuations in glassforming liquids . we motivate the approach by starting from a density functional approach to liquids and using an equilibrium replica theory to derive the mean field theory as starting point of our calculation . next we analyze fluctuations of the configurational entropy in bulk , summarize the effective potential approach of ref.@xcite and the instanton calculation that yields the mean barrier @xmath10 . finally we demonstrate how higher moments of the barrier distribution can be derived within the same formalism . at the end of the section we present our results for physical observables . the paper is concluded with a summarizing section . we start our description of glassy systems from the point of view that there exists a density functional , @xmath53 $ ] , that describes a supercooled liquid undergoing a mean field glass transition . initially , such an approach was used in ref.@xcite where it was shown that it allows one to describe the emergence of a metastable amorphous solid , fig . [ fig2 ] . following the classical approach to freezing into ordered crystalline states@xcite , the density was assumed to be @xmath54 . here , @xmath55 determines the lindemann length , @xmath56 over which particles are localized . in distinction to crystallization , the mean positions , @xmath57 , of an amorphous solid were taken to be those of a random hard sphere packing@xcite , instead of the periodic crystal lattice positions . the free energy of this amorphous solid was shown to have a global minimum at @xmath58 and a local minimum for finite @xmath55 . if @xmath59 , the particles are delocalized and the system in an ergodic liquid state with homogeneous density @xmath60 . finite @xmath61 corresponds to a frozen amorphous solid i.e. a glassy state . for @xmath62 the amorphous solid is metastable with respect to the liquid , and higher in free energy by @xmath63 . it is of course always metastable with respect to the crystalline solid . essentially the same mean field physics can be described in terms of a formally more precise replica approach , introduced by monasson@xcite . here one determines the partition function in the presence of a bias configuration @xmath64:@xmath65 = \int d\rho e^{-\beta \phi \left [ \rho \right ] -g\int d^{d}x\left ( \rho \left ( x\right ) -\widehat{\rho } \left ( x\right ) \right ) ^{2}}\]]here , @xmath66 corresponds to the statistical sum over all density configurations of the system . the free energy of a bias configuration is @xmath67 = -t\log z\left [ \widehat{\rho } \right ] .\]]physically @xmath68 $ ] can be interpreted as the free energy for a metastable amorphous configuration of atoms , for example with density @xmath69 being a sum over gaussians discussed above . in the replica formalism , no specific configuration like this needs to be specified in order to perform the calculation . rather , the assumption is made that the probability distribution for metastable configurations is determined by @xmath68 $ ] according @xmath70 \propto \exp \left ( -\beta _ { \mathrm{eff}}f% \left [ \widehat{\rho } \right ] \right ) \label{distrms}\]]and is characterized by the effective temperature @xmath71 . this allows one to determine the mean free energy @xmath72 p\left [ \widehat{\rho } \right]\]]and the corresponding mean configurational entropy @xmath73 \log p% \left [ \widehat{\rho } \right ] . \label{sc}\ ] ] the approach of ref.@xcite was successfully used to develop a mean field theory for glass formation in supercooled liquids@xcite yielding results in detailed agreement with earlier , non - replica approaches@xcite . as shown by monasson@xcite , the mean values @xmath74 and @xmath75 can be determined from a replicated partition function @xmath76 = \int d^{m}\rho e^{-\beta \sum_{a=1}^{m}\phi \left [ \rho _ { a}\right ] + g\sum_{a , b=1}^{m}\int d^{d}x\rho _ { a}\left ( x\right ) \rho _ { b}\left ( x\right ) } \label{z(m)}\]]via @xmath77 and @xmath78 with @xmath79and replica index @xmath80 . it follows from the replicated free energy @xmath81 : @xmath82these results are in analogy to the usual thermodynamic relations between free energy ( @xmath83 ) , internal energy ( @xmath84 ) , entropy ( @xmath85 ) and temperature ( @xmath86 ) , see also ref.@xcite . if the liquid gets frozen in one of the many metastable states , the system can not anymore realize its configurational entropy , i.e. the mean free energy of frozen states is @xmath87 , higher by @xmath88 if compared to the equilibrium free energy of the liquid , fig . [ fig2 ] . if the replica solution is taken to be marginally stable , so that the lowest eigenvalue of the fluctuation spectrum beyond mean field solution vanishes , it has been shown@xcite that @xmath89 agrees with the result obtained from the generalized fluctuation - dissipation theorem in the dynamic description of mean field glasses@xcite . typically , the assumption of marginality is appropriate for early times right after a rapid quench from high temperatures . in this case @xmath90 for @xmath8 , i.e. the distribution function of the metastable states is not in equilibrium on the time scales where mode coupling theory or the requirement for marginal stability applies . above the kauzmann temperature it is however possible also to consider the solution where @xmath91 , i.e. where the distriburion of metastable states has equilibrated with the external heat bath of the system . since we are interested in a system where ergodicity is restored for @xmath92 we use @xmath91 . technically this is a supercooled liquid and not a non - equilibrium aging glass . below the kauzmann temperature this assumption can not be made any longer ( at least within the mean field theory ) as it leads to a negative configurational entropy , inconsistent with the definition eq.[sc ] . the physically intuitive analogy between effective temperature , mean energy , @xmath87 and configurational entropy to thermodynamic relations , eq.[thermo ] , suggest that one should analyze the corresponding _ configurational heat capacity _ @xmath93using the distribution function eq.[distrms ] we find , as expected , that @xmath94 is a measure of the fluctuations of the energy and configurational entropy of glassy states . we obtain for the configurational heat capacity @xmath95where @xmath96 and @xmath97 . here the mean values @xmath87 and @xmath98 are determined by eq.[thermo ] . the fluctuations of the configurational entropy and frozen state energy are then determined by @xmath99 \log ^{2}p\left [ \widehat{\rho } \right ] $ ] and @xmath100 f\left [ \widehat{\rho } \right ] ^{2}$ ] , respectively . both quantities can be expressed within the replica formalism in terms of a second derivative of @xmath101 with respect to @xmath102 . for example it follows : @xmath103it is then easy to show that eq.[ffluctc ] holds . with the introduction of the configurational heat capacity into the formalism we have a measure for the fluctuation of the number of available metastable states from its mean value . the analogy of these results to the usual fluctuation theory of thermodynamic variables@xcite further suggests that @xmath104 also determines fluctuations of the effective temperature with mean square deviation : @xmath105since @xmath104 is extensive , fluctuations of intensive variables , like @xmath106 , or densities , like @xmath107 , vanish for infinite systems . however , they become relevant if one considers finite subsystems or small droplets . in the context of glasses this aspect was first discussed by donth@xcite . using standard techniques of many body theory we can proceed by expressing the replica free energy @xmath101 as a functional of the liquid correlation functions of the problem ( see e.g. ref . @xcite):@xmath108 = \frac{t}{2m}\left ( \text{tr}\left ( \chi _ { 0}^{-1}\chi \right ) + t\text{tr}\ln \chi + \phi \left [ \chi \right ] \right ) . \label{lw}\]]here the mean field result of @xmath101 is determined by the stationary point of the effective hamiltonian : @xmath109 with replicated density - density correlation function @xmath110the luttinger - ward functional @xmath111 $ ] is determined by the nature of the density - density interaction and determines the self energy @xmath112 , see also refs.mc,bouchaud96 . for example , expanding the density functional , @xmath113 , into a taylor series in @xmath114 , yields to lowest order in @xmath114 : @xmath115where @xmath116 . in what follows we will formulate the theory in terms of the collective variable @xmath117 , instead of the original density fluctuations and write @xmath118where @xmath119 if @xmath120 . @xmath121 depends on two spatial variables @xmath122 and @xmath123 . we simplify the problem by assuming that its fourier transformation , @xmath124 , with respect to the relative coordinates @xmath125 can be written as@xmath126with liquid structure factor , @xmath127 and @xmath128 is the center of mass coordinate . the important collective variable of our theory is @xmath129 , which plays the role of a spatially varying debye - waller factor . within mean field theory we expect below the temperature @xmath3 that @xmath130 is replica symmetric with @xmath131 of order unity , while @xmath131 vanishes in the equilibrium liquid state for @xmath7 . physically the debye - waller factor @xmath131 contains the same information as does the localization parameter @xmath132 discussed above . finally , it is important to stress that a replica symmetric approach of @xmath133 in the present formalism is equivalent to one step replica symmetry breaking in the conventional replica language@xcite . in order to keep our calculation transparent we will not determine @xmath134 $ ] from an explicit microscopic calculation for supercooled liquids . it was demonstrated by franz and parisi@xcite that this is possible . however , in order be able to make simple calculations beyond the mean field theory explicitly including droplet fluctuations , we start from a simpler landau theory in the same universality classgross85,dzero05 : @xmath135 -% \frac{u}{3}\sum_{c}q_{ab}q_{bc}q_{ca}\ \right ) \label{ham1}\]]with @xmath136 = \frac{a_{0}^{2}}{2}\left ( \nabla q_{ab}\right ) ^{2}+% \frac{t}{2}q_{ab}^{2}-\frac{u+w}{3}q_{ab}^{3}+\frac{y}{4}q_{ab}^{4 } \label{ham2}\]]and replica index @xmath137,@xmath138 . @xmath139 is a length scale of the order of the first peak in the radial distribution function of the liquid and @xmath140 is a typical energy of the problem that determines the absolute value of @xmath141 . in addition the problem is determined by the dimensionless variables @xmath142 , @xmath143 , @xmath144 and @xmath145 , which are in principle all temperature dependent . we assume that the primary @xmath146-dependence is that of the quadratic term , where @xmath147 . making this assumption , along with the polynomial form of the lagrangian , requires constraints between the other landau parameters to ensure thermodynamic consistency from the fluctuation formulas . in appendix a we give estimates for the parameters of eqs.[ham1 ] and [ ham2 ] obtained from fits to experimental data for o - terphenyl ( otp ) , a well studied fragile glass forming material . formally , eqs.[ham1 ] and [ ham2 ] can be motivated as the taylor expansion of eq.[lw ] together with the functional @xmath148 of eq.[phi_kb ] . in what follows we further simplify the notation and measure all energies in units of @xmath140 and all length scales in units of @xmath139 . the mean field analysis of this model hamiltonian is straightforward . inserting a replica symmetric ansatz @xmath149 into @xmath150 and minimizing w.r.t @xmath131 yields @xmath151 or @xmath152nontrivial solutions exits for @xmath153 with @xmath154 , which determines the mode coupling temperature @xmath155 . inserting @xmath131 of eq.[qstar ] into @xmath156 $ ] yields for the replica free energy : @xmath157 the mean configurational entropy density , as determined by @xmath158 is given by @xmath159inserting @xmath131 of eq.[qstar ] yields the result that @xmath160 vanishes at @xmath161 with @xmath162 . close to @xmath163 it follows that @xmath164as expected . at @xmath165 one finds @xmath166 . @xmath167 we next determine the configurational heat capacity of the mean field theory discussed above . from eq.[f(m)mon ] we obtain@xmath168if @xmath91 ( corresponding again to the equilibrium behavior between @xmath9 and @xmath3 ) it follows that @xmath169 , so that comparing it with the expression for the configurational entropy ( [ scmon ] ) at @xmath170 we can write@xmath171notice that @xmath172 must equal @xmath173 for complete thermodynamic consistency . for temperatures close to but above @xmath9 the energy fluctuations decrease as @xmath174as the temperature approaches @xmath3 the configurational heat capacity decreases faster than linearly . it reaches a value @xmath175 signalling that the system becomes unstable close to @xmath165 for @xmath176 . this is analogous to the higher order replica symmetry breaking suggested by tarzia and mooretarzia07 on the basis of the same landau action . @xmath177 approaches @xmath178 from above with an infinite slope:@xmath179 the full temperature dependences of the mean field overlap @xmath180 and configurational heat capacity are shown in fig . while @xmath131 or @xmath160 did not depend on the parameter @xmath143 of the hamiltonian , eqs.[ham1 ] and [ ham2 ] , the configurational heat capacity does . we may therefore consider @xmath143 as a parameter that determines the strength of entropy fluctuations . in order to investigate the transition state for the decay of a specific metastable frozen state , it is necessary to have a technique that agrees with the mean field replica approach discussed in the previous section in case of a homogeneous solution @xmath181 , but that allows one to study the behavior for arbitrary values of the overlap between two states , including inhomogeneous droplet solutions . this is possible within the effective potential approach introduced in refs.@xcite . we use this technique to formulate our replica instanton and barrier fluctuation theory . we are interested in the regime above @xmath9 ( i.e. for @xmath182 ) and consider the partition function of a system with constrained overlap between the particle density configurations @xmath183 and @xmath184 : @xmath185 } \notag \\ & & \times \prod\limits_{\mathbf{x}}\delta \left ( p_{c}\left ( \mathbf{r}% \right ) -\rho _ { 2}\left ( \mathbf{r}\right ) \rho _ { 1}\left ( \mathbf{r}\right ) \right ) . \label{zprho}\end{aligned}\]]the configuration @xmath186 is assumed to be in equilibrium and unaffected by @xmath187 . averaging the free energy , @xmath188 , weighted with the equilibrium probability for @xmath186 yields the corresponding mean energy for the overlap @xmath189 : @xmath190 & = & -t\left\langle \log z_{p_{c},\rho _ { 2}}\right\rangle _ { \rho _ { 2 } } \notag \\ & = & -t\frac{\int d\rho _ { 2}e^{-\beta \phi \left [ \rho _ { 2}\right ] } \log z_{p_{c},\rho _ { 2}}}{z}.\end{aligned}\]]here @xmath191 is the equilibrium partition function @xmath192 } $ ] . using a replica approach @xmath193 $ ] can be written as@xcite @xmath194 = -t\lim_{n\rightarrow 0}\lim_{m\rightarrow 0}% \frac{1}{m}\left ( z^{\left ( n , m\right ) } -1\right ) , \]]with@xmath195 -\beta \sum_{b=1}^{m}\phi \left [ \widehat{\rho } _ { b}\right ] } \notag \\ & & \times \prod\limits_{r , b=1}^{m}\delta \left ( p_{c}\left ( r\right ) -\rho _ { 1}\left ( r\right ) \widehat{\rho } _ { b}\left ( r\right ) \right ) .\end{aligned}\]]@xmath193 $ ] can be determined by analyzing an @xmath196 times replicated problem with additional constraint for the overlap between certain replicas . in complete analogy to the previous paragraph one can introduce a field theory of the overlap , but now with order parameter@xmath197where @xmath198 is an @xmath199 matrix , @xmath200 an @xmath201 matrix and the @xmath202 matrix @xmath203 must obey the additional constrained that @xmath204 for @xmath205 . we have the same effective hamiltonian , eqs.[ham1 ] and [ ham2 ] only with @xmath200 replaced by @xmath206 , i.e.@xmath207 } \prod% \limits_{b=1}^{m}\delta \left ( p_{1b}-p_{c}\right ) .\]]following franz and parisi@xcite we use @xmath208 and and @xmath209 and it follows@xmath210 & = & \int d^{d}r\left ( 2mh\left [ p_{c}\right ] + \sum_{ab}h\left [ q_{ab}\right ] \right . \notag \\ & & \left . -\frac{u}{3}\sum_{abc}\ q_{ab}q_{bc}q_{ca}-up_{c}^{2}\sum_{bc}q_{bc}\right ) . \label{heffp}\end{aligned}\]]the remaining @xmath133-dependent problem is formaly similar to the original one of eqs.[ham1 ] and [ ham2 ] , but in an external field @xmath211 . using these results we find that the effective potential can be written as @xmath194 = 2\int d^{d}rh\left [ p_{c}\right ] + \widetilde{% \omega } \left [ p_{c}\right ] \label{epott}\]]with @xmath212 = -\left . \frac{\partial } { \partial m}% \int dqe^{-\beta h_{p_{c}}\left [ q\right ] } \right\vert _ { m\rightarrow 0}. \label{intf}\ ] ] we start with a homogeneous replica symmetric ansatz for @xmath213 with equal off diagonal elements , @xmath200 , and zero diagonal elements . it was demonstrated that this replica symmetric ansatz is again equivalent to one step replica symmetry breaking in the conventional replica language@xcite . we take the limit @xmath214 and find @xmath215if we extremize this with respect to @xmath200 , i.e. perform the integration with respect to @xmath200 at the level of a homogeneous saddle point approximation , it follows that @xmath216depending on the value for the overlap @xmath189 , @xmath200 is in general different from the value @xmath131 obtained within the replica approach of the previous section . if we however insert @xmath200 back into @xmath217 and determine the effective potential @xmath193 = h_{p_{c}}\left ( q\left ( p_{c}\right ) \right ) $ ] we find that it has minima @xmath218 } { \partial p_{c}}=0 $ ] for @xmath219 and for @xmath220 , with @xmath131 of eq.qstar , in complete agreement with the approach of the previous section . the value of the effective potential for the metastable minimum @xmath221 is identical to the mean configurational entropy discussed above @xmath222 = \overline{s_{c}}.\]]in order to analyze the local stability of the solution at the minimum of @xmath223 we determine the replicon eigenvalue of the problem . following ref.@xcite gives @xmath224for the lowest eigenvalue . inserting @xmath200 from eq . ( [ mfclo ] ) yields close to @xmath163 a positive eigenvalue : @xmath225 , while the replicon eigenvalue vanishes at @xmath165 as @xmath226 . this behavior is completely consistent with the results obtained in ref.cwilich89 for the random potts model . the homogeneous solution is marginal at @xmath3 and stable below @xmath3 . except for temperatures close to @xmath3 , the mean field approach is locally stable with respect to small fluctuations . the effective potential approach is not restricted to overlaps @xmath189 that minimize @xmath193 $ ] . in general @xmath227 is given by the solution of eq.[mfclo ] . inserting this solution into @xmath228 gives @xmath229 . the result is shown as broken line in fig.([fig5 ] ) . it turns out that for intermediate values of @xmath189 , between @xmath219 and @xmath230 an additional replica symmetry breaking of @xmath133 occurs@xcite . for the present model this effect was also studied in ref.@xcite . this is related to the well known effect that an external field can cause replica symmetry breaking above @xmath9@xcite , yielding one step replica symmetry breaking for @xmath133 . the result for @xmath231 shown in fig . [ fig4 ] has been obtained , including this additional replica symmetry breaking . physically this additional replica symmetry breaking as @xmath189 `` goes over the hill '' was discussed in ref.dzero05 . we also note that the generic behavior for @xmath231 as shown in fig . [ fig4 ] is very similar to @xmath232 of eq.[scmon ] , if we simply plot this function for arbitrary @xmath131 . as required both expressions exactly agree at saddle points and minima . in summary , we showed that on the level of spatially homogeneous mean field solutions , the effective potential approach of refs.franz95,franz97,barrat97 yields results in complete agreement with the physically transparent approach of ref.@xcite . since it allows for arbitrary overlap between two states , we use it as starting point for the determination of activated events within an instanton calculation . at the mean field level a glass at @xmath8 is frozen in metastable states and characterized by the overlap @xmath131 of eq.[qstar ] between configurations at distant times . for temperatures @xmath233 above the kauzmann temperature , the free energy of such a frozen state is higher by the configurational entropy @xmath234 compared to the liquid state that is characterized by @xmath151 . thus , for @xmath92 the mean field glass is locally stable ( the replicon eigenvalue @xmath235 is positive ) but globally unstable with respect to the ergodic liquid . the effective potential @xmath229 shown in fig . 3 suggests that the decay modes for the frozen state are droplet excitations , similar to the nucleation of an unstable phase close to a first order transition . this situation was analyzed in ref.@xcite . using this approach , an analysis of the effective droplet size for a kac - type model was performed in ref.@xcite in agreement with the rfot theory@xcite , the driving force for nucleation is the configurational entropy , leading to the notion of entropic droplets . instanton solutions for entropic droplets are determined from@xmath236 } { \delta p_{c}\left ( \mathbf{r}% \right ) } = 0 \label{inststat}\]]where we allow for spatial variations of the overlap @xmath237 . recently , franz has given a transparent dynamical interpretation for the solutions of eq.[inststat]@xcite . assuming that the effective potential is characterized by spatially anisotropic but replica symmetric solutions @xmath238 , we perform the integral over the @xmath239 in eq.[intf ] at the saddle point level . we find a solution @xmath240 with@xmath241where @xmath242eq.[inst1 ] admits an exact solution in the thin wall limit @xmath243@xmath244 , \label{pcthinwall}\]]where the integration constant @xmath245 is a function of @xmath246 and @xmath145 . @xmath16 is the droplet radius and @xmath247 is the interface width given by @xmath248the final expression for the function @xmath249 is more involved and is given in appendix c. ( @xmath16 is the size of the droplet ) together with the temperature dependence of interface width @xmath250 ( inset).,width=268 ] inserting the solution eq.[pcthinwall ] into the expression for the effective potential eq.[epott ] we calculate the value of the mean barrier . the latter is determined by optimizing the energy gain due to creation of a droplet and energy loss due to the surface formation . as a result for the mean barrier we find ( reintroducing the energy scale @xmath140 and length scale @xmath139 ) @xmath251the droplet radius @xmath252is determined from the balance between the interface tension and the entropic driving force for nucleation . the barrier energy @xmath253 of eq.[fdagger ] determines the relaxation time @xmath254 of eq.[meanr ] . furthermore , @xmath255 is the order parameter at the kauzmann temperature . when temperature approaches the @xmath9 the radius of the droplet as well as the mean barrier diverge . one finds @xmath256 and @xmath257 . since the droplet interface @xmath247 remains finite as @xmath258 , the thin wall approximation is well justified close to the kauzmann temperature . on the other hand , @xmath16 and @xmath259 become comparable for temperatures close to @xmath3 and the thin wall approximation breaks down . we see that replica landau functional calculation predicts a rather diffuse droplet near the laboratory @xmath260 . combining @xmath261 and @xmath262 , we obtain @xmath20 for the exponent that relates the droplet size @xmath16 and the configurational entropy density : @xmath263 . in ref . it was demonstrated that for a similar model the interface free energy is exponentially small for a large system and in any finite dimension the one step replica symmetry breaking state does not exist . as was already pointed out in ref . , our approach is very different in scope and in its conclusions . while ref . is concerned with the absence of replica symmetry breaking in the ultimate equilibrium state , our conclusions are relevant for the nonequilibrium sitation . since we obtain a finite barrier height of the frozen mean field solution , our result offers a mechanism for equilibration on time scales @xmath264 demonstrating that our approach and the conclusions of ref . are consistent . to analyze the barrier fluctuations we start from the barrier for a given density configuration @xmath265 : @xmath266where @xmath267 is determined by the constrained partition function of eq.[zprho ] and @xmath268 the corresponding energy with homogeneous overlap , i.e. @xmath269 . when we analyze variations of activation barriers we need to keep in mind that both , ground state and transition state contributions to @xmath270 are statistically fluctuating and are correlated in general . fluctuations in the first term of eq.[fbarr ] correspond to variations of the free energy of the localized instanton . on the other hand , fluctuations in the second term correspond to variations of the homogeneous background . in principle cancellations between both occur that are properly included in the analysis of @xmath271 defined above . barrier fluctuations are then characterized by the second moment @xmath272where the average is , just as for the analysis of the mean barrier , with respect to the density configuration @xmath265 . the second term in eq.bfl is the square of the mean activation barrier @xmath10 and was determined in the previous section . thus , we can concentrate on first term . using eq.[fbarr ] it follows that the first term in eq.bfl consists of three terms:@xmath273 in what follows we consider these three contributions separately . the first two terms correspond to independent fluctuations of the droplet and the homogeneous background , while the last term measures their mutual correlations . we find that this last term in eq.[ff ] is proportional to the surface area of the instanton . correlations between droplet and homogeneous background terms result from their mutual interface . the detailed analysis of the three contributions to @xmath274 of eq.[ff ] is summarized in appendix b. in what follows we summarize the results of this derivation . for the first two terms in eq.[ff ] it follows that @xmath275 \notag \\ \overline{f_{p_{c},\rho } ^{2 } } & = & \overline{f_{p_{c},\rho } } ^{2}-2t\int d^{3}% \mathbf{r}h[p_{c}\left ( \mathbf{r}\right ) ] . \label{fttb}\end{aligned}\ ] ] the first expression gives fluctuations of the configurational entropy for the homogeneous problem , while the second one describes the energy fluctuations of the configurations with spatially heterogeneous overlap . the result for the homogeneous problem , @xmath276 can alternatively be obtained from our analysis of the configurational heat capacity given in eq.[ffluctc ] . using the replica formulation of eq.repsecd together with the explicit result of eq.[f(m)mon ] yields @xmath277where @xmath4 is a volume . if we now recall the definition of @xmath278 $ ] eq.ham2 , one readily observes that the first equation in eq.[fttb ] coincides with eq . ( [ flmon ] ) . the derivation of the third contribution @xmath279 to @xmath280 in eq.[ff ] , which describes the energy correlations between the instanton and environment , is also performed in the appendix b. the result is : @xmath281\right . \notag \\ & & \left . + up_{c}^{3}(\mathbf{r})-uq^{\ast } p_{c}({\mathbf{r}})\right ) . \label{flcorr}\end{aligned}\]]combining eqs.[fttb ] with eq.[flcorr ] , we finally obtain for the second moment of the barrier fluctuations : @xmath282-h[q^{\ast } ] + up_{c}^{3}({\mathbf{r}})-uq^{\ast } p_{c}^{2}({\mathbf{r% } } ) \right ) . \label{barfl}\]]this result for the second moment of the barrier fluctuations depends on the value of the homogeneous overlap @xmath131 of eq.[qstar ] as well as the instanton solution @xmath283 . the thin wall limit result for @xmath283 is given in eq.[pcthinwall ] . inserting these expressions into the eq.[barfl ] yields @xmath284where @xmath16 is the radius of the droplet . in the thin - wall approximation , the explicit expressions for the coefficient @xmath285 is @xmath286 , \ ] ] where the value of @xmath287 is computed using an effective potential @xmath288 , i.e. it is the turning point of the intstanton , see fig.3 . deriving ( [ at ] ) we also made the following choice for the parameters @xmath289 ( see appendix a for more details ) . this choice is of particular convenience as it allows one to express the resulting expressions in terms of the ratio @xmath290 and @xmath143 . the length scale @xmath291 can be compactly as follows:@xmath292 , \end{split}\]]where @xmath293 are functions of the ratio @xmath290 only and are defined via @xmath294dz\ ] ] we have computed these integrals numerically for @xmath295 and @xmath296 . temperature dependence of the surface length scale @xmath297 is shown on fig . [ fig5 ] . before we proceed with the analysis of our results , it is instructive to write down the expressions for the coefficient @xmath285 and @xmath297 for temperatures close to @xmath298 . when @xmath170 , one readily finds @xmath299 , so that expanding expression in the square brackets ( [ at ] ) up to the linear order in @xmath300 we have @xmath301+{\cal o}((t - t_k)^2)$ ] with @xmath302 where once again we employed @xmath303 ( see appendix a ) . thus we find that close to @xmath298 the bulk contribution to the barrier fluctuations is governed solely by the value of the phenomenological parameter @xmath143 . expression for @xmath304 can be derived straightforwardly using our results above together with the expansion for the functions @xmath293 and @xmath305 . here we provide the expression for the value of @xmath306 : @xmath307,\ ] ] where we have used the notations @xmath308 . we have found that the coefficients @xmath309 have the following values : @xmath310 , @xmath311 and @xmath312 . describing the contribution to the barrier fluctuations due to surface energy fluctuations of the structural droplets . we also show the temperature dependence of the ratio @xmath313 ( @xmath16 is the size of the droplet).,width=268 ] ( bottom inset ) . the latter describes the contribution to the barrier fluctuations due to surface energy fluctuations of the structural droplets.,width=268 ] governed by the fluctuations of the isolated droplets together with exponent @xmath314 corrected by the effects of the droplet - droplet interactions.,width=268 ] thus , in addition to the bulk term , which behaves in a similar way as the formulation in ref.@xcite ( see eq.[xw ] ) , the replica theory yields a surface term , resulting from fluctuations in the interface . in the inset of fig.[fig5 ] we show the @xmath146-dependence of @xmath167the length scale @xmath291 , demonstrating that the surface term becomes gradually more important as the temperature increases . the value of @xmath315 remains non - singlar , so that the volume term is dominant . this is consistent with the view that the droplet interface close to @xmath9 is smaller then the droplet radius , while both are comparable as @xmath146 approaches @xmath3 , similar to the behavior close to a spinodal@xcite . the resulting temperature dependence of @xmath297 is shown on fig . [ fig5 ] , @xmath51 is shown in fig . [ fig6 ] and the exponent of the streched exponential relaxation @xmath316 shown in fig . [ fig7 ] . the plots are constructed by setting the parameters of the theory to re - produce experimentally relevant values of @xmath9 , @xmath1 and @xmath317 as well as configurational entropy above the glass transition using @xmath318terphenyl as the example of the glass transition ( see appendix a for details ) . there is one free parameter left to be fixed , either @xmath145 or @xmath319 . we have determined the value of @xmath145 in the comparison with the experimental value of the mean barrier at @xmath320 given by @xmath321 with the theoretically derived expression ( [ fdagger ] ) . this procedure yields @xmath322 , @xmath323 and @xmath324 . as we can see from fig . [ fig7 ] , @xmath325 , a value significantly reduced compared to the exponential behavior where @xmath326 . this value for @xmath316 is lower than the experimentally found values of @xmath327 . this discrepancy may signal that the present approach overestimates the strength of the energy barrier fluctuations an effect that may be related with non - gaussian fluctuations of the activation barriers . it is however clear that these may be described as instanton interaction effects as discussed by xia and wolynes@xcite . our analysis yields a gaussian distribution of barriers@xmath328which yields the following estimatefor the exponent of the streched exponential relaxation @xmath329our results for the temperature dependence of @xmath330 are shown in fig.[fig7 ] . the calculation that led to this result was based on the assumption of single instanton events , i.e. entropic droplets were assumed to be diluted . obviously this is not addressing the fact that distict droplets interact . the mosaic picture underlying the rfot theory is clearly based upon the view that such droplet - droplet interactions occur and are in fact crucial . the impact of droplet - droplet interactions for the distribution of barriers was analyzed in ref.@xcite . here it was pointed out that , as soon as the droplet size becomes larger than the size @xmath16 of a mean droplet ( eq.55 in our theory ) , boundary effects will limit the size of a droplet , leading to a cut off of the distribution function . this leads to the modified distribution function @xmath331which leads to a reduction of the mean square deviation of the barriers by a factor @xmath332 , compared to the gaussian distribution . this corrects the exponent of the streched exponential relaxation to@xmath333where @xmath51 is still the result , eq.63 , of the dilute droplet calculation . in fig.[fig7 ] we compare @xmath330 and @xmath52 . obviously , the reduction of the barrier fluctuation width due to droplet - droplet coupling leads to an increase of the exponent @xmath52 . we mention however the resulting value of @xmath314 at the glass transition is @xmath334 . this suggests the present landau functional indeed over - estimates the fluctuation effects . this probably implies that the diffuseness of the droplet from the landau functional is also overestimated . the fluctuations do vanish at @xmath335 , consistent with the emergence of simple exponential dynamics at a significantly high temperature . we did not attempt to vary the parameters of our model to reach perfect agreement with experiments for otp , as the simplified landau functional is clearly an oversimplification of the physics of real glasses . it does however demonstrate the generic and qualitative trends for barrier fluctuations in the rfot theory . finally , in fig . [ fig8 ] we show the dielectric finction @xmath336 as finction of frequency for different temperatures . the peak in @xmath337 dramatically broadens as temperature approaches the glass transition temperature , @xmath1 , as predicted from the static barrier distributions . for this figure we used the second moment @xmath51 of the barrier distribution function and then assumed that @xmath338 is gaussian . in case of a gaussian distribution , all moments can be expressed in terms of the second moment @xmath339where @xmath340 for @xmath341 odd and @xmath342 for @xmath341 even . in particular it holds for higher order moments @xmath343 , @xmath344 , @xmath345 , and @xmath346 etc . we can explicitly analyze higher moments of this distribution and thereby demonstrate explicitly that @xmath34 is gaussian . to be precise , we have only shown this for the first six moments but strongly suspect this to be true in general . the calculation of higher moments are quite cumbersome but they formally are a straightforward generalization of the method we used to determine the second moment @xmath51 . in case of the @xmath341-th moment the block structure of the matrix @xmath206 of eq.[eq8 ] consists of @xmath347 blocks where the total dimension of @xmath206 is @xmath348 . the rather tedious matrix algebra is then most easily analyzed using computer algebra software , such as mathematica . thus , we only list the results here . we analyzed all moments@xmath349up to @xmath350 of . we find @xmath351 in case of the fourth moment we analyze:@xmath352-% \overline{{f}^{\ddagger } } ) ^{4 } } \notag \\ & = & \overline{f^{\ddagger 4}[\rho _ { 2}]}-4\overline{f^{\ddagger 3}[\rho _ { 2}]}% \overline{{f}^{\ddagger } } + 3\overline{{f}^{\ddagger } } + 6\overline{{f}% ^{\ddagger } } \cdot \overline{{f}^{\ddagger 2}},\end{aligned}\]]it follows @xmath353 -% \mathcal{h}_{inh}\left [ p_{c}\right ] -\mathcal{h}_{hom}(q^{\ast } ) ] ^{2}\]]where @xmath354 and @xmath355 are defined below eq . evaluating this expression yields @xmath356we further find that the fifth moment vanishes as well@xmath357and that the sixth moment is given as @xmath358thus , up to the sixth moment the barrier distribution of droplets is gaussian . thus , dilute entropic droplets should indeed have a gaussian distribution of barriers , strongly suggesting that the observed non - gaussian behavior of the effective barrier distribution results from droplet - droplet interactions . in summary , we have shown how the energy barrier fluctuations @xmath359 as well as higher moments of the static free energy barrier distribution function @xmath34 can be computed using the replica landau theory of ref.@xcite and use the replica instanton theory of refs.@xcite . we have generalized the replica formalism to determine barrier fluctuations in addition to the determination of the mean barrier @xmath10 . barrier fluctuations @xmath51 consist of a dominating bulk contribution at temperatures close to the kauzmann temperature along with a contribution stemming from fluctuations in the interface . the latter dominate the barrier fluctuations well above the kauzmann temperature . fluctuations of the surface energy enter prominently in the recent work of biroli et al . @xcite on directly computing point - to - set correlations in liquids . it is very nice to see that the effect they found naturally emerges from the replica instanton calculation framework . it is interesting to speculate that these surface energy fluctuations could account for the observed system specific deviations from the xia - wolynes prediction of a direct relation between the stretching exponent and @xmath360 . chemical trends in these deviations may provide clues on this . @xmath51 increases as the typical droplet size increases , i.e. as the temperature is lowered towards @xmath9 . clearly , our theory only applies until the system falls out of equilibrium at the laboratory glass temperature @xmath1 with @xmath361 . in the framework of the present instanton approach , the energy barrier distribution function turns out to be gaussian . the skewness of the observed relaxation time distribution is , in our view , an effect due to the interaction of spatially overlapping instantons . it can be captured , at least in spirit , by an extended mode coupling approach , incorporating instantons . this research was supported by the ames laboratory , operated for the u.s . department of energy by iowa state university under contract no . de - ac02 - 07ch11358 ( m.d . and j. s. ) , a fellowship of the institute for complex adaptive matter ( m.d . ) , and the national science foundation grants che-0317017 ( p. g. w. ) and dmr 0605935 ( m.d . ) the authors also thank the aspen center for physics , where part of this work was performed . the landau functional is quite convenient for formal use but there are variety of ways that it may be mapped onto real fluids . in this appendix we show how the model parameters of the landau replica potential , eqs.[ham1 ] and [ ham2 ] , can be found by fitting the results of the mean field theory to experimental data . we use the concrete example o - terphenyl ( otp ) , a well known glassforming material . values for the effective dynamical freezing temperature and the kauzmann temperature of otp are @xmath362@xcite and @xmath363@xcite . on the other hand , a typical value for the glass temperature is@xcite @xmath364 here , the glass transition temperature is the temperature at which the viscosity reaches a value of @xmath365 , which is also where the mean relaxation time reaches values of @xmath366 . the melting temperature of otp is @xmath367 with an entropy jump at melting of @xmath368 . the configurational entropy at @xmath1 and @xmath3 are @xmath369 and @xmath370 , respecticely@xcite , where @xmath371 is the molar gas constant . to obtain the entropy per spherical object or bead one has to divide these results by @xmath372 , the number of beads as defined by the procedure in ref.@xcite . this yields @xmath373 as well ws @xmath374 . the heat capacity change at the glass transition is@xcite @xmath375where we used @xmath376 . the specific and molar volume of the system are@xcite : @xmath377the typical volume per particle @xmath378 is then characterized by the length @xmath379 . it follows @xmath380 , such that @xmath381 . in comparison , the van der waals radius of otp was given as@xcite @xmath382 . if we introduce the energy and length scales into the problem , the mean field theory results obtained from the hamiltonian eqs.[ham1 ] and [ ham2 ] are @xmath383the order parameter at @xmath9 should be large of order unity . we chose @xmath384 which gives @xmath385 . using the relation @xmath386 this yields @xmath387if we further make the reasonable choice of @xmath388for the length scale @xmath139 , @xmath389 of eq.[expr ] gives precisely the experimental value listed above . we can also determine the change in heat capacity @xmath390 from the same parameters . . [ ew ] yields the result that @xmath391so that we obtain for the temperature dependence of the configurational entropy : @xmath392where the mean field result for the slope is @xmath393 . this compares reasonably well with the experimenal value @xmath394 . note , however , that @xmath49 as given by eq.[dcp ] is different from the configurational heat capacity @xmath395 , eq . ( [ confheatcap ] ) . it holds @xmath396 , i.e. @xmath397 , i.e. both are comparable provided @xmath398 , a value we will assume in what follows thus ensuring thermodynamic consistency from the two fluctuation formulas . in summary , the existing thermodynamic measurements can determine several parameters and allow a consistent description of several thermodynamic data . still , one free parameter , either @xmath144 or @xmath145 remains since only the ratio @xmath144 to @xmath145 is determined . we use this freedom to obtain the correct value of the mean barrier at the glass temperature . finally , we note that the above result also imply that @xmath399 i.e. about half of @xmath400 . it is worth mentioning that the values of the landau parameters are consistent with the random first order framework as assumed . .parameters for the landau functional [ cols="^,^,^,^,^,^ " , ] [ summary ] finally , we summarize our results for the model parameters by fitting the experimentally relevant quantities to the data obtained for otp in table [ summary ] . in this appendix we summarize the derivation of the various contributions to the barrier fluctuation given in eq.[ff ] . we start our calculation with the third term in eq.[ff ] . the other two terms are simpler to determine and can be obtained as specific limits of the third one . it holds : @xmath401}\log { z}_{p_{c},\rho } \log { z}% _ { q^{\ast } , \rho } } { \int \mathcal{d}\rho { e}^{-\beta { h}[\rho ] } } \notag \\ & = & \lim\limits_{m_{1,2}\rightarrow { 0}}\left ( \mathcal{y}_{12}-\mathcal{y}% _ { 1}-\mathcal{y}_{2}\right ) . \label{eq5}\end{aligned}\]]where we introduced @xmath402}z_{p_{c},\rho } ^{m_{1}}z_{q^{\ast } , \rho } ^{m_{2}}+z}{m_{1}m_{2}z}\ \notag \\ \mathcal{y}_{1 } & = & \ \frac{t^{2}\int \mathcal{d}\rho { e}^{-\beta { h}[\rho ] } z_{p_{c},\rho } ^{m_{1}}}{m_{1}m_{2}z } \notag \\ \mathcal{y}_{2 } & = & \ \frac{t^{2}\int \mathcal{d}\rho { e}^{-\beta { h}[\rho ] } z_{q^{\ast } , \rho } ^{m_{2}}}{m_{1}m_{2}\int \mathcal{d}\rho { e}^{-\beta { h}% [ \rho ] } } \end{aligned}\]]where @xmath403}$ ] . to analyze the first term in eq.[eq5 ] we use again the replica trick and write : @xmath404}z_{p_{c},\rho } ^{m1}z_{q^{\ast } , \rho } ^{m_{2}}+z% } { m_{1}m_{2}\left ( \int \mathcal{d}\varphi { e}^{-\beta { h}[\varphi ] } \right ) ^{n-1 } } \notag \\ & = & \lim\limits_{m_{1},m_{2}\rightarrow { 0}}\lim\limits_{n\rightarrow { 0}}% \frac{\mathcal{i}_{p_{c};q^{\ast } } ^{(n;m_{1},m_{2})}}{m_{1}m_{2 } } \label{eq6}\end{aligned}\]]here we introduced a quantity @xmath405 that can be determined from an @xmath406 replicated problem with order parameter @xmath407 : @xmath408it holds @xmath409}\prod\limits_{{\mathbf{x},}a}\delta \left ( p_{c}(% \mathbf{x})-p_{1a}({\mathbf{x}})\right ) \notag \\ & & \times \prod\limits_{{\mathbf{x},}\alpha } \delta \left ( q_{0}-q_{1\alpha } ( % { \mathbf{x}})\right ) , \end{aligned}\]]@xmath410 $ ] has the same form as eqs.[ham1 ] and [ ham2 ] if one uses the matrix @xmath407 of eq.[eq8 ] instead of the original replica variable @xmath200 . the sub - matrices of @xmath407 are given as follows : @xmath198 is an @xmath411 matrix , @xmath203 is an @xmath412 matrix , @xmath413 is an @xmath414 matrix , @xmath143 is an @xmath415 matrix , @xmath416 is an @xmath417 matrix , and @xmath418 is an @xmath419 matrix . the matrices @xmath203 and @xmath413 obey the additional constraints : @xmath420which is enforced through the @xmath421-function above . we assume for a replica symmetric instanton solution that and the elements of @xmath422 are all the same : @xmath423similary we assume @xmath424 . the analysis of @xmath410 $ ] tedious but straightforward and yields that it can be written as a sum of three terms:@xmath425 & = & m_{1}\mathcal{h}_{inh}[p_{c};u]+m_{2}\mathcal{h}% _ { hom}[q_{0};v ] \notag \\ & & + m_{1}m_{2}\mathcal{v}[\psi ] . \label{hhh}\end{aligned}\]]here @xmath426 & = & \int d^{3}\mathbf{x}\left\ { 2h\left [ q^{\ast } \right ] + \left ( m_{2}-1\right ) h\left [ v\right ] \right . \\ & & \left . -\left ( m_{2}-1\right ) vq^{\ast 2}-\frac{1}{3}\left ( m_{2}-1\right ) v^{3}\right\}\end{aligned}\]]refers to the homogeneous problem . in addition , @xmath427 & = & \int { d^{3}\mathbf{x}}\left\ { 2h[p_{c}]\ + ( m_{1}-1)h[u]\right . \\ & & -(m_{1}-1)u(\mathbf{x})p_{c}^{2}(\mathbf{x } ) \\ & & \ \left . -\frac{1}{3}(m_{1}-1)(m_{1}-2)u^{3}(\mathbf{x})\right\}\end{aligned}\ ] ] is the contribution of the inhomogeneous instanton . finally , the coupling term between the two is given as @xmath428 & = & \int { d^{3}\mathbf{x}}\{2h[\psi ] + ( m_{1}-1)u_{0}(% \mathbf{x } ) \\ & & + ( m_{2}-1)v]\psi ^{2}(\mathbf{x})-2p_{c}(\mathbf{x})q^{\ast } \psi ( \mathbf{% x})\}.\end{aligned}\]]in order to determine the various matrix elements of @xmath407 defined in eq.[eq8 ] we perform saddle point approximations with respect to the numerous variables . we find @xmath429 which has been calculated earlier . similarly it follows @xmath430 . the saddle point equation for @xmath418 is then @xmath431\psi ( \mathbf{x } ) \notag\end{aligned}\]]iand it follows that the solution of this equation is @xmath432 . with these results we are in the position to calculate the barrier fluctuations . first we re - write expression eq.(b1 ) by employing the saddle - point solution as follows : @xmath433}\right . \notag \\ & & \left . -e^{-m_{1}\beta \mathcal{h}_{inh}[p_{c};r_{0},r_{1}]}+\right . \notag \\ & & \left . e^{-m_{1}\beta \mathcal{h}_{inh}[p_{c};r_{0},r_{1}]-m_{2}\beta \mathcal{h}_{hom}[q^{\ast } , q^{\ast } ] -m_{1}m_{2}\beta \mathcal{v}% [ p_{c}]}\right ) . \label{cross}\end{aligned}\]]expanding the exponents up to the second order in @xmath434 s and taking @xmath435 yields the expression ( 57 ) in the text . equations ( 55 ) can be obtained using the same steps which lead us to ( [ cross ] ) . for example , we have : @xmath436}% \right . \notag \\ & & \left . -e^{-m_{1}\beta \mathcal{h}_{inh}[p_{c};r_{0},r_{1}]}\right . \notag \\ & & \left . e^{-(m_{1}+m_{2})\beta \mathcal{h}% _ { inh}[p_{c};r_{0},r_{1}]-m_{1}m_{2}\beta \mathcal{v}_{inh}[p_{c}]}\right ) , \end{aligned}\]]where we introduced @xmath437=\int d^{3}{\mathbf{x}}h[p_{c}]$ ] . upon taking the limit @xmath438 and using the replica symmetry @xmath439 we recover the second expression in eq.[ff ] * * in this section we provide an expression for the integration constant @xmath440 which enters into eq . ( [ pcthinwall ] ) for the interface profile . first , we define the following functions : @xmath441 , ~\varphi(t)=\sqrt{\frac{3q_m^2(t)-1}{2 } } \end{split}\ ] ] in terms of these functions the expression for @xmath249 reads : @xmath442.\ ] ] finally , we remark that the integral , which determines the length scale @xmath297 ( [ rho ] ) can be computed exactly using relations above . the resulting expression is quite cumbersome and will not be listed here .

we study the origin of absorption features on the blue side of the broad emission line of the large - separation lensed quasar sdss j1029 + 2623 at.197 . the quasar images , produced by a foreground cluster of galaxies , have a maximum separation angle of 22.5 . the large angular separation suggests that the sight - lines to the quasar central source can go through different regions of outflowing winds from the accretion disk of the quasar , providing a unique opportunity to study the structure of outflows from the accretion disk , a key ingredient for the evolution of quasars as well as for galaxy formation and evolution . based on medium- and high - resolution spectroscopy of the two brightest images conducted at the subaru telescope , we find that each image has different intrinsic levels of absorptions , which can be attributed either to variability of absorption features over the time delay between the lensed images , 744 days , or to the fine structure of quasar outflows probed by the multiple sight - lines toward the quasar . while both these scenarios are consistent with the current data , we argue that they can be distinguished with additional spectroscopic monitoring observations .

1212.6689

quasars are routinely used as background sources to study the gaseous phases of intervening objects via absorption - line diagnostics . these absorption lines have their origins both in _ intervening _ objects ( i.e. , foreground galaxies , inter - galactic medium ( igm ) , quasar host galaxies ) and in sources that are _ intrinsic _ to the quasars . one of the most promising candidates for the intrinsic absorbers is an outflowing wind from the accretion disk of quasar central engines . the outflow is accelerated by magnetocentrifugal forces @xcite , radiation pressure in lines and continuum @xcite , and/or by thermal pressure force ( e.g. , @xcite ) . the outflowing winds play an important role in three ways : 1 ) the extraction of angular momenta from disks allows accretions to proceed ( e.g. , @xcite ) , leading to growth of black holes , 2 ) the disk outflow also provides energy and momentum feedback to interstellar media of host galaxies and to intergalactic media ( igm ) , and inhibits star formation activity ( e.g. , @xcite ) , 3 ) outflowing winds may induce the metal enrichment of the igm ( e.g. , @xcite ) . thus , the physical conditions of the outflow not only promote the evolution of quasars themselves , but greatly impact the surrounding environments . among absorption lines , broad features ( hereafter , bals ; fwhm @xmath4 2000 ) are easily identified as being intrinsic , because it is almost impossible for foreground objects to produce very broad and smooth line profiles . bals are detected in about 10 20% of optically selected quasars ( e.g. , @xcite ) , and their detection rate is slightly higher in radio - quiet quasars ( e.g. , @xcite ) . bals are thought to originate in the outflowing winds when our sight - line intersects this component . this idea is supported by the fact that there are no significant differences in the properties of quasars with bals ( bal qsos ) and those without bals ( non - bal qsos ) @xcite . moreover , quasar spectra tend to be redder when bal profiles , especially those with low ionization absorption lines ( i.e. , lobals and felobals ) , are observed , which is probably caused by dust reddening in the outflows ( e.g. , @xcite ) . thus , quasars with and without bals may intrinsically be a single population , although the evolutionary phase of quasars could also affect the detection rate of bals ( e.g. , @xcite ) . the only difference is whether the sight - line passes through the outflowing wind or not in this orientation scheme . thus , bal qsos have traditionally been the best targets for the study of outflows ( e.g. , @xcite ) . in addition to bals , a fraction of narrow absorption lines ( hereafter nals ; fwhm @xmath5 500 ) and mini - bals ( an intermediate category between bals and nals ) have also been suggested to be physically associated with quasars . however , the origin of nals and mini - bals are still under debate . each of the above classes of absorption features could represent either different lines of sight through the outflowing wind to the quasar continuum source or different stages in the evolution of the absorbing gas parcels ( e.g. , @xcite ) . the observed fraction of optically - selected quasars hosting bals ( @xmath01020 ) and nals ( @xmath02050% ; e.g. , @xcite ) constrains the solid angle subtended by the dense portion of the wind to the central engine . however , such a statistical treatment requires the assumption that all quasars are identical . this remains an assumption because in past studies , it has only been possible to trace _ single _ sight - lines ( i.e. , probe a one dimension profile ) toward the nucleus of each quasar . gravitationally lensed images of background quasars are our only tools for the study of the three - dimensional geometry of the absorbers ( e.g. , @xcite ) . indeed , this technique has frequently been applied for investigating intervening absorption lines , produced by galaxies or igm on scales of up to several kilo - parsecs . however , these traditional quasar lenses are not useful for multi - angle studies of the agns themselves because their separation angles ( @xmath1 @xmath0 a few arcsec ) are too small to separate internal structures in the vicinity of the quasars . on the other hand , @xcite discovered a lensed quasar , sdss j1004 + 4112 , with a maximum image separation of @xmath1 @xmath0 [email protected] , which is the first example of a quasar lensed by a cluster of galaxies rather than a single massive galaxy . if there exists outflow gas at a distance of @xmath01 kpc from the continuum source as measured for some specific quasars @xcite , the large separation angle of sdss j1004 + 4112 translates into a physical distance of @xmath60.1 pc , with which we may trace outflowing winds with different physical properties unless their transverse sizes are much larger than sub - parsec scale . taking advantage of this large image separation , @xcite proposed that the differences in emission lines between the lensed images seen in their follow - up spectra can be explained by differential absorptions along each sight - line , although the spectra do not show explicit intrinsic absorption features . the second large - separation lensed quasar sdss j1029 + 2623 @xcite has an image separation of @xmath1 @xmath0 [email protected] , and therefore is the largest quasar lens currently known . @xcite identified the third image of this lens system ( see figure [ f1 ] ) . as shown in figure [ f2 ] , all the spectra of quasar images a , b , and c have absorption features on the blue side of the emission lines , which implies that each sight - line passes through the outflowing wind . in what follows , we call this feature `` the associated absorption line '' . moreover , we see a clear difference in the absorption features between image a and images b / c . the former has a broader and shallower profile and an additional redshifted absorption line at @xmath04990 , while the latter two have narrower and deeper profiles ( figure [ f2 ] ) . after correcting for magnification , the eddington ratio of the quasar is estimated to be @xmath7 @xmath0 0.11 . this value is consistent with the typical eddington ratio of bright quasars , @xmath7 @xmath0 0.07 1.6 @xcite . in this paper , we discuss the origin of these differential absorption profiles in sdss j1029 + 2623 . the likely scenarios include ( i ) time - variation of a single absorber covering both sight - lines toward the quasar between the time - delay of images a and b / c , which is @xmath0744 days in the observed frame according to monitoring observations by @xcite ( hereafter , * scenario i * ) , ( ii ) each sight - line penetrating different absorbers or different regions of a single absorber toward the quasar as proposed by @xcite ( * scenario ii * ) , and ( iii ) micro - lensing ( * scenario iii * ) . among these , the last scenario has been already rejected because each image shows common ratios between radio , optical , and x - ray fluxes @xcite , which is not expected for micro - lensing . thus we concentrate on the viability of scenarios i and ii , based on our medium- and high - resolution spectroscopic observations of sdss j1029 + 2623 . the structure of this paper is as follows . in 2 , we describe the observations and data reduction . we present results in 3 , and discuss the results in 4 . we summarize our results in 5 . we adopt @xmath8 as the emission redshift of the quasar , which was estimated using broad uv emission lines by @xcite . time intervals between observations are given in the observed frame throughout the paper , unless otherwise noted . we conducted spectroscopic observation of the two brightest lensed images ( i.e. , images a and b ) of sdss j1029 + 2623 with subaru / focas on 2009 april 4 , using a similar resolution power to previous observations ( @xmath9 @xmath0 500 ) taken about 23 years before . this time separation is comparable to the measured time delay between image a and images b / c ( @xmath0744 days ; * ? ? ? we used the grating with l600 filter to cover the wavelength range of @xmath10 = 3700 6000 . we also used the @xmath11 slit without pixel sampling ( @xmath01.34 per pixel ) , which enables us to compare our spectra with the previous observations from the literature @xcite including sdss @xcite by performing convolution where needed . the sky conditions were moderate although affected by thin clouds . the total integration time was 1200 s for each image , and the data quality of the final spectra are s / n @xmath08.5 pix@xmath12 and @xmath07.0 pix@xmath12 for images a and b , respectively . we also obtained high - resolution spectra of images a and b with subaru / hds on 2010 february 10 to resolve the associated absorption line completely and extract physical parameters such as column density , line width , ejection velocity from the quasar , and the covering factor ( the fraction of flux from the background source that is geometrically covered by the foreground absorber ) using voigt profile fitting . we used the standard setup , std - bc , with a @xmath13 slit width ( @xmath9=30,000 ) , covering the wavelength range of 33904210 on the blue ccd and 42805110 on the red ccd . this configuration covers ly@xmath14 , , as well as the absorption line at @xmath0 . the ccd is binned every 4 pixels in both spatial and dispersion directions ( i.e. , @xmath00.05 per pixel and each resolution element contains three pixels around 4500 ) . the total integration times were 14400 s and 14200 s for images a and b , respectively , and the final s / n ratios are about 13 pix@xmath12 for both the images . we reduced both the focas and hds data in a standard manner with the software iraf . as for the hds spectra , we divided them by a flat frame before normalization in order to remove the instrumental blaze pattern . wavelength calibration was performed using a th - ar lamp . the observation log is summarized in table [ t1 ] . we examined the suite of subaru / focas spectra to search for variability in the lines . if absorption in both the images show time - variations , it would imply that the profiles are variable and that the scenario i is more favorable . if this is the case , we have monitored a single absorber in five epochs on 2006 february 28 + 744 days , 2006 june 29 + 744 days , 2009 april 4 + 744 days ( toward image a ) , and on 2007 december 15 , 2009 april 4 ( toward image b ) . if neither images show time variation , the scenario ii will instead be favored . the associated absorption line of the image a was spectroscopically observed three times on 2006 february 28 @xcite , on 2006 june 29 @xcite , and on 2009 april 4 ( this paper ) , as shown in figure [ f2 ] . the first one was part of the sdss , while the others were taken with subaru / focas . for inter - comparison , we adjust the spectral resolution of all observations to that lowest available ( i.e. , @xmath9 @xmath0 500 ) by resampling and convolution . the emission and absorption features show a so - called p - cygni profile , which implies that the absorption feature is physically associated to the quasar itself . an additional weak absorption feature is seen on the red side of the emission line at @xmath10 @xmath0 4990 . however , this is not , but @xmath102600 at @xmath0 0.9187 , because there exists a corresponding doublet at the same redshift . we do not see any substantial variation of the absorption profile ( see figure [ f2 ] ) . the associated absorption line of image b was spectroscopically observed only twice : once on 2007 december 15 with keck / lris by @xcite , and once on 2009 april 4 by us with subaru / focas . @xcite also obtained a spectrum of image c. images b and c show profiles that are almost identical near the absorption / emission lines , which is probably because the separation angle between them , @xmath15 @xmath0 [email protected] , is much smaller than the separations between images a and b ( @xmath16 @xmath0 [email protected] ) and images a and c ( @xmath17 @xmath0 [email protected] ) . our focas spectrum of image b is consistent with those of images b and c taken about 15 months ( @xmath04.7 month in the rest - frame ) before . under scenario i , time - variation is to be expected . instead , we confirm that the absorption profile of each image is stable ( see figure [ f2 ] ) , suggesting that scenario i is less likely to be true . however , this conclusion based on our medium - resoluton spectroscopy will have to be modified significantly , following investigation of the high - resolution spectra , as we will see in the next subsection . from our high - resolution ( @xmath9 = 30,000 ) spectra of images a and b , we find that the associated absorption line consists of multiple narrower components ( figure [ f3 ] ) . thus , the system is not a bal ( or a mini - bal ) but nal , which means that the features are not necessarily arising at the outflow . if the corresponding absorber is the ism of the host galaxy , igm , or a foreground galaxy ( i.e. , intervening absorption ) , both the differential absorption profiles and the lack of time - variation toward images a and b are reasonable because their sight lines pass through completely different regions separated by @xmath0kpc or @xmath0mpc scales with respect to each other . in order to classify the associated absorption feature as an intrinsic or intervening feature , we performed a voigt profile fitting to measure line parameters . we used the line fitting software package minfit @xcite , with which we can fit absorption profiles using redshift ( @xmath18 ) , column density ( @xmath19 in ) , doppler parameter ( @xmath20 in ) , and covering factor ( , defined later in 3.2.1 ) as free parameters . the revised version of the code can be applied to self - blending features , because it fits profiles to the blue and red members of doublets _ simultaneously _ , by multiplying the contributions from the two doublet members @xcite . the fitting results are summarized in table [ t2 ] . column ( 1 ) is a line identification ( i d ) number . columns ( 2 ) and ( 3 ) give the absorption redshift and the ejection velocity ( negative corresponds to blueshifted ) from the quasar emission redshift , @xmath02.197 . columns ( 4 ) and ( 5 ) are the column density and its 1@xmath21 error , columns ( 6 ) and ( 7 ) , the doppler parameter and its 1@xmath21 error , and columns ( 8) , ( 9 ) , and ( 10 ) , the covering factors and its upper / lower 1@xmath21 error . we also applied line fitting to the doublet . the doublet can not be fitted in both the images , because the red member of the doublet , @xmath101403 , is severely affected by line contamination . covering factor , , is the fraction of coverage of the absorber over the continuum source and the broad emission line region ( belr ) along the sight - line . the covering factor can be systematically evaluated in an unbiased manner by considering the optical depth ratio of resonant , rest - frame uv doublets of lithium - like species ( e.g. , , , and ) , using the equation @xmath22 where @xmath23 and @xmath24 are the residual ( i.e. , unabsorbed ) fluxes of the blue and red members of the doublets in the continuum normalized spectrum ( e.g. , @xcite ) . a value less than unity indicates that a portion of the background source is not occulted by the absorber . this , in turn , means that the doublet is probably produced by an intrinsic absorber because intervening absorbers like igm , ism , and foreground galaxies usually have physical scales with a few kilo - parsecs much larger than the background sources ( e.g. , @xcite ) . in general , the continuum source and the belr have different covering factors , as discussed in @xcite . the derived coverage fractions can be significantly affected by uncertainty of the continuum level , providing unphysical values ( i.e. , @xmath25 0 or 1 @xmath25 ) , especially for very weak components whose values are close to 1 . therefore , if minfit gives unphysical values of for some components in the first trial , we rerun the fit assuming = 1 , following the procedure detailed in @xcite . the fitting results for and doublets are shown in figures [ f4 ] and [ f5 ] , and fitting parameters are summarized in table [ t2 ] . in our high - resolution spectrum , the associated absorption line is deblended into at least 16 narrow components at @xmath0 @xmath261600 94 from the quasar emission redshift at @xmath0 2.197 , as shown in the @xmath101548 panel of figures [ f3 ] and [ f4 ] . among the 16 components , all have corresponding ly@xmath14 absorption , while only components at @xmath4 @xmath26846 have corresponding doublets . those at @xmath0 @xmath261400 have only subtle doublets . it is difficult to tell whether doublets corresponding to the system exist or not because the @xmath101394 blends with @xmath101527 at = 1.9115 and 1.9141 , and because the @xmath101403 is also severely contaminated by doublets at = 1.8909 1.9141 . the absorption component at @xmath0 800 in the @xmath101548 panel is not but instead @xmath102344 at @xmath0 0.9187 . we also confirm that a strong absorption feature at @xmath10 @xmath0 4900 only seen in image a of figure [ f2 ] is @xmath102600 at the same redshift . it follows that the differential absorption profile of the associated absorption line between images a and b seen in low - resolution spectra ( figure [ f2 ] ) is mainly due to these physically unrelated lines . the associated line is deblended into 12 narrow components at @xmath0 @xmath261467 9.4 from the quasar emission redshift as shown in the @xmath101548 panel of figures [ f3 ] and [ f4 ] . their general profile is similar to that of the image a. among the 12 components , all have corresponding ly@xmath14 , while only those at @xmath27 @xmath26789 have corresponding doublets . on the other hand , only two components at @xmath0 @xmath261467 and @xmath261401 have doublets . the ionization condition of the components at @xmath27 @xmath26789 is clearly higher than those at @xmath0 @xmath261467 and @xmath261401 . the @xmath101403 lines blend with a clustering of lines at @xmath0 1.90 . absorption features at @xmath0 @xmath26700 in the @xmath101394 panel and at @xmath0 200 in the @xmath101243 panel are @xmath101527 at = 1.9118 and @xmath101335 at = 1.9788 , respectively . because our spectra cover a wide range of wavelength , we also identify all , , and doublets from 3800 4210 on blue ccd chip and from 4280 5110 on red ccd chip , avoiding some unusable regions ( 4155 4160 and 4620 4740 ) , ly@xmath14 forest , and bad ccd columns . toward image a , we identify 11 doublets at = 1.6151 , 1.8909 , 1.8956 , 1.8983 , 1.9115 , 1.9133 , 1.9141 , 1.9322 , 2.1084 , 2.1270 , 2.1350 and 2 doublets at = 1.9115 and 1.9138 with @xmath4 5@xmath21 detection in addition to the associated system . we also find one system identified with @xmath102344 , @xmath102374 , @xmath102383 , @xmath102587 , @xmath102600 at @xmath0 0.9187 as described above . toward image b , we identified 15 doublets at = 1.6148 , 1.6556 , 1.6925 , 1.7065 , 1.8910 , 1.8950 , 1.8972 , 1.9018 , 1.9118 , 1.9322 , 1.9788 , 2.1078 , 2.1085 , 2.1210 , 2.1284 and 4 doublets at = 1.8941 , 1.8949 , 1.8975 , 1.9018 with @xmath4 5@xmath21 detection in addition to the associated system . in this section , we discuss the origin ( i.e. , intrinsic or intervening ) of the associated absorption lines ( [ sec4.1 ] ) , possible geometries toward images a and b ( [ sec4.2 ] ) , and models of single / multiple sight - line(s ) ( [ sec4.3 ] ) , respectively . while bals have high probability of being physically associated to the quasars , nals are difficult to classify as intrinsic or intervening . with our high - resolution spectra , the associated absorption line is deblended into multiple narrow components . as for and doublets ( whose absorption features are clearly detected without being affected by any data defects or line blending ) , we measure their physical parameters by performing voigt profile fitting with minfit . we have several observational results that support an intrinsic origin for these features , itemized below in order of relevance . * the associated nals show clear evidence of partial coverage ( see figure [ f4 ] and table [ t2 ] ) . covering factors are quite important for ascertaining the location of the absorbers . among 16 and 12 components in the spectra of images a and b , 6 and 6 components show partial coverage at the 4@xmath21 level ( i.e. , @xmath28 4@xmath21 ( ) @xmath29 1.0 ) , respectively . this supports the physical sizes of the absorbers being comparable to or even smaller than the size of the background source , suggesting that the system is intrinsic to the quasar itself ( e.g. , @xcite ) . interestingly , nals have higher values and are consistent with full coverage with only a few exceptions . covering factors are not necessarily identical for all ions from the same absorber , both in bals ( e.g. , @xcite ) and nals ( e.g. , @xcite ) . in the literature , ions in higher ionization states usually have larger , as is also the case for the associated system in sdss j1029 + 2623 . * the associated nals show the so - called line - locking phenomenon ( i.e. , blue components of doublets are aligned with red ones of the following doublets ; see figures [ f3 ] and [ f4 ] ) . chance coincidence of such alignments is negligibly small . this is naturally explained by radiative acceleration ( e.g. , @xcite ) . although several mechanisms have been proposed to explain the acceleration of outflowing winds , radiative acceleration almost always contributes substantially . line locking requires that the sight - lines be approximately parallel to the gas motion . * the velocity distribution of the associated and nals shows values beyond 1,000 , which is too large for an origin in igm , foreground galaxy , and ism of the host galaxy . intervening absorption lines are typically clustered in @xmath30 @xmath25 400 both for metal lines ( e.g. , @xcite ) and for lines ( e.g. , @xcite ) . the ism of most galaxies show velocities below this limit . a distribution with velocities @xmath4 1,000 strongly supports an intrinsic origin due to the outflow wind . * these nals have small ejection velocities from the quasar emission redshift , @xmath31 @xmath5 1600 . @xcite found that a high fraction ( @xmath021% ) of the associated nals within = 5,000 from the quasars , is time - variable ( i.e. , intrinsic lines ) . @xcite discovered an excess number density of nals in quasar vicinities , and concluded that almost half of absorbers at @xmath31 @xmath5 12000 are intrinsic to the quasars , with a peak value of @xmath0 80 % at @xmath31 @xmath0 2000 . based on multiple medium - resolution spectra , we confirmed that the associated absorption lines toward both images a and b retain their profiles unchanged , while their profiles are clearly different each other . this result was originally expected to reject the time - variation scenario ( i.e. , scenario i ) . however , with the high - resolution spectrograph , the associated absorption lines are deblended into multiple narrow components , of which two are foreground absorption lines of @xmath102344 and @xmath102600 at @xmath0 0.9187 . we confirmed that the differential absorption profiles between the lensed images seen in medium - resolution spectra are mainly due to these physically unrelated lines . thus , we can not distinguish different scenarios with medium - resolution spectra alone . high - resolution spectra taken with subaru / hds provide us with several important clues regarding the origin of the associated lines . first of all , the general profiles of the absorption lines toward images a and b are very similar to each other . this means that the size of the absorber must be larger than the transverse distance of the two sight - lines . in the extreme case , the absorber s location would be very close to the background flux source so as to make the separation angle almost negligible . second , we see a clear difference of absorption profiles between images a and b in various structures . for example , at @xmath0 0 200 , there exist additional absorption components in , , ly@xmath14 panels of figure [ f6 ] only in image a. this small difference , discovered only after taking high - resolution spectra , reminds us of the original question : what is the source of this difference ? finally , the most important clue is that a substantial fraction of absorption components show partial coverage , which means that the physical sizes of the absorbers are smaller than or comparable to the background source such as the continuum source and the belr . while some intrinsic absorption lines cover only the continuum source ( e.g. , @xcite ) , the covering factors toward the continuum source and the belr usually take specific values , depending on the relative strengths of their fluxes @xcite . in our case , the residual flux at @xmath10 @xmath0 4948 , at which the emission line peaks , is almost zero . this means that absorbers as a whole cover both the continuum source and the belr significantly . therefore , size estimation of the flux sources is important . we estimate the size of the belr ( @xmath32 ) using the empirical relation between @xmath32 and quasar luminosity . this relation was originally discovered through reverberation mapping @xcite and then extended to brighter quasars with monochromatic luminosities of @xmath2710@xmath33 ergs s@xmath12 to redshifts @xmath27 0.7 ( eq.[a4 ] in @xcite ) . the monochromatic luminosity of image a at @xmath34 = 3000 in the quasar rest - frame is measured in @xcite as @xmath35 = 46.21 ergs s@xmath12 . after correcting the magnification factor of image a , @xmath36 = 10.4 @xcite , we estimate @xmath32 to be @xmath0 0.09@xmath37 pc , where the main source of 1@xmath21 uncertainty comes from the scatter in the empirical relation in @xcite . as to the size of the continuum source ( @xmath38 ) , we take five times the schwarzschild radius , @xmath39 = 10@xmath40 , following @xcite , where @xmath41 and @xmath42 are the schwarzschild radius and mass of the central black hole . the virial mass of the central black hole is already calculated in @xcite using the luminosity and fwhm of broad emission lines of image a. again , after correcting for magnification , we obtain @xmath43 = 8.72 , and then @xmath38 is evaluated to be @xmath0 2.54@xmath44@xmath45 pc . this is almost 300 times smaller than @xmath32 . the size of the background flux source ( @xmath9 ) and the absorber s distance from the central flux source ( @xmath46 ) decide geometry , i.e. , a single sight - line or multiple sight - lines . the transverse distance of the two sight - lines ( @xmath47 ) can be calculated as @xmath47 = @xmath48 , where @xmath1 is the separation angle of the two sight - lines seen from the flux source ( which is very similar to the separation angle seen from us , @xmath1 @xmath0 [email protected] ) . at a distance , where @xmath9 @xmath5 @xmath47 (= @xmath48 ) is satisfied , the two sight - lines become fully separated with no overlap . we call this distance a boundary distance ( @xmath49 ) , hereafter ( see figure [ f7]a ) . the boundary distance is @xmath49 @xmath02.3 pc if only the continuum source is the flux source , and @xmath0788 pc if the belr is also the background source . the latter is comparable to the distance of outflowing gas measured by @xcite and @xcite for other quasars . the geometry also depends on the absorber s size ( @xmath50 ) . if the size of the absorber ( whose internal sub - structure is ignored here ) is much larger than the transverse distance ( i.e. , @xmath50 @xmath51 @xmath52 ) , a single sight - line scenario is applicable regardless of the absorber s distance from the flux source ( see figure [ f7]b ) . however , this geometry can not be applied to our case because the partial coverage we found in [ sec3.2 ] requires the size of each absorber to be comparable or smaller than the size of background source . on the other hand , if @xmath50 @xmath5 @xmath52 ( figure [ f7]c ) , we should analyze the condition in more detail . in the case of @xmath50 @xmath5 @xmath52 @xmath53 @xmath9 ( i.e. , the absorber s distance is smaller than the boundary distance ) , a substantial fraction of both sight - lines is covered by a single absorber although the fraction depends on the distance and the absorber s size . small differences of absorption profiles , seen in the high - resolution spectra of images a and b , can be explained either by i ) a different part of the absorber s outskirts covering each sight - line ( _ quasi-_scenario ii ) , or by ii ) time - variation ( scenario i ) . in the case of @xmath50 @xmath5 @xmath9 @xmath5 @xmath48 ( i.e. , the absorber s distance is larger than the boundary distance ) , this will be _ bona - fide _ scenario ii . one shortcoming of this model is that we can not reproduce common absorption profiles seen in the two sight - lines because the same absorber can not cover both sight - lines . we will discuss this problem later . @xcite claimed that nal absorbers should be located in a low gravity environment far from the central bh in order to maintain kinematic stability . their result suggests that the ejection velocity of nal absorbers is larger than the escape velocity so that the absorber is gravitationally unbound . if the nal absorbers in our target quasar are also in a similar kinematic condition , their radial distance from the central bh should be @xmath46 @xmath4 1.79 pc . this is close to the boundary distance in the case that only the continuum source is the background flux source . as a result , we do not have any conclusive evidence to either accept or reject scenarios i or ii . therefore , we will discuss the absorber s physical condition further assuming both scenarios in turn below . if scenario i represents reality , there are two possible origins of time variability , i ) a change of ionization condition of the absorber ( e.g. , @xcite ) and ii ) the absorber moving across our sight lines to change the covering factor ( e.g. , @xcite ) . neither situation is applicable for the case of intervening absorbers as discussed in @xcite . if a change in ionization is the origin of variation , we can place constraints on the electron density and the distance from the flux source following the procedure by @xcite and @xcite . we can not evaluate the specific ionization condition because a wide range of ionic species ( which is necessary for photoionization modeling ) are not detected in our spectra . therefore , by adopting the following assumptions ; i ) the gas is very close to ionization equilibrium , ii ) the change of ionizing flux is small , and iii ) is the dominant ionization stage of carbon , we estimate the electron density to be @xmath54 @xmath0 @xmath55 @xmath0 1.78@xmath4410@xmath56 @xmath57 by assuming a variation timescale in the quasar rest frame ( @xmath0233 days ) as a recombination time ( @xmath58 ) , where we use the recombination coefficient of @xmath59 in gas temperature of 20,000 k @xcite . because the variation timescale is an upper limit of the recombination time , we should regard the electron density as a lower limit . we can also evaluate the distance of the absorber from the flux source using the prescription of @xcite . the ionization parameter ( @xmath60 ) depends on the bolometric luminosity of the quasar , the continuum shape , distance from the flux source , and electron density . by adopting the continuum shape of @xcite , we estimate the distance of the absorber to be @xmath46 @xmath50.44 kpc , with an ionization parameter of @xmath60 = 0.02 ( in which is the dominant ionization stage ; @xcite ) , and the bolometric luminosity of the quasar of @xmath61 = 45.87 erg s@xmath12 @xcite . the limits on the electron density and the absorber s distance are consistent with the literature ( e.g. , @xcite ) . if the gas motion ( i.e. , crossing of our sight - line ) is the origin of variation , we can simply estimate the average crossing velocity ( @xmath62 @xmath0 @xmath63 ) to be @xmath0390 if the continuum source is the only background source . again , this is the minimum velocity because @xmath3 is an upper limit on the variability timescale . the belr should not be the flux source , because the corresponding crossing velocity , @xmath01.32@xmath4410@xmath64 , is comparable to the speed of light . thus , we conclude that the absorber covers primarily the uv continuum source with only a small part of the belr if the gas motion is the origin of variation . the main problem of the _ bona - fide _ scenario ii model , in which the absorber s size is smaller than the transverse distance of the two sight - lines , is that we can not reproduce closely similar absorption profiles seen in the two sight - lines because the same absorbers can not cover both sight - lines . a possible solution for this is that there exists a number of clumpy absorbers or fluctuations in gas density ( whose sizes are smaller than the background source ) constituting a filamentary ( or a sheet - like ) structure that covers both sight - lines ( figure [ f7]c ) . such a structure is well reproduced above the main body of the outflow by hydrodynamical simulations ( e.g , @xcite ) . here , we calculate the radial mass - outflow rate ( @xmath65 ) toward the two sight - lines summing contributions from all absorption components as @xmath66 where @xmath67(@xmath68 ) and @xmath69(@xmath68 ) are the total ( @xmath28 ) hydrogen column density and the radial depth of the @xmath68-th absorption component toward a sight - line . @xmath9 and @xmath70 are the size of background source and the proton mass . in [ sec3.2 ] , we measured ejection velocities of all and absorption components by applying voigt profile fitting . however , their absolute values could be underestimated because emission redshifts determined from broad uv emission lines are systematically blueshifted from the systemic redshift , as measured by narrow , forbidden lines ( see , e.g. , @xcite ) . @xcite find a mean blueshift of the broad uv emission lines relative to the systemic redshift of about 260 and that 90% of the blueshifts are between 0 and 650 . we therefore add 260 to the values given in table [ t2 ] for the calculation herein . because our spectra do not detect a wide range of ionic species ( which are necessary for photoionization modeling , as described in [ sec4.3.1 ] ) , we can measure neither the ionization condition nor the total hydrogen column density @xmath71(h ) . these parameters decide an absorber s metallicity and radial depth . therefore , we simply calculate @xmath65 assuming all absorption components have the same ionization condition , metallicity , and radial depth toward both the sight - lines as follows , @xmath72 where @xmath73 ( a metallicity - ionization correction ) for and are defined as follows , respectively , @xmath74 @xmath75 based on the and absorption lines , a difference of radial mass - outflow rates toward images a and b is only @xmath030 35% , which suggests that internal fluctuation of the nal absorber is almost negligible within the angular distance of @xmath1 @xmath0 [email protected] . because @xmath020 50% of quasars have at least one intrinsic nal absorber in their spectra @xcite , the global covering factor of the central flux source surrounded by nal absorbers can have values of @xmath76 @xmath5 0.8@xmath77 2@xmath77 in terms of solid angle . thus , small areas within @xmath00.002% 0.005% of the total solid angle for nal absorbers produce similar absorption profiles with the same order of total column densities . in other words , the outflow wind corresponding to the nal absorbers can be divided into @xmath520,000 50,000 small zones with common physical conditions . we carried out medium - resolution spectroscopy of the two brightest images of the quasar sdss j1029 + 2623 to see whether the associated absorption lines are variable or not , and also high - resolution spectroscopy in order to measure line parameters and place strict constraints on the absorber s physical conditions . our main results are as follows : 1 . the associated absorption profile does not show clear time - variation toward either sight - lines in medium - resolution spectra . with high - resolution spectra , we confirmed that the differential line profile is mainly due to the effect of @xmath102344 and @xmath102600 lines which arise in a physically unrelated system at @xmath0 0.9187 . 2 . in high - resolution spectra , the associated absorption line is deblended into more than 10 narrow components , showing i ) partial coverage , ii ) line - locking , iii ) large velocity distribution , and iv ) small ejection velocity . all of these results support the associated absorption line being physically related to the quasar ( i.e. , _ intrinsic _ absorption line ) . the associated absorber probably covers both the continuum source and the belr because the residual flux at @xmath10 @xmath0 4948 , at which the emission line peaks , is almost zero . the size of the continuum source and the belr are evaluated to be @xmath38 @xmath0 2.54@xmath44@xmath45 pc and @xmath78 @xmath0 0.09@xmath37 pc , respectively . if the associated absorber covers both the flux sources toward our sight - lines , their typical scale should be comparable to or smaller than @xmath32 because they show partial coverage . 4 . the geometry of the associated absorber depends on the sizes of the background flux source ( @xmath9 ) and the absorber ( @xmath50 ) , and the absorber s distance from the center ( @xmath46 ) . if a single sight - line model represents reality , the differential line profile impies time variation of the absorber . if a change of ionization level is the origin of the variation , we can place constraints on the absorber s electron density , @xmath54 @xmath5 1.78@xmath4410@xmath56 @xmath57 , and the absorber s distance from the continuum source , @xmath46 @xmath5 0.44 kpc . if gas motion is the origin of variation , the continuum source is the only background source because the estimated crossing velocity is too large ( comparable to the speed of light ) if the belr is also the flux source . 5 . if multiple sight - line model is the real scenario , the following condition should be satisfied ; the absorber s size ( @xmath50 ) @xmath5 the size of the background flux source ( @xmath9 ) @xmath5 transverse distance of the two sight - lines ( @xmath47 ) . however , such a condition can not allow a single absorber to cover both sight - lines , in order to reproduce similar absorption profiles as seen in the high - resolution spectra . a filamentary ( or a sheet - like ) structure made of multiple clumpy gas clouds could solve this issue . indeed , such structures are reproduced by hydro - dynamical simulations . in order to assess the origin of the associated absorber further , we need to obtain high - resolution spectra again . we have shown that monitoring observations with medium - resolution spectra are not useful because the associated absorption line in the image a is significantly contaminated with physically unrelated foreground absorption lines . additional high - resolution spectra will provide us with stronger evidence for scenario i if we see clear time - variations or for scenario ii if the images do not show variability . tm acknowledges support from the special postdoctoral research program of riken . the research was partially supported by the japan society for the promotion of science through grant - in - aid for scientific research 23740148 and shinshu university research grant for exploratory research by young scientists . this work was also supported in part by the first program `` subaru measurements of images and redshifts ( sumire ) '' , world premier international research center initiative ( wpi initiative ) , mext , japan , and grant - in - aid for scientific research from the jsps ( 23740161 ) . we also thank kentaro aoki for his comments about data analysis , and chris churchill for providing us with minfit . we also thank the anonymous referee for helpful comments and suggestions . arav , n. , becker , r. h. , laurent - muehleisen , s. a. , gregg , m. d. , white , r. l. , brotherton , m. s. , & de kool , m. 1999 , , 524 , 566 balsara , d. s. , & krolik , j. h. 1993 , , 402 , 109 barlow , t. a. , & sargent , w. l. w. , 1997 , , 113 , 136 becker , r. h. , gregg , m. d. , hook , i. m. , mcmahon , r. g. , white , r. l. , & helfand , d. j. 1997 , , 479 , l93 becker , r. h. , et al . 2001 , , 135 , 227 blandford , r. d. , & payne , d. g. 1982 , , 199 , 883 brotherton , m. s. , wills , b. j. , steidel , c. c. , & sargent , w. l. w. 1994 , , 423 , 131 capellupo , d. m. , hamann , f. , shields , j. c. , rodrguez hidalgo , p. , & barlow , t. a. 2012 , , 422 , 3249 chartas , g. 2000 , , 531 , 81 chelouche , d. , & netzer , h. 2005 , , 625 , 95 churchill , c. w. , vogt , s. s. , & charlton , j. c. 2003 , , 125 , 98 churchill , c. w. , & vogt , s. s. 2001 , , 122 , 679 churchill , c. w. 1997 , ph.d . thesis , university of california , santa cruz corbin , m. r. 1990 , , 357 , 346 crenshaw , d. m. , kraemer , s. b. , boggess , a. , maran , s. p. , mushotzky , r. f. , & wu , c .- c . 1999 , , 516 , 750 crotts , a. p. s. , & fang , y. 1998 , , 502 , 16 de kool , m. , arav , n. , becker , r. h. , et al . 2001 , , 548 , 609 de kool , m. , & begelman , m. c. 1995 , , 455 , 448 elvis , m. 2000 , , 545 , 63 emmering , r. t. , blandford , r. d. , & shlosman , i. 1992 , , 385 , 460 everett , j. e. , 2005 , , 631 , 689 fohlmeister , j. , kochanek , c. s. , falco , e. e. , et al . 2012 , arxiv:1207.5776 foltz , c. b. , weymann , r. j. , morris , s. l. , & turnshek , d. a. 1987 , , 317 , 450 gabel , j. r. , arav , n. , & kim , t .- s . 2006 , , 646 , 742 ganguly , r. , eracleous , m. , charlton , j. c. , & churchill , c. w. , 1999 , , 117 , 2594 gibson , r. r. , jiang , l. , brandt , w. n. , et al . 2009 , , 692 , 758 gibson , r. r. , brandt , w. n. , schneider , d. p. , & gallagher , s. c. 2008 , , 675 , 985 green , p. j. 2006 , , 644 , 733 hamann , f. , kanekar , n. , prochaska , j. x. , et al . 2011 , , 410 , 1957 hamann , f. , kaplan , k. f. , rodrguez hidalgo , p. , prochaska , j. x. , & herbert - fort , s. 2008 , , 391 , l39 hamann , f. , & sabra , b. 2004 , agn physics with the sloan digital sky survey , 311 , 203 hamann , f. w. , barlow , t. a. , chaffee , f. c. , foltz , c. b. , & weymann , r. j. 2001 , , 550 , 142 hamann , f. , barlow , t. a. , junkkarinen , v. , & burbidge , e. m. , 1997a , , 478 , 80 hamann , f. , barlow , t. a. , & junkkarinen , v. , 1997b , , 478 , 87 hamann , f. , barlow , t. a. , beaver , e. a. , et al . 1995 , , 443 , 606 hewett , p. c. , & foltz , c. b. 2003 , , 125 , 1784 inada , n. , oguri , m. , morokuma , t. , et al . 2006 , , 653 , l97 inada , n. , oguri , m. , pindor , b. , et al . 2003 , , 426 , 810 kaspi , s. , smith , p. s. , netzer , h. , et al . 2000 , , 533 , 631 konigl , a. , & kartje , j. f. 1994 , , 434 , 446 krolik , j. h. , & kriss , g. a. 2001 , , 561 , 684 lopez , s. , reimers , d. , gregg , m. d. , et al . 2005 , , 626 , 767 lu , l. , sargent , w. l. w. , barlow , t. a. , churchill , c. w. , & vogt , s. s. 1996 , , 107 , 475 marziani , p. , sulentic , j. w. , dultzin - hacyan , d. , calvani , m. , & moles , m. 1996 , , 104 , 37 mclure , r. j. , & dunlop , j. s. 2004 , , 352 , 1390 misawa , t. , eracleous , m. , charlton , j. c. , & kashikawa , n. 2007b , , 660 , 152 misawa , t. , charlton , j. c. , eracleous , m. , ganguly , r. , tytler , d. , kirkman , d. , suzuki , n. , & lubin , d. 2007a , , 171 , 1 misawa , t. , eracleous , m. , charlton , j. c. , & tajitsu , a. , 2005 , , 629,115 misawa , t. , tytler , d. , iye , m. , et al . 2004 , , 128 , 2954 murray , n. , chiang , j. , grossman , s. a. , & voit , g. m. , 1995 , , 451 , 498 narayanan , d. , hamann , f. , barlow , t. , burbidge , e. m. , cohen , r. d. , junkkaribeb , v. , & lyons , r. , 2004 , , 601 , 715 nestor , d. , hamann , f. , & rodriguez hidalgo , p. 2008 , , 386 , 2055 netzer , h. , lira , p. , trakhtenbrot , b. , shemmer , o. , & cury , i. 2007 , , 671 , 1256 ohsuga , k. , mori , m. , nakamoto , t. , & mineshige , s. 2005 , , 628 , 368 oguri , m. , schrabback , t. , jullo , e. , et al . 2012 , arxiv:1209.0458 oguri , m. , ofek , e. o. , inada , n. , et al . 2008 , , 676 , l1 ota , n. , oguri , m. , dai , x. , et al . 2012 , arxiv:1202.1645 penton , s. v. , stocke , j. t. , & shull , j. m. 2004 , , 152 , 29 perry , j. j. , burbidge , e. m. , & burbidge , g. r. 1978 , , 90 , 337 petitjean , p. , & bergeron , j. 1994 , , 283 , 759 pichon , c. , scannapieco , e. , aracil , b. , et al 2003 , , 597 , l97 proga , d. , stone , j. m. , & kallman , t. r. , 2000 , , 543 , 686 proga , d. , stone , j. m. , & drew , j. e. 1999 , , 310 , 476 proga , d. , stone , j. m. , & drew , j. e. 1998 , , 295 , 595 rauch , m. , sargent , w. l. w. , & barlow , t. a. 1999 , , 515 , 500 reichard , t. a. , et al . 2003a , , 125 , 1711 sargent , w. l. w. , young , p. j. , boksenberg , a. , & tytler , d. 1980 , , 42 , 41 shen , y. , richards , g. t. , strauss , m. a. , et al . 2011 , , 194 , 45 sprayberry , d. , & foltz , c. b. , 1992 , , 390 , 39 springel , v. , di matteo , t. , & hernquist , l. 2005 , , 620 , l79 stocke , j. t. , morris , s. l. , weymann , r. j. , & foltz , c. b. 1992 , , 396 , 487 tytler , d. , & fan , x .- m . 1992 , , 79 , 1 vestergaard , m. 2003 , , 599 , 116 voit , g. m. , weymann , r. j. , & korista , k. t. , 1993 , , 413 , 95 wampler , e. j. , chugai , n. n. , & petitjean , p. 1995 , , 443 , 586 weymann , r. j. , morris , s. l. , foltz , c. b. , & hewett , p. c. , 1991 , , 373 , 23 weymann , r. j. , carswell , r. f. , & smith , m. g. 1981 , , 19 , 41 weymann , r. j. , williams , r. e. , peterson , b. m. , & turnshek , d. a. 1979 , , 234 , 33 wise , j. h. , eracleous , m. , charlton , j. c. , & ganguly , r. , 2004 , , 613 , 20 yamamoto , t. m. , & vanseviius , v. 1999 , , 51 , 405 york , d. g. , adelman , j. , anderson , j. e. , jr . , et al . 2000 , , 120 , 1579 young , p. , sargent , w. l. w. , & boksenberg , a. 1982 , , 48 , 455 cccccccc image a & 2006 february 28 & 2 & sdss & 1800 & & 9.1 & 1 + & 2006 june 29 & 3 & subaru / focas & 500 & 600 & 19 & 2 + & 2009 april 4 & 5 & subaru / focas & 500 & 1200 & 8.5 & 3 + image b & 2007 december 15 & 1 & keck / lris & 1000 & 800 & 43 & 4 + & 2009 april 4 & 4 & subaru / focas & 500 & 1200 & 7.0 & 3 + image c & 2007 december 15 & 1 & keck / lris & 1000 & 800 & 11 & 4 + image a & 2010 february 10 & & subaru / hds & 30000 & 14400 & 13 & 3 + image b & 2010 february 10 & & subaru / hds & 30000 & 14200 & 13 & 3 + cccccccccc & & + + & & & & & & & & + 1 & 2.1800 & @xmath261599.5 & 13.52 & 0.54 & 6.30 & 2.45 & 0.20 & 0.13 & @xmath260.13 + 2 & 2.1810 & @xmath261505.2 & 13.80 & 0.29 & 25.52 & 3.01 & 0.38 & 0.25 & @xmath260.18 + 3 & 2.1818 & @xmath261429.7 & 14.08 & 0.06 & 18.51 & 1.19 & 0.82 & 0.05 & @xmath260.05 + 4 & 2.1822 & @xmath261392.0 & 13.85 & 0.06 & 13.16 & 0.83 & 0.96 & 0.08 & @xmath260.08 + 5 & 2.1829 & @xmath261326.0 & 12.97 & 0.98 & 14.35 & 37.5 & 1.00@xmath79 & ... & ... + 6 & 2.1880 & @xmath26845.7 & 13.39 & 0.75 & 16.40 & 3.62 & 0.62 & 2.88 & @xmath260.88 + 7 & 2.1886 & @xmath26789.3 & 14.28 & 0.15 & 16.08 & 2.18 & 0.49 & 0.05 & @xmath260.05 + 8 & 2.1894 & @xmath26714.0 & 15.03 & 0.34 & 25.44 & 6.00 & 0.62 & 0.04 & @xmath260.04 + 9 & 2.1901 & @xmath26648.2 & 14.28 & 0.12 & 19.18 & 2.61 & 0.88 & 0.06 & @xmath260.06 + 10 & 2.1910 & @xmath26563.6 & 14.28 & 0.09 & 44.19 & 9.87 & 1.00@xmath79 & ... & ... + 11 & 2.1942 & @xmath26262.9 & 14.01 & 0.29 & 143.71 & 113.3 & 1.00@xmath79 & ... & ... + 12 & 2.1944 & @xmath26244.1 & 14.82 & 0.10 & 21.34 & 1.63 & 0.89 & 0.04 & @xmath260.04 + 13 & 2.1953 & @xmath26159.6 & 14.84 & 0.09 & 34.58 & 5.17 & 0.85 & 0.06 & @xmath260.06 + 14 & 2.1964 & @xmath2656.3 & 15.65 & 0.20 & 32.62 & 2.83 & 0.82 & 0.04 & @xmath260.04 + 15 & 2.1975 & 46.9 & 14.37 & 0.12 & 16.02 & 2.14 & 0.46 & 0.06 & @xmath260.06 + 16 & 2.1980 & 93.8 & 14.46 & 0.06 & 34.85 & 2.60 & 0.48 & 0.04 & @xmath260.04 + + & & & & & & & & + 1 & 2.1889 & @xmath26761.1 & 14.18 & 0.60 & 77.80 & 73.1 & 1.00@xmath79 & ... & ... + 2 & 2.1894 & @xmath26714.0 & 13.94 & 0.85 & 27.88 & 31.5 & 1.00@xmath79 & ... & ... + 3 & 2.1901 & @xmath26648.2 & 14.52 & 0.06 & 31.83 & 3.25 & 0.92 & 0.07 & @xmath260.07 + 4 & 2.1908 & @xmath26582.4 & 14.99 & 0.39 & 14.50 & 3.05 & 0.88 & 0.07 & @xmath260.07 + 5 & 2.1914 & @xmath26526.0 & 14.13 & 0.16 & 26.67 & 10.92 & 1.00@xmath79 & ... & ... + 6 & 2.1938 & @xmath26300.4 & 15.25 & 4.07 & 12.93 & 23.71 & 0.11 & 0.07 & @xmath260.07 + 7 & 2.1943 & @xmath26253.5 & 15.03 & 0.20 & 18.61 & 2.16 & 0.96 & 0.06 & @xmath260.06 + 8 & 2.1952 & @xmath26169.0 & 14.92 & 0.11 & 36.29 & 8.38 & 0.94 & 0.07 & @xmath260.07 + 9 & 2.1962 & @xmath2675.1 & 15.11 & 0.15 & 39.00 & 15.02 & 0.90 & 0.06 & @xmath260.06 + 10 & 2.1969 & @xmath269.4 & 14.41 & 0.07 & 12.90 & 1.91 & 1.00@xmath79 & ... & ... + 11 & 2.1978 & 75.1 & 14.69 & 0.08 & 46.82 & 3.58 & 0.61 & 0.07 & @xmath260.07 + + & & + + & & & & & & & & + 1 & 2.1814 & @xmath261467.4 & 13.37 & 0.41 & 16.46 & 18.15 & 1.00@xmath79 & ... & ... + 2 & 2.1821 & @xmath261401.4 & 14.20 & 0.04 & 13.75 & 0.37 & 0.99 & 0.04 & @xmath260.04 + 3 & 2.1827 & @xmath261344.9 & 13.81 & 0.32 & 21.73 & 3.05 & 0.31 & 0.17 & @xmath260.15 + 4 & 2.1886 & @xmath26789.3 & 15.96 & 1.03 & 17.73 & 4.40 & 0.26 & 0.05 & @xmath260.05 + 5 & 2.1896 & @xmath26695.2 & 14.45 & 0.17 & 38.57 & 5.69 & 0.76 & 0.05 & @xmath260.05 + 6 & 2.1901 & @xmath26648.2 & 14.18 & 0.14 & 18.49 & 2.90 & 0.77 & 0.14 & @xmath260.14 + 7 & 2.1910 & @xmath26563.6 & 14.09 & 0.18 & 37.21 & 11.44 & 1.00@xmath79 & ... & ... + 8 & 2.1944 & @xmath26244.1 & 14.67 & 0.07 & 22.64 & 1.35 & 0.85 & 0.04 & @xmath260.04 + 9 & 2.1952 & @xmath26169.0 & 15.51 & 3.21 & 6.63 & 5.29 & 0.74 & 0.06 & @xmath260.06 + 10 & 2.1959 & @xmath26103.2 & 15.03 & 0.06 & 81.86 & 6.01 & 0.77 & 0.06 & @xmath260.06 + 11 & 2.1970 & 0.0 & 14.29 & 0.15 & 26.03 & 4.26 & 0.54 & 0.14 & @xmath260.14 + 12 & 2.1971 & 9.4 & 15.52 & 0.23 & 67.71 & 6.19 & 0.27 & 0.04 & @xmath260.04 + + & & & & & & & & + 1 & 2.1884 & @xmath26808.1 & 13.44 & 0.78 & 20.68 & 37.5 & 1.00@xmath79 & ... & ... + 2 & 2.1897 & @xmath26685.8 & 14.89 & 0.04 & 51.32 & 1.50 & 0.84 & 0.05 & @xmath260.05 + 3 & 2.1909 & @xmath26573.0 & 15.19 & 0.40 & 13.57 & 2.11 & 0.90 & 0.05 & @xmath260.05 + 4 & 2.1915 & @xmath26516.6 & 13.95 & 0.20 & 19.20 & 10.0 & 1.00@xmath79 & ... & ... + 5 & 2.1943 & @xmath26253.5 & 14.58 & 0.13 & 20.22 & 1.92 & 0.99 & 0.06 & @xmath260.06 + 6 & 2.1946 & @xmath26225.3 & 15.30 & 1.85 & 9.69 & 7.26 & 0.86 & 0.07 & @xmath260.07 + 7 & 2.1952 & @xmath26169.0 & 14.65 & 0.07 & 24.92 & 5.53 & 0.98 & 0.06 & @xmath260.06 + 8 & 2.1957 & @xmath26122.0 & 14.54 & 0.17 & 14.51 & 4.85 & 0.93 & 0.08 & @xmath260.08 + 9 & 2.1965 & @xmath2646.9 & 15.34 & 0.29 & 34.20 & 8.34 & 0.91 & 0.05 & @xmath260.05 + 10 & 2.1972 & 18.8 & 14.92 & 0.35 & 20.74 & 3.18 & 0.65 & 0.08 & @xmath260.08 + 11 & 2.1980 & 93.8 & 14.87 & 0.35 & 15.78 & 3.11 & 0.35 & 0.05 & @xmath260.05 + @xmath0 500 ) : ( top ) zoomed spectra of images a , b , and c of sdss j1029 + 2623 around the associated absorption line @xcite . fluxes are arbitrary and normalized to compare the absorption profiles of the three images . an additional absorption line at @xmath04990 present only in the image a spectrum is @xmath102600 at @xmath0 0.9187 . ( middle ) spectra of image a , taken in 2006 twice @xcite and in 2009 by us . ( bottom ) same as the above panel , but for image b , taken in 2007 @xcite and in 2009 by us.[f2],width=453 ] 500 , it is self - blended . the horizontal axis denotes relative velocity from the quasar emission redshift at = 2.197 . top and middle panels show the profiles of the blue and red members of a doublet , with model profiles output by minfit superposed as dashed lines . the positions of the kinematic components are marked with upward arrows at the bottom of each panel . the bottom panel shows the two profiles together , along with the resulting covering factors and their 1@xmath21 uncertainties with green open circles.[f4],title="fig:",width=302 ] 500 , it is self - blended . the horizontal axis denotes relative velocity from the quasar emission redshift at = 2.197 . top and middle panels show the profiles of the blue and red members of a doublet , with model profiles output by minfit superposed as dashed lines . the positions of the kinematic components are marked with upward arrows at the bottom of each panel . the bottom panel shows the two profiles together , along with the resulting covering factors and their 1@xmath21 uncertainties with green open circles.[f4],title="fig:",width=302 ] and meshed ovals / circles with a size of @xmath50 denote the central flux source and absorbers . the radial distance @xmath49 is the boundary distance . a region surrounded by two curves in ( c ) is a filamentary / sheet - like structure.[f7],width=453 ]

the formation of chondrules is one of the oldest unsolved mysteries in meteoritics and planet formation . recently an old idea has been revived : the idea that chondrules form as a result of collisions between planetesimals in which the ejected molten material forms small droplets which solidify to become chondrules . pre - melting of the planetesimals by radioactive decay ofal would help producing sprays of melt even at relatively low impact velocity . in this paper we study the radiative cooling of a ballistically expanding spherical cloud of chondrule droplets ejected from the impact site . we present results from a numerical radiative transfer models as well as analytic approximate solutions . we find that the temperature after the start of the expansion of the cloud remains constant for a time and then drops with time approximately as^{-5/3}$ ] for . the time at which this temperature drop starts depends via an analytical formula on the mass of the cloud , the expansion velocity and the size of the chondrule . during the early isothermal expansion phase the density is still so high that we expect the vapor of volatile elements to saturate so that no large volatile losses are expected

1501.05791

the formation of chondrules is one of the fundamental questions of meteoritics and planet formation . chondrules are the 0.1@xmath51 mm - size once - molten silicate spherules found abundantly in primitive meteorites known as chondrites ( jones , grossman & rubin 2005 ; sears 2004 ; davis et al . most of these chondrites in fact consist predominantly of these chondrules , so the melt - producing events that created them must have been extremely common during the first few million years of the solar system . yet there is confusing and conflicting evidence as to what these events might have been . boss ( 1996 ) published an overview of the status of the discussion at that time , though significant new developments have occurred since then . many theories have been put forward over the last half a century . one theory involves impact melt sprays . put forward by urey ( 1953 ) and refined by kieffer ( 1975 ) this model states that high - velocity impacts ( @xmath6 3@xmath55 km / s ) could lead to sprays of impact melt that produce droplets which solidify into chondrules . while these required collision velocities are high , a small fraction of impacts might acquire such velocities ( bottke et al . perhaps the so - called `` grand tack''-scenario of walsh et al . ( 2011 ) , in which jupiter temporarily entered the asteroid belt region before migrating back outward , might produce such high velocities . however , taylor et al . ( 1983 ) put forward a number of arguments against the impact - melt scenario , some of which were based the differences between chondrites and lunar impact regolith . zook ( 1980 ) suggested that if the interiors of the colliding bodies are already in a molten state due to @xmath0al decay , then the required impact velocities to create sprays of melt are much lower , and thus more consistent with expectations of the average relative velocities between planetesimals . sanders & taylor ( 2005 ) , hevey & sanders ( 2006 ) and sanders & scott ( 2012 ) follow up on this idea of pre - molten impactors and back it up with models and meteoritic evidence . sanders & scott ( 2012 ) argue that this scenario is hard to avoid , given that in the first 2.5 million years most planetesimals were internally nearly fully molten by @xmath0al decay heat , and that collisions between planetesimals were extremely common . each collision would almost certainly release substantial amounts of molten rock into the nebula in the form of sprays of lava droplets , and it is natural to assume that these may be chondrules . they argue that the near - solar composition of chondrules can be explained by the vigorous convection inside the molten planetesimals that may slow down iron / nickel - core formation , thus keeping the melt solar . some degree of differentiation would then in fact explain the low iron - content l and ll chondrites . recently , asphaug et al . ( 2011 ) performed smooth particle hydrodynamics simulations for such impacts to demonstrate the dynamics of this scenario . in addition they proposed a simple way to calculate the melt droplet size by equating the released enthalpy after the collision to the surface energy of the droplets . there are many alternative scenarios proposed in the literature . perhaps the most popular model is the flash heating by nebular shocks ( hood & hornyi , 1991 ) . desch & connolly ( 2002 ) and ciesla & hood ( 2002 ) developed detailed 1-d radiative shock models with dust particles and chondrules interacting both radiatively and frictionally with the gas . they showed how the radiative shocks exhibit temperature spikes of mere tens of seconds ( with cooling rates @xmath7 k / hr ) that would be good candidates for chondrule - forming events . after the main shock temperature spike their model exhibits a further cooling at intermediate cooling rate ( @xmath850 k / hr ) , which cools the chondrules to sufficiently low temperature in a sufficiently short amount of time . this model seems to produce the flash - heating events required for turning `` dust bunnies '' into chondrules . however , so far the shock scenario still has several unresolved issues ( e.g. , desch et al . 2012 ; boley et al . 2013 ; stammler & dullemond 2014 ) . also some issues are raised about whether large scale shocks in the optically thick solar nebula can explain the time scales involved in chondrule formation ( stammler & dullemond 2014 ) . it is noteworthy that there exist several other nebular flash heating models , most notably flash heating by nebular lightning ( gibbard et al . 1997 ) and flash heating by energy dissipation in current sheets forming in mhd turbulence ( hubbard et al . 2012 ) . one important constraint on chondrule formation models is that these heating events were very short - lived compared to any other nebular time scales . by comparing textures of chondrule to those obtained in furnace experiments it can be inferred that chondrules must have cooled from the liquidus temperature down to below the solidus temperature in a matter of hours ( hewins 1983 ; hewins & connolly 1996 ; see also references in morris & desch 2010 and the excellent review paper of desch et al . in other words : the chondrule forming events must have been flash - events . on the other hand , under optically thin conditions a molten chondrule would radiatively cool in about a second , which would be too fast . another constraint that a chondrule formation model must fulfill is the retention of volatile elements such as fe , mg , si , na and k. chondrules are not observed to have low abundances of these elements . in particular for the highly volatile elements na and k this is a puzzle , because any heating event that heats a ( proto-)chondrule above @xmath8 1700 k and keeps these chondrules above that temperature for more than a few minutes will cause most of the na and k to evaporate out of the chondrule . alexander et al . ( 2008 ) argue that this means that chondrules must have formed in regions that are extremely dense in solids , much more so that typical dust concentration mechanisms in the protoplanetary disk can achieve . according to their calculations such regions are so dense in solids that they must be self - gravitating . morris et al . ( 2012 ) instead propose that chondrules formed behind the bow shock of a fast moving planetary embryo , and that the embryo s atmosphere is rich in such volatile elements caused by outgassing from the embryo s interior , thus providing the necessary vapor pressure to keep the chondrules volatile - rich . in this paper we will focus on the impact splash hypothesis , either the low - velocity pre - molten planetesimal version or the high - velocity impact - melt version . we assume that after the collision between two planetesimals a cloud of molten droplets of magma was released . since soon after the ejection of this cloud of magma droplets the internal pressure of the cloud would have dropped to near - zero ( there is , to good approximation , only vacuum between the droplets ) , the cloud would simply expand ballistically . initially the density of the cloud is so high that the cloud is completely optically thick and no radiation can escape from its interior . there is also no adiabatic cooling because of the lack of pressure . so the temperature of the magma droplets will initially stay roughly constant in time . as the cloud expands , however , the optical depth drops and eventually the cloud will start to cool radiatively . during the early stages the density is very high and the evaporation of volatile elements will be saturated ( as was argued by alexander et al . once the temperature drops below the solidus , volatiles can no longer escape . in this first paper we intend to compute how the temperature behaves as a function of time after the impact . to do this we set up a simple model : that of a homologously ballistically expanding homogeneous spherical cloud of lava droplets . we compute the temperature of the chondrules as a function of time @xmath2 and comoving radial location inside the cloud @xmath9 by solving the time - dependent radiative transfer equation . we also present an analytic approximate solution . while the spherical cloud model is not an accurate model of the complex shape of an impact splash , it can be regarded as a model of _ part _ of the impact splash . as such we believe that it will give reasonable estimates of the radiative cooling behavior of the impact splash as a function of model parameters such as the total mass of the cloud and its expansion velocity . the total mass of the cloud tells something about the masses of the two colliding bodies while the expansion velocity of the cloud tells us something about the impact velocity . when lava droplets are produced in a collision between two planetesimals , they will disperse away from the impact site in a ballistic way . let us call this the `` impact splash '' . if the impact velocity is larger than the escape velocity of the two colliding planetesimals this ballistic ( pressureless ) expansion will be linear ( i.e. the velocity of expansion will not change with time ) . we regard the impact splash as an expanding cloud of lava droplets ( chondrules - to - be ) . this cloud is not necessarily centered on the impact site ; it can also move away from the impact site . the radius of the cloud increases linearly with time @xmath2 and thus the density of the cloud will drop as @xmath10 . depending on the complexity of the geometry of the impact splash ( see e.g. asphaug et al . 2011 ) we can also consider the splash to be multiple smaller clouds . in either case , each of these clouds will expand linearly . the scenario is pictographically shown in fig . [ fig - pictogram ] . if the impact is at low velocity , then the gravitational pull of the surviving ( or merged ) body can cause some ( or most ) of the material to re - accrete . in this case the @xmath10 expansion will cease and turn into recompression . initially , however , the linear expansion and @xmath10 density drop will still be a good description . to compute the temperature of the droplets in these clouds as a function of time after impact we must solve the time - dependent equation of radiative transfer . this is a very non - local problem , because a cooling droplet at location @xmath11 radiates away some heat into the form of infrared radiation , which can then be absorbed by another droplet at location @xmath12 which is then heated . the droplets are thus radiatively coupled over large distances . solving this problem in 3-d for complex cloud geometries is somewhat challenging . we believe , however , that much can already be learned from a simple spherical cloud model , for which the radiative transfer problem can be solved to high precision and reliability with the 1-d tangent - ray variable eddington factor method . we will describe the method in section [ sec - radiative - transfer ] . the three main parameters of the model are : @xmath13 where @xmath14 is the total mass worth of chondrules ( lava droplets ) in our spherical cloud , @xmath15 is the initial temperature of the chondrules and @xmath16 is the expansion velocity of the outer radius of the cloud . to get a feeling for the numbers it is more convenient to express the mass of the cloud in terms of the radius @xmath17 of a ball of magma of the same mass @xmath14 : @xmath18 where @xmath19 is the material density of the chondrule droplets and thus , by definition , the material density of the hypothetical ball of magma . for a linearly expanding cloud the radius of the cloud is @xmath20 where the time @xmath2 is the time since the impact and the ejection of the cloud of melt droplets , assuming perfectly ballistic ( pressureless ) expansion and of course assuming large enough @xmath2 that @xmath21 so that the droplets are clearly separated from each other . the velocity profile inside the cloud is @xmath22 which is of course only valid for @xmath23 . let us define a coordinate @xmath24 centered on the center of the spherical cloud . we assume that the density within the cloud is constant and outside of the cloud is zero . the density of the cloud as a function of time is then @xmath25 from here on we write only @xmath26 instead of @xmath27 . the time - dependence of @xmath26 is @xmath28 let us consider chondrules ( lava droplets ) of radius @xmath29 and material density @xmath19 . we take @xmath30 g/@xmath31 for our model . the mass of a chondrule is then @xmath32 the number density of chondrules is then @xmath33 the geometric cross section of a chondrule is @xmath34 . let us assume that the chondrule has zero albedo . then the absorption cross section equals the geometric cross section . the opacity ( i.e. the cross section per gram ) is then @xmath35 we will assume that the opacity of any possible vapor between the chondrules will be low compared to that of the chondrules so that it can be ignored . the optical depth from the center of the cloud to the edge is @xmath36 the rate of cooling will strongly depend on this optical depth . in this paper we will present our results based on a set of fiducial models as well as parameter scans . the parameters of the fiducial models are listed in table [ tab - model - param ] . .[tab - model - param]the model parameters ( first three columns ) for the main ( `` fiducial '' ) models presented in this paper . the cloud mass @xmath14 can be derived from @xmath37 ( cf . eq . [ eq - def - rmelt0 ] ) . the last two columns are computed from these parameters : the cooling time @xmath1 ( the time after the impact when the cloud starts to cool , see eq . [ eq - tcool ] ) and the optical depth at that time @xmath38 ( see eq . [ eq - taucool ] ) . [ cols="^,^,^,^,^,^",options="header " , ] as the cloud expands it can start to radiate away energy . a fully fledged time - dependent radiative cooling computation will yield temperature as a function of time _ and _ space : @xmath39 . we will compute this in section [ sec - radiative - transfer ] . a first estimate of the cooling behavior of the cloud can already be made with pen and paper . the cloud will initially be optically thick : @xmath40 . the luminosity of the cloud is then @xmath41 where @xmath42 is the stefan - boltzmann constant and @xmath43 is the effective surface temperature at @xmath44 . in the above equation , and from here onward , we omit the @xmath45 for aesthetic reasons , but the time - dependence is still assumed . during the cooling phase the effective surface temperature will be lower than the central temperature . a commonly used estimation of this effect is : @xmath46 where @xmath47 is now the central temperature and @xmath48 is the correction factor to account for the lower temperature of the surface @xmath43 , and is based on radiative diffusion theory which states that @xmath49 ( for @xmath50 ) . to compute the change of the temperature as a result of the radiative loss given by eq . ( [ eq - lum - tau ] ) we must use the heat capacity formula . the total thermal energy stored in the chondrule cloud ( approximating it , for the moment , as an isothermal cloud ) is : @xmath51 where @xmath52 is the mass - weighted specific heat of the lava droplet . a value of @xmath53 erg g@xmath54 k@xmath54 is a reasonable value which we will adopt here . the radiative loss time scale can now be defined as @xmath55 this time scale varies with time : it is very long at early times because the cloud is then still extremely optically thick . at late times the cloud is optically thinner and the radiation can more freely escape ( eq . [ eq - lum - tau ] ) and @xmath56 becomes smaller . it is therefore to be expected that at early times the temperature of the chondrules remains constant . note that since the cloud consists of liquid drops that stay at a constant volume and move away from each other , there is no adiabatic cooling involved here . any potential vapor may adiabatically cool , but we will ignore this effect in this paper , assuming that the total mass in vapor is always small compared to the mass in droplets . at some point in time , however , @xmath56 becomes smaller than @xmath2 and the chondrules start to cool . based on the results of the true radiative transfer calculations of section [ sec - radiative - transfer ] it turns out that near the center of the cloud the transition from the initial constant temperature phase to the temperature - decline phase occurs roughly at a time @xmath1 defined by @xmath57 inserting eqs.([eq - trad],[eq - lum - tau],[eq - cloud - energy],[eq - tau - t ] ) into eq . ( [ eq - tcool - def ] ) yields @xmath58 note that in eq . ( [ eq - tcool - def ] ) the factor of 5 is purely empirical . the estimate we make here is essentially a dimensional analysis in which proportionality factors have to be calibrated against more exact calculations ( in our case the full radiative transfer calculation ) . roughly for @xmath59 the temperature stays constant while for @xmath4 the temperature drops with time . from now on we shall define _ the _ @xmath1 as the one calculated for the center of the cloud . near the surface the cooling sets in earlier . the temperature decline with time can also be estimated , at least up until the point where the cloud becomes optically thin , after which our approximation breaks down . the way to make this estimate is to solve the cooling equation @xmath60 where @xmath61 is the cooling time given by eq . ( [ eq - tcool ] ) but with @xmath15 replaced by @xmath62 . the reasoning behind this is that if we would instead use @xmath63 ( which might look more reasonable at first sight ) we will quickly cool to temperatures for which condition @xmath64 is again no longer fulfilled . this is because @xmath65 and thus very rapidly rises with declining temperature . if we insert eq . ( [ eq - tcool ] ) with @xmath15 replaced by @xmath47 into eq . ( [ eq - temp - cool - analyt ] ) we obtain a differential equation that can be solved by separation of variables . the solution is @xmath66^{-5/3}\ ] ] where here @xmath1 is again the original one with @xmath15 . we assume that for @xmath59 we have : @xmath67 the solution eqs . ( [ eq - t - analytic],[eq - t - analytic - zero ] ) holds true for the central temperature . but most of the mass of a homogeneous sphere resides in the outer parts . for the temperature at @xmath68 and @xmath69 ( roughly the radii of half mass and of 75% mass respectively ) the best fitting solution to the real solutions of section [ sec - radiative - transfer ] are similar to eq . ( [ eq - t - analytic ] ) but with the term @xmath70 replaced by @xmath71 and @xmath72 respectively . these analytic estimates of the temperature as a function of time are plotted in fig . [ fig - temp - afo - time - analytic ] . as one can see , although the time after the impact at which the cooling stars is different depending on whether you look at the center or at the edge of the cloud , the typical cooling rate after the cooling begins is similar throughout the cloud . the temperature evolution of the expanding cloud model according to the analytical estimate of eqs . ( [ eq - t - analytic],[eq - t - analytic - zero ] ) , shown in black . in grey are , for comparison , the full radiative transfer solutions from section [ sec - radiative - transfer ] for the fiducial model f1 ( see table [ tab - model - param ] ) . the temperatures are shown at three positions in the cloud : the center , at 80% of the radius and at 90% of the radius ( near the surface of the cloud ) . the time scale is scaled to @xmath1 ( eq . [ eq - tcool ] ) . the analytic solutions are only valid as long as @xmath73 ( cf . eq . [ eq - taucool]).,scaledwidth=50.0% ] it is important to remind ourselves that the cooling normally starts well before the cloud becomes optically thin . optically thin cooling would be extremely fast : a single chondrule would cool within about 1 second . the protracted cooling over a time scale of hours occurs because the high optical depth ( even after @xmath74 ) acts as a kind of blanket that keeps the chondrules warm , albeit a blanket that is thinning over time as the optical depth drops . to get a feeling for this we can calculate the optical depth at the time @xmath74 from eq . ( [ eq - tau - t ] ) : @xmath75 so eqs . ( [ eq - t - analytic],[eq - t - analytic - zero ] ) , which are based on the assumption of the `` blanket effect '' of high optical depth , are expected to be valid for the case @xmath73 . in table [ tab - model - param ] one can see that @xmath76 is well above 1 for all models , except the most extreme model f4 which has anyway a much too small cooling time to be consistent with chondrule textures . the analytic model of eqs . ( [ eq - t - analytic],[eq - t - analytic - zero ] ) tells us that for @xmath4 the temperature drops off suddenly on a typical time scale of @xmath1 . the temperature decay rate at @xmath74 is : @xmath77 this is plotted , together with @xmath1 , in fig . [ fig - tcool - afo - params ] . it is also useful to define another kind of cooling time scale @xmath78 which is the time between the start of the cooling at @xmath79 and the time when the temperature has dropped below the solidus temperature 1400 k. to first order we can write @xmath80 the quantities @xmath81 and @xmath78 can be compared to the constraints coming from the analysis of textures of chondrules , which put time limits on the cooling process of order of hours , or more precisely : cooling rates in the range @xmath82 k / hr ( see morris & desch 2010 and desch et al . 2012 and references therein , though see also miura & yamamoto 2014 for a different view based on theoretical modeling of crystallization ) , indicated by the grey area in fig . [ fig - tcool - afo - params ] . according to eq . ( [ eq - tcool ] ) @xmath1 goes as @xmath83 and as @xmath84 . there is not very much freedom of choice of @xmath15 : it must lie somewhere between @xmath85 and @xmath86 k ( see morris & desch 2010 and references therein ) . the @xmath87 does not have too much wiggle room either : the radii of chondrules are known . this means that constraints on @xmath1 from textural analysis directly set limits on the ratio @xmath88 or equivalently on the ratio @xmath89 . if , for example , we choose @xmath90 cm and @xmath91 k we obtain @xmath92 @xmath93/g . if we require , for instance , the cooling time scale to be @xmath94 minutes , then we find that @xmath95 g s@xmath96 @xmath97 this would be fulfilled e.g. for @xmath98 km / s and @xmath99 g ( a mass corresponding to a ball of magma of roughly @xmath1002 km radius ) . it would also be fulfilled by a smaller mass and smaller velocity : e.g. for @xmath101 m / s and @xmath102 g ( @xmath1000.1 km radius ) . such small mass could either mean that upon impact only a small fraction of the debris is in the form of chondrule droplets , or it could simply mean that the largest closed unit of the droplet splash ( subcloud ) is so small , but many such expanding cloudlets of chondrules are ejected . the elongated streams of debris found in asphaug et al . ( 2011 ) could be regarded as a string of such smaller mass cloudlets . this is shown more directly in fig . [ fig - rmelt - afo - vexp ] which shows the parameter space of the model , with the grey area again showing the models which give cooling rates between 10 and 3000 k / hr . the cooling time @xmath1 ( eq . [ eq - tcool ] ) as a function of parameters of the model for @xmath91 k and @xmath90 cm . note that the time to cool down from 2000 k to 1400 k ( @xmath78 ) is , according to eq . ( [ eq - t1400-definition ] ) , about four times smaller than @xmath1 . on the right axis the corresponding cooling rate @xmath103 in k / hr is plotted . the grey area is the observed range of chondrule cooling rates ( see main text ) . note that the values of @xmath1 are ( by definition ) for the center of the cloud . the regions of the cloud more closely to the surface will start to cool earlier , but the cooling rate is almost the same.,scaledwidth=50.0% ] the parameter space of the model . the grey area shows where the solutions have cooling rates in the range observed for chondrules . in the bottom - right the region is shown where @xmath104 , where the analytic solution becomes invalid.,scaledwidth=50.0% ] so far we have only made an estimate of the cooling behavior of the cloud of liquid chondrules . let us now calculate the temperature profile with a full treatment of time - dependent radiative transfer . the time - dependence comes in only due to the time it takes the lava droplets to convert their heat into radiation , not in the form of light - travel time ( which typically only plays a role for media at temperatures above @xmath105k ) . this can be easily verified by comparing the radiative energy density @xmath106 to the material energy density @xmath107 , which in our case always satisfies @xmath108 . therefore the radiative transfer equation itself is stationary , and the time - dependence is in the equation for heating / cooling of the chondrules . the formal radiative transfer equation is ( see e.g. mihalas & mihalas 1999 ) : @xmath109 where @xmath110 is the frequency - integrated mean intensity in erg s@xmath54 @xmath97 ster@xmath54 , @xmath111 is the position vector in space , @xmath112 is the unit vector of direction of the radiation , and finally @xmath113 the frequency - integrated planck function given by @xmath114 we will use frequency - integrated quantities because the opacity of a chondrule is expected to be roughly equal to the geometric cross section , independent of frequency , because the chondrules are much larger than the wavelength of infrared radiation . in this case the frequency - dependence of the radiative transfer does not need to be considered . ( [ eq - formal - rt - eq ] ) is called the `` formal transfer equation '' , and describes the transport of radiation along rays of direction @xmath112 . it is easy to integrate _ if _ the value of @xmath115 is known everywhere , but since we want to calculate the temperature @xmath116 , this function is part of the thing we want to solve . in our 1-d spherically symmetric setting eq . ( [ eq - formal - rt - eq ] ) can be written as @xmath117 where we shortened @xmath118 here @xmath119 where @xmath120 is the angle between the ray along which the radiation is followed and the radially outward pointing unit vector . the eddington factor method of solving this equation relies on the definition of the first three angular moments of the intensity at each location @xmath24 : @xmath121 here @xmath122 is called the mean intensity and @xmath123 is the flux divided by @xmath124 . by integrating eq . ( [ eq - formal - rt - eq - spher ] ) over @xmath125 after multiplying it by @xmath126 and by @xmath127 respectively one obtains the first two moment equations : @xmath128 these are two equations with three unknowns . to close this set of equations we write @xmath129 where @xmath130 is called the `` eddington factor '' . if @xmath131 is known , then the two moment equations are in closed form and can be solved for @xmath132 and @xmath133 . from the moment equations alone we can not know what @xmath131 is for all @xmath24 , but for certain limiting cases we know their values : for optically thick media @xmath134 . outside of the cloud for @xmath135 we will get @xmath136 , which is the free - streaming limit . however , for general @xmath24 and for general values of the optical depth we must employ another method of calculating @xmath131 . we do this by integrating , for the current value of @xmath137 ( and thus @xmath138 ) , the formal transfer equation eq . ( [ eq - formal - rt - eq - spher ] ) using the `` tangent ray method '' ( see e.g. mihalas & mihalas 1999 ) . by integrating the resulting intensities over angle we calculate the current estimates of @xmath132 and @xmath139 . let us call these @xmath140 and @xmath141 ( where `` fte '' stands for `` formal transfer equation '' ) . our current estimate of @xmath131 is then @xmath142 this is then the @xmath131 function we stick into eq . ( [ eq - k - afo - j - eddington ] ) , so that the moment equations can be solved for @xmath132 and @xmath133 . we also need to impose boundary conditions at the inner and outer edge . at the inner edge we take @xmath143 ( zero flux ) . at the outer edge we could set @xmath144 , which is valid if the outer edge of our computational domain is at @xmath135 . for finite outer radius we set instead @xmath145 note that we can choose to set the outer edge of our computational domain @xmath146 at @xmath147 , but we can also set it at @xmath148 , as long as we properly set the boundary condition at @xmath149 according to eq . ( [ eq - outer - bc ] ) . we can now combine the two moment equations into the following form : @xmath150 \right\}=j-\frac{\sigma}{\pi}t^4\ ] ] this can be solved , together with the boundary conditions , using a matrix equation . the details of this are discussed in appendix [ app - rt - eq - discrete ] . next we must compute how the temperature @xmath137 reacts to the radiation field , or in other words , how @xmath137 radiatively cools with time . the energy equation is @xmath151 where @xmath152 is given by eq . ( [ eq - rho - afo - time ] ) . the @xmath153 operator is the comoving derivative , i.e. the time derivative is computed for a given chondrule . ( [ eq - dtemp - dt ] ) is a local equation at each location . the non - locality of the heating / cooling is only through the non - locality of the equation for @xmath132 ( eq . [ eq - secorder - momeq ] ) . the way we solve these equations numerically is by setting up a radial grid of @xmath154 gridpoints @xmath155 ( with @xmath156 ) upon which we define the mean intensity @xmath157 and the temperature @xmath158 . since the cloud is expanding we let the gridpoints move along with the material ( lagrangian approach ) so that the location of grid point @xmath159 at time @xmath2 is @xmath160 where @xmath161 is the dimensionless radial coordinate @xmath162 the values of @xmath163 do not change with time . each chondrule stays at constant @xmath161 , and so the comoving time derivative @xmath153 used in e.g. eq . ( [ eq - dtemp - dt ] ) becomes simply the time derivative of that quantity at a given @xmath163 grid point . in principle one could imagine a simple time - dependent numerical integration method for the combination of eqs.([eq - secorder - momeq],[eq - dtemp - dt ] ) . we know @xmath164 and @xmath165 at time @xmath166 , and we can solve for @xmath167 by taking an explicit euler integration step of eq . ( [ eq - dtemp - dt ] ) . then we can solve the moment equations ( because now we have @xmath168 ) and obtain @xmath169 . the problem with this scheme is that the required time steps can become very small due to numerical stiffness . a more robust method is to integrate the complete system eqs.([eq - secorder - momeq],[eq - dtemp - dt ] ) using an implicit integration scheme . this is discussed in appendix [ app - rt - eq - discrete ] . the results of the time - dependent radiative transfer calculations are shown in fig . [ fig - tprofile-1 ] . + these results confirm the estimates made in section [ sec - estim - cooling ] . we find that the shape of the cooling curves is mostly the same for all models , except for the time scaling , which we know from the analytically expression of @xmath1 ( eq . [ eq - tcool ] ) . only for the rather extreme ( and presumably unrealistic ) case such as fiducial model f4 with @xmath170 km and @xmath171 m / s we find from the full radiative transfer model that the cooling is slower than the analytic solution this is because in that case the condition that @xmath76 ( eq . [ eq - taucool ] ) is @xmath172 is broken and the cooling time scale will then not be determined by the optical depth decline but by the effectiveness by which a chondrule can convert its heat into radiation . these cases , however , imply typically time scales of seconds rather than hours , and these are anyway too short to be consistent with chondrule textures . it seems that the analytic solutions are fairly accurate when @xmath173 , which is always true for the interesting regime of parameter space . from figs . [ fig - temp - afo - time - analytic ] and [ fig - tprofile-1 ] we see the model systematically predicts a period of constant temperature with a duration similar to the later cooling phase . this is true at least near the center of the cloud . near the surface this temperature plateau is shorter , but the cooling time is similar . in fig . [ fig - t - afo - r ] we show , for model f1 , how the temperature looks as a function of radial position in the cloud , for different times after the collision . since the cloud is expanding , which would make it difficult to compare the temperature profiles at different times , we use the radial coordinate relative to the outer cloud radius . one sees that the outer regions start cooling earlier because the optical depth to the surface is smaller . also one sees that a negative radial temperature gradient is produced which causes the radiative diffusive energy transport to the surface . the cooling curve from our model is significantly different from the ones predicted from shock heating ( e.g. morris & desch 2010 , morris et al . the shock heating models predict a radiative preheating phase , a temperature spike with rapid cooling , and depending on geometric conditions followed by a slower cooling phase . in our model the temperature is high from the start ( or for high - speed collisions : upon impact ) , stays relatively constant for some time and then drops over a similar time scale . the radial temperature temperature structure of fiducial model f1 for different times . the horizontal axis is the radius relative to the outer cloud radius . note that the temperature profile extends beyond the outer radius of the cloud because even if that region is empty , a chondrule residing there would have a well - defined temperature . for our purposes , however , only the temperature within the cloud is relevant.,scaledwidth=50.0% ] the scenario of splashes of melt droplets originating from colliding pre - molten planetesimals has been discussed at length in several papers by sanders and coworkers ( e.g. sanders & scott 2012 ) and the paper by asphaug et al . these papers discuss how this scenario holds up agains the many meteoritic constraints . we will not repeat these arguments here , but refer for those discussions to those papers . instead , we will focus on the discussion of aspects related to our splash cooling model . our model describes a simple spherical expanding cloud as a model for ( part of ) the splash resulting from colliding bodies . it is obvious that this is a drastic simplification . the sph model of asphaug et al . ( 2011 ) shows a much more realistic geometry involving , among other complexities , tube - like geometries . a proper 3-d time - dependent radiative transfer simulation done in concert with the gravito - hydrodynamic simulation of the impact splashing is required to know precisely what the cooling behavior is . nevertheless we predict that our result can be used to estimate the results of those detailed computations : the tube - like geometry can be approximated as a homologously expanding cylinder , expanding not only in the 2 perpendicular directions but also in the longitudinal direction . its properties are then not too much different from a chain of spherically expanding clouds with diameters similar to the diameter of the tube , except that these spherical cloud can not cool in all directions ( @xmath124 ) but only in part of the sky ( in the other part it will heat its neighbors and be heated by its neighbors ) . we would then expect that the cooling time @xmath1 will be a bit longer but not by a large factor . also , one can expect some parts of the splash to be ejected at slow speed compared to the escape speed , so that it may fall back to the remaining ( unsplashed ) bodies and accrete on them . this evidently breaks the assumption of homologous expansion , and as a result will also deform any initially spherical sub - cloud into a more pancake - like shape . the idea of regarding the splash cloud as a collection of spherical sub - clouds is then clearly no longer valid , not even in an approximative way . in such cases a full 3-d radiation - gravito - hydrodynamic simulation of the splash is unavoidable . however , multi - dimensional radiation - hydrodynamics is an enormously challenging problem , in particular for problems involving such extremely complex geometries . it is an interesting challenge for the future . for getting order - of - magnitude estimates we may be forced to stick , for the time being , with simple models such as the one presented in this paper . according to the model by asphaug et al . ( 2011 ) , by which the lava drop size is calculated from the initial pressure under which the magma was in the pre - molten planetesimals before the collision , the planetesimal sizes must have been 10 km or larger . since our cloud of lava droplets can be one of a number of such clouds ejected from the impact , the @xmath17 of our model can not be directly compared to that @xmath174 10 km size lower limit from asphaug . but it does give an indication that , if we adopt their model of the chondrule sizes , it seems reasonable to be looking to values of roughly the order of @xmath175 km or larger . to elucidate our reasoning : if we collide two fully molten planetesimals of 20 km radius , and we model the splash with 10 spherically expanding droplet clouds , then we would get @xmath176 km . in fig . [ fig - tcool - afo - params ] this means we are looking at the region of the diagram above the dotted line . there is a caveat , however : what tells us that it is just 10 clouds , not a 1000 ? this brings us to the complicated issue of the actual geometry of the splash . if we `` decompose '' the splash geometry into many small spherical clouds we must assure that these clouds can radiatively cool to the outside . in other words : we can not consider a big spherical cloud as consisting of many smaller sub - clouds , because they will irradiate each other and not cool much . but if the cloud geometry is a long extended streak , it can be considered as consisting of a chain of spherical subclouds . we are here confronted with the limitation of our spherical expanding cloud model . the reason why pre - melting seems to be necessary for the impact splash scenario to work is that very high speed collisions ( @xmath177 km / s ) needed to produce melt from cold planetesimals are presumably rare . however , pre - melting may lead to differentiation of the planetesimals , causing the iron and nickel to sink into the core and producing a basaltic mantle . chondrules formed from such objects would be very non - solar in composition . although sanders & scott ( 2012 ) argue that convection reduces the efficiency of the differentiation and that some amount of differentiation is in fact consistent with e.g. l and ll chondrites , it seems that this issue is not yet conclusively solved and more detailed and quantitative modeling of the differentiation process in these molten planetesimals may need to be done . however , a middle way is if planetesimals were pre - heated to elevated temperatures below the melting point . this is to be expected in particular for later generation planetesimals ( i.e. planetesimals formed at a time when a substantial fraction of the @xmath0al has already decayed ) . lower velocity collisions might then be sufficient to add the remaining energy needed to produce melt . the problem with this scenario is that it requires fine - tuning : the planetesimal must have formed at a time when the remaining abundance of @xmath0al is low enough not to melt the planetesimal but high enough to heat it to temperatures only slightly below the melting temperature . this might be too much coincidence . on the other hand , we know that some planetesimals were strongly differentiated : the parent bodies of iron and basaltic meteorites . if chondrules formed from collisions between planetesimals , one would expect some chondrules to have been formed from collisions with such differentiated planetesimals . some chondrules should have compositions that are very iron poor , presumably more iron - poor than ll - chondrites . this raises the question where are these ? one possible answer is that chondrules produced in these very early phases ( @xmath178 1 million years after cai formation ) were either accreted into the sun or reincorporated into other bodies , where they were again molten . but if this process of elimination is so efficient , why did some cais survive and get incorporated into the ( later ) chodrites ? this shows that while the impact splashing model is appealing , there is more work to be done to answer such questions . in our homologously expanding cloud model the melt droplets move away from their nearest neighbor at extremely low velocity ( millimeters per minute ) . small inhomogeneities ( which are of course unavoidable in such a messy process as an impact ) would lead to some droplets actually colliding . when they do so during the initial constant temperature phase , they presumably coalesce and form a larger melt droplet . however , when they collide during the cooling phase , they may have become cool enough to be plastic but not liquid anymore . they will then produce a compound chondrule . it would be useful if we could quantify the percentage of chondrules that become compound in this way . however , at present we see no way how we can estimate the random relative velocities between adjacent chondrules in the expanding clouds . many compound chondrules , however , clearly show one chondrule having remained rigid , while the other having been indented . so clearly they were of different temperature . this is difficult to achieve in our simple spherical expanding cloud model . for this it will be necessary to have other debris ( e.g. another cloud - fragment emerging from the impact site ) to interdisperse with our cloud of droplet , so as to have chondrules from different thermal environments to mix . and this must happen on a time scale shorter than the time scale of thermal equilibration between the chondrules from these different origins . to make this more quantitative would be , however , a challenge . it would likely require us to contemplate complex splash geometries . metzler ( 2012 ) found that some chondrites in fact contain entire large scale clusters of compound chondrules . a possible explanation could be that some of the impact debris may have fallen back onto the remainder of the biggest of the two colliding planetesimals . if this happens quickly enough , so that this re - accretion occurs before the chondrules have cooled below the temperature at which the chondrules are still sticky , then it is reasonable to assume that such clusters may form as chondrules simply fall onto each other . another problem with our simple spherically symmetric expanding plume model is that it does not account for any possible `` pollution '' of the melt droplets with other ( non - molten ) debris . some chondrules contain relict grains ( jones 1996 ) , which must have been inserted while the chondrule was still liquid , but the chondrule must have cooled very soon thereafer . as in the case with the compound chondrules ( section [ sec - comp - chond ] ) this would require debris from a different part of the impact debris cloud to interdisperse with our cloud of droplets . this foreign debris must have originated from non - molten parts of the original planetesimals , e.g. the regolith or pristine dust covering the crust of the original planetesimals . only those grains from that non - molten debris that entered late enough can enter the chondrules shortly before they cool below the solidus , and thus survive . like with the compound chondrules , however , getting quantitative estimates of this process is challenging . the impact model may provide a plausible scenario for the _ formation _ of chondrules , but how do these chondrules ( and matrix ) accrete to form another planetesimal ( the parent body of the later chondritic meteorite ) ? a nice aspect of the low - velocity collision of pre - molten planetesimals model is that much of the debris may , in fact , reaccrete onto one of the surviving original planetesimals ( or what remains of it ) . the idea is that due to the high gas density in the protoplanetary disk the stochastic planetesimal motions are of low velocity so that when they collide , they collide essentially at velocities that are only just above their mutual escape speed . since much of the energy is dissipated in the inelastic impact , most of the debris will not have enough kinetic energy to escape the system and will ( after perhaps a flyby or two ) eventually accrete onto a single body or a binary body . however , there is evidence that inside a single chondrite there exist chondrules of different ages ( villeneuve et al . and some chondrules experienced multiple heating events . somehow chondrules produced in multiple chondrule - forming impacts must have mixed together and form a single chunk of rock that later is observed as a chondritic meteorite . the immediate re - accretion scenario may make this somewhat difficult because once the collision releases most of the magma , and much of this is reaccreted as chondrules ( for the molten part ) and matrix ( for the non - molten part ) the remaining body will no longer be a sphere of magma . if , however , the collisions would be high - speed enough that much of the debris escapes , the newly formed chondrules will disperse into the protoplanetary disk . there they mix with the pool of other chondrules of various ages , and then later accrete into a chondrite parent body . an important question in this scenario is how aerodynamic drag and the resulting radial drift affects this : will the chondrules stay long enough in the asteroid - belt - region before they dift away or not ? jacquet et al . ( 2012 ) study this question in detail . perhaps more likely is a combination of the two . in particular the cluster chondrites of metzler ( 2012 , see section [ sec - comp - chond ] ) can be easily understood in terms of the reaccretion scenario while the multiple ages seems more to point toward a dispersion and later accretion scenario . one of the main characteristics of most chondrules is their `` normal '' abundance of volatile elements such as na , k etc . if a chondrule would stay at a high temperature ( e.g. 2000k ) for more than a few minutes , these elements would have evaporated out of the chondrule and leave a volatile - depleted chondrule behind . we believe that the impact splash scenario may naturally resolve this issue , because during the hot phases of the expanding impact splash the density of the cloud is so high that the vapor quickly reaches the saturation pressure and evaporation stops . however , this idea must be verified by actual modeling : will the evaporation indeed saturate ? will the temperature drop below the evaporation temperature before the density of the expanding cloud becomes too low to saturate the pressure ? answering these questions with explicit calculations is the topic of paper ii in this series . so far we have not made any distinction between different kinds of chondrules or chondrites . given that chondrule properties vary considerably between chondritic classes , the natural question arises : do all chondrules form through the same mechanism or are different chondrule formation scenarios responsible for different kinds of chondrules ? and for which kind , if at all , could the impact splash scenario of this paper be applicable ? a special class of chondrites and chondrules is the cb and ch class . it has been recognized for some time that these chondrules may indeed have originated from colliding planetesimals or planets ( see desch et al . 2012 for a discussion ) . the evidence for this includes ( 1 ) the cryptocrystalline structure of cb chondrules , ( 2 ) their strong depletion of volatile elements , ( 3 ) zoned metal spherules that appear to have been condensed out of the gas phase in a highly energetic single - staged process ( krot et al . 2005 ) and ( 4 ) their young age , 5 myr after cais . one can interpret these properties as meaning ( 1 ) that these chondrules must have cooled more rapidly ( seconds to minutes ) than `` normal '' chondrules ( hours ) to explain their textures , ( 2 ) that the vapor - melt plume of the impact must have rapidly become of low enough density to allow volatile elements to get out of the chondrules and not recondense , ( 3 ) the impact must have been of very high velocity ( several km / s ) to produce metal vapor , ( 4 ) which appears consistent with the young age which means low gas densities in the protoplanetary disk and thus allows for high - speed impacts . arguments ( 1 ) and ( 2 ) seem to require a rapid reduction of the density and optical depth , which can be achieved by a high impact velocity ( also required for ( 3 ) and consistent with ( 4 ) ) and a low mass of the impactor . krot et al . ( 2005 ) argue , instead , for a very high mass collision to explain cb chondrules . perhaps the adiabatic cooling of the impact - shock - compressed rock is enough to explain the fast cooling , or only a fraction of the mass from the giant impact actually gets converted into a plume of chondrules . in constrast to the case for the cb / ch chondrules our model may explain the properties of `` normal '' chondrules : ( 1 ) barred or porphyritic textures indicating slower cooling ( hours ) , ( 2 ) little depletion of volatiles ( see section [ sec - volatile - elements ] ) , ( 3 ) possibly lower temperatures and ( 4 ) earlier times after cais . the slower cooling rates come naturally out of our model because the optical depth of the cloud keeps the chondrules warm for a while . the little depletion of volatiles may be achieved by the high densities ( though a quantitative analysis has to wait until paper ii in this series ) . and the lower temperatures are natural for lower impact velocities which are expected in the earlier phases of the protoplanetary disk . the melt is the molten interior of the @xmath0al - heated planetesimals , which also is expected during the earlier phases . we might also be able to apply our radiatively cooling expanding cloud model to the high impact velocity scenario for cb / ch chondrules , but it would require us to include more physics , in particular a proper equation of state for the shock - heated rock and the subsequent adiabatic expansion and adiabatic cooling . among the `` normal '' chondrules there is also the issue of radial and barred textures versus porphyritic textures . the latter are the majority ( about 85% of chondrules , gooding & keil 1981 ) . radial textures require that no nucleation sites were present in the melt , so that crystallization in the supercooled melt droplet starts from a single site and progresses radially outward from that site . in the splash scenario this means that the melt in the original pre - molten planetesimals was well above the liquidus , so that any previously existing crystals were molten . conversely , for porphyritic chondrules numerous nucleation sites must have been present in the melt , and thus must have presumably remained present in the magma of the colliding planetesimals , meaning that that magma was likely slightly below the liquidus temperature . the impact may have heated the melt through shock - heating above the liquidus , but if the cooling sets in quickly enough some of these nucleation sites may survive . the question of porphyritic textures and the origin of the nucleation sites in the splash scenario remains , however , a tricky one . sanders & scott ( 2012 ) propose instead that these nucleation sites may have entered into the melt droplets as pollution from e.g. the surface regolith that may have been mixed in with the chondrules . the problem with this ( as with the relict grain issue of section [ sec - relict - grains ] ) is that the timing of mixing must be ideal : too early and the temperature is still above the liquidus and these regolith particles will also melt ; too late and the chondrules have already formed radial or barred textures . finally there is the issue of fe - content of chondrules : the h , l and ll classes . if the pre - molten planetesimals have experienced strong differentiation , it is possible that the mantles of these planetesimals are poor in iron and nickel . if the masses of the planetesimals are low enough and convection might be present , then perhaps all too strong loss of fe in the mantle can be avoided . the issue of h , l and ll chondrites in the context of the splash scenario is further discussed in sanders & scott ( 2012 ) . this paper studies the radiative cooling of an expanding cloud of lava droplets originating from the low - velocity collision between two partly or fully pre - molten planetesimals . the model is also valid for high - speed collisions between cold planetesimals , in which the melt is produced by the collision itself . the main result of this paper is a self - consistent cooling profile for chondrules based on the initial conditions of the expanding plume of melt droplets ( chondrules ) . these cooling profiles are characterized by a time scale @xmath1 given by eq . ( [ eq - tcool ] ) and are well approximated by a constant temperature for @xmath59 and a powerlaw dropoff ( eq . [ eq - t - analytic ] ) for @xmath4 . the cooling rate is given by eq . ( [ eq - temp - cool - analyt ] ) . the cooling time scale @xmath1 depends on the total mass of the plume and on its expansion velocity . both parameters are related to the conditions of the collision between the two planetesimals . a large total mass of the expanding cloud of chondrules means that the colliding planetesimals must have been at least of that mass , but presumably considerably more massive since not all of the matter is likely to end up in a single expanding plume . the expansion velocity of the droplet cloud is related to the collision velocity of the two planetesimals : it is unlikely that it is higher than the impact velocity . note that the expansion velocity is different from ( and presumably much less than ) the velocity by which the cloud moves away from the impact site . to relate both parameters ( @xmath14 and @xmath16 ) to paramters of the colliding planetesimals it is necessary to perform 3-d hydrodynamic simulations of the impact , such as those performed by asphaug et al . ( 2011 ) . the results of our model , when compared to known cooling time scales of chondrules , put constraints on @xmath14 and @xmath16 , and thus on the conditions under which planetesimal collisions can produce chondrules . * acknowledgements : * we would like to thank andreas pack , knut metzler , alessandro morbidelli , david lundberg , addi bischoff , neal turner , stephan henke and mark swain for useful discussions and feedback . this paper benefitted a lot from detailed comments and suggestions by an anonymous referee , for which we are very grateful . s. stammler is funded by the deutsche forschungsgemeinschaft ( dfg ) grant du 414/12 - 1 . 100 alexander , c.m.o . 1996 , chondrules and the protoplanetary disk , 233 alexander , c.m.o . , ebel , d.s . , 2012 , meteoritics and planetary science , 47 , 1157 alexander , c.m.o . , grossman , j.n . , ebel , d.s . , ciesla , f.j . , 2008 , science , 320 , 1617 asphaug , e. , jutzi , m. , movshovitz , n. , 2011 , earth and planetary science letters , 308 , 369 boss , a. 1996 in proceedings `` chondrules and the protoplanetary disk '' , pp 257 . bottke , w.f . 1994 , icarus , 107 , 255 ciesla , f.j . , hood , l.l . , 2002 , icarus , 158 , 281 connelly , j.n . , bizzarro , m. , krot , a.n . , nordlund , a. , wielandt , d. , ivanova , m.a . , 2012 , science , 338 , 651 cuzzi , j. , alexander , c.m.o . , 2006 , nature , 441 , 483 davis , a.m. , alexander , c.m . od . , ciesla , f.j . , gounelle , m. , krot , a.n . , petaev , m.i . , stephan , t. , 2014 , in protostars and planets vi , eds . beuther , klessen , dullemond , henning . desch , s.j . , connolly , h.c . , 2002 , meteoritics and planetary science , 37 , 183 desch , s.j . , morris , m.a . , connolly , h.c . & boss , a.p . 2012 , meteoritics and planetary science , 47 , 11391156 gibbard , s. g. , levy , e. h. & morfill , g. e. 1997 , icarus , 130 , 517 gooding , j.l , keil , k. , 1981 , meteoritics , 16 , 17 hevey , p.j . , sanders , i.s . , 2006 , meteoritics and planetary science , 41 , 95 hewins , r.h . , 1983 , in : chondrules and their origins , houston , pp 122 hewins , r.h . & connolly , h. c. , jr . 1996 , in chondrules and the protoplanetary disk ( cambridge : cambridge univ . press ) , hood , l.l . , horanyi , m. , 1991 , icarus , 93 , 259 hubbard , a. , mcnally , c. p. & mac low , m .- m . 2012 , apj , 761 , 58 jacquet , e. , gounelle , m. & fromang , s. 2012 , icarus , 220 , 162 jones , r.h . , grossman , j.n . , rubin , a.e . 2005 , in chondrites and the protoplanetary disk , asp conference series , vol 341 , eds : krot , scott & reipurth kieffer , s. , 1975 , science , 189 , 333 krot , a. n. , amelin , y. , cassen , p. & meibom , a. 2005 , nature , 436 , 989992 lambrechts , m. , & johansen , a. , 2012 , a&a , 544 , a32 libourel , g. , chaussidon , m. , krot , a.n . , 2008 , in 39th lunar and planetary science conference , vol . 39 , pp 2017 metzler , k. , 2012 , meteoritics and planetary science , 47 , 2193 metzler , k. , bischoff , a. & stoeffler , d. 1992 , geochimica et cosmochimica acta , 56 , 2873 mihalas , d. & mihalas , b , `` foundations of radiation hydrodynamics '' courier dover publications , 1999 miura , h. & yamamoto , t. 2014 , the astronomical journal , 147 , 54 morris , m. a. & desch , s. j. 2010 , the astrophysical journal , 722 , 1474 morris , m. a. , boley , a. c. , desch , s. j. & athanassiadou , t. 2012 , apj , 752 , 27 neumann , w. , breuer , d. & spohn , t. , 2012 , a&a , 543 , 141 ormel , c. w. , & klahr , h. h. , 2010 , a&a , 520 , a43 sanders , i.s . , scott , e.r.d . , 2012 , meteoritics and planetary science , 47 , 2170 sanders , i.s . , taylor , g.j . , 2005 , chondrites and the protoplanetary disk , 341 , 915 sears , d. , 2004 `` the origin of chondrules and chondrites '' , cambridge university press stammler , s.m . & dullemond , c.p . 2014 , icarus in press . taylor , g. j. , scott , e. r. d. & keil , k. 1983 , in : chondrules and their origins ( a85 - 26528 11 - 91 ) . houston , 262 villeneuve , j. , chaussidon , m. , libourel , g. , 2009 , science , 325 , 985 urey , h.c . , craig , h. 1953 , geochimica et cosmochimica acta , 4 , 36 zook , h.a . , 1980 , meteoritics , 15 , 390 solving the radiative transfer moment equations coupled to the time - dependent heating / cooling equation requires an implicit integration scheme . the two coupled equations to solve are : @xmath179 \right\}-j+\frac{\sigma}{\pi}t^4\ ] ] and @xmath180 the inner boundary condition is : @xmath181 where @xmath182 is the smallest radius of our grid , which we take @xmath183 . this condition translates into @xmath184_{r = r_{\mathrm{in } } } = 0\ ] ] the outer boundary condition is @xmath185 with @xmath186 , which translates into @xmath187_{r = r_{\mathrm{out } } } = 0\ ] ] the only time - derivative in the equations is the one on the temperature . the radiation field is assumed to immediately adapt . we put @xmath188 and @xmath39 on a spatial grid @xmath189 . so for time step @xmath190 we have the values @xmath191 and @xmath192 . the above equations are also to be evaluated at these grid points : @xmath193 and @xmath194 . let us define @xmath195 and @xmath196 the objective of the time - integration of these equations is to find the values of @xmath197 : the values of @xmath198 at the next time step . for numerical stability we express all instances of @xmath198 in the equations @xmath199 in their future form : @xmath197 , except when a time - derivative of @xmath198 is used , which is written as @xmath200 and where the present value of @xmath201 is required . we thus get as our set of equations : @xmath202 which we need to solve for @xmath197 . if @xmath199 were linear in @xmath197 then one could write the vector @xmath199 as a matrix multiplication with the vector @xmath197 plus a contant vector @xmath203 : @xmath204 then the solution to @xmath205 would be just a matrix inversion : @xmath206 however , @xmath199 is non - linear in @xmath197 because the planck function @xmath207 . and so we have a non - linear function @xmath208 . so let us start with an initial guess @xmath209 . we typically then have @xmath210 ideally we want to find a @xmath211 for which @xmath212 , but we can only use newton s method by approximating @xmath213 using first order taylor expansion : @xmath214 or more in general for the @xmath215-th iteration : @xmath216 if we define @xmath217 we can solve for @xmath218 : @xmath219 and then we obtain the new : @xmath220 now let us write out the discrete equations @xmath221 and @xmath222 explicitly : @xmath223 for the gas equation we get : @xmath224 boundary conditions only have to be applied to the first equation . the inner boundary condition : @xmath225 the outer boundary condition : @xmath226 the energy equation can be rescaled to : @xmath227 the matrix coefficients then become : @xmath228 the matrix coefficients for the boundary conditions become : @xmath229

a new version of the barcelona - catania - paris ( bcp ) energy functional is applied to a study of nuclear masses and other properties . the functional is largely based on calculated ab initio nuclear and neutron matter equations of state . compared to typical skyrme functionals having 10 - 12 parameters apart from spin - orbit and pairing terms , the new functional has only 2 or 3 adjusted parameters , fine tuning the nuclear matter binding energy and fixing the surface energy of finite nuclei . an energy rms value of 1.58 mev is obtained from a fit of these three parameters to the 579 measured masses reported in the audi and wapstra 2003 compilation . this rms value compares favorably with the one obtained using other successful mean field theories , which range from 1.5 - 3.0 mev for optimized skyrme functionals and 0.7 - 3.0 for the gogny functionals . the other properties that have been calculated and compared to experiment are nuclear radii , the giant monopole resonance , and spontaneous fission lifetimes .

1210.1321

the physics of atomic nuclei and nuclear systems is very rich but extremely complicated to describe . it unites many of the toughest aspects of the quantum many body problem in a single spot . to name only a few , we can mention the existence of four different types of fermions ( proton / neutron - spin up / down ) , the relevance of pairing and quartet correlations in the dynamics , or the role of the mean field approach as a first order approximation while quantum fluctuations are very important as a consequence of the smallness of nuclei . moreover , the nucleon - nucleon ( nn ) force is very complicated with its hard core , tensor ( dipolar ) components and strong spin orbit interactions . it is remarkable that nuclear theory has nowadays mastered a great deal of those challenges and many of the experimental data are very well described either on a phenomenological basis , or , more rarely , with microscopic approaches . in this work we try to progress with the latter strategy . theories aiming towards a global description of low energy nuclear data always start with a mean field description and a semi - phenomenological nuclear energy density functional . in a recent paper @xcite , we proposed an approach slightly different from the usual for the establishment of a nuclear energy density functional . we tried to follow as much as possible the kohn - sham ( ks ) @xcite density functional theory ( dft ) , which is based on the hohenberg - kohn theorem @xcite , and is of common use , e.g. , in atomic and molecular physic @xcite . the ks method introduces a set of auxiliary single - particle states and takes for the kinetic energy the slater form in this basis . in the ks method the energy density functional is written as a sum of two terms . one of them corresponds to the uncorrelated kinetic energy density plus the direct contribution derived from the underlying two body ( coulomb ) interaction . the other piece contains an unknown term for the exchange and correlation energies . for a practical implementation of the method , the latter term has to be guessed and therefore many strategies have been developed along the years to describe different physical phenomena . a popular strategy in atomic and molecular physics is to use accurate theoretical calculations ( mainly by monte - carlo techniques ) as a function of the density in simplified models where finite size effects are neglected . the results of those calculations are used to make an educated guess of the unknown exchange and correlation part of the functional in the real system . in this way the minimization procedure gives hartree - like equations for the single - particle states , where the potential part includes in an effective way the overall exchange and correlation contribution . the application of the ks - dft scheme to self - bound systems like atomic nuclei is not straightforward , because of the absence of a external confining potential . the usual procedure used with skyrme functionals has been justified recently @xcite by reformulating the problem in the intrinsic " system . in nuclear physics to build up the exchange and correlation contribution amounts to a detailed determination of the ground state of symmetric nuclear matter and of neutron matter to be used as a guide for finite nuclei energy density functionals . accurate nuclear matter calculations are very complicated and not as much advanced as for electron systems of condensed matter . nonetheless , in recent years , quite some progress has been achieved and we will build our ks - dft approach on the equation of state ( eos ) of baldo et al . , developed in refs . @xcite , which is based on the well known hole line expansion . this eos is calculated from realistic two body and three body interactions treated at the brueckner hartree - fock level . it is compatible with phenomenological constraints coming from data of heavy ion reactions @xcite as well as from the analysis of astrophysical observations @xcite . it is worth mentioning that only two works exist which tried to follow the same strategy as here , one by fayans @xcite and the other by steiner et al @xcite . however , they are based on a somewhat older eos @xcite . the functional obtained will be used to describe finite nuclei at the mean field level concentrating on the binding energy systematic . therefore , it is important to include beyond mean field effects having a strong impact on binding energies and not present in nuclear matter . the most obvious one is the correlation energy associated to symmetry restoration that will be accounted for in the case of translational and rotational invariance . a different approach to nuclear energy density functional ( edf ) was developed on the basis of the chiral two- and three body forces @xcite . the final form of the edf contains a set of contact interactions , similar to the skyrme functionals , with a corresponding set of about a dozen of parameters . this part is supplemented by a contribution coming from pion exchanges , which is non - local and treated through the density matrix expansion procedure . preliminary calculations @xcite on a selected set of nuclei show a slight improvement with respect to the standard skyrme functionals and open a promising prospect for large - scale fitting procedures . the reason for such an intense focus on an accurate theoretical description of bulk properties of nuclei all over the nuclide chart is discussed for instance in @xcite : many physical scenarios involve neutron rich nuclei which are far from reach from an experimental point of view and therefore reliable theoretical predictions are the only possible option at present . the present status is that good agreement between different theoretical models is to be expected as long as we do not move towards the neutron drip line , a region of relevance in stellar nucleo - synthesis processes . therefore , a new model based on somewhat different ideas closer to the ks - dft approach than the more traditional skyrme - like edf ( inspired by contact central potentials ) , can help to clarify the uncertainties associated to present - day models . in the present work , a ks approach based on the same microscopic bulk input as in @xcite will be used for the particle- hole channel . however , we will be able to reduce the number of parameters by two without losing accuracy . as before , we will consider an additional bulk parameter for a precise adjustment of the numerically obtained e / a value . however , for the surface , we will reduce the number of parameters from three to one by the condition that , in infinite symmetric and infinite pure neutron matter , the strength parameters of the finite range term reproduce the coefficients of the quadratic terms of the bulk energy density obtained from the microscopic calculation . therefore , no subtraction term in the finite range part as in @xcite is necessary . the range @xmath0 of the surface term remains the only adjustable parameter , see the more detailed discussion below . although there are , together with the strength of the spin - orbit term , three parameters in the particle - hole channel , we want to emphasize that the most relevant are the ones from bulk and surface . we stress this point , because it seems to us a reduction to two basic physical inputs to the binding energy of nuclei : energy per particle of infinite matter and surface energy ! that this is possible is as much surprising as it is gratifying ! it also should be mentioned that the adjustment of those two parameters is extremely sensitive : both have to be fine tuned to the order of @xmath1 ! this sensitivity points to a well defined physical content of the parameters . it should be pointed out , however , that ks - dft addresses only the ground state and is , in principle , not tailored to describe excited states of the system . nevertheless , we will also use it for the description of the giant monopole resonances ( gmr s ) . the paper is organized as follows : the following section is devoted to briefly recall our previous bcp functional . in the third section , the new energy density functional built up in this paper is discussed . the results obtained with this improved functional , that we will call bcpm ( barcelona - catania - paris - madrid ) , are also presented in the same section . the predictive power of the bcpm functional regarding other observable such as charge radii , quadrupole and octupole deformations and fission barriers are discussed in the fourth section . the ability of the bcpm functional for describing the isoscalar giant monopole resonance is the subject of the fifth section . finally , the summary and conclusions are given in the last section . the former bcp functional was proposed in ref . this and subsequent refinements @xcite are based on the kohn - sham density functional theory ( ks - dft ) where the one body density @xmath2 plays a central role . in the ks - dft theory an auxiliary set of @xmath3 orthonormal wave functions @xmath4 , where @xmath3 is the mass number , is introduced to express formally the density as if it were obtained from a slater determinant as a sum of the product of single particle wave functions @xmath5 with the @xmath6 s determined from the minimization of the ground state energy . in condensed matter and atomic physics the energy density functional @xmath7 $ ] is usually split into two parts @xcite : @xmath8 + w[\rho].\ ] ] the first piece @xmath9 $ ] corresponds to the uncorrelated kinetic energy written in the usual manner as @xmath10 the other piece , @xmath11 $ ] , contains the potential energy and the correlated part of the kinetic energy . in the bcp family of functionals this part is given as the sum of spin - orbit energy , the coulomb part and the nuclear energy term which depends on the neutron and proton densities @xmath12 and @xmath13 , respectively . in nuclear physics , contrary to the situation in condensed matter and atomic physics , the contribution of the spin - orbit interaction to the energy functional is very important . non - local contributions have been included in dft in several ways long ago ( see ref . @xcite for a review of this topic ) . in line with this , we split the spin - orbit part into an uncorrelated part @xmath14 plus a remainder not treated explicitly here . the form of the uncorrelated spin - orbit part is taken exactly as in typical skyrme @xcite or gogny forces @xcite . another piece that we explicitly split off is the coulomb part @xmath15 . we treat this contribution at lowest order , i.e. the direct term plus the exchange contribution in the slater approximation , that is @xmath16 and @xmath17 with @xmath18 . the nuclear energy functional contribution @xmath19 $ ] contains the nuclear potential energy as well as additional correlations . this contribution is divided into a finite range term @xmath20 $ ] to account for correct surface properties and a bulk correlation part @xmath21 $ ] that we take from one of the most advanced microscopically determined eos existing so far in the literature @xcite as mentioned before . collecting all these contributions our final ks - dft functional reads @xmath22 the functional is supplemented by a density dependent pairing interaction and some beyond mean field corrections ( see below ) . one of the prominent results of our recent work within the ks - dft bcp density functional scheme @xcite was that , with a very reduced number of adjustable parameters to finite nuclei , rms values for binding energies and radii came out to be of the same quality as the ones of some skyrme @xcite or gogny @xcite functionals . the reduction of the number of parameters stemmed from the fact that , prior to a fit of the functional to nuclear masses , the bulk part of the functional @xmath23 was adjusted to the microscopic eos obtained in refs . @xcite by fitting polynomials in the total density @xmath24 to microscopic results in symmetric and pure neutron matter , followed by a quadratic interpolation in the asymmetry parameter @xmath25 between these two limits . the curve obtained in the polynomial fit can be slightly changed without changing significantly the mean square deviation to the microscopic eos . on the other hand , although the microscopic eos is state - of - the - art , the uncertainties in the underlying interaction and the ones attributed to the approximations involved translate to minor uncertainties in the computed points of the microscopic eos that will eventually lead to minor changes in the polynomial curve . we will exploit this limited freedom to produce polynomial fits with slightly different values of e / a at saturation ( around e / a=16 mev ) . the resulting edfs are then used in the finite nuclei fit ( see below ) and the one leading to the smallest binding energy rms is chosen . only changes of the order of a few hundred of kev in e / a are considered . the additional surface term in @xcite was of the form @xmath26 = & \frac{1}{2 } & \sum_{t , t'}\!\!\iint \!\ ! d { \bf r } d { \bf r}'\rho_{t}({\bf r})v_{t , t'}({\bf r}-{\bf r}')\rho_{t'}({\bf r } ' ) \\ \nonumber & - & \frac{1}{2}\sum_{t , t'}\gamma_{t , t'}\int d { \bf r}\rho_{t}({\bf r})\rho_{t'}({\bf r})\end{aligned}\ ] ] where the index @xmath27 is the label for neutron and proton , i.e. @xmath28 , and @xmath29 the volume integral of @xmath30 . the subtraction term was introduced not to contaminate the bulk part , determined from the microscopic infinite matter calculation . for the finite range form factor @xmath30 a simple gaussian ansatz @xmath31 was made . the strength parameters were chosen to distinguish only between like and unlike particles , i.e. @xmath32 so that the finite range term contained three adjustable parameters : @xmath33 , and @xmath34 . together with the adjustment of the bulk and the spin - orbit strength , the functional in @xcite has all in all five adjustable parameters , even though they are not all to be considered on the same level of significance . we will come back to these considerations later in the text where we also will show how to avoid the subtraction term in the finite range part of the energy . in order to describe properly open shell nuclei one has to introduce pairing correlations . the formal way of including pairing within the dft is also known since long from the generalization of the hohenberg - kohn theorem to paired systems @xcite . in @xcite we have introduced pairing in our functional through the bcs approach using a zero - range density dependent interaction adjusted to reproduce the neutron gaps of the gogny force in symmetric nuclear matter with @xmath35 @xcite . the free parameters of the energy density functional reported in @xcite were fitted either to reproduce the binding energies of a few spherical nuclei or binding energies and charge radii of the same set of spherical nuclei . we called these functionals bcp1 or bcp2 ( barcelona - catania - paris ) , respectively . the predictive power of these functionals was illustrated by computing the binding energies and charge radii of 161 spherical nuclei between @xmath36ne and @xmath37u . we found that the rms in energy and charge radii are similar to the ones obtained with successful mean - field models such us the gogny d1s force @xcite , the skyrme sly4 interaction @xcite and the relativistic mean field parametrization nl3 @xcite . in subsequent publications we explored the properties of bcp1 and bcp2 in describing quadrupole and octupole deformation ground state properties , fission , excited octupole states , etc @xcite . in dealing with deformed nuclei the most relevant beyond mean field effect , namely the rotational energy correction has been considered in an approximate way @xcite . also some other parametrizations of the functional based on fitting to the binding energy of deformed nuclei instead of the fitting to spherical nuclei , have been considered @xcite . we would like to point out that for other energy density functionals , like the ones of the skyrme , gogny , or relativistic mean field type , the number of parameters seems much higher , typically more than ten . however , many of those parameters are implicitly used to get a reasonable nuclear matter and neutron matter equation of state . the advantage of our ks - dft procedure is that one clearly separates the tasks of reproducing the nuclear matter eos from the most prominent finite size effects , namely the surface and coulomb energies and the spin - orbit potential . following @xcite we write the bulk part of the energy density functional as @xmath38=\int\ ! d{\bf r } \big[p_{s}(\rho)(1-\beta^{2 } ) + p_{n}(\rho)\beta^{2}\big]\rho\label{eq : eq4a},\ ] ] where the interpolating polynomials for symmetric and pure neutron matter , @xmath39 and @xmath40 respectively , read @xmath41 where the reference densities are @xmath42=0.16 @xmath43 and @xmath44=0.155 @xmath43 , respectively . as compared to the former version of the bcp energy density functional , the fit of the microscopic eos has been redone in order to avoid some rather strong oscillatory behavior in the density - dependent incompressibility of the nuclear matter ( see section iv below ) when plotted as a function of density . to cure this unwanted effect , the number of theoretical points is increased and a fifth order polynomial in the density is chosen for fitting the microscopic eos in symmetric nuclear matter in a wider range of densities , up to 0.625 @xmath43 , instead of the prescription given in @xcite , see also @xcite . furthermore special care has been paid to the smoothness not only of the fitted eos but also of the density - dependent incompressibility . this has been achieved by considering only polynomial fits with smooth high order derivatives with respect to the density . in fig . [ fig : eos ] the microscopic eos for nuclear and neutron matter as well as the corresponding polynomial fits are shown as function of the density . in the lower panel , the symmetry energy computed using the polynomial fits as a function of the density is also displayed . as discussed in @xcite , the low density behavior of the microscopic nuclear matter eos has a characteristic trend , usually not reproduced by skyrme and gogny functionals ( see also ref . @xcite ) , missing there quite a substantial part of binding . since we want to construct the ks - dft functional on the basis of the microscopic calculations , the bulk part @xmath23 of the functional , directly related to a realistic eos , is determined once and for all as in eq . ( [ eq : eq4a ] ) together with the local density approximation . however , as mentioned before slightly different polynomial interpolations with different values of e / a at saturation have been considered to improve the finite nuclei results for the binding energies . the saturation density has been kept fixed at the nominal value of 0.16 @xmath43 . for e / a at saturation we have explored the interval between 15.97 and 16.03 mev in steps of 0.01 mev . for each value of e / a the complete fitting process for finite nuclei has been performed ( see below ) . the optimal value for the saturation energy per particle turns out to be 15.98 mev . the fine tuning of e / a at saturation is to be expected because of the extreme sensitivity of the mass of heavy nuclei to this parameter . in addition , as the fine tuning involves finite nuclei calculations where the surface term is included it can be considered as a way to account for the coupling of volume and surface in finite nuclei . .parameters ( in mev ) of the polynomial fits , @xmath45 and @xmath46 for symmetric and neutron matter , respectively . the reference densities are @xmath47 @xmath43 and @xmath48 @xmath43 . two sets of parameters for symmetric matter are given , they correspond to eos with a minimum at the given values of @xmath49 . the parameters are given with six digits , enough to obtain binding energies with an accuracy of a couple of ev . the value @xmath50 mev has been used . [ tab : polpar ] [ cols="^,^,^,^",options="header " , ] experimental information about the excitation energy of isoscalar giant resonances in medium and heavy nuclei is obtained through inelastic scattering of @xmath51 particles of few hundreds of mev measured at extremely forward angles . the multipole decomposition analysis allows to extract from the scattering data the strength associated to different multipolarities , in particular those corresponding to monopole and quadrupole . from these strengths , the position and width of the isgmr and isgqr peaks can be determined @xcite . we analyze first the excitation energies of isgmr ( @xmath52 ) of several nuclei for which the experimental values are known @xcite . we compare these values with the theoretical predictions @xmath53 and @xmath54 provided by the bcpm energy density functional . the relevant information about these excitations energies is collected in tables [ t1 ] and [ t2 ] . as far as we are including in our study tin and cadmium ( neglecting the effects of the small deformation ) isotopes and @xmath55sm , for which pairing correlations are essential to describe accurately the ground state , we discuss first the role of pairing in the sum rule approach used to estimate @xmath52 . the generalization of rpa to the case of nuclei with open shells when pairing correlations are active , is the quasi - particle rpa ( qrpa ) . the so - called dielectric theorem , which justifies constrained calculations within qrpa to obtain the @xmath56 sum rule , has been proven recently by col and collaborators in @xcite . also khan et al have shown the validity of the thouless theorem for the energy weighted sum rule in the case of self - consistent qrpa based on hfb @xcite . as far as a simple generalization of the thouless theorem allows to obtain the @xmath57 sum rule @xcite , it can be expected that this also will be the case in the qrpa framework . in addition we have also numerically checked that the virial theorem , i.e. the stability against scale transformations of the wave - functions in eq . ( [ eq2 ] ) , is also fulfilled in the hfb approximation . from this discussion we may conclude that the sum rule approach can also be applied in the qrpa case . in table [ t1 ] we display the theoretical scaled and constrained estimates ( including pairing ) of @xmath52 of the nuclei @xmath55sm , @xmath58pb , @xmath59zr , @xmath60sn , and @xmath61cd together with the corresponding experimental values . these estimates are given by @xmath53 and @xmath54 , respectively , defined previously , which , as mentioned , fulfill @xmath62 ( see appendix ) . the escape width ( [ eq21 ] ) lies in a window of 2 - 3 mev , pointing out that the monopole strength of the considered nuclei is mainly concentrated around well defined peaks . the influence of pairing correlations on @xmath52 has been discussed in earlier literature ( see @xcite and references therein ) . we have repeated the scaled and constrained estimates of the @xmath52 switching off pairing and the corresponding results are reported in table [ t2 ] . comparing tables [ t1 ] and [ t2 ] we notice that , when pairing correlations are taken into account ( table [ t1 ] ) , the @xmath54 and @xmath53 values in open shell nuclei slightly decrease as compared with the values obtained without pairing ( table [ t2 ] ) which in agreement with the results of refs . the effect of pairing correlations is , in general , more important in the constrained calculations than in the scaled ones , especially in half - filled major shell nuclei as it is the case of @xmath63sn , @xmath64sn , @xmath65sn , @xmath66cd , @xmath67cd , @xmath63cd . when the occupancy of the shell increases , as it happens in @xmath68sn , @xmath69sn , @xmath70sn , @xmath71sn , @xmath64cd and @xmath65cd , the influence of pairing is actually very small as it can be seen in table [ t2 ] . in order to compare with the experiment , we will use as representative the constrained energy @xmath54 , which estimates the energy of the centroid in a very precise way at least in the case of the nucleus in @xmath58pb @xcite ( see also table iv of @xcite in this respect ) . from table [ t1 ] , we see that the experimental @xmath52 in @xmath58pb is very well reproduced by @xmath54 obtained using the bcpm energy density functional . the experimental excitation energies of @xmath59zr and @xmath55sm are slightly overestimated by the theoretical constrained calculation . however , our theoretical calculation are less successful in the case of the cadmium and tin isotopes . the predicted @xmath54 values are shifted by around 1 mev up as compared with the experimental results @xcite . this is a general tendency shown by mean - field models of different nature @xcite . several attempts to explain this behavior have been proposed in the past such as the density dependence of the incompressibility in neutron rich matter @xcite or different types of pairing interactions @xcite . however , no clear explanation has been found to date and the question why sn ( and cd ) isotopes are so soft is still an open problem . it may be that nonlinear effects coming from particle vibration coupling play a role in such nuclei as sn isotopes , since we reproduce the stiff doubly magic nucleus of @xmath58pb and not the semi - magic ones as , e.g. , the sn isotopes . from the previous discussion it seems that the @xmath52 estimated with the bcpm functional are somewhat high as compared with the experiment in spite of its relatively low incompressibility modulus @xmath72=216 mev ( see table [ nmatter ] ) . the fact that effective mean field models with different values of @xmath72 can give similar values of @xmath52 in @xmath58pb has actually been a general puzzle in the past ( see @xcite and references therein ) . the reason is the proportionality between @xmath52 and the square root of @xmath73 found by blaizot @xcite for gogny forces . for example rmf approaches systematically lead to @xmath52 comparable to the ones obtained with non - relativistic theories like skyrme or gogny , in spite of the fact that @xmath73 values are appreciably larger in relativistic than in non relativistic approaches . some time ago col et al . @xcite showed that one can construct skyrme functionals that give reasonable values for @xmath52 but with relatively high values of @xmath73 i.e. values which approach the ones of rmf theories . as discussed in @xcite , the underlying reason of this fact can be attributed to the different density dependence of the symmetry energy at , and around , saturation . in the same reference @xcite it is concluded that within non - relativistic models there is not a unique relation between the value of @xmath73 associated to a given model and the energy and the @xmath52 predicted by the same model . to clarify this latest problem a very recent proposal of khan , margueron and vidaa @xcite may be useful . these authors analyzed a relatively large set of skyrme , gogny and rmf interactions finding that the corresponding density dependent incompressibilities @xmath74 defined in eq . ( [ incomp ] ) show a crossing point at a density of about 75% of the saturation density . this crossing density corresponds roughly to the average density of a heavy nucleus defined as _ av = ^2 ( ) d . the authors claim that the excitation energy of finite nuclei is more properly correlated with the slope of incompressibility @xmath74 computed at the crossing density than with the incompressibility modulus computed at saturation density @xmath72 . this scenario is similar to the one found for the nuclear symmetry energy in bulk matter , which also shows a crossing point for a similar value of the density ( @xmath750.11 @xmath43 ) @xcite . in this case it is also found @xcite that the slope of the symmetry energy at the crossing density is linearly correlated with the neutron skin of a heavy nucleus , e.g. @xmath58pb . in general , the existence of a crossing point is actually related to the fact that the parameters of successful mean- field models have been fitted to reproduce properties of finite nuclei whose densities are , on the average , smaller than the saturation density . therefore , if an observable measured in finite nuclei is related to some property of nuclear matter , the correlation is better described at the average ( crossing ) density than at the saturation one @xcite . following the suggestion of ref . @xcite , we have analyzed the behavior of @xmath52 of @xmath58pb and @xmath59zr calculated using different mean field models as a function of the parameter @xmath76 defined as m = 3 _ _ av , which is directly related to the slope of the density dependent incompressibility @xmath74 at the crossing density @xmath77 . in our study in addition of the bcpm energy density functional , we have also considered the d1 and d1s gogny forces , the skm@xmath78 and t6 skyrme interactions and the rmf parametrizations nl1 , nl3 and nlsh . using these models the correlation between the scaled and constrained estimates of @xmath52 s and the parameter @xmath76 are displayed in fig . [ f2 ] for the two aforementioned nuclei . the data of the @xmath52 predicted by gogny forces and rmf parametrizations are taken from refs . @xcite and @xcite respectively . from this figure it can be seen that the @xmath52 follows a linear trend with the parameter @xmath76 ( the correlation parameter is r=0.985 in all the considered cases ) with high accuracy , in agreement with the suggestion of @xcite . consequently , and as it is claimed in @xcite , the density dependence of @xmath74 can be constrained by the @xmath52 measured in finite nuclei . the same correlation is also fulfilled by the nucleus @xmath59zr with @xmath76 evaluated at the same average density . this fact points out that the dependence on the mass number of the average density is actually rather weak for medium and heavy nuclei @xcite . from the lower panel of this figure we also see explicitly that considering globally mean field models of different nature , the linearity between @xmath52 and the square root of the incompressibility modulus at saturation @xcite is lost in some cases , in agreement with the conclusion of ref . @xcite . to investigate this aspect in more detail , we plot in the upper panel of fig . [ f3 ] the bulk incompressibility of eq . ( [ incomp ] ) as a function of the density . in the lower panel we show , magnified , the relevant region of the upper panel . for densities close to 0.72@xmath42 the density dependent incompressibility @xmath74 computed with skyrme and gogny forces and the rmf takes similar values around 40 mev . however , the bcpm functional predict relatively larger values of @xmath74 at this average density . the reason of that lies , probably , in the fact that @xmath74 grows more linearly with the density than the other considered models . as far as the slope of @xmath74 has a value consistent with @xmath52 , as it can be seen in fig . [ f2 ] , the value of @xmath74 at 0.72 @xmath42 is larger than the value predicted by the other considered mean field models . from the analysis of fig . [ f3 ] one derives two important conclusions . first , the crossing density actually disappears when one considers mean field models of different nature . what is actually relevant , is to compute the slope of @xmath74 at _ the average density_. the same happens in the analysis of the symmetry energy using different mean field models @xcite . second , the small differences of @xmath74 at the average density may explain why @xmath79 and @xmath52 are not always linearly correlated . this happens , e.g. , with the nl1 parametrization with @xmath80=211 mev , a value similar to the one of the skm@xmath78 force but predicting @xmath52 clearly smaller , or the case of bcpm discussed previously . in what concerns the isgqr , we are well aware of the fact that a theory with @xmath81 , as is the case of the bcpm functional , underestimates the collective quadrupole excitation energy . in table [ t1 ] we display the scaled energies of the quadrupole excitation computed with bcpm for several nuclei together with the available experimental values @xcite . insofar as the focus with a ks - dft approach is on ground state properties , we may not worry about the failure in the estimate of the isgqr . in general , for the description of excitation energies , the ks - dft should be generalized to include non - localities . attempts in this direction exist in condensed matter . it may be worth to look into that in the future also in the nuclear context . in this work we further elaborated on the previously established kohn - sham dft functional , bcp @xcite , based on advanced microscopic nuclear and neutron matter calculations . the bcp contained by large 5 adjustable parameters , three for finite size , one for spin - orbit and one for bulk . we reduced this number by approximately a factor of two coming up with a ks - dft functional containing only three adjustable parameters without deteriorating the excellent results . one of these parameters , the strength of the spin - orbit force , has the usual value , the results being quite insensitive to its variation within a relatively large margin . we would like to call this a rather weak , i.e. almost predetermined parameter ( in addition it may be possible to extract it from the g - matrix @xcite ) . the other two parameters , on the contrary , are fixed within extremely tight margins of the order of 10@xmath82 . they concern the bulk energy per particle , e / a , on the one hand and the surface energy on the other hand . indeed in the polynomial fit to the microscopically determined eos , a single fine tuning parameter had to be introduced in order to get a minimal rms value for nuclear masses as well as a drift free ( i.e. scatter around zero ) difference of theoretical and experimental binding energies as a function of mass number . this parameter turned out to pin down the binding energy to the value , e / a = 15.98 mev . we emphasize that the fourth digit is significant here . for the second parameter , we made the simplifying ansatz to introduce a finite range gaussian convoluting the two densities in the @xmath83 terms of nuclear and neutron matter fits , that is @xmath84 . the pre - factor is determined from the bulk , so that the only free parameter is the range @xmath34 which evidently is responsible for the surface energy . the minimization of the rms of masses ( see main text for details ) again is determined within the very narrow margin of 10@xmath82 and turned out to have the value @xmath85 fm . with this , a rms value for the masses of 1.58 mev for 579 nuclei and of 0.027 fm for the radii with 313 nuclei is obtained . those values are as good as , e.g. , the ones of the gogny d1s and d1 m forces for the same set of nuclei . this very encouraging result calls for some comments . it is , indeed , quite surprising but on the other hand very satisfying , that the two basic nuclear quantities , energy per particle of the bulk and surface energy essentially suffice to determine a ks - dft functional with excellent ground state properties of nuclei . we should point out that the ks - dft approach is tailored for the ground state and that a time dependent extension is not evident . we nevertheless tried to determine at least the gmr energies with the sum rule technique and found that the results are in the same ball park as the ones of other skyrme , relativistic , or gogny type approaches . on the other hand the energies of the gqr seem to be underestimated systematically by about 1 mev . this is not so surprising , since a particularity of the ks - dft approach is that the bare mass is used while it is known that the gqr is sensitive to the value of the effective mass . future developments concern eventually the introduction of an effective density dependent mass @xmath86 . whether this is possible with results as satisfying as here , remains to be seen . a further point which needs to be improved is pairing . at the moment we treat it at a rather elementary level with a zero range force and cut off . a finite range force of the gogny type in the pairing channel would certainly be preferable . promising developments in this direction have recently been published @xcite . at the end let us mention that our theory fulfills all the eleven criteria put forward in @xcite besides one . our functional does not agree with the constraint concerning the value of the volume isospin incompressibility coefficient that appears in the leptodermous expansion of the finite nucleus incompressibility @xcite . this quantity is defined as @xmath87 the bcpm functional predicts @xmath88=-195.2 mev which is outside of the range proposed in @xcite , -760 @xmath89 -372 mev . however , two comments are in order . first the extraction of this coefficient from the experimental excitation energies of the gmr in finite nuclei @xcite is not a simple task at all and may give erroneous estimates of the coefficients of the leptodermous expansion of the finite nucleus incompressibility as it has been discussed in the past ( see for example refs . second , there are other estimates of the @xmath88 coefficients available in the literature , as for instance in ref . @xcite , that predict -370 @xmath90 120 mev which embraces the prediction of the bcpm functional . in general , the criteria and their limitations in @xcite can certainly be subjected to debate . for example there is no reason to exclude functionals with @xmath35 as we have demonstrated in this work . the fact that our results comply with the criteria of @xcite is an indication that nuclear density functionals should be based as much as possible on a solid microscopic input . in this respect it is to be noted that out of the five functionals which survived the criteria in @xcite , three are also based on a microscopic input . we are grateful to p. danielewicz for useful discussions . we would like to thank n. schunck for a careful reading of the manuscript and many useful suggestions . work supported in part by micinn grants nos . fpa2009 - 08958 , fis2009 - 07277 , and fpa2008 - 03865-e / in2p3 and by the consolider - ingenio 2010 program cpan csd2007 - 00042 and multidark csd2009 - 00064 . x. v. also acknowledges the support from fis2008 - 01661 ( spain and feder ) and 2009sgr-1289 ( spain ) . support by compstar , a research networking programme of the european science foundation is also acknowledged . references m. baldo , p. schuck and x. vias , phys . * b663*,390 ( 2008 ) . w. kohn and l.j . sham , phys . rev . * 140 * , a1135 ( 1965 ) . p. hohenberg and w. kohn , phys . rev . * 136 * , b864 ( 1964 ) . r. o. jones and o. gunnarsson , rev . phys . * 61 * , 689 ( 1989 ) . h. eschrig , _ the fundamentals of density functional theory _ ( b. g. teubner , stuttgart , 1996 ) . j. p. perdew and k.schmidt , _ density functional theory and its applications to material _ , v. van doren et al . cp577 ( american institute of physics , melville ny , 2001 ) . a. e. mattsson , science * 298 * , 759 ( 2002 ) . jianmin tao , j. p. perdew , v. n. staroverov , and g. e. scuseria , phys . lett . * 91 * , 146401 ( 2003 ) . jianmin tao and j. p. perdew , phys . rev . lett.*95 * , 196403 ( 2005 ) . j. messud , m. bender , and e. suraud phys . rev . * c80 * , 054314 ( 2009 ) . j. engel , phys . rev . * c75 * , 014306 ( 2007 ) . n. barnea , phys . c76 , 067302 ( 2007 ) . m. baldo , i. bombaci and g.f . burgio , astronomy and astrophysics * 328 * , 274 ( 1997 ) . m. baldo , a. fiasconaro , h.q . song , g. giansiracusa , and u. lombardo , phys . * c 65 * , 017303 ( 2001 ) . m. baldo , g. giansiracusa , u. lombardo and h.q . song , phys . * b 473 * , 1 ( 2000 ) . zhou , g.f . burgio , u. lombardo , h .- j . schulze and w. zuo , phys . * c 69 * , 018801 ( 2004 ) . li , u. lombardo , h .- j . schulze , w. zuo , l.w . chen and h.r . ma , phys . rev . * c74 * , 047304 ( 2006 ) . m. baldo , c. maieron , p. schuck and x. vias , nucl a736 * , 241 ( 2004 ) . m. baldo and c. maieron , j. of phys . * g34 * , r1 ( 2007 ) . m. baldo and g.f . burgio , rep.prog.phys . * 75 * , 026301 ( 2012 ) . g. taranto , m. baldo and g.f . burgio phys . rev . * c87 * , 045803 ( 2013 ) . s.a . fayans , jetp lett . * 68 * , 169 ( 1998 ) . steiner , m. prakash , j.m . lattimer and p.j . ellis , phys . rep . * 411 * , 325 ( 2005 ) b. friedman and v.r . pandharipande , nuc . phys . * a361 * , 502 ( 1981 ) ; r.b . wiringa , v. fiks and a. fabrocini , phys . rev . * c38 * , 1010 ( 1988 ) . holt , n. kaiser , and w. weise , phys . rev . * c 81 * , 024002 ( 2010 ) . holt , n. kaiser and w. weise , eur . j. * a47 * , 128 ( 2011 ) . n. kaiser , eur . j * a48 * , 36 ( 2012 ) . b. gebremariam , t. duguet , and s. k. bogner phys . rev . * c 82 * , 014305 ( 2010 ) . m. stoitsov , m. kortelainen , s. k. bogner , t. duguet , r. j. furnstahl , b. gebremariam , and n. schunck phys . rev . * c 82 * , 054307 ( 2010 ) . j. erler , n. birge , m. kortelainen , w. nazarewicz , e. olsen a. perhac , and m. stoitsov , nature * 486 * , 509 ( 2012 ) . m. baldo , l. robledo , p. schuck and x. vias , j. phys . g , nucl . and part . physics * 37 * , 064015 ( 2010 ) . j. engel , lecture notes in physics * 620 * , 56 ( 2003 ) , c. fiolhais , f. nogueira and m. marques eds . ( springer - verlag , berlin , 2003 ) . e. chabanat , p. bonche , p. haensel , j. meyer and r. schaeffer , nucl . phys . * a 635 * , 231 ( 1998 ) ; nucl . phys . * a 643 * , 441 ( 1998 ) . s s goriely , n n chamel and j m jm pearson , phys . lett . * 102 * , 152503 ( 2009 ) . j. decharg , d. gogny , phys . rev . * c21 * , 1568 ( 1980 ) . j.f . berger , m. girod , d. gogny , nucl . phys , * a428 * , 23c ( 1984 ) . oliveira , e.k.u . gross and w. khon , phys . lett . * 60 * , 2430 ( 1988 ) . e. garrido , p. sarriguren , e. moya de guerra , and p. schuck , phys . rev . * c60 * , 064312 ( 1999 ) . g.a . lalazissis , j. konig , and p. ring , phys . rev . * c55 * , 540 ( 1997 ) . robledo , m. baldo , p. schuck , and x. vias , phys . rev . * c77 * , 051301(r ) ( 2008 ) . x. vias , l.m robledo , p. schuck , and m. baldo , intl . j. of mod . e18 * , 935 ( 2009 ) . robledo , m. baldo , p. schuck , and x. vias , phys . * c81 * , 034315 ( 2010 ) . j.l egido and l.m . robledo , lecture notes in physics * 641 * , 269 ( 2004 ) . a. akmal , v. r. pandharipande and d. g. ravenhall , phys * c58 * , 1804 ( 1998 ) . robledo , intl . j. of mod . e16 * 337 ( 2007 ) . robledo , jour . of phys.*g37 * 064020 ( 2010 ) . j. margueron , e. van dalen and c. fuchs , phys . rev . * c76 * , 034309 ( 2007 ) . e. khan , j. margueron , g. col , k. hagino and h. sagawa phys . rev . * c82 * , 024322 ( 2010 ) . w.d . myers and w.j . swiatecki , ann . ( n.y . ) * 55 * , 395 ( 1969 ) . o. bohigas , a.m. lane and j. martorell , phys . rep . * * 5**1 , 267 ( 1979 ) . blaizot , phys . rep . * 64 * 171 ( 1980 ) . m. dutra , o. loureno , j.s . s martins , a. delfino , j.r . stone and p.d . stevenson , phys . rev . * c85 * , 035201 ( 2012 ) . tsang et al . , arxiv 1204.0166 . brown , phys . lett . * 85 * , 5296 ( 2000 ) . m. warda , x. vias , x. roca - maza and m. centelles , phys . rev . * c80 * 024316 ( 2009 ) . m. centelles , x. roca - maza , x. vias and m. warda , phys . lett . * 102 * , 122502 ( 2009 ) . nayak , j.m . pearson , m. farine , p. gleissl , and m. brack , nucl . a516 * , 62 ( 1990 ) . chossy and w. stocker , phys . rev . * c56 * , 2518 ( 1997 ) . t. niksic , d. vretenar and p. ring phys . rev . * c78 * , 034318 ( 2008 ) . p. ring and p. schuck , _ the nuclear many body problem _ , ( springer , berlin ) ( 1980 ) . s. hilaire , m. girod , eur . j * a33 * , 237 ( 2007 ) . robledo and g.f . bertsch , phys . rev c84 , 014312 ( 2011 ) . c. titin - schnaider and p. quentin , phys * b * 49 , 397 ( 1974 ) . m. anguiano , j. l. egido , and l. m. robledo , nucl . phys . * a * 683 , 227 ( 2001 ) . butler , d.w.l . sprung , j. martorell nucl . phys . * a * 422 , 157 ( 1984 ) . soubbotin , v.i . tselyaev , x. vias phys . rev . * c 67 * , 014324 ( 2003 ) . g. audi and a.h . wapstra , nucl . phys . * a729 * , 337 ( 2003 ) . m.centelles , m.del estal , and x.vinas , nucl.phys . * a635 * 193 ( 1998 ) . gogny d1s web mass table _ http://www-phynu.cea.fr/hfb-gogny.htm _ g. f. bertsch , b. sabbey , and m. uusnkki , phys . rev . * c 71 * , 054311 ( 2005 ) . f. chappert , m. girod , and s. hilaire , phys . lett . * b668 * , 420 ( 2008 ) . s. goriely , s. hilaire , m. girod , and s. per , phys . rev . lett . * 102 * , 242501 ( 2009 ) . robledo and g.f . bertsch , phys . rev * c84 * , 054302 ( 2011 ) . m. kortelainen , t. lesinski , j. mor , w. nazarewicz , j. sarich , n. schunck , m. v. stoitsov , and s. wild , phys . rev . * c 82 * , 024313 ( 2010 ) . n. nikolov , n. schunck , w. nazarewicz , m. bender , and j. pei , phys . rev . * c 83 * , 034305 ( 2011 ) . m. kortelainen , j. mcdonnell , w. nazarewicz , p .- reinhard , j. sarich , n. schunck , m. v. stoitsov , and s. m. wild , phys . rev . * c 85 * , 024304 ( 2012 ) . p. klpfel , p .- reinhard , t.j . brvenich , and j. a. maruhn phys . rev * c79 * , 034310 ( 2009 ) . reinhard and b.k . agrawal int . jour . of mod . e20 * , 1379 ( 2011 ) . i. angeli , at . data nucl . data tables * 87 * , 185 ( 2004 ) . f. buchinger et _ al . rev . * c41 * , 2883 ( 1999 ) . p. campbell et _ al . lett . * 89 * , 082501 ( 2002 ) . c. amsler et _ al . _ ( particle data group ) , phys b667 * , 1 ( 2008 ) . m. dupuis , s. karataglidis , e. bauge , j.p . delaroche , and d.gogny , phys . rev . * c 73 * , 014605 ( 2006 ) . mang , phys . rep . * 18 * , 325 ( 1975 ) . robledo , m. baldo , p. schuck and x. vias , jour . of phys . : conference series * 321 * , 012015 ( 2011 ) . m. warda , j.l . egido , l.m . robledo and k. pomorski , phys . rev . * c66 * , 014310 ( 2002 ) . m. hunyadi et _ al _ , phys b505 * , 27 ( 2001 ) . youngblood , y .- w . lui , h.l . clark , b. john , y. tokimoto and x. chen , phys . rev . * c69 * , 034315 ( 2004 ) . d.h . youngblood , y .- w . lui , b. john , y. tokimoto , h.l . clark and x. chen , phys . rev . * c69 * , 054312 ( 2004 ) . t. li et al . , lett . * 99 * , 162503 ( 2007 ) ; phys . rev . * c81 * , 034309 ( 2010 ) . d. patel et al . , arxiv:1209.0681 . l. capelli , g. col and jun li , phys . rev . * c79 * , 054329 ( 2009 ) . e. khan , n. sandulescu , m. grasso and n. v. giai , phys . rev . * c66 * , 024309 ( 2002 ) . v. tselyaev , j. speth , s. krewald , e. litvinova , s. kamerdzhiev , n. lyutorovich , a. avdeenkov and f. grmmer , phys . rev . * c79 * , 034309 ( 2009 ) . g. col , n. v. giai , j. meyer , k. bennaceur and p. bonche , phys . rev . * c70 * , 024307 ( 2004 ) . e. khan , phys . rev . * c80 * , 011307 ( 2009 ) . g. col and n. v. giai , nucl . phys . * a731 * , 15 ( 2004 ) . d. vretenar , t. niksic , and p. ring phys . rev . * c68 * , 024310 ( 2003 ) . j. piekarewicz and m. centelles , phys . rev . * c79 * , 054311 ( 2009 ) . blaizot , j.f . berger , j. decharg and m. girod , nucl . a591 * , 435 ( 1995 ) . e. khan , j. margueron and i. vidaa , phys . lett . * 109 * , 092501 ( 2012 ) . brown , phys . lett . * 85 * , 5296 ( 2000 ) . m. centelles , x. roca - maza , x. vias and m. warda , phys . lett . * 102 * , 122502 ( 2009 ) . soubbotin , v.i . tselyaev and x. vias , phys . rev . * c69 * , 064312 ( 2004 ) . patra , x. vias , m. centelles and m. del estal , nucl . a703 * , 240 ( 2002 ) . lui , d. h. youngblood , y. tokimoto , h. l. clark , and b. john phys . rev . * c69 * , 034611 ( 2004 ) . brieva , j.r . rook , nucl . phys . * a 297 * 206 ( 1978 ) . m. kohno , arxiv:1209.5048 . j. dukelsky , s. lerma h. , l. m. robledo , r. rodriguez - guzman , and s. m. a. rombouts phys . c 84 , 061301 ( 2011 ) . patra , m. centelles , x. vias , and m. del estal , phys . rev . * 65 * , 044304 ( 2002 ) . l.w . chen , b.j . cai , c.m . ko , bao - an li , c. shen , and j. xu , phys . rev . * 80 * , 014322 ( 2009 ) . m. centelles , x. vias , s.k . patra , j.n . de and tapas sil , phys . rev . * c72 * , 014304 ( 2005 ) . p. gleissl , m. brack , j. meyer and p. quentin , ann . ( n.y . ) * 197 * , 205 ( 1990 ) . in this appendix we briefly summarize the basic theoretical aspects of the sum rule approach to obtain moments of the rpa strength functions . for more details we refer the reader to refs . @xcite . in the sum rule approach it is enough to know few low energy weighted moments of @xmath91 ( the so - called sum rules ) to estimate the excitation energies of the giant resonances . the sum rules are defined as m_k = _ n e_n^k | n |q| 0 |^2 , [ eq01 ] where @xmath92 is the operator associated to the excitation , @xmath93 and @xmath94 are the exact ground and excited states respectively and @xmath95 the excitation energies . if @xmath96 is an odd integer , the sum rules @xmath97 can be computed as the expectation values in the ground state @xmath98 of some commutators which involve the hamiltonian @xmath99 and the operator @xmath92 ( assumed to be a hermitian and one - body operator ) @xcite , as for example m_1 = 0 | | 0 [ eq06 ] and m_3 = 0 | | 0 , [ eq07 ] from where an average excitation energy can be estimated as the full calculation of the sum rules is still a complicated task because the exact ground - state wavefunction is , in general , unknown . however , if the moments are computed within the 1p1h rpa , it is possible to replace the exact ground - state wavefunction by the uncorrelated hf one to obtain the sum rules ( [ eq06 ] ) and ( [ eq07 ] ) @xcite . in spite of all these simplifications , the calculation of the sum rules is only easy to handle in few particular cases , namely the isgmr and isgqr ones . in these cases the sum rules needed to estimate the average excitation energy of these resonances can be simplified as we will discuss below . for the isoscalar monopole and quadrupole oscillations the corresponding excitation operators are taken as q_m = _ i=1^a r_i^2 q_q = _ i=1^a ( r_i^2 -3 z_i^2 ) . [ eq02 ] the underlying effective interaction associated to the bcpm energy density functional commutes with the operator @xmath92 and , therefore , the only contribution to the commutator @xmath100 $ ] in eq . ( [ eq06 ] ) comes from the kinetic energy operator . consequently @xcite m_1 = a r^2 ( l=0 ) m_1 = a r^2 ( l=2 ) , [ eq09b ] where the expectation value of the operator @xmath101 is calculated with the hf wave functions . the direct evaluation of the commutators entering in eq . ( [ eq07 ] ) to compute the @xmath57 sum rule is a rather cumbersome task , that in the case of the monopole and quadrupole oscillations can be avoided with the help the so - called scaling method @xcite . we start from the scaled ground - state wave function that in the case of the monopole reads _ ^m = ^3 /2 _ 0(x , y , z ) , [ eq2 ] where @xmath102 is an arbitrary scaling parameter . in this case the @xmath57 sum rule can be expressed by means of the second derivative of the scaled ground - state energy @xcite , that is , m_3 = _ = 0 . [ eq3 ] in the scaling approach the bcpm energy density functional reads e ( ) = t ( ) + e^_int ( ) + e^fr_int ( ) + e_c ( ) + e^s.o ( ) , [ eq6a ] where the scaled kinetic , spin - orbit and coulomb energies can be found in ref . the scaled bulk contribution can be easily obtained from the scaling of the particle density ( @xmath103 ) . the scaling of the finite range contribution has been described in refs . it is easy to show that the scaling of the hartee term reduces to a renormalization of the range @xmath104 of the interaction , or in other words that the scaled relative coordinate is @xmath105 . in this way , the second derivative with respect to the scaling parameter @xmath102 , needed to compute the @xmath57 sum rule , finally reads @xcite : _ = 1 = d d ( ) ( ) [ eq6d ] in the bcpm energy density functional , the form factor @xmath106 is of gaussian type with a range @xmath104 . in this case the second derivative of the scaled finite range contribution can be recast as @xcite _ = 1 = ^2 . [ eq6e ] finally the @xmath57 for the monopole oscillation computed with the bcpm energy density functional can be written as @xmath107\end{aligned}\ ] ] as explained in detail in ref . @xcite , a pure hartree term scales in the quadrupole case as e^h_q ( ) = d d ( ) ( ) v(s_q , ) , [ eq7d ] where @xmath108 . after some algebra the surface contribution to the @xmath57 sum rule in the scaling approach for the quadrupole oscillation can be written as ( see ref.@xcite for details ) _ = 1 = d d ( ) ( ) , that in the case of the bcpm functional owing to the gaussian character of the form factor @xmath109 can also be written as : _ = 1 = . this is due to the fact of the volume conservation in the quadrupole scaling , the bulk interaction part of the bcp energy density functional does not contribute to the @xmath57 sum rule in this case . the contribution from kinetic and spin - orbit energies can be easily obtained from the scaling transformation of the single - particle wave functions assuming a spherically symmetric ground - state @xcite . the coulomb contribution can also be obtained from eq . ( [ eq7d ] ) using the coulomb form factor @xmath110 @xcite . collecting all these contributions , the @xmath57 sum rule for the quadrupole oscillation can finally be written as @xmath111 \right].\end{aligned}\ ] ] once the @xmath57 sum rule has been obtained through the scaling method , an estimate of the excitation energy of the isgmr and isgqr can be calculated with the help of eqs . ( [ eq04 ] ) and([eq09b ] ) e^s_m = . [ eq19a ] let us now consider a nucleus , described by a hamiltonian @xmath99 , under the action of a weak one - body field @xmath112 . assuming @xmath113 sufficiently small , so that the perturbation theory holds , the variation of the expectation value of @xmath92 and @xmath99 are directly related to the @xmath56 moment @xcite , m_-1 = _ n = - _ = 0 = _ = 0 . [ eq12b ] therefore at rpa level the average energy of the giant resonances can also be estimated performing constrained hf calculations , i.e. looking for the hf solutions of the constrained hamiltonian h ( ) = h - q [ eq17 ] where @xmath92 is the collective monopole ( quadrupole ) operator defined previously ( see eq . ( [ eq02 ] ) . from the constrained hf wave - function @xmath114 solution of ( [ eq17 ] ) , the rpa @xmath56 moment can be written as @xmath115_{\eta=0 } \\ \nonumber & = & \frac{1}{2 } \left [ \frac{\partial^2}{\partial \eta^2 } \langle \phi(\eta)\vert h \vert \phi(\eta ) \rangle \right]_{\eta=0},\end{aligned}\ ] ] from where another estimate of the average energy of the isoscalar giant resonances is given by |e_1 = . [ eq19 ] due to the fact that the rpa moments fulfill @xmath116 @xcite , the average energies @xmath53 and @xmath54 values are an upper and lower bound of the mean energy of the resonance and their difference is related with the variance of the strength function ( resonance width ) = . [ eq21 ]

the tropical variety of a-dimensional prime ideal in a polynomial ring with complex coefficients is a pure-dimensional polyhedral fan . this fan is shown to be connected in codimension one . we present algorithmic tools for computing the tropical variety , and we discuss our implementation of these tools in the grbner fan software ` gfan ` . every ideal is shown to have a finite tropical basis , and a sharp lower bound is given for the size of a tropical basis for an ideal of linear forms .

math0507563

every ideal in a polynomial ring with complex coefficients defines a tropical variety , which is a polyhedral fan in a real vector space . the objective of this paper is to introduce methods for computing this fan , which coincides with the `` logarithmic limit set '' in george bergman s seminal paper @xcite . given any polynomial @xmath1 $ ] and a vector @xmath2 , the initial form @xmath3 is the sum of all terms in @xmath4 of lowest @xmath5-weight ; for instance , if @xmath6 then @xmath7 and @xmath8 . the _ tropical hypersurface _ of @xmath4 is the set @xmath9 equivalently , @xmath10 is the union of all codimension one cones in the inner normal fan of the newton polytope of @xmath4 . note that @xmath10 is invariant under dilation , so we may specify @xmath10 by giving its intersection with the unit sphere . for the linear polynomial @xmath11 above , @xmath12 is a two - dimensional fan with six maximal cones . its intersection with the @xmath13-sphere is the complete graph on the four nodes @xmath14 , @xmath15 , @xmath16 and @xmath17 . a finite intersection of tropical hypersurfaces is a _ tropical prevariety _ @xcite . if we pick the second linear form @xmath18 then @xmath19 is a graph with two vertices connected by three edges on the @xmath13-sphere , and @xmath20 consists of three edges of @xmath12 which are adjacent to @xmath17 . in particular , the tropical prevariety @xmath20 is not a tropical variety . tropical varieties are derived from ideals . namely , if @xmath21 is an ideal in @xmath22 $ ] then its _ tropical variety _ @xmath23 is the intersection of the tropical hypersurfaces @xmath10 where @xmath4 runs over all polynomials in @xmath21 . theorem [ thm : tropical bases ] below states that every tropical variety is actually a tropical prevariety , i.e. , the ideal @xmath21 has a finite generating set @xmath24 such that @xmath25 if this holds then @xmath26 is called a _ tropical basis _ of @xmath21 . for instance , our ideal @xmath27 has the tropical basis @xmath28 , and we find that its tropical variety consists of three points on the sphere : @xmath29 our main contribution is a practical algorithm , along with its implementation , for computing the tropical variety @xmath30 from any generating set of its ideal @xmath21 . the emphasis lies on the geometric and algebraic features of this computation . we do not address issues of computational complexity , which have been studied by theobald @xcite . our paper is organized as follows . in section 2 we give precise specifications of the algorithmic problems we are dealing with , including the computation of a tropical basis . we show that a finite tropical basis exists for every ideal @xmath21 , and we give tight bounds on its size for linear ideals , thereby answering the question raised in ( * ? ? ? * , page 13 ) . in section 3 we prove that the tropical variety @xmath23 of a prime ideal @xmath21 is connected in codimension one . this result is the foundation of algorithm [ alg : traverse ] for computing @xmath23 . section 4 also describes methods for computing tropical bases and tropical prevarieties . our algorithms have been implemented in the software package gfan @xcite . in section 5 we compute the tropical variety of several non - trivial ideals using gfan . the tropical variety @xmath23 is a subfan of the grbner fan of @xmath21 ( defined in section 2 ) . the grbner fan is generally much more complicated and harder to compute than @xmath23 . in section 6 we compare these two fans , and we exhibit a family of curves for which the tropical variety of each member consists of four rays but the number of one - dimensional cones in the grbner fan grows arbitrarily . a note on the choice of ground field is in order . in this paper we will work with varieties defined over @xmath31 . in the implementation of our algorithm ( section [ sec : examples ] ) , we have required our polynomials to have rational coefficients , but our algorithms do not use any particular properties of @xmath32 . it is important , however , that we work over a field of characteristic @xmath33 , as our proof of correctness uses the kleiman - bertini theorem in the proof of theorem [ codimone ] . in most papers on tropical algebraic geometry ( cf . @xcite ) , tropical varieties are defined from polynomials with coefficients in a field @xmath34 with a non - archimedean valuation . these tropical varieties are not fans but polyhedral complexes . we close the introduction by illustrating how our algorithms can be applied to this situation . consider the field @xmath35 of rational functions in the unknown @xmath36 . then @xmath37 is a subfield of the algebraically closed field @xmath38 of puiseux series with real exponents , which is an example of a field @xmath34 as in the above cited papers . suppose we are given an ideal @xmath21 in @xmath39 $ ] . let @xmath40 $ ] be the ideal generated by @xmath21 . the tropical variety @xmath41 , in the sense of the papers above , is a finite polyhedral complex in @xmath42 which usually has both bounded and unbounded faces . to study this complex , we consider the polynomial ring in @xmath43 variables , @xmath44\,$ ] and we let @xmath45 denote the intersection of @xmath21 with this subring of @xmath46 $ ] . generators of @xmath45 are computed from generators of @xmath21 by clearing denominators and saturating with respect to @xmath36 . the tropical variety of @xmath47 is related to the tropical variety of @xmath45 as follows . [ northern ] a vector @xmath48 lies in the polyhedral complex @xmath41 if and only if the vector @xmath49 lies in the polyhedral fan @xmath50 . thus the tropical variety @xmath41 equals the restriction of @xmath50 to the northern hemisphere of the @xmath51-sphere . note that if @xmath21 is a prime ideal then so are @xmath47 and @xmath45 . einsiedler , kapranov and lind @xcite have shown that if @xmath47 is prime , then @xmath41 is connected . our connectivity results in section 3 ( which use the result of @xcite ) imply the following result which was conjectured in @xcite . if @xmath21 is an ideal in @xmath52 $ ] whose radical is prime of dimension @xmath0 , then the tropical variety @xmath30 is a pure @xmath0-dimensional polyhedral complex which is connected in codimension one . on the algorithmic side , we conclude that the polyhedral complex @xmath41 can be computed by restricting the flip algorithm of section 4 to maximal cones in the fan @xmath50 which intersect the open northern hemisphere in @xmath53 . for all algorithms in this paper we fix the ambient ring to be the polynomial ring over the complex numbers , @xmath54 : = { \mathbb{c}}[x_1 , \ldots , x_n]$ ] . the most basic computational problem in tropical geometry is the following : [ prob : tropical_prevariety ] given a finite list of polynomials @xmath55 $ ] , compute the tropical prevariety @xmath56 in @xmath42 . the geometry of this problem is best understood by considering the newton polytopes @xmath57 of the given polynomials . by definition , @xmath58 is the convex hull in @xmath42 of the exponent vectors which appear with non - zero coefficient in @xmath59 . the tropical hypersurface @xmath60 is the @xmath61-skeleton of the inner normal fan of the polytope @xmath62 . our problem is to intersect these normal fans . the resulting tropical prevariety can be a fairly general polyhedral fan . its maximal cones may have different dimensions . the tropical variety of an ideal @xmath21 in @xmath54 $ ] is the set @xmath63 . equivalently , @xmath64 where @xmath65 is the initial ideal of @xmath21 with respect to @xmath5 . bieri and groves @xcite proved that @xmath30 is a @xmath0-dimensional fan when @xmath0 is the krull dimension of @xmath54/i$ ] . the fan is pure if @xmath21 is unmixed . in section 3 we shall prove that @xmath30 is connected in codimension one if @xmath21 is prime . we first note that it suffices to devise algorithms for computing tropical varieties of homogeneous ideals . let @xmath66 $ ] be the homogenization of an ideal @xmath21 in @xmath54 $ ] and @xmath67 the homogenization of @xmath68 $ ] . [ lemma : homogeneous ] fix an ideal @xmath69 $ ] and a vector @xmath48 . the initial ideal @xmath70 contains a monomial if and only if @xmath71 contains a monomial . suppose @xmath72 . then @xmath73 for some @xmath74 . the @xmath75-weight of a term in @xmath67 equals the @xmath5-weight of the corresponding term in @xmath4 . hence @xmath76 where @xmath77 is some non - negative integer . conversely , if @xmath78 then @xmath79 for some @xmath80 . substituting @xmath81 in @xmath4 gives a polynomial in @xmath21 . the @xmath75-weight of any term in @xmath4 equals the @xmath5-weight of the corresponding term in @xmath82 . since @xmath83 is a monomial , only one term in @xmath4 has minimal @xmath75-weight . this term can not be canceled during the substitution . hence it lies in @xmath70 . our main goal in this paper is to solve the following problem . [ prob : tropical_variety ] given a finite list of homogeneous polynomials @xmath84 $ ] , compute the tropical variety @xmath85 of their ideal @xmath86 . it is important to note that the two problems stated so far are of a fundamentally different nature . problem [ prob : tropical_prevariety ] is a problem of polyhedral geometry . it involves only polyhedral computations : no algebraic computations are required . problem [ prob : tropical_variety ] , on the other hand , combines the polyhedral aspect with an algebraic one . to solve problem [ prob : tropical_variety ] we must perform algebraic operations ( e.g. grbner bases ) with polynomials . in problem [ prob : tropical_prevariety ] we do not assume that the input polynomials @xmath87 are homogeneous as the polyhedral computations can be performed easily without this assumption . [ homogtwoways ] let @xmath21 be an ideal in @xmath54 $ ] and let @xmath48 . the following are equivalent : 1 . the ideal @xmath21 is @xmath5-homogeneous ; _ i.e. _ @xmath21 is generated by a set @xmath88 of @xmath5-homogeneous polynomials , meaning that @xmath89 for all @xmath90 . the initial ideal @xmath91 is equal to @xmath21 . if @xmath21 has a @xmath5-homogeneous generating set then @xmath92 . any maximal @xmath5-homogeneous component of @xmath93 is in @xmath21 . in particular @xmath94 . conversely , the ideal @xmath91 is generated by @xmath5-homogeneous elements by definition so , if @xmath95 , then @xmath21 is generated by @xmath5-homogeneous elements . the set of @xmath48 for which the above equivalent conditions hold is a vector subspace of @xmath42 . its dimension is called the _ homogeneity _ of @xmath21 and is denoted @xmath96 . this space is contained in every cone of the fan @xmath30 and can be computed from the newton polytopes of the polynomials that form any reduced grbner basis of @xmath21 . passing to the quotient of @xmath42 modulo that subspace and then to a sphere around the origin , @xmath30 can be represented as a polyhedral complex of dimension @xmath97 . here @xmath98 and @xmath99 are the codimension and dimension of @xmath21 . in what follows , @xmath30 is always presented in this way , and every ideal @xmath21 is presented by a finite list of generators together with the three numbers @xmath100 , @xmath101 and @xmath102 . [ sym4x4 ] let @xmath21 denote the ideal which is generated by the @xmath103-minors of a symmetric @xmath104-matrix of unknowns . this ideal has @xmath105 , @xmath106 and @xmath107 . hence @xmath30 is a two - dimensional polyhedral complex . we regard @xmath30 as the tropicalization of the secant variety of the veronese threefold in @xmath108 , i.e. , the variety of symmetric @xmath104-matrices of rank @xmath109 , applying our gfan implementation ( see example [ 3minors of nbynsymmetric ] ) , we find that @xmath30 is a simplicial complex consisting of @xmath110 triangles , @xmath110 edges and @xmath111 vertices . our next problem concerns tropical bases . a finite set @xmath112 is a tropical basis of @xmath21 if @xmath113 [ prob : find tropical basis ] compute a tropical basis of a given ideal @xmath114 $ ] . a priori , it is not clear that every ideal @xmath21 has a finite tropical basis , but we shall prove this below . first , here is one case where this is easy : [ rem : principal ] if @xmath115 is a principal ideal , then @xmath116 is a tropical basis . in @xcite it was claimed that any universal grbner basis of @xmath21 is a tropical basis . unfortunately , this claim is false as the following example shows . [ ex : universalgb ] let @xmath21 be the intersection of the three linear ideals @xmath117 , @xmath118 , and @xmath119 in @xmath120 $ ] . then @xmath21 contains the monomial @xmath121 , so @xmath23 is empty . a minimal universal grbner basis of @xmath21 is @xmath122 and the intersection of the four corresponding tropical surfaces in @xmath123 is the line @xmath124 . thus @xmath125 is not a tropical basis of @xmath21 . we now prove that every ideal @xmath69 $ ] has a tropical basis . by lemma [ lemma : homogeneous ] , one tropical basis of a non - homogeneous ideal @xmath21 is the dehomogenization of a tropical basis for @xmath126 . hence we shall assume that @xmath21 is a homogeneous ideal . tropical bases can be constructed from the _ grbner fan _ of @xmath21 ( see @xcite , @xcite ) which is a complete finite rational polyhedral fan in @xmath42 whose relatively open cones are in bijection with the distinct initial ideals of @xmath21 . two weight vectors @xmath127 lie in the same relatively open cone of the grbner fan of @xmath21 if and only if @xmath128 . the closure of this cell , denoted by @xmath129 , is called a _ grbner cone _ of @xmath21 . the @xmath51-dimensional grbner cones are in bijection with the reduced grbner bases , or equivalently , the monomial initial ideals of @xmath21 . every grbner cone of @xmath21 is a face of at least one @xmath51-dimensional grbner cone of @xmath21 . if @xmath130 is not a monomial ideal , then we can refine @xmath5 to @xmath131 by breaking ties in the partial order induced by @xmath5 with a fixed term order @xmath132 on @xmath54 $ ] . let @xmath133 denote the reduced grbner basis of @xmath21 with respect to @xmath131 . the grbner cone of @xmath133 , denoted by @xmath134 , is an @xmath51-dimensional grbner cone that has @xmath129 as a face . the tropical variety @xmath135 consists of all grbner cones @xmath129 such that @xmath70 does not contain a monomial . from the description of @xmath30 as @xmath136 it is clear that @xmath30 is closed . thus we deduce that @xmath30 is a closed subfan of the grbner fan . this endows the tropical variety @xmath23 with the structure of a polyhedral fan . [ thm : tropical bases ] every ideal @xmath69 $ ] has a tropical basis . let @xmath137 be any finite generating set of @xmath21 which is not a tropical basis . pick a grbner cone @xmath129 whose relative interior intersects @xmath138 non - trivially and whose initial ideal @xmath139 contains a monomial @xmath140 . compute the reduced grbner basis @xmath133 for a refinement @xmath131 of @xmath5 , and let @xmath141 be the normal form of @xmath140 with respect to @xmath133 . let @xmath142 . since the normal form of @xmath140 with respect to @xmath143 is @xmath33 and @xmath141 is the normal form of @xmath140 with respect to @xmath133 , every monomial occurring in @xmath141 has higher @xmath5-weight than @xmath140 . moreover , @xmath141 depends only on the reduced grbner basis @xmath133 and is independent of the particular choice of @xmath5 in @xmath129 . hence for any @xmath144 in the relative interior of @xmath129 , we have @xmath145 . this implies that the polynomial @xmath146 is a _ witness _ for the cone @xmath129 not being in the tropical variety @xmath135 . we now add the witness @xmath4 to the current basis @xmath137 and repeat the process . since the grbner fan has only finitely many cones , this process will terminate after finitely many steps . it removes all cones of the grbner fan which violate the condition for @xmath137 to be a tropical basis . we next show that tropical bases can be very large even for linear ideals . let @xmath147 be the ideal in @xmath54 $ ] generated by @xmath0 linear forms @xmath148 where @xmath149 and @xmath150 is an integer @xmath151 matrix of rank @xmath0 . the tropical variety @xmath23 depends only on the matroid associated with @xmath21 , and it is known as the _ bergman fan _ of that matroid . the results on the bergman fan proved in @xcite imply that the circuits in @xmath21 form a tropical basis . a _ circuit _ of @xmath21 is a non - zero linear polynomial @xmath74 of minimal support . the following result answers the question which was posed in @xcite . for any @xmath152 , there is a linear ideal @xmath21 in @xmath22 $ ] such that any tropical basis of linear forms in @xmath21 has size at least @xmath153 . suppose that all @xmath154-minors of the coefficient matrix @xmath150 are non - zero . equivalently , the matroid of @xmath21 is uniform . there are @xmath155 circuits in @xmath21 , each supported on a different @xmath156-subset of @xmath157 . since the circuits form a tropical basis of @xmath21 and each circuit has support of size @xmath158 , the tropical variety @xmath23 consists of all vectors @xmath48 whose smallest @xmath159 components are equal . the latter condition is necessary and sufficient to ensure that no single variable in a circuit becomes the initial form of the circuit with respect to @xmath5 . consider any vector @xmath48 satisfying @xmath160 since @xmath161 , any tropical basis of linear forms in @xmath21 contains an @xmath4 such that @xmath162 . this implies that @xmath4 is one of the @xmath0 circuits whose support contains the @xmath163 variables @xmath164 with @xmath165 . the support of each circuit has size @xmath158 , hence contains @xmath158 distinct @xmath166-subsets . there are @xmath167 @xmath168-subsets of @xmath157 to be covered . hence any tropical basis consisting of linear forms has size at least @xmath169 . let @xmath170 . the bergman fan @xmath23 corresponds to the line in tropical projective @xmath171-space which consists of the five rays in the coordinate directions . we have @xmath172 . hence this line is not a complete intersection of three tropical hyperplanes , but it requires four . in this section we assume that @xmath21 is a prime ideal of dimension @xmath0 in @xmath22 $ ] . then its tropical variety @xmath30 is called _ irreducible_. it is a subfan of the grbner fan of @xmath21 and , by the bieri - groves theorem @xcite , all facets of @xmath30 are cones of dimension @xmath0 . a cone of dimension @xmath173 in @xmath23 is called a _ ridge _ of the tropical variety @xmath30 . a _ ridge path _ is a sequence of facets @xmath174 such that @xmath175 is a ridge for all @xmath176 . our objective is to prove the following result , which is crucial for the algorithms . [ codimone ] any irreducible tropical variety @xmath30 is connected in codimension one , i.e. , any two facets are connected by a ridge path . the proof of this theorem will be based on the following important lemma . [ intersect ] ( transverse intersection lemma ) let @xmath21 and @xmath45 be ideals in @xmath22 $ ] whose tropical varieties @xmath30 and @xmath50 meet transversally at a point @xmath48 . then @xmath177 . by `` meet transversely '' we mean that if @xmath178 and @xmath179 are the cones of @xmath30 and @xmath50 which contain @xmath5 in their relative interior , then @xmath180 . this lemma implies that any transverse intersection of tropical varieties is a tropical variety . in particular , any transverse intersection of tropical hypersurfaces is a tropical variety , and such a tropical variety is defined by an ideal which is a complete intersection in the commutative algebra sense . for any two ideals @xmath21 and @xmath45 in @xmath22 $ ] we have @xmath181 equality holds if the latter intersection is transverse at every point except the origin and the two fans meet in at least one point other than the origin . we have @xmath182 . clearly , this contains @xmath183 . if @xmath23 and @xmath184 intersect transversally and @xmath5 is a point of @xmath185 other than the origin then the preceeding lemma tells us that @xmath177 . thus @xmath186 contains every point of @xmath185 except possibly the origin . in particular , @xmath186 is not empty . every nonempty fan contains the origin , so we see that the origin is in @xmath186 as well . we first derive theorem [ codimone ] from lemma [ intersect ] , which will be proved later . we must at this point address an annoying technical detail . the subset @xmath187 depends only on the ideal @xmath188 $ ] generated by @xmath21 in the laurent polynomial ring @xmath189 $ ] . ( this is easy to see : if @xmath190 and @xmath191 generate the same ideal in @xmath192 $ ] and @xmath193 then there is a polynomial @xmath194 such that @xmath195 is a monomial . there is some monomial @xmath196 such that @xmath197 , then @xmath198 is a monomial and @xmath199 . ) from a theoretical perspective then , it would be better to directly work with ideals in @xmath192 $ ] . one reason is the availability the symmetry group @xmath200 of the multiplicative group of monomials . the action of this group transforms @xmath30 by the obvious action on @xmath42 . this symmetry will prove invaluable for simplifying the arguments in this section . therefore , in this section , we will work with ideals in @xmath192 $ ] . computationally , however , it is much better to deal with ideals in @xmath54 $ ] as it is for such ideals that grbner basis techniques have been developed and this is the approach we take in the rest of the paper . note that , if @xmath69 $ ] is prime then so is the ideal it generates in @xmath192 $ ] . we will signify an application of the @xmath200 symmetry by the phrase `` making a multiplicative change of variables '' . the polyhedral structure on @xmath30 induced by the grbner fan of @xmath21 may change under a multiplicative change of variables of @xmath201 $ ] in @xmath192 $ ] , but all of the properties of @xmath30 that are of interest to us depend only on the underlying point set . _ proof of theorem [ codimone ] . _ as discussed , we replace @xmath21 by the ideal it generates in @xmath192 $ ] and , by abuse of notation , continue to denote this ideal as @xmath21 . the proof is by induction on @xmath202 . if @xmath203 then the statement is trivially true . we now explain why the result holds for @xmath204 . by a multiplicative change of coordinates , it suffices to check that @xmath205 is connected . let @xmath34 be the puiseux series field over @xmath31 . let @xmath206 $ ] be the prime ideal generated by @xmath21 via the inclusion @xmath207 \to k$ ] . by lemma [ northern ] , the tropical variety of @xmath47 is @xmath205 . in @xcite it was shown that the tropical variety of @xmath47 is connected whenever @xmath47 is prime . we conclude that @xmath205 is connected , so our result holds for @xmath204 . we now suppose that @xmath208 . let @xmath178 and @xmath209 be facets of @xmath30 . we can find @xmath210 such that @xmath211 are relatively prime integers , both @xmath212 and @xmath213 are cones of dimension @xmath173 , and @xmath214 intersects every cone of @xmath30 except for the origin transversally . to see this , select rays @xmath5 and @xmath144 in the relative interiors of @xmath178 and @xmath209 . by perturbing @xmath5 and @xmath144 slightly , we may arrange that the span of @xmath5 and @xmath144 does not meet any ray of @xmath30 here it is important that @xmath215 . now , taking @xmath214 to be the span of @xmath5 , @xmath144 and a generic @xmath216-plane , we get that @xmath214 also does not contain any ray of @xmath30 and hence does not contain any positive dimensional face of @xmath30 . so @xmath214 is transverse to @xmath30 everywhere except at the origin . since @xmath212 and @xmath213 are positive - dimensional ( as @xmath217 ) , the hyperplane @xmath214 does intersect @xmath30 at points other than just the origin . the hyperplane @xmath214 is the tropical hypersurface of a binomial , namely , @xmath218 , where @xmath219 and @xmath220 is an arbitrary point in the algebraic torus @xmath221 . our transversality assumption regarding @xmath214 and lemma [ intersect ] imply that @xmath222 since @xmath21 is prime of dimension @xmath0 , and @xmath223 , the ideal @xmath224 has dimension @xmath173 by krull s principal ideal theorem ( * ? ? ? * theorem 10.1 ) . if @xmath224 were a prime ideal then we would be done by induction . indeed , this would imply that there is a ridge path between the facets @xmath212 and @xmath213 in the @xmath225-dimensional tropical variety @xmath226 . since @xmath227 , the @xmath225- and @xmath228-dimensional faces of @xmath229 arise uniquely from the intersections of @xmath214 with @xmath0- and @xmath225-dimensional faces of @xmath30 . hence this path is also a ridge path considered as a path in @xmath30 . let @xmath230 denote the subvariety of the algebraic torus @xmath231 defined by an ideal @xmath232 $ ] . the tropical variety in ( [ hct ] ) depends only on the subvariety of @xmath221 defined by our ideal @xmath233 . this subvariety is @xmath234 here @xmath235 denotes the identity element of @xmath221 . for generic choices of the group element @xmath236 , the intersection ( [ tropvar ] ) is an irreducible subvariety of dimension @xmath173 in @xmath221 . this follows from kleiman s version of bertini s theorem ( * ? ? ? * theorem iii.10.8 ) , applied to the algebraic group @xmath221 . hence ( [ hct ] ) is indeed an irreducible tropical variety of dimension @xmath173 , defined by the prime ideal @xmath237 . this completes the proof by induction . _ proof of lemma [ intersect ] : _ again , we replace @xmath69 $ ] by the ideal it generates in @xmath192 $ ] . let @xmath178 be the cone of @xmath30 which contains @xmath5 in its relative interior and @xmath179 the cone of @xmath50 which contains @xmath5 in its relative interior . our hypothesis is that @xmath178 and @xmath179 meet transversally at @xmath5 , that is , @xmath238 we claim that the ideal @xmath70 is homogeneous with respect to any weight vector @xmath239 or , equivalently ( see proposition [ homogtwoways ] ) , that @xmath240 . according to proposition 1.13 in @xcite , for @xmath36 a sufficiently small positive number , @xmath241 . the vector @xmath242 is in the relative interior of @xmath178 so @xmath243 . by the same argument , the ideal @xmath244 is homogeneous with respect to any weight vector in @xmath245 . after a multiplicative change of variables in @xmath246\,$ ] we may assume that @xmath247 , @xmath248 and @xmath249 . we change the notation for the variables as follows : @xmath250 the homogeneity properties of the two initial ideals ensure that we can pick generators @xmath251 @xmath252 for @xmath70 and generators @xmath253 for @xmath244 . since @xmath70 is not the unit ideal , the laurent polynomials @xmath254 have a common zero @xmath255 , and likewise the laurent polynomials @xmath256 have a common zero @xmath257 . next we consider the following general chain of inclusions of ideals : @xmath258 the product of two ideals which are generated by ( laurent ) polynomials in disjoint sets of variables equals the intersection of the two ideals . since the set of @xmath259-variables is disjoint from the set of @xmath260-variables , it follows that the first ideal in ( [ cdotcap ] ) equals the last ideal in ( [ cdotcap ] ) . in particular , we conclude that @xmath261 we next claim that @xmath262 the left hand side is an ideal which contains both @xmath70 and @xmath244 , so it contains their sum . we must prove that the right hand side contains the left hand side . consider any element @xmath263 where @xmath74 and @xmath264 . let @xmath265 and @xmath266 . we have the following representation for some integer @xmath267 and non - zero polynomial @xmath268 : @xmath269 if @xmath270 then we conclude @xmath271 if @xmath272 then @xmath273 lies in @xmath274 . in view of ( [ goodnews ] ) , there exists @xmath275 with @xmath276 . then @xmath277 and replacing @xmath4 by @xmath278 and @xmath279 by @xmath280 puts us in the same situation as before , but with @xmath77 reduced by @xmath281 . by induction on @xmath77 , we conclude that @xmath282 is in @xmath283 , and the claim ( [ betternews ] ) follows . for any constant @xmath284 , the vector @xmath285 is a common zero in @xmath221 of the ideal ( [ betternews ] ) . we conclude that @xmath286 is not the unit ideal , so it contains no monomial , and hence @xmath177 . in this section we describe algorithms for solving the computational problems raised in section [ sec : algorithmic problems ] . the emphasis is on algorithms leading to a solution of problem [ prob : tropical_variety ] for prime ideals , taking advantage of theorem [ codimone ] . recall that we only need to consider the case of homogeneous ideals in @xmath54 $ ] . in order to state our algorithms we must first explain how polyhedral cones and polyhedral fans are represented . a polyhedral cone is represented by a canonical minimal set of inequalities and equations . given arbitrary defining linear inequalities and equations , the task of bringing these to a canonical form involves linear programming . representing a polyhedral fan requires a little thought . we are rarely interested in all faces of all cones . a set @xmath88 of polyhedral cones in @xmath42 is said to _ represent _ a fan @xmath287 in @xmath42 if the set of all faces of cones in @xmath88 is exactly @xmath287 . a representation may contain non - maximal cones , but each cone is represented minimally by its canonical form . a grbner cone @xmath288 is represented by the pair @xmath289 of marked reduced grbner bases , where @xmath132 is some globally fixed term order . in a _ marked _ grbner basis the initial terms are distinguished . the advantage of using marked grbner bases is that the weight vector @xmath290 need not be stored we can deduce defining inequalities for its cone from the marked reduced grbner bases themselves , see example [ ex : hankel4 ] . this is done as follows ; see ( * ? ? ? * proof of proposition 2.3 ) : [ lem : cone inequalities ] let @xmath291 $ ] be a homogeneous ideal , @xmath132 a term order and @xmath292 a vector . for any other vector @xmath293 : @xmath294 our first two algorithms perform polyhedral computations , and they solve problem [ prob : tropical_prevariety ] . by the _ support _ of a fan we mean the union of its cones . recall that , for a polynomial @xmath4 , the tropical hypersurface @xmath295 is the union of the normal cones of the edges of the newton polytope @xmath296 . the first algorithm computes these cones . tropical hypersurface[alg : hypersurface ] + * input : * @xmath297 $ ] . + * output : * a representation @xmath88 of a polyhedral fan whose support is @xmath295 . + @xmath298 + @xmath299 ; + for every vertex @xmath300 + @xmath298 + compute the normal cone @xmath301 of @xmath302 in @xmath296 ; + @xmath303 ; + @xmath304 + @xmath304 let @xmath305 and @xmath306 be polyhedral fans in @xmath42 . their _ common refinement _ is @xmath307 to compute a common refinement we simply run through all pairs of cones in the fan representations and bring their intersection to canonical form . the canonical form makes it easy to remove duplicates . common refinement[alg : common refinement ] + * input : * representations @xmath308 and @xmath309 for polyhedral fans @xmath305 and @xmath306 . + * output : * a representation @xmath88 for the common refinement @xmath310 . + @xmath298 + @xmath299 ; + for every pair @xmath311 + @xmath312 ; + @xmath304 if refinements of more than two fans are needed , algorithm [ alg : common refinement ] can be applied successively . note that the intersection of the support of two fans is the support of the fans common refinement . hence algorithm [ alg : common refinement ] can be used for computing intersections of tropical hypersurfaces . this solves problem [ prob : tropical_prevariety ] , but the output may be a highly redundant representation . recall ( from the proof of theorem [ thm : tropical bases ] ) that a witness @xmath93 is a polynomial which certifies @xmath313 . computing witnesses is essential for solving problems [ prob : tropical_variety ] and [ prob : find tropical basis ] . the first step of constructing a witness is to check if the ideal @xmath314 contains monomials , and , if so , compute one such monomial . the check for monomial containment can be implemented by saturating the ideal with respect to the product of the variables ( cf . * lemma 12.1 ) ) . knowing that the ideal contains a monomial , a simple way of finding one is to repeatedly reduce powers of the product of the variables by applying the division algorithm until the remainder is @xmath33 . monomial in ideal[alg : monomial ] + * input : * a set of generators for an ideal @xmath291 $ ] . + * output : * a monomial @xmath315 if one exists , * no * otherwise . + @xmath298 + if @xmath316 return * no * ; + @xmath317 ; + while @xmath318 @xmath319 ; + return @xmath196 ; + @xmath304 to pick the smallest monomial in @xmath21 with respect to a term order , we first compute the largest monomial ideal contained in @xmath21 using ( * ? ? ? * algorithm 4.2.2 ) and then pick the smallest monomial generator of this ideal . constructing a witness from a monomial was already explained in the proof of theorem [ thm : tropical bases ] . we only state the input and output of this algorithm . witness[alg : witness ] + * input : * a set of generators for an ideal @xmath291 $ ] and a vector @xmath292 with @xmath314 containing a monomial . + * output : * a polynomial @xmath93 such that the tropical hypersurface @xmath320 and the relative interior of @xmath288 have empty intersection . combining algorithm [ alg : monomial ] and algorithm [ alg : witness ] with known methods ( e.g. ( * ? ? ? * algorithm 3.6 ) ) for computing grbner fans , we can now compute the tropical variety @xmath135 and a tropical basis of @xmath21 . this solves problem [ prob : tropical_variety ] and problem [ prob : find tropical basis ] . this approach is not at all practical , as shown in section 6 . we will present a practical algorithm for computing @xmath135 when @xmath21 is prime . an ideal @xmath69 $ ] is said to define a _ tropical curve _ if @xmath321 . our problems are easier in this case because a tropical curve consists of only finitely many rays and the origin modulo the homogeneity space . tropical basis of a curve [ alg : tropical basis of curve ] + * input : * a set of generators @xmath322 for an ideal @xmath21 defining a tropical curve . + * output : * a tropical basis @xmath323 of @xmath21 . + @xmath298 + compute a representation @xmath88 of @xmath324 ; + for every @xmath325 + @xmath298 + let @xmath290 be a generic relative interior point in @xmath301 ; + if ( @xmath314 contains a monomial ) + then add a witness to @xmath322 and restart the algorithm ; + @xmath304 + @xmath326 ; + @xmath304 _ proof of correctness . _ the algorithm terminates because @xmath21 has only finitely many initial ideals and at least one is excluded in every iteration . if a vector @xmath290 passes the monomial test ( which verifies @xmath327 ) then @xmath301 has dimension @xmath33 or @xmath281 modulo the homogeneity space since we are looking at a curve and @xmath290 is generic in @xmath301 . any other relative interior point of @xmath301 would also have passed the monomial test . ( this property fails in higher dimensions , when @xmath135 is no longer a tropical curve ) . hence , when we terminate only points in the tropical variety are covered by @xmath88 . thus @xmath323 is a tropical basis . in the curve case , combining algorithms [ alg : hypersurface ] and [ alg : common refinement ] with algorithm [ alg : tropical basis of curve ] we get a reasonable method for solving problem [ prob : tropical_variety ] . this method is used as a subroutine in algorithm [ alg : neighbours ] below . in the remainder of this section we concentrate on providing a better algorithm for problem [ prob : tropical_variety ] in the case of a prime ideal the idea is to use connectivity to traverse the tropical variety . the next algorithm is an important subroutine for us . we only specify the input and output . this algorithm is one step in the grbner walk @xcite . lift [ alg : lift ] + * input : * marked reduced grbner bases @xmath328 and @xmath329 where @xmath330 is an unspecified vector and @xmath132 and @xmath331 are unspecified term orders . + * output : * the marked reduced grbner basis @xmath332 . + we now suppose that @xmath21 is a monomial - free prime ideal with @xmath333 , and @xmath132 is a globally fixed term order . we first describe the local computations needed for a traversal of the @xmath0-dimensional grbner cones contained in @xmath23 . neighbors [ alg : neighbours ] + * input : * a pair @xmath289 such that @xmath314 is monomial - free and @xmath288 has dimension @xmath0 . + * output : * the collection @xmath334 of pairs of the form @xmath335 where one @xmath336 is taken from the relative interior of each @xmath0-dimensional grbner cone contained in @xmath135 that has a facet in common with @xmath288 . + @xmath298 + @xmath337 ; + compute the set @xmath287 of facets of @xmath288 ; + for each facet @xmath338 + @xmath298 + compute the initial ideal @xmath339 + where @xmath340 is a relative interior point in @xmath178 ; + use algorithm [ alg : tropical basis of curve ] and algorithm [ alg : common refinement ] to produce a relative + interior point @xmath341 of each ray in the curve @xmath342 ; + for each such @xmath341 + @xmath298 + compute @xmath343 ; + apply algorithm [ alg : lift ] to @xmath332 and @xmath344 to get @xmath345 ; + @xmath346 ; + @xmath304 + @xmath304 + @xmath304 _ proof of correctness . _ facets and relative interior points are computed using linear programming . figure [ fig : neighbours ] illustrates the choices of vectors in the algorithm . the initial ideal @xmath347 is homogeneous with respect to the span of @xmath178 . hence its homogeneity space has dimension @xmath173 . the krull dimension of @xmath54/\myinit_\u(i)$ ] is @xmath0 . hence @xmath347 defines a curve and @xmath348 can be computed using algorithm [ alg : tropical basis of curve ] . the identity @xmath349 for small @xmath350 , see ( * ? ? * proposition 1.13 ) , implies that we run through all the desired @xmath351 where @xmath352 for small @xmath350 . the lifting step can be carried out since @xmath353 . traversal of an irreducible tropical variety [ alg : traverse ] + * input : * a pair @xmath289 such that @xmath314 is monomial free and @xmath288 has dimension @xmath0 . + * output : * the collection @xmath354 of pairs of the form @xmath335 where one @xmath336 is taken from the relative interior of each @xmath0-dimensional grbner cone contained in @xmath135 . the union of all the @xmath355 is @xmath135 . + @xmath298 + @xmath356 ; + @xmath357 ; + while ( @xmath358 ) + @xmath298 + @xmath359 ; + @xmath360 ; + @xmath304 + @xmath304 _ proof of correctness . _ by @xmath361 we mean the union of all the output of algorithm [ alg : neighbours ] applied to all pairs in @xmath354 . the algorithm computes the connected component of the starting pair . since @xmath21 is a prime ideal , theorem [ codimone ] implies that the union of all the computed @xmath355 is @xmath135 . to use algorithm [ alg : traverse ] we must know a starting @xmath0-dimensional grbner cone contained in the tropical variety . one inefficient method for finding one would be to compute the entire grbner fan . instead we currently use heuristics , which are based on the following probabilistic recursive algorithm : starting cone [ alg : starting cone ] + * input : * a marked reduced grbner basis @xmath322 for an ideal @xmath21 whose tropical variety is pure of dimension @xmath333 . a term order @xmath132 for tie - breaking . + * output : * two marked reduced grbner bases : * one for an initial ideal @xmath351 without monomials , where the homogeneity space of @xmath351 has dimension @xmath0 . the term order is @xmath362 . * a marked reduced grbner basis for @xmath21 with respect to @xmath362 . @xmath298 + if ( @xmath363 ) + return @xmath364 ; + if not + @xmath298 + repeat + @xmath298 + compute a random reduced grbner basis of @xmath21 ; + compute a random extreme ray @xmath290 of its grbner cone ; + @xmath304 + until ( @xmath314 is monomial free ) ; + compute @xmath332 ; + @xmath365:= starting cone(@xmath329 ) ; + apply algorithm [ alg : lift ] to @xmath332 and @xmath366 + to get a marked reduced grbner basis @xmath323 for @xmath21 ; + return @xmath367 ; + @xmath304 + @xmath304 we implemented the algorithms of section [ sec : algorithms ] in the software package ` gfan ` @xcite . ` gfan ` uses the library ` cddlib ` @xcite for polyhedral computations such as finding facets and extreme rays of cones and bringing cones to canonical form . we require our ideals to be generated by polynomials in @xmath368 $ ] . exact arithmetic is done with the library ` gmp ` @xcite . this is needed both for polyhedral computations and for efficient arithmetic in @xmath368 $ ] . in this section we illustrate the use of ` gfan ` in computing various tropical varieties . [ ex : hankel4 ] we consider the prime ideal @xmath369 \,$ ] which is generated by the @xmath103 minors of the generic _ hankel matrix _ of size @xmath104 : @xmath370 its tropical variety is a @xmath171-dimensional fan in @xmath371 with @xmath13-dimensional homogeneity space . its combinatorics is given by the graph in figure [ fig : hankel ] . to compute @xmath135 in ` gfan ` , we write the ideal generators on a file hankel.in : .... % more hankel.in { -c^3 + 2*b*c*d - a*d^2-b^2*e+a*c*e ,- c^2*d+b*d^2+b*c*e - a*d*e - b^2*f+a*c*f , -c*d^2+c^2*e+b*d*e - a*e^2-b*c*f+a*d*f ,- d^3 + 2*c*d*e - b*e^2-c^2*f+b*d*f , -c^2*d+b*d^2+b*c*e - a*d*e - b^2*f+a*c*f ,- c*d^2 + 2*b*d*e - a*e^2-b^2*g+a*c*g , -d^3+c*d*e+b*d*f - a*e*f - b*c*g+a*d*g ,- d^2*e+c*e^2+c*d*f - b*e*f - c^2*g+b*d*g , -c*d^2+c^2*e+b*d*e - a*e^2-b*c*f+a*d*f ,- d^3+c*d*e+b*d*f - a*e*f - b*c*g+a*d*g , -d^2*e+2*c*d*f - a*f^2-c^2*g+a*e*g ,- d*e^2+d^2*f+c*e*f - b*f^2-c*d*g+b*e*g , -d^3 + 2*c*d*e - b*e^2-c^2*f+b*d*f ,- d^2*e+c*e^2+c*d*f - b*e*f - c^2*g+b*d*g , -d*e^2+d^2*f+c*e*f - b*f^2-c*d*g+b*e*g ,- e^3 + 2*d*e*f - c*f^2-d^2*g+c*e*g } .... we then run the command .... gfan_tropicalstartingcone < hankel.in > hankel.start .... which applies algorithm [ alg : starting cone ] to produce a pair of marked grbner bases . this represents a maximal cone in @xmath135 , as explained prior to lemma [ lem : cone inequalities ] . .... % more hankel.start { c*f^2-c*e*g , b*f^2-b*e*g , b*e*f+c^2*g , b*e^2+c^2*f , b^2*g - a*c*g , b^2*f - a*c*f , b^2*e - a*c*e , a*f^2-a*e*g , a*e*f+b*c*g , a*e^2+b*c*f } { c*f^2+e^3 - 2d*e*f+d^2*g - c*e*g , b*f^2+d*e^2-d^2*f - c*e*f+c*d*g - b*e*g , b*e*f+d^2*e - c*e^2-c*d*f+c^2*g - b*d*g , b*e^2+d^3 - 2c*d*e+c^2*f - b*d*f , b^2*g+c^2*e - b*d*e - b*c*f+a*d*f - a*c*g , b^2*f+c^2*d - b*d^2-b*c*e+a*d*e - a*c*f , b^2*e+c^3 - 2b*c*d+a*d^2-a*c*e , a*f^2+d^2*e-2c*d*f+c^2*g - a*e*g , a*e*f+d^3-c*d*e - b*d*f+b*c*g - a*d*g , a*e^2+c*d^2-c^2*e - b*d*e+b*c*f - a*d*f } .... using lemma [ lem : cone inequalities ] we can easily read off the canonical equations and equalities for the corresponding grbner cone @xmath288 . for example , the polynomials @xmath372 and @xmath373 represent the equation @xmath374 and the inequalities @xmath375 at this point , we could run algorithm [ alg : traverse ] using the following command : .... gfan_tropicaltraverse < hankel.start > hankel.out .... however , we can save computing time and get a better idea of the structure of @xmath135 by instructing ` gfan ` to take advantage of symmetries of @xmath21 as it produces cones . the only symmetries that can be used in ` gfan ` are those that simply permute variables . the output will show which cones of @xmath135 lie in the same orbit under the action of the symmetry group we provide . our ideal @xmath21 is invariant under reflecting the @xmath104-matrix along the anti - diagonal . this reverses the variables @xmath376 . to specify this permutation , we add the following line to the bottom of the file ` hankel.start ` : .... { ( 6,5,4,3,2,1,0 ) } .... we can add more symmetries by listing them one after another , separated by commas , inside the curly braces . ` gfan ` will compute and use the group generated by the set of permutations we provide , and it will return an error if we input any permutation which does not keep the ideal invariant . after adding the symmetries , we run the command .... gfan_tropicaltraverse --symmetry < hankel.start > hankel.out .... to compute the tropical variety . we show the output with some annotations : .... % more hankel.out ambient dimension : 7 dimension of homogeneity space : 2 dimension of tropical variety : 4 simplicial : true order of input symmetry group : 2 f - vector : ( 16,28 ) .... a short list of basic data : the dimensions of the ambient space , of @xmath135 , and of its homogeneity space , and also the face numbers ( @xmath4-vector ) of @xmath135 and the order of symmetry group specified in the input . .... modulo the homogeneity space : { ( 6,5,4,3,2,-1,0 ) , ( 5,4,3,2,1,0,-1 ) } .... a basis for the homogeneity space . the rays are considered in the quotient of @xmath371 modulo this @xmath13-dimensional subspace . .... rays : { 0 : ( -1,0,0,0,0,0,0 ) , 1 : ( -5,-4,-3,-2,-1,0,0 ) , 2 : ( 1,0,0,0,0,0,0 ) , 3 : ( 5,4,3,2,1,0,0 ) , 4 : ( 2,1,0,0,0,0,0 ) , 5 : ( 4,3,2,1,0,0,0 ) , 6 : ( 0,-1,0,0,0,0,0 ) , 7 : ( 6,5,4,3,2,0,0 ) , 8 : ( 3,2,1,0,0,0,0 ) , 9 : ( 0,0,-1,0,0,0,0 ) , 10 : ( 0,0,0,0,-1,0,0 ) , 11 : ( 0,0,0,-1,0,0,0 ) , 12 : ( -6,-4,-3,-3,-1,0,0 ) , 13 : ( -3,-2,-2,-1,-1,0,0 ) , 14 : ( 3,2,2,1,1,0,0 ) , 15 : ( 3,2,2,0,1,0,0 ) } .... the direction vectors of the tropical rays . since the homogeneity space is positive - dimensional , the directions are not uniquely specified . for instance , the vectors @xmath377 and @xmath378 represent the same ray . note that ` gfan ` uses negated weight vectors . .... rays incident to each dimension 2 cone : { { 2,6 } , { 3,7 } , { 2,4 } , { 3,5 } , { 4,9 } , { 5,10 } , { 4,8 } , { 5,8 } , { 8,11 } , { 0,12 } , { 1,12 } , { 0,1 } , { 1,6 } , { 0,7 } , { 1,9 } , { 0,10 } , { 0,13 } , { 1,13 } , { 6,14 } , { 7,14 } , { 9,13 } , { 10,13 } , { 6,10 } , { 7,9 } , { 6,7 } , { 11,12 } , { 11,15 } , { 14,15 } } .... the cones in @xmath135 are listed from highest to lowest dimension . each cone is named by the set of rays on it . there are 28 two - dimensional cones , broken down into 11 orbits of size 2 and 6 orbits of size 1 . the further output , which is not displayed here , shows that the 16 rays break down into 5 orbits of size 2 and 6 orbits of size 1 . using the same procedure , we now compute several more examples . [ 3minors of hankel5by5 ] let @xmath21 be the ideal generated by the @xmath103 minors of the generic @xmath379 hankel matrix . we again use the symmetry group @xmath380 . the tropical variety is a graph with vertex degrees ranging from 2 to 7 . .... ambient dimension : 9 dimension of homogeneity space : 2 dimension of tropical variety : 4 simplicial : true f - vector : ( 28,53 ) .... [ 3minors of 3by5 ] let @xmath21 be the ideal generated by the @xmath103 minors of a generic @xmath381 matrix . we use the symmetry group @xmath382 , where @xmath383 acts by permuting the columns and @xmath384 by permuting the rows . .... ambient dimension : 15 dimension of homogeneity space : 7 dimension of tropical variety : 12 simplicial : true f - vector : ( 45,315,930,1260,630 ) .... [ 3minors of nbynsymmetric ] let @xmath21 be the ideal generated by the @xmath103 minors of a generic @xmath104 symmetric matrix . we use the symmetry group @xmath385 which acts by simultaneously permuting the rows and the columns . .... ambient dimension : 10 dimension of homogeneity space : 4 dimension of tropical variety : 7 simplicial : true f - vector : ( 20,75,75 ) .... if we take the @xmath103 minors of a generic @xmath379 symmetric matrix then we get .... ambient dimension : 15 dimension of homogeneity space : 5 dimension of tropical variety : 9 simplicial : true f - vector : ( 75 , 495 , 1155 , 855 ) .... [ 2by2 commuting ] let @xmath21 be the prime ideal of a pair of commuting @xmath386 matrices . that is , @xmath387 $ ] is defined by the matrix equation @xmath388 the tropical variety is the graph @xmath389 , which ` gfan ` reports as follows : .... ambient dimension : 8 dimension of homeogeneity space : 4 dimension of tropical variety : 6 simplicial : true f - vector : ( 4,6 ) .... .1 cm if @xmath21 is the ideal of @xmath103 commuting symmetric matrices then we get : .... ambient dimension : 12 dimension of homeogeneity space : 2 dimension of tropical variety : 9 simplicial : false f - vector : ( 66,705,3246,7932,10888,8184,2745 ) .... in this paper we developed tools for computing the tropical variety @xmath23 of a @xmath0-dimensional homogeneous prime ideal @xmath21 in a polynomial ring @xmath54 $ ] . we took advantage of the fact that , since @xmath21 is homogeneous , the set @xmath23 has naturally the structure of a polyhedral fan , namely , @xmath23 is the collection of all cones in the grbner fan of @xmath21 whose corresponding initial ideal is monomial - free . a naive algorithm would be to compute the grbner fan of @xmath21 and then retain only those @xmath0-dimensional cones who survive the monomial test ( algorithm [ alg : monomial ] ) . the software gfan also computes the full grbner fan of @xmath21 , and so we tested this naive algorithm . we found it to be too inefficient . the reason is that the vast majority of @xmath0-dimensional cones in the grbner fan of @xmath21 are typically not in the tropical variety @xmath23 . consider the ideal @xmath21 in example [ ex : hankel4 ] which is generated by the @xmath103-minors of a generic @xmath104-hankel matrix . let @xmath390 be its initial ideal with respect to the first vector @xmath5 in the list of rays . the initial ideal @xmath45 defines a tropical curve consisting of five rays and the origin . the curve is a subfan of the much more complicated grbner fan of @xmath45 . the grbner fan is full - dimensional in @xmath371 with @xmath391 being three - dimensional . its f - vector equals @xmath392 . of the @xmath393 rays only @xmath394 are in the tropical variety . the grbner fan of @xmath45 is the link of the grbner fan of @xmath21 at @xmath5 . we were unable to compute the full grbner fan of @xmath21 . * toric ideals . * [ ti ] let @xmath395 be the toric ideal of a matrix @xmath396 of rank @xmath0 . the ideal @xmath21 is a prime of dimension @xmath0 . the tropical variety @xmath23 coincides with the homogeneity space @xmath397 which is just the row space of @xmath398 . hence @xmath23 modulo @xmath397 is a single point . yet , the grbner fan of @xmath21 can be very complicated , as it encodes the sensitivity information for an infinite family of integer programs ( * ? ? ? * chapter 7 ) . _ sketch of proof : _ the ideal @xmath403 is prime . its variety is the parametric curve @xmath405 . the poles and zeros of this map are @xmath406 . the tropical variety of @xmath403 consists of the four rays defined by the valuations at these points . these rays are generated by the columns of @xmath407 we examine the grbner fan around the ray @xmath408 . the initial ideal @xmath409 equals the toric ideal @xmath410 . to see this , we note that the two generators of @xmath403 form a grbner basis with respect to the underlined leading terms and @xmath409 is generated by @xmath411 for each @xmath412 in this grbner basis since @xmath5 lies in this grbner cone . the grbner fan of @xmath409 is the link at @xmath5 of the grbner fan of @xmath403 . to prove the theorem we show that the grbner fan of @xmath409 has at least @xmath413 distinct grbner cones . this implies , by euler s formula , that the grbner fan of @xmath409 has at least @xmath414 rays and hence so does the grbner fan of @xmath403 . to argue that the grbner fan of @xmath409 has at least @xmath413 distinct grbner cones we use the methods in @xcite . more specifically , this involves first showing that the binomials @xmath415 for @xmath416 are all in the universal grbner basis of @xmath417 . each monomial in a binomial in the universal grbner basis of a toric ideal contributes a minimal generator to some initial ideal of the toric ideal . thus there exist reduced grbner bases of @xmath409 in which the binomials @xmath418 are elements with leading term @xmath419 for @xmath420 . this implies that these reduced grbner bases are all distinct , which completes the proof . while the grbner fan is a fundamental object which has had a range of applications ( the grbner walk @xcite , integer programming ( example [ ti ] ) ) , many computer algebra experts do not like it . their view is that _ the grbner fan is a combinatorial artifact which is marginal to the real goal of computing the variety of @xmath21_. while this opinion has some merit , the story is entirely different for the subfan @xmath23 of the grbner fan . in our view , _ the tropical variety the variety of @xmath21_. every point on @xmath23 furnishes the starting system for a _ numerical homotopy _ towards the complex variety of @xmath21 , see ( * ? ? ? * chapter 3 ) . thus computing @xmath23 is not only much more efficient than computing the grbner fan of @xmath21 , it is also geometrically more meaningful . * acknowledgments . * we thank komei fukuda for answering polyhedral computation questions and for customizing cddlib for our purpose , and we thank eth zrich for hosting anders jensen and bernd sturmfels during the summer of 2005 . we acknowledge partial financial support by the university of aarhus , the danish research training council ( forskeruddannelsesrdet , fur ) , the institute for operations research at eth , the swiss national science foundation project 200021 - 105202 , and the us national science foundation ( dms-0456960 , dms-0401047 and dms-0354131 ) .

symmetrical multilevel diversity coding ( smdc ) is a classical model for coding over distributed storage . in this setting , a simple separate encoding strategy known as superposition coding was shown to be optimal in terms of achieving the minimum sum rate ( roche , yeung , and hau , 1997 ) and the entire admissible rate region ( yeung and zhang , 1999 ) of the problem . the proofs utilized carefully constructed induction arguments , for which the classical subset entropy inequality of han ( 1978 ) played a key role . this paper includes two parts . in the first part the existing optimality proofs for classical smdc are revisited , with a focus on their connections to subset entropy inequalities . first , a new sliding - window subset entropy inequality is introduced and then used to establish the optimality of superposition coding for achieving the minimum sum rate under a weaker source - reconstruction requirement . second , a subset entropy inequality recently proved by madiman and tetali ( 2010 ) is used to develop a new structural understanding to the proof of yeung and zhang on the optimality of superposition coding for achieving the entire admissible rate region . building on the connections between classical smdc and the subset entropy inequalities developed in the first part , in the second part the optimality of superposition coding is further extended to the cases where there is either an additional all - access encoder ( smdc - a ) or an additional secrecy constraint ( s - smdc ) .

1111.0689

symmetrical multilevel diversity coding ( smdc ) is a classical model for coding over distributed storage , which was first introduced by roche @xcite and yeung @xcite . in this setting , there are a total of @xmath0 _ independent _ discrete memoryless sources @xmath1 , where the importance of the source @xmath2 is assumed to decrease with the subscript @xmath3 . the sources are to be encoded by a total of @xmath0 _ randomly accessible _ encoders . the goal of encoding is to ensure that the number of sources that can be nearly perfectly reconstructed grows with the number of available encoder outputs at the decoder . more specifically , denote by @xmath4 the set of accessible encoders . the realization of @xmath5 is _ unknown _ a priori at the encoders . however , the sources @xmath6 need to be nearly perfectly reconstructed whenever @xmath7 at the decoder . the word symmetrical " here refers to the fact that the sources that need to be nearly perfectly reconstructed depend on the set of accessible encoders only via its cardinality . the rate allocations at different encoders , however , can be different and are not necessarily symmetrical . a natural strategy for smdc is to encode the sources separately at each of the encoders ( no coding across different sources ) known as _ superposition coding _ @xcite . to show that the natural superposition coding strategy is also optimal , however , turned out to be rather nontrivial . the optimality of superposition coding in terms of achieving the _ minimum sum rate _ was established by roche , yeung , and hau @xcite . the proof used a carefully constructed induction argument , for which the classical subset entropy inequality of han @xcite played a key role . later , the optimality of superposition coding in terms of achieving the _ entire admission rate region _ was established by yeung and zhang @xcite . their proof was based on a _ new _ subset entropy inequality , which was established by carefully combining han s subset inequality with several highly technical results on the analysis of a sequence of linear programs ( which are used to characterize the performance of superposition coding ) . this paper includes two parts . in the first part ( section [ sec : smdc ] ) , the optimality proofs of @xcite and @xcite are revisited in light of two new subset entropy inequalities : * first , a new _ sliding - window _ subset entropy inequality is introduced , which not only implies the classical subset entropy inequality of han @xcite in a trivial way , but also leads to a new proof of the optimality of superposition encoding for achieving the minimum sum rate under a _ weaker _ source - reconstruction requirement . * second , a subset entropy inequality recently proved by madiman and tetali @xcite is leveraged to provide a new _ structural _ understanding to the subset entropy inequality of yeung and zhang @xcite . based on this new understanding , a _ conditional _ version of the subset entropy inequality of yeung and zhang @xcite is further established , which plays a key role in extending the optimality of superposition coding to the case where there is an additional secrecy constraint . in the second part of the paper ( section [ sec : ext ] ) , two extensions of classical smdc are considered : * the first extension , which we shall refer to as smdc - a , features an _ all - access encoder _ , in addition to the @xmath0 randomly accessible encoders in the classical setting , whose output is available at the decoder _ at all time_. this model is mainly motivated by the proliferation of mobile computing devices ( laptop computers , tablets , smart phones etc . ) , which can access both remote storage nodes via unreliable wireless links and local hard disks which are always available but are of limited capacity . it is shown that in this setting , superposition coding remains optimal in terms of achieving the entire admissible rate region . key to our proof is to identify the supporting hyperplanes that define the superposition coding rate region and then apply the subset entropy inequality of yeung and zhang @xcite . * the second extension , which we shall refer to as s - smdc , extends the problem of smdc to the _ secure _ communication setting . the problem was first introduced in @xcite , where the optimality of superposition coding for achieving the minimum sum rate was established via the classical subset entropy inequality of han @xcite . through the conditional version of the subset entropy inequality of yeung and zhang @xcite established in the first part , here we show that superposition coding can , in fact , achieve the entire admissible rate region of the problem , resolving the conjecture of @xcite by positive . independent discrete memoryless sources @xmath1 are to be encoded by a total of @xmath0 encoders . the decoder , which has access to a subset @xmath5 of the encoder outputs , needs to nearly perfectly reconstruct the sources @xmath8 no matter what the realization of @xmath5 is . ] as illustrated in figure [ fig : smdc ] , the problem of smdc consists of : * a total of @xmath0 _ independent _ discrete memoryless sources @xmath9\}_{t=1}^{\infty}$ ] , where @xmath10 and @xmath11 is the time index ; * a set of @xmath0 encoders ( encoder @xmath12 to @xmath0 ) ; * a decoder which can access a nonempty subset @xmath13 of the encoder outputs . the realization of @xmath5 is _ unknown _ a priori at the encoders . however , no matter which @xmath5 actually materializes , the decoder needs to nearly perfectly reconstruct the sources @xmath6 whenever @xmath7 . formally , an @xmath14 code is defined by a collection of @xmath0 encoding functions : @xmath15 and @xmath16 decoding functions : @xmath17 a nonnegative rate tuple @xmath18 is said to be _ admissible _ if for every @xmath19 , there exits , for sufficiently large block - length @xmath20 , an @xmath14 code such that : * ( rate constraints at the encoders ) @xmath21 * ( asymptotically perfect reconstructions at the decoder ) @xmath22 where @xmath23\}_{t=1}^{n}$ ] , @xmath24 is the output of encoder @xmath3 , and @xmath25 . the _ admissible rate region _ @xmath26 is the collection of _ all _ admissible rate tuples @xmath18 . the _ minimum sum rate _ @xmath27 is defined as @xmath28 as mentioned previously , a natural strategy for smdc is _ superposition coding _ , i.e. , to encode the sources separately at the encoders and there is _ no _ coding across different sources . formally , the problem of encoding a single source @xmath29 can be viewed as a special case of the general smdc problem , where the sources @xmath30 are deterministic for all @xmath31 . in this case , the source @xmath29 needs to be nearly perfectly reconstructed whenever the decoder can access at least @xmath32 encoder outputs . thus , the problem is essentially to transmit @xmath29 over an _ erasure _ channel , and the following simple source - channel separation scheme is known to be optimal @xcite : * first compress the source sequence @xmath33 into a source message @xmath34 using a _ lossless _ source code . it is well known ( * ? ? ? 5 ) that the rate of the source message @xmath34 can be made arbitrarily close to the entropy rate @xmath35 for sufficiently large block - length @xmath20 . * next , the source message @xmath34 is encoded at encoders @xmath12 to @xmath0 using a _ maximum distance separable _ code @xcite . it is well known @xcite that the source message @xmath34 can be perfectly recovered at the decoder whenever @xmath36 for sufficiently large block length @xmath20 , where @xmath37 denotes the collection of all subsets of @xmath38 of size @xmath32 . combining the above two steps , we conclude that the admissible rate region for encoding a single source @xmath29 is given by the collection of all nonnegative rate tuples @xmath18 satisfying @xmath39 by definition , the superposition coding rate region @xmath40 for encoding the sources @xmath1 is given by the collection of all nonnegative rate tuples @xmath18 such that @xmath41 for some nonnegative @xmath42 , @xmath10 and @xmath43 , satisfying @xmath44 in principle , an explicit characterization of the superposition coding rate region @xmath40 can be obtained by eliminating @xmath42 , @xmath10 and @xmath43 , via a fourier - motzkin elimination from and . however , the elimination process is _ unmanageable _ even for moderate @xmath0 , as there are simply too many equations involved . on the other hand , note that the superposition coding rate region @xmath40 is a convex polyhedron with polyhedral cone being @xmath45 , so an equivalent characterization is to characterize the supporting hyperplanes : @xmath46 where @xmath47 clearly , the above optimization problem can be separated into the following @xmath0 sub - optimization problems : @xmath48 where @xmath49 and follows from the strong _ duality _ for linear programs . for any @xmath50 and any @xmath10 , let @xmath51 then , we have @xmath52 and hence @xmath53 for any @xmath50 . substituting into , we conclude that the superposition coding rate region @xmath40 is given by the collection of nonnegative rate tuples @xmath18 satisfying @xmath54 for a general @xmath55 , the linear program does not admit a _ closed - form _ solution . however , for @xmath56 it can be easily verified that @xmath57 where @xmath58 is an _ optimal _ solution to the linear program , and we thus have @xmath59 for any @xmath10 . hence , the minimum sum rate that can be achieved by superposition coding is given by @xmath60 to show that superposition coding is optimal in terms of achieving the entire admissible rate region , we need to show that for any @xmath50 we have @xmath61 in particular , to show that superposition coding is optimal in terms of achieving the minimum sum rate , we need to show that @xmath62 note that for any admissible rate tuple @xmath63 and @xmath19 , by the rate constraints we have @xmath64 for sufficiently large block - length @xmath20 . furthermore , by the asymptotically perfect reconstruction requirement and the well - known fano s inequality we have @xmath65 for any @xmath66 and @xmath10 , where @xmath67 in the limit as @xmath68 and @xmath69 . thus , for any @xmath70 we have @xmath71 where follows from the facts that all sources are independent so @xmath72 and that @xmath73 therefore , starting with and applying _ iteratively _ may lead us towards a proof of and . note , however , that to apply iteratively we shall need to bound from below @xmath74 in terms of @xmath75 for some @xmath66 . the key observation of @xcite and @xcite is that such bounds exist , not for an arbitrary individual pair of @xmath5 and @xmath76 , but rather at the level of an appropriate _ averaging _ among @xmath70 and @xmath66 . more specifically , @xcite considered the classical subset entropy inequality of han @xcite , which can be written as follows . [ thm : han ] for any collection of @xmath0 jointly distributed random variables @xmath77 , we have @xmath78 for any @xmath79 . iteratively applying and , we may obtain @xmath80 for any @xmath81 . in particular , let @xmath82 , and we have @xmath83 substituting into and dividing both sides of the inequality by @xmath20 , we have @xmath84 finally , letting @xmath68 and @xmath69 completes the proof of , i.e. , superposition coding can achieve the minimum sum rate for the general smdc problem . to prove that superposition coding can in fact achieve the entire admissible rate region , yeung and zhang @xcite proved the following key subset entropy inequality . [ thm : yz ] for any @xmath50 , there exists a function @xmath85 such that : * for each @xmath10 , @xmath86 is an _ optimal _ solution to the linear program ; and * for each @xmath79 , @xmath87 for any collection of @xmath0 jointly distributed random variables @xmath77 . iteratively applying and , we may obtain @xmath88 for any @xmath81 . in particular , let @xmath82 , and note that for @xmath89 the optimal solution to the linear program is unique and is given by @xmath90 we have @xmath91 substituting into and dividing both sides of the inequality by @xmath20 , we have @xmath92 finally , letting @xmath68 and @xmath69 completes the proof of , i.e. , superposition coding can achieve the entire admissible rate region for the general smdc problem . in this section , we prove a new _ sliding - window _ subset entropy inequality and then use it to provide an alternative proof of the optimality of superposition coding for achieving the minimum sum rate . when the integers @xmath93 are circularly placed ( clockwise ) based on their natural order . ] for any integer @xmath3 let @xmath94 and for any @xmath43 and @xmath10 let @xmath95 as illustrated in figure [ fig : sw ] , @xmath96 represents a _ sliding window _ of length @xmath32 starting with @xmath3 when the integers @xmath93 are circularly placed ( clockwise or counter clockwise ) based on their natural order . we have the following _ sliding - window _ subset entropy inequality . [ thm : sw ] for any collection of @xmath0 jointly distributed random variables @xmath77 , we have @xmath97 for any @xmath79 . the equalities hold when @xmath98 are mutually independent of each other . consider a proof via an induction on @xmath32 . first , for @xmath99 we have @xmath100 where follows from the independence bound on entropy . next , assume that the inequality holds for @xmath101 for some @xmath102 , i.e. , @xmath103 we have @xmath104\\ & \geq & \frac{1}{2}\sum_{l=1}^{l}\left[h(x_{w_l^{(r+1)}})+h(x_{w_{\langle l+1\rangle}^{(r-1)}})\right]\label{l4 - 2}\\ & = & \frac{1}{2}\sum_{l=1}^{l}h(x_{w_l^{(r+1)}})+\frac{1}{2}\sum_{l=1}^{l}h(x_{w_{\langle l+1\rangle}^{(r-1)}})\\ & = & \frac{1}{2}\sum_{l=1}^{l}h(x_{w_l^{(r+1)}})+\frac{1}{2}\sum_{l=1}^{l}h(x_{w_l^{(r-1)}})\label{l4 - 3}\\ & \geq & \frac{1}{2}\sum_{l=1}^{l}h(x_{w_l^{(r+1)}})+\frac{1}{2}\cdot\frac{r-1}{r}\sum_{l=1}^{l}h(x_{w_l^{(r)}})\label{l4 - 4}\end{aligned}\ ] ] where follows from the submodularity of entropy ( * ? ? ? 14.a ) @xmath105 for @xmath106 and @xmath107 so @xmath108 and @xmath109 , and follows from the induction assumption . moving the second term on the right - hand side of to the left and multiplying both sides by @xmath110 , we have @xmath111 we have thus proved that the inequality also holds for @xmath112 . finally , note that when @xmath98 are mutually independent , we have @xmath113 this completes the proof of theorem [ thm : sw ] . note that for @xmath114 , the classical subset entropy inequality of han and the sliding - window subset entropy inequality are equivalent , and both can be equivalently written as @xmath115 for a _ general _ @xmath32 , the classical subset entropy inequality of han can be derived from the sliding - window subset entropy inequality via a simple _ permutation _ argument as follows . let @xmath116 be a permutation on @xmath38 . for any @xmath43 and @xmath10 , let @xmath117 by theorem [ thm : sw ] , we have @xmath118 for any @xmath79 . averaging over all possible permutations @xmath116 , we have @xmath119 \geq \frac{1}{l!}\sum_{\pi}\left[\frac{1}{\alpha}\sum_{l=1}^{l}h(x_{w_{\pi , l}^{(\alpha)}})\right ] . \label{eq : sw3}\ ] ] note that for any @xmath10 , @xmath120 substituting into and dividing both sides of the inequality by @xmath0 establish the classical subset entropy inequality of han . the sliding - window subset entropy inequality can be used to provide an alternative proof of the optimality of superposition coding for achieving the minimum sum rate as follows . let us first show that @xmath121 for any @xmath81 . consider a proof via an induction on @xmath122 . when @xmath123 , can be written as @xmath124 which can be obtained via a uniform averaging of for @xmath99 and @xmath125 for @xmath43 . now assume that the inequality holds for @xmath126 for some @xmath127 . we have @xmath128 where follows from the sliding - window subset entropy inequality with @xmath101 . letting @xmath112 and @xmath129 in , we have @xmath130 substituting into gives @xmath131 this completes the proof of the induction step and hence . now let @xmath82 , and we have @xmath132 substituting into and dividing both sides of the inequality by @xmath20 , we have @xmath84 finally , letting @xmath68 and @xmath69 completes the proof of , i.e. , superposition coding can achieve the minimum sum rate for the general smdc problem . note that unlike the original proof of @xcite , which uses the classical subset entropy inequality of han @xcite and hence involves _ all _ nonempty subsets @xmath5 of @xmath38 , our proof relies on the sliding - window subset entropy inequality and hence only involves the subsets @xmath5 of a sliding - window type , i.e. , @xmath133 for some @xmath43 and @xmath10 . therefore , based on our proof , the converse result remains to be true even if we weaken the asymptotically perfect reconstruction requirement to @xmath134 this is the _ definitive _ advantage of our proof over that based on the classical subset entropy inequality of han @xcite . in this section , we revisit the subset entropy inequality of yeung and zhang , which played a key in their proof @xcite of the optimality of superposition coding for achieving the entire admissible rate region of the problem . as mentioned previously , in @xcite the subset entropy inequality was proved by combining the classical subset entropy inequality of han @xcite and a number of analysis results on the sequence of linear programs . however , the inequality , as stated in theorem [ thm : yz ] , does not even directly imply the classical subset entropy inequality of han @xcite . the reason is that theorem [ thm : yz ] merely asserts the existence of a set of optimal solutions @xmath135 , @xmath10 , that satisfies the subset entropy inequality , rather than providing a sufficient condition for the inequality to hold . below , we shall use a subset entropy inequality recently proved by madiman and tetali @xcite to summarize the analysis results of @xcite on the sequence of linear programs into a succinct sufficient condition for the subset entropy inequality to hold . consider a _ @xmath136 where @xmath5 is a finite ground set and @xmath137 is a collection of subsets of @xmath5 . a function @xmath138 is called a _ fractional cover _ of @xmath136 if it satisfies @xmath139 [ thm : mt ] let @xmath136 be a hypergraph , and let @xmath140 be a fractional cover of @xmath136 . then @xmath141 for any collection of jointly distributed random variables @xmath142 . the following corollary provides a chain " form of the subset entropy inequality . let @xmath143 be a positive integer , and let @xmath144 be a finite ground set . let @xmath145 be a collection of subsets of @xmath144 for each @xmath146 . assuming that @xmath145 , @xmath147 , are _ mutually exclusive _ , then @xmath148 _ induces _ a collection of hypergraphs @xmath149 where @xmath150 we shall term each subset @xmath151 a child " of @xmath5 . for convenience , we shall also define @xmath152 and term each subset @xmath153 a parent " of @xmath76 . [ cor : mt ] let @xmath154 . for any @xmath155 , if there exists a collection of functions @xmath156 for which each @xmath157 is a fractional cover of @xmath158 and such that @xmath159 we have @xmath160 for any collection of jointly distributed random variables @xmath161 . fix @xmath162 . for any @xmath163 , @xmath157 is a fractional cover of @xmath158 . by the subset entropy inequality of madiman and tetali , we have @xmath164 multiplying both sides of by @xmath165 and summing over @xmath163 , we have @xmath166 note that @xmath167 where follows . substituting into completes the proof of the corollary . specifying @xmath168 , @xmath169 , and @xmath170 for @xmath10 , the subset entropy inequality of madiman and tetali can be used to provide a _ unifying _ proof for both the subset entropy inequality of han and the subset entropy inequality of yeung and zhang . note that the choice @xmath171 is _ regular _ in that each subset @xmath66 has exactly @xmath32 children in @xmath172 , and each subset @xmath70 has exactly @xmath173 parents in @xmath37 . to see how the subset entropy inequality of madiman and tetali implies the subset entropy inequality of han , let @xmath174 and @xmath175 for any @xmath79 and @xmath66 , @xmath176 so @xmath157 is a _ uniform _ fractional cover of @xmath158 . furthermore , for any @xmath79 and @xmath70 we have @xmath177 substituting into immediately gives the subset entropy inequality of han . to see how the subset entropy inequality of madiman and tetali implies the subset entropy inequality of yeung and zhang , we shall need the following result , which is a synthesis of the analytical results on the sequence of linear programs established in @xcite . ( for completeness , a sketched proof based on the results of @xcite is included in appendix [ app : a ] . ) [ thm : yz2 ] for any @xmath50 , any @xmath79 , and any @xmath135 which is an optimal solution to the linear program with the optimal value @xmath178 , there exists a collection of functions @xmath179 for which each @xmath157 is a fractional cover of @xmath158 and such that @xmath180 where @xmath181 is an optimal solution to the linear program with @xmath32 replaced by @xmath182 . now fix @xmath50 , and consider the following construction of @xmath183 . for @xmath114 , choose @xmath184 to be an arbitrary _ optimal _ solution to the linear program . for @xmath185 , construct @xmath135 iteratively as follows . suppose that @xmath135 is already in place for some @xmath79 such that @xmath135 is an optimal solution to the linear program . if the optimal value @xmath178 , construct @xmath186 according to so @xmath186 is an optimal solution to the linear program with @xmath32 replaced by @xmath182 . moreover , by corollary [ cor : mt ] @xmath186 and @xmath135 satisfy the subset entropy inequality of yeung and zhang . if , on the other hand , @xmath187 , we have @xmath188 for _ all _ @xmath66 . in this case , choose @xmath186 to be an arbitrary _ optimal _ solution to the linear program with @xmath32 replaced by @xmath182 , and @xmath186 and @xmath135 will trivially satisfy the subset entropy inequality of yeung and zhang . we have thus constructed for any @xmath50 , a sequence of @xmath135 , @xmath10 , such that each @xmath135 is an optimal solution to the linear program , and the subset entropy inequality of yeung and zhang holds for each @xmath79 . we mention here that even though both the subset entropy inequality of han and the subset entropy inequality of yeung and zhang can be directly established from the subset entropy inequality of madiman and tetali , this is _ not _ the case for the sliding - window subset entropy inequality except for @xmath99 and @xmath0 . this can be seen as follows . let @xmath168 , @xmath169 , and @xmath189 for @xmath10 . note that for any @xmath185 , each sliding window @xmath96 represents a _ subset for different @xmath3 . ( for @xmath114 , all sliding windows @xmath190 , @xmath43 , represent the _ same _ subset @xmath38 . ) furthermore , for any @xmath191 each sliding window @xmath96 has only two children : @xmath192 and @xmath193 , and each sliding window @xmath192 has only two parents : @xmath96 and @xmath194 . now consider the elements @xmath3 and @xmath195 from @xmath96 . note that among the two children @xmath192 and @xmath193 of @xmath96 , @xmath3 belongs only to @xmath192 , and @xmath195 belong only to @xmath193 . thus , _ any _ fractional cover @xmath196 of the hypergraph @xmath197 must satisfy @xmath198 now let @xmath199 for all @xmath43 and @xmath185 . we have @xmath200 for any @xmath201 . we thus conclude that for any @xmath202 , the sliding - window subset entropy inequality can not be directly inferred from the subset entropy inequality of madiman and tetali . we conclude this section by providing a conditional extension of the subset entropy inequality of yeung and zhang , which will play a key role in proving the optimality of superposition coding for achieving the entire admissible rate region of the general s - smdc problem . we shall start with the following generalization of corollary [ cor : mt ] . let @xmath144 be a finite ground set , and let @xmath145 , @xmath147 , be a collection of subsets of @xmath144 . as before , we shall assume that the collections @xmath145 , @xmath147 , are mutually exclusive , so @xmath203 induces a hypergraph @xmath158 for every @xmath204 . for each @xmath205 let @xmath206 be a collection of subsets of @xmath144 , and let @xmath207 . for @xmath208 , define @xmath209 iteratively as follows . suppose that @xmath210 is already in place for some @xmath155 . let @xmath211 where @xmath212 [ prop : cmt ] for each @xmath213 , let @xmath214 . for any @xmath155 , if there exists a collection of functions @xmath156 for which each @xmath157 is a fractional cover of @xmath158 and such that @xmath215 we have @xmath216 for any collection of jointly distributed random variables @xmath161 . fix @xmath162 . for any @xmath163 , @xmath157 is a fractional cover of @xmath158 . by the subset entropy inequality of madiman and tetali , we have @xmath217 multiplying both sides of by @xmath218 and summing over @xmath219 and @xmath163 , we have @xmath220 note that @xmath221 where follows from . substituting into completes the proof of the proposition . [ thm : cyz ] for any @xmath50 and @xmath222 , there exists for each @xmath223 a collection of subsets @xmath206 of @xmath38 such that : @xmath224 such that : * for each @xmath225 , @xmath226 where @xmath227 is an _ optimal _ solution to the linear program ; and * for each @xmath228 , @xmath229 for _ any _ collection of @xmath0 jointly distributed random variables @xmath77 . fix @xmath50 and @xmath230 , and let @xmath168 , @xmath231 , and @xmath170 for @xmath225 . consider the following construction of @xmath210 and @xmath232 , @xmath225 . for @xmath233 , let @xmath234 where @xmath235 , i.e. , each @xmath206 contains a single subset @xmath236 of size @xmath237 and such that @xmath238 . furthermore , let @xmath239 be an _ optimal _ solution to the linear program for @xmath233 , and let @xmath240 since by construction each @xmath206 , @xmath241 , contains a _ single _ subset @xmath236 , we trivially have @xmath242 for @xmath243 , let us construct @xmath210 and @xmath244 iteratively as follows . suppose that @xmath210 and @xmath244 are already in place for some @xmath228 such that @xmath245 and @xmath246 for any @xmath66 and @xmath219 , and @xmath226 where @xmath247 is given by is an _ optimal _ solution to the linear program . first , construct @xmath248 according to . based on this construction , for any @xmath70 and @xmath249 we have @xmath219 for some @xmath250 . therefore , by the induction assumption we must have @xmath245 and @xmath251 for any @xmath70 and @xmath249 . next , construct @xmath252 as follows . if the optimal value @xmath253 , by theorem [ thm : yz2 ] there exists a collection of functions @xmath179 for which each @xmath157 is a fractional cover of @xmath158 and such that @xmath180 where @xmath254 is given by is an _ optimal _ solution to the linear program with @xmath32 replaced by @xmath182 . in this case , let @xmath255 where @xmath256 thus , for each @xmath70 we have @xmath257\\ & = \sum_{u \in \mathcal{u}_v}g_u(v)\left[\sum_{a \in \mathcal{a}_u}s_{\boldsymbol{\lambda}}(u , a)\right]\\ & = \sum_{u \in \mathcal{u}_v}g_u(v)c_{\boldsymbol{\lambda}}(u)\\ & = c_{\boldsymbol{\lambda}}(v)\end{aligned}\ ] ] furthermore , by proposition [ prop : cmt ] @xmath252 and @xmath252 satisfy the subset entropy inequality . if , on the other hand , @xmath258 , we have @xmath259 for _ all _ @xmath66 and @xmath219 . in this case , choose an arbitrary @xmath252 such that @xmath180 where @xmath260 is an optimal solution to the linear program with @xmath32 being replaced by @xmath182 , and @xmath252 and @xmath244 will trivially satisfy the subset entropy inequality . we have thus constructed for any @xmath50 and @xmath222 , a sequence of @xmath210 and @xmath135 , @xmath225 , such that all conditions of theorem [ thm : cyz ] are met simultaneously . this completes the proof of the theorem . randomly accessible encoders @xmath12 to @xmath0 . a total of @xmath0 independent discrete memoryless sources @xmath261 are to be encoded at the encoders . the decoder , which has access to encoder 0 and a subset @xmath5 of the randomly accessible encoders , needs to nearly perfectly reconstruct the sources @xmath262 no matter what the realization of @xmath5 is . ] as illustrated in figure [ fig : model ] , the problem of smdc - a consists of : * a total of @xmath0 _ independent _ discrete memoryless sources @xmath9\}_{t=1}^{\infty}$ ] , where @xmath10 and @xmath11 is the time index ; * a set of @xmath263 encoders ( encoder @xmath264 to @xmath0 ) ; * a decoder who has access to a subset @xmath265 of the encoder outputs for some nonempty @xmath13 . the realization of @xmath5 is _ unknown _ a priori at the encoders . however , no matter which @xmath5 actually materializes , the decoder needs to nearly perfectly reconstruct the sources @xmath266 whenever @xmath7 . formally , an @xmath267 code is defined by a collection of @xmath263 encoding functions @xmath268 and @xmath16 decoding functions @xmath269 a nonnegative rate tuple @xmath270 is said to be _ admissible _ if for every @xmath19 , there exits , for sufficiently large block length @xmath20 , an @xmath267 code such that : * ( rate constraints at the encoders ) @xmath271 * ( asymptotically perfect reconstructions at the decoder ) @xmath272 where @xmath23\}_{t=1}^{n}$ ] , @xmath24 is the output of encoder @xmath3 , and @xmath273 . the _ admissible rate region _ @xmath26 is the collection of _ all _ admissible rate tuples @xmath270 . similar to classical smdc , a natural strategy for smdc - a is superposition coding , i.e. , to encode the sources separately at the encoders and there is no coding across different sources . formally , the problem of encoding a single source @xmath29 can be viewed as a special case of the general problem where the sources @xmath30 are deterministic for all @xmath31 . in this case , the source @xmath29 needs to be nearly perfectly reconstructed whenever the decoder can access at least @xmath32 randomly accessible encoders in addition to the all - access encoder @xmath264 . the following scheme is a natural extension of the simple source - channel separation scheme considered previously for classical smdc : * first compress the source sequence @xmath33 into a source message @xmath34 using a _ lossless _ source code . it is well known ( * ? ? ? 5 ) that the rate of the source message @xmath34 can be made arbitrarily close to the entropy rate @xmath35 for sufficiently large block length @xmath20 . * next , divide the source message @xmath34 into two independent sub - messages @xmath274 and @xmath275 so we have @xmath276 the sub - message @xmath274 is stored at the all - access encoder 0 _ without _ any coding , which requires @xmath277 the sub - message @xmath275 is encoded by the randomly accessible encoders 1 to @xmath0 using a _ maximum distance separable _ code @xcite . clearly , the sub - message @xmath275 can be perfectly recovered at the decoder whenever @xmath278 for sufficiently large block length @xmath20 . eliminating @xmath279 and @xmath280 from , we conclude that the source message @xmath34 can be perfectly recovered at the decoder whenever @xmath281 combining the above two steps , we conclude that the rate region that can be achieved by the above source - channel separation scheme is given by the collection of all nonnegative rate tuples @xmath270 satisfying @xmath282 following the same footsteps as those for classical smdc @xcite , it is straightforward to show that the above rate region is in fact the admissible rate region for encoding the single source @xmath29 . by definition , the superposition coding rate region @xmath40 for smdc - a is given by the collection of all nonnegative rate tuples @xmath270 such that @xmath283 for some nonnegative @xmath42 , @xmath10 and @xmath284 , satisfying @xmath285 similar to classical smdc , the superposition coding rate region @xmath40 for smdc - a is a _ polyhedron _ with polyhedral cone being the nonnegative orthant in @xmath286 and hence can be completely characterized by the supporting hyperplanes @xmath287 where @xmath288 clearly , the above optimization problem can be separated into the following @xmath0 sub - optimization problems : @xmath289 where @xmath290 here , follows from the strong duality for linear programs . for any @xmath291 , @xmath50 , and @xmath10 , let @xmath292 we have @xmath293 and hence @xmath294 for any @xmath291 and @xmath50 . note that in the optimization problem , if the constraint @xmath295 is inactive , it can be removed from the program . in this case the optimal value @xmath296 , where @xmath297 is the optimal value of the linear program . on the other hand , if the constraint @xmath298 is active , the optimal value @xmath299 . combing these two cases , we have @xmath300 substituting and into , we conclude that the superposition coding rate region @xmath40 for smdc with an all - access encoder is given by the collection of all nonnegative rate tuples @xmath270 satisfying @xmath301 as mentioned previously , the superposition coding rate region @xmath40 is a polyhedron , so among all @xmath291 and @xmath50 , most of the inequalities in are _ redundant_. identifying those which define the _ faces _ of the superposition coding rate region @xmath40 appears to be very difficult . note , however , that for any given @xmath302 and @xmath50 , the left - hand side of is a linear , nondecreasing function of @xmath303 , and the right - hand side of is a piecewise linear , nondecreasing , and concave function of @xmath303 . thus , the left - hand side of will dominate the right - hand side for _ every _ @xmath291 if and only if it dominates the right - hand side at its boundary points @xmath304 , @xmath81 , between the adjacent line segments . see figure [ fig : obs ] for an illustration . for a fixed @xmath302 and @xmath50 . ] formally , we have the following proposition , which plays a key role next in proving the optimality of superposition coding for achieving the entire admissible rate region of smdc - a . [ prop : sc ] the superposition coding rate region @xmath40 for smdc - a is given by the collection of all nonnegative rate tuples @xmath270 satisfying @xmath305 where @xmath297 is the optimal value of the linear program . let us first recall the following results from @xcite : for any @xmath306 we have @xmath307 it follows that @xmath308 it remains to show that for any given @xmath50 , the set of inequalities over all @xmath291 is dominated by that over @xmath304 for @xmath81 . fix @xmath50 , and consider the following three cases separately . case 1 : @xmath309 . by , we have @xmath310 and hence @xmath311 for any @xmath312 . for @xmath123 , the inequality can be written as @xmath313 which implies that @xmath314 for any @xmath309 . case 2 : @xmath315 . by , we have @xmath316 and hence @xmath317 for any @xmath312 . for @xmath82 , the inequality can be written as @xmath318 which implies that @xmath319 for any @xmath315 . case 3 : @xmath320 for some @xmath321 . by , we have @xmath316 and hence @xmath317 for @xmath322 , and @xmath323 and hence @xmath311 for @xmath324 . for @xmath325 and @xmath326 , the inequality can be written as @xmath327 respectively , which together imply that @xmath328 for any @xmath320 . combining these three cases completes the proof of the proposition . the main result of this section is that superposition coding remains optimal in terms of achieving the entire admissible rate region for smdc - a , as summarized in the following theorem . [ thm : smdc - a ] for the general smdc - a problem , the admissible rate region @xmath329 based on the discussions from section [ sec : ext - smdca - sc ] , we naturally have @xmath330 . thus , to show @xmath331 we only need to show that @xmath332 . in light of proposition [ prop : sc ] , it is sufficient to show that _ any _ admissible rate tuple @xmath270 must satisfy @xmath333 for _ all _ @xmath81 and @xmath50 . let @xmath270 be an admissible rate tuple . by definition , for any sufficiently large block - length @xmath20 there exists an @xmath267 code satisfying the rate constraints for the admissible rate tuple @xmath270 and the asymptotically perfect reconstruction requirement . fix @xmath50 , and let @xmath334 be a set of _ optimal _ solutions that satisfies the subset entropy inequality of yeung and zhang . note that for @xmath89 , the optimal solution for the linear program is _ unique _ and is given by @xmath335 we thus have for any @xmath81 @xmath336 where follows from the rate constraint , and are due to the fact that @xmath337 and @xmath338 are optimal so we have @xmath339 follows from the subset entropy inequality of yeurng and zhang so we have @xmath340 and follows from the independence bound on entropy . for any @xmath341 and @xmath81 , by the asymptotically perfect reconstruction requirement and the well - known fano s inequality we have @xmath342 where @xmath343 in the limit as @xmath68 and @xmath69 . by the chain rule for entropy , @xmath344 where is due to the fact that @xmath1 are independent memoryless sources , and follows from . substituting into , we have @xmath345 next , we show , via an induction on @xmath122 , that for any @xmath81 we have @xmath346 first consider the base case with @xmath82 . in this case , the inequality is trivial as the right - hand side of the inequality is zero . next , assume that the inequality holds for @xmath347 for some @xmath348 , i.e , @xmath349 for any @xmath350 , we have @xmath351 where follows from the fact that conditioning reduces entropy , follows the fact that @xmath352 and follows from the fact that @xmath1 are independent memoryless sources . multiplying both sides of the inequality by @xmath247 and summing over all @xmath350 , we have @xmath353 where follows from the induction assumption . finally , by the subset entropy inequality of yeung and zhang we have @xmath354 this proves that the inequality also holds for @xmath355 and hence completes the proof of . substituting into and dividing both sides of the inequality by @xmath20 , we have @xmath356 letting @xmath68 and @xmath69 , we have from that @xmath357 for any admissible rate tuple @xmath270 . this proves that @xmath332 and hence completes the proof of the theorem . in this section , we conclude our discussion on smdc - a by focusing on a _ greedy _ rate allocation policy at the all - access encoder . based on our previous discussion in section [ sec : ext - smdca - sc ] , the output of the all - access encoder 0 consists of only _ uncoded _ information bits for the source messages @xmath358 . hence , its storage efficiency is the same for each of the information sources @xmath1 . on the other hand , for the randomly accessible encoders 1 to @xmath0 , @xmath359 has the highest reconstruction requirement and hence is the least efficient source to encode , and @xmath360 has the lowest reconstruction requirement and hence is the most efficient source to encode . therefore , intuitively , the greedy policy that assigns the remaining rate budget of the all - access encoder 0 to the least efficient source should be optimal . more specifically , suppose that the rate budget @xmath361 of the all - access encoder 0 satisfies @xmath362 for some @xmath363 . the greedy policy stores the source messages @xmath364 in their entireties ( without any coding ) at the all - access encoder 0 , and the residual rate budget @xmath365 is then committed in full to the source message @xmath366 . the residual source messages are @xmath366 , with a residual rate @xmath367 and @xmath368 with respective rates @xmath369 . the residual source messages are encoded at the randomly accessible encoders using superposition coding , and the corresponding rate region @xmath370 is given by @xmath371 of course , when @xmath372 all source messages @xmath358 can be stored at the all - access encoder 0 ( without any coding ) , and there is no need to use the randomly access encoders 1 to @xmath0 . in this case , we have @xmath373 . to show that the aforementioned greedy rate allocation policy at the all - access encoder 0 is optimal , we need to show that @xmath370 _ matches _ the @xmath361-slice of the superposition coding rate region @xmath374 for _ all _ @xmath375 . by proposition [ prop : sc ] , for any @xmath375 the @xmath361-slice of the superposition coding rate region can be written as @xmath376 where @xmath377 for any @xmath378 , it is straightforward to calculate that @xmath379 by , @xmath380 for any @xmath378 and @xmath50 . thus , when @xmath381 for some @xmath363 , we have @xmath382 and hence @xmath383 for all @xmath384 , and @xmath385 and hence @xmath386 for all @xmath387 . we conclude that in this case , @xmath388 for any @xmath50 and hence @xmath389 . when@xmath390 , we have @xmath385 and hence @xmath386 for all @xmath378 . in this case , @xmath391 for any @xmath50 , and we once again have @xmath389 . we summarize the above results in the following theorem . greedy rate allocation at the all - access encoder combined with superposition coding at the randomly accessible encoders can achieve the entire admissible rate region for the general smdc - a problem . randomly accessible encoders @xmath12 to @xmath0 . a total of @xmath392 independent discrete memoryless sources @xmath393 are to be encoded at the encoders . the legitimate receiver , which has access to a subset @xmath5 of the encoder outputs , needs to nearly perfectly reconstruct the sources @xmath394 whenever @xmath395 . the eavesdropper has access to a subset @xmath396 of the encoder ouputs . all sources @xmath393 need to be kept perfectly secret from the eavesdropper whenever @xmath397 . ] let @xmath0 be a positive integer , and let @xmath230 . let @xmath398,\ldots , s_{l - n}[t]\}_{t=1}^{\infty}$ ] be a collection of @xmath392 _ independent _ discrete memoryless sources with time index @xmath11 , and let @xmath399,\ldots , s_\alpha[n])$ ] for @xmath225 . as illustrated in figure [ fig : ssmdc ] , an @xmath400 s - smdc problem consists of a set of @xmath0 encoders , a legitimate receiver who has access to a subset @xmath5 of the encoder outputs , and an eavesdropper who has access to a subset @xmath396 of the encoder outputs . which subsets of the encoder outputs are available at the legitimate receiver and the eavesdropper are _ unknown _ a priori at the encoders . however , no matter which subsets @xmath5 and @xmath396 actually occur , the legitimate receiver must be able to asymptotically perfectly reconstruct the sources @xmath266 whenever @xmath401 , and all sources @xmath393 must be kept perfectly secure from the eavesdropper as long as @xmath402 . formally , an @xmath14 code is defined by a collection of @xmath0 encoding functions @xmath403 and @xmath404 decoding functions @xmath405 where @xmath406 is the key space accessible to all @xmath0 encoders . a nonnegative rate tuple @xmath18 is said to be _ admissible _ if for every @xmath19 , there exits , for sufficiently large block length @xmath20 , an @xmath14 code such that : * ( rate constraints ) @xmath407 * ( asymptotically perfect reconstruction at the legitimate receiver ) @xmath408 where @xmath409 is the output of the @xmath3th encoder , and @xmath410 is the secret key shared by all @xmath0 encoders ; and * ( perfect secrecy at the eavesdropper ) @xmath411 i.e. , observing the encoder outputs @xmath412 does not provide _ any _ information regarding to the sources @xmath413 . the _ admissible rate region _ @xmath414 is the collection of _ all _ admissible rate tuples @xmath415 . a simple strategy for s - smdc is to encode each of the @xmath392 sources separately without coding across different sources . formally , the problem of encoding a single source @xmath29 can be viewed as a special case of the general s - smdc problem with @xmath416 for all @xmath31 . when @xmath89 , the problem of encoding the single source @xmath359 is the well - known @xmath417 _ threshold secret sharing _ problem , for which the admissible rate region was characterized in the classical works @xcite . for the general case with @xmath418 , the admissible rate region for encoding the single source @xmath29 was characterized in @xcite via a connection to the problem of threshold _ ramp - type _ secret sharing @xcite and utilizing some basic polyhedral structure of the admissible rate region . the result is summarized in the following proposition . [ prop : s - ssdc ] let @xmath419 be the collection of all admissible rate tuples for encoding the single source @xmath29 . then , @xmath419 is given by the collection of all nonnegative tuples @xmath420 such that @xmath421 by definition , the superposition coding rate region @xmath40 for encoding the sources @xmath422 is given by the collection of nonnegative rate tuples @xmath18 such that @xmath423 note that @xmath419 is _ identical _ to the admissible rate region for encoding the single source @xmath29 in classical smdc ( even though the reconstruction and secrecy requirements are different between these two settings ) . we thus conclude that the superposition coding rate region @xmath40 for s - smdc is given by the collection of nonnegative rate tuples @xmath18 satisfying @xmath424 where @xmath297 is the optimal value of the linear program . in @xcite , it was shown that superposition coding can achieve the minimum sum rate for the general s - smdc problem . the proof was based on the trivial conditional version of the subset entropy inequality of han . the main result of this section is to show that superposition coding can , in fact , achieve the entire admissible region for the general s - smdc problem . our main technical tool is the conditional extension of the subset entropy inequality of yeung and zhang proved in theorem [ thm : cyz ] . [ thm : s - smdc ] for the general s - smdc problem , the admissible rate region @xmath329 based on the discussions from section [ sec : ext - ssmdc - sc ] , we naturally have @xmath330 . thus , to show @xmath331 we only need to show that @xmath332 , i.e. , _ any _ admissible rate tuple @xmath18 must satisfy . let @xmath18 be an admissible rate tuple . by definition , for any sufficiently large block - length @xmath20 there exists an @xmath14 code satisfying the rate constraints for the admissible rate tuple @xmath18 , the asymptotically perfect reconstruction requirement , and the perfect secrecy requirement . fix @xmath50 , and choose @xmath210 and @xmath244 , @xmath225 , to satisfy all the requirement of theorem [ thm : cyz ] . first , let us show that @xmath425 for any @xmath66 , @xmath219 , and @xmath225 , where @xmath426 in the limit as @xmath68 and @xmath69 . fix @xmath66 , @xmath219 , and @xmath225 . by construction @xmath427 , @xmath245 , and @xmath246 , so we have @xmath428 . by the asymptotically perfect reconstruction requirement and the well - known fano s inequality , we have @xmath429 where @xmath426 in the limit as @xmath68 and @xmath69 . furthermore , by the perfect secrecy requirement we have @xmath430 we thus have @xmath431 where follows from , follows from , follows from the fact that conditioning reduces entropy , follows from the fact that the sources @xmath6 are mutually independent , and follows from the fact that the source @xmath29 is memoryless . moving @xmath432 to the right - hand side of the inequality completes the proof of . next , let us we show that @xmath433 for any @xmath434 . consider a proof via an induction on @xmath122 . first consider the base case with @xmath123 . we have @xmath435\label{eq : c1}\\ & & \hspace{20pt } = \ ; nf_1(\boldsymbol{\lambda})h(s_1)-nf_1(\boldsymbol{\lambda})\delta_n^{(1)}+\sum_{u \in \omega_l^{(1)}}\sum_{a \in \mathcal a_u}s_{\boldsymbol{\lambda}}(u , a)h(x_u|s_1^n , x_a)\label{eq : c2}\end{aligned}\ ] ] where follows from with @xmath89 . next , assume that the inequality holds for @xmath436 for some @xmath437 , i.e. , @xmath438 we have @xmath439\label{eq : c4}\\ \hspace{-20pt } & & \hspace{20pt } \geq \ ; nf_k(\boldsymbol{\lambda})h(s_k)-nf_k(\boldsymbol{\lambda})\delta_n^{(k)}+\sum_{u \in \omega_l^{(k)}}\sum_{a \in \mathcal{a}_u}s_{k}(u , a)h(x_u|s_1^n,\ldots , s_k^n , x_a)\label{eq : c5}\end{aligned}\ ] ] where follows from , and follows from with @xmath440 . substituting into gives @xmath441 i.e. , the inequality also holds for @xmath442 . this completes the induction step and hence the proof of . finally , note that for @xmath89 the optimal solution for the linear program is _ unique _ and is given by @xmath335 we thus have @xmath443h(x_u)-nf_1(\boldsymbol{\lambda})\epsilon\\ & = & \sum_{u \in \omega_l^{(1)}}\sum_{a\in \mathcal a_u}s_{\boldsymbol{\lambda}}(u , a)h(x_u)-nf_1(\boldsymbol{\lambda})\epsilon\\ & \ge & \sum_{u\in\omega_l^{(1)}}\sum_{a\in \mathcal a_u}s_{\boldsymbol{\lambda}}(u , a)h(x_u|x_a)-nf_1(\boldsymbol{\lambda})\epsilon\label{pf - ssmdc-3}\\ & \geq & \left[n\sum_{\alpha=1}^{l - n}f_\alpha(\boldsymbol{\lambda})h(s_\alpha)-n\sum_{\alpha=1}^{l - n}f_\alpha(\boldsymbol{\lambda})\delta_n^{(\alpha)}+\right.\nonumber\\ & & \left.\sum_{u\in\omega_l^{(l - n)}}\sum_{a\in \mathcal a_u}s_{\boldsymbol{\lambda}}(u , a)h(x_u|s_1^n,\ldots , s_{l - n}^n , x_a)\right]-nf_1(\boldsymbol{\lambda})\epsilon\label{pf - ssmdc-4}\\ & \geq & n\sum_{\alpha=1}^{l - n } f_\alpha(\boldsymbol{\lambda})h(s_\alpha)-n\sum_{\alpha=1}^{l - n}f_\alpha(\boldsymbol{\lambda})\delta_n^{(\alpha)}-nf_1(\boldsymbol{\lambda})\epsilon\label{pf - ssmdc-5}\end{aligned}\ ] ] where follows from the rate constraint , follows from the fact that @xmath337 is optimal so @xmath444 , follows from the fact that conditioning reduce entropy , and follows from with @xmath445 . divide both sides of by @xmath20 and let @xmath68 and @xmath69 . note that @xmath446 in the limit as @xmath68 and @xmath69 for all @xmath225 . we have thus proved that holds for any admissible rate tuple @xmath18 . this completes the proof of the theorem . smdc is a classical model for coding over distributed storage . in this setting , a simple separate encoding strategy known as superposition coding was shown to be optimal in terms of achieving the minimum sum rate @xcite and the entire admissible rate region @xcite of the problem . the proofs utilized carefully constructed induction arguments , for which the classical subset entropy inequality of han @xcite played a key role . this paper includes two parts . in the first part the existing optimality proofs for classical smdc were revisited , with a focus on their connections to subset entropy inequalities . first , a new sliding - window subset entropy inequality was introduced and then used to establish the optimality of superposition coding for achieving the minimum sum rate under a weaker source - reconstruction requirement . second , a subset entropy inequality recently proved by madiman and tetali @xcite was used to develop a new structural understanding to the proof of yeung and zhang @xcite on the optimality of superposition coding for achieving the entire admissible rate region . building on the connections between classical smdc and the subset entropy inequalities developed in the first part , in the second part the optimality of superposition coding was further extended to the cases where there is either an additional all - access encoder ( smdc - a ) or an additional secrecy constraint ( s - smdc ) . finally , we mention here that an asymmetric " setting of the multilevel diversity coding problem was considered in the recent work @xcite , where the sources that need to be asymptotically perfectly reconstructed depend on , not only the cardinality , but the actual subset of the encoder outputs available at the decoder . unlike the symmetrical setting considered in @xcite and in this paper , as demonstrated in @xcite for the case with three encoders , coding across different sources is generally needed to achieve the entire admissible rate region of the problem . consider a proof via an induction on the total number of encoders @xmath0 . fix @xmath50 . without loss of generality , let us assume that @xmath447 first consider the base case with @xmath448 . in this case , the optimal solution to the linear program is unique and is given by @xmath449 when @xmath450 , it is straightforward to verify that @xmath451 is a fractional cover of @xmath452 and such that @xmath453 now , assume that the theorem holds for @xmath454 for some integer @xmath455 . fix @xmath456 , and let @xmath135 be an optimal solution to the linear program to with the optimal value @xmath178 . next , we show that we can always find a collection of functions @xmath179 for which each @xmath157 is a fractional cover of @xmath158 and such that @xmath180 where @xmath254 is given by is an optimal solution to the linear program with @xmath32 replaced by @xmath182 . we shall consider the following three cases separately . case 1 : @xmath457 . in this case , it is sufficient to consider for any @xmath458 , the _ uniform _ fractional cover @xmath459 for the hypergraph @xmath158 so we have @xmath460 by ( * ? ? ? * eq . ( 39 ) ) , @xmath186 constructed as such is an optimal solution to the linear program with @xmath32 replaced by @xmath182 . case 2 : @xmath461 . in this case , by ( * ? ? ? * lemma 6 ) @xmath462 implies that @xmath463 . furthermore , by ( * ? ? ? * lemma 8) @xmath464 where @xmath465 is an optimal solution to the linear program @xmath466 with the optimal solution @xmath467 . thus , by the induction assumption there exists a collection of functions @xmath468 such that each @xmath469 is a fractional cover of @xmath470 and @xmath471 where @xmath472 is an optimal solution to the linear program @xmath473 for any @xmath474 such that @xmath463 , let @xmath475 , and let @xmath476 for any @xmath474 such that @xmath477 , let us choose @xmath157 to be an _ arbitrary _ fractional cover of @xmath158 . then , for any @xmath478 such that @xmath479 we have @xmath480 where @xmath481 , and for any @xmath478 such that @xmath482 @xmath483 by ( * ? ? ? * eq . ( 46 ) ) , @xmath186 constructed as such is an optimal solution to the linear program with @xmath32 replaced by @xmath182 . it remains to show that @xmath157 is a fractional cover of @xmath158 for any @xmath474 such that @xmath463 . fix @xmath474 such that @xmath463 . for any @xmath484 , we have @xmath485 and @xmath486 this completes the proof of case 2 . case 3 : @xmath487 . in this case , we shall need the following notations . for any @xmath458 and @xmath488 , denote by @xmath489 the _ smallest _ positive integer @xmath3 such that @xmath490 let @xmath491 so @xmath492 . for each @xmath458 , @xmath493 , and @xmath494 , let @xmath495 where @xmath496 let @xmath497 consider the collection of functions @xmath498 where @xmath499 this gives @xmath500 by ( * ? ? ? * eq . ( 55 ) ) , @xmath186 constructed as such is an optimal solution to the linear program with @xmath32 replaced by @xmath182 . it remains to show that @xmath157 is a fractional cover of @xmath158 for any @xmath474 note that for any @xmath501 , @xmath502 and @xmath503 where follows from ( * ? ? ? * eq . ( 66 ) ) . combing and gives @xmath504 we thus conclude that @xmath157 as defined in is indeed a fractional cover of @xmath158 for any @xmath458 . this completes the proof of case 3 . tie liu would like to thank dr . jihong chen for discussions that have inspired some ideas of the paper . a. balasubramanian , h. d. ly , s. li , t. liu , and s. l. miller , secure symmetrical multilevel diversity coding , " _ ieee trans . inf . theory _ , submitted for publication . available online at http://arxiv.org/abs/1201.1935 h. yamamoto , secret sharing system using @xmath506 threshold scheme , " _ ieice trans . fundamentals ( japanese edition ) _ , vol . j68-a , pp . 945952 , sept . 1985 ( english translation : scripta technica , inc . , electronics and comm . in japan , part i , vol . 69 , pp . 4654 , 1986 )

the aim of these notes is to provide a new proof of the mattila - melnikov - verdera theorem on the uniform rectifiability of an ahlfors - david regular measure whose associated cauchy transform operator is bounded . they are based on lectures given by the second author in the analysis seminars at kent state university and tel - aviv university .

1307.1156

the purpose of these notes is to provide a new proof of mattila , melnikov , and verdera s theorem . the exposition is self - contained , relying only on a knowledge of basic real analysis . [ thm1]@xcite an ahlfors - david regular measure @xmath0 whose associated cauchy transform operator is bounded in @xmath1 is uniformly rectifiable . the precise statement of this theorem is given in section [ statementsec ] . the scheme employed to prove theorem [ thm1 ] in these notes is quite different from that in @xcite , and relies upon a characterization of _ reflectionless measures_. in this regard , one may compare the proof to that of mattila s theorem @xcite : _ suppose that @xmath0 is a finite borel measure satisfying @xmath2 for @xmath0-a.e . if the cauchy transform of @xmath0 exists @xmath0-a.e . in the sense of principal value , then @xmath0 is rectifiable_. mattila s proof of this theorem uses a characterization of _ symmetric measures _ , the reader may consult chapter 14 of the book @xcite for more information . subsequently , mattila s theorem was generalized to the case of singular integrals in higher dimensions by mattila and preiss in @xcite . to find the analogous generalization of the proof we carry out here would answer a longstanding problem of david and semmes @xcite . very recently , nazarov , tolsa , and volberg @xcite completed the solution of the problem of david and semmes in the case of singular integral operators of co - dimension 1 . they proved that if @xmath0 is a @xmath4-dimensional ahlfors - david regular measure in @xmath5 , then the boundedness of the @xmath4-dimensional riesz transform in @xmath1 implies that one of the criteria for uniform rectifiability given in @xcite is satisfied . see @xcite for more details , and for further references and history about this problem . throughout this paper , we shall only consider ahlfors - david regular measures . for closely related results without this assumption , see the paper of lger @xcite . we shall adopt the following notation : * @xmath6 denotes the open disc centred at @xmath7 with radius @xmath8 . * for a square @xmath9 , we set @xmath10 to be the centre of @xmath9 , and @xmath11 to denote the side - length of @xmath9 . * we shall denote by @xmath12 the standard lattice of dyadic squares in the complex plane . a dyadic square is any square of the form @xmath13 for @xmath14 and @xmath15 in @xmath16 . * we define the lipschitz norm of a function @xmath17 by @xmath18 * we denote by @xmath19 the space of compactly supported functions with finite lipschitz norm . the continuous functions with compact support are denoted by @xmath20 . * for @xmath21 , we set @xmath22 in particular , note that we are taking the _ pointwise everywhere _ supremum here . * the closure of a set @xmath23 is denoted by @xmath24 * the support of a measure @xmath0 is denoted by @xmath25 . * for a line @xmath26 , we write the one dimensional hausdorff measure restricted to @xmath26 by @xmath27 . if @xmath28 , we instead write @xmath29 . * we will denote by @xmath30 and @xmath31 various positive absolute constants . these constants may change from line to line within an intermediate argument . the constant @xmath30 is thought of as large ( at the very least greater than @xmath32 ) , while @xmath31 is thought of as small ( certainly smaller than @xmath32 ) . we shall usually make any dependence of a constant on a parameter explicit , unless it is clear from the context what the dependencies are . let @xmath33 for @xmath34 . for a measure @xmath35 , the _ cauchy transform _ of @xmath35 is formally defined by @xmath36 in general , the singularity in the kernel is too strong to expect the integral to converge absolutely on @xmath37 . it is therefore usual to introduce a regularized cauchy kernel . for @xmath38 , define @xmath39 then the @xmath40-regularized cauchy transform of @xmath35 is defined by @xmath41 before we continue , let us introduce a very natural condition to place upon @xmath0 . a measure @xmath0 is called @xmath42-_nice _ if @xmath43 for any disc @xmath44 . if @xmath0 is a @xmath42-nice measure , then for any @xmath45 and @xmath3 , we have that @xmath46 is bounded in absolute value in terms of @xmath40 , @xmath42 , and @xmath47 . to see this , we shall need an elementary tail estimate , which we shall refer to quite frequently in what follows : [ tailest ] suppose @xmath0 is @xmath42-nice measure . for every @xmath48 and @xmath8 , we have @xmath49 the proof of lemma [ tailest ] is a standard exercise , and is left to the reader . with this lemma in hand , we return to our claim that @xmath50 is bounded . first apply the cauchy - schwarz inequality to estimate @xmath51 but now @xmath52 the first term on the right hand side of this inequality is at most @xmath53 , and the second term is no greater than @xmath54 by lemma [ tailest ] . we therefore see that @xmath55 . in particular , we have @xmath56 . one conclusion of this discussion is that for any nice measure @xmath0 , it makes sense to ask if @xmath57 is a bounded operator from @xmath1 to @xmath1 . we say that @xmath0 is @xmath42-_good _ if it is @xmath42-nice and @xmath58 by definition , the cauchy transform operator associated with @xmath0 is bounded in @xmath1 if @xmath0 is good . the two dimensional lebesgue measure restricted to the unit disc is good . however , this measure is not supported on a @xmath32-rectifiable set and so such measures should be ruled out in a statement such as theorem [ thm1 ] . to this end , we shall deal with ahlfors - david ( ad ) regular measures . [ adreg ] a nice measure @xmath0 is called ad - regular , with regularity constant @xmath59 , if @xmath60 for any disc @xmath61 with @xmath62 . a set @xmath63 is called _ uniformly rectifiable _ if there exists @xmath64 such that for any dyadic square @xmath65 , there exists a lipschitz mapping @xmath66\rightarrow \mathbb{c}$ ] with @xmath67 and @xmath68)$ ] . we can altnernatively say that @xmath23 is uniformly rectifiable if there exists @xmath64 such that for any dyadic square @xmath65 , there is a rectifiable curve containing @xmath69 of length no greater than @xmath70 . a measure @xmath0 is uniformly rectifiable if the set @xmath71 is uniformly rectifiable . we may now restate theorem [ thm1 ] in a more precise way . a good ad - regular measure @xmath0 is uniformly rectifiable . the definition of a good measure does not immediately provide us with a workable definition of the cauchy transform on the support of @xmath0 . in this section , we rectify this matter by defining an operator @xmath72 as a weak limit of the operators @xmath73 as @xmath74 . this idea goes back to mattila and verdera @xcite . we fix a @xmath42-good measure @xmath0 . note that if @xmath75 , then @xmath17 is bounded in absolute value by @xmath76 . fix @xmath77 . then for any @xmath38 , we may write @xmath78d\mu(z)d\mu(\xi).\ ] ] let @xmath79 $ ] . it will be useful to denote by @xmath80 the expression @xmath81 now , note that if @xmath82 , it is clear that @xmath83 . in addition @xmath84 is lipschitz in @xmath85 with lipschitz norm no greater than @xmath86 . to see this , first observe that @xmath87 , whenever @xmath88 . by using this inequality twice , we see that @xmath89,\ ] ] and the claim follows since @xmath90 . since @xmath91 , this lipschitz bound immediately yields @xmath92 as a result of this bound on the absolute value of @xmath84 , there exists a constant @xmath93 such that @xmath94 on the other hand , since @xmath0 is a nice measure , the set @xmath95 is @xmath96 null , and so for @xmath96 almost every @xmath97 , the limit as @xmath98 of @xmath99 is equal to @xmath100 . as a result , the dominated convergence theorem applies to yield @xmath101 this limit will be denoted by @xmath102 . moreover , there is a quantitative estimate on the speed of convergence : @xmath103 since @xmath0 is @xmath42-nice , @xmath104 can be bounded in terms of the diameters of the supports of @xmath17 and @xmath105 , and we see that @xmath106 . we have now justified the existence of an operator @xmath72 acting from the space of compactly supported lipschitz functions to its dual with respect to the standard pairing @xmath107 . since @xmath0 is @xmath42-good , for any @xmath38 we have @xmath108 to the case when @xmath17 and @xmath105 are @xmath1 functions . to do this , we first pick functions @xmath17 and @xmath105 in @xmath1 . let @xmath48 . using the density of @xmath109 in @xmath1 , we write @xmath110 and @xmath111 , where @xmath112 and @xmath113 are compactly supported lipschitz functions , and the norms of @xmath114 and @xmath115 in @xmath1 are as small as we wish ( say , less than @xmath116 ) . we know that @xmath117 as @xmath74 . consequently , for each @xmath48 , @xmath80 can be written as a sum of two terms , the first of which ( namely @xmath118 ) has a finite limit , and the second term ( which is @xmath119 ) has absolute value no greater than @xmath120 . it follows that the limit as @xmath74 of @xmath80 exists . we define this limit to be @xmath102 . from ( [ idelt ] ) , we see that @xmath121 . therefore we may apply the riesz - fisher theorem to deduce the existence of a ( unique ) bounded linear operator @xmath122 such that @xmath123 having defined an operator @xmath72 for any good measure @xmath0 , we now want to see what weak continuity properties this operator has . we say that the sequence @xmath124 tends to @xmath0 weakly if , for any @xmath125 , @xmath126 we now recall a standard weak compactness result , which can be found in chapter 1 of @xcite ( or any other book in real analysis ) . [ bookconv ] let @xmath127 be a sequence of measures . suppose that for each compact set @xmath63 , @xmath128 . then there exists a subsequence @xmath129 and a measure @xmath0 such that @xmath130 converges to @xmath0 weakly . an immediate consequence of this lemma is that any sequence @xmath127 of @xmath42-nice measures has a subsequence that converges weakly to a measure @xmath0 . the next lemma shows that the various regularity properties of measures that we consider are inherited by weak limits . [ wlnicegood ] suppose that @xmath124 converges to @xmath0 weakly . if each measure @xmath124 is @xmath42-good with @xmath131 regularity constant @xmath132 , then the limit measure @xmath0 is also @xmath42-good with @xmath131 regularity constant @xmath132 . we shall first check that @xmath0 is ad regular . let @xmath133 , @xmath8 , and choose @xmath134 . consider a smooth non - negative function @xmath17 , supported in the disc @xmath135 , with @xmath136 on @xmath137 . then @xmath138 . hence , for all sufficiently large @xmath139 , @xmath140 . for all such @xmath139 , @xmath141 , and so there exists @xmath142 satisfying @xmath143 . as a result , @xmath144 . now let @xmath145 be nonnegative and supported in @xmath146 , satisfying @xmath147 and @xmath148 on @xmath149 . then @xmath150 letting @xmath151 , we arrive at the desired ad regularity . the property that @xmath0 is @xmath42-nice is easier and left to the reader ( it also follows from standard lower - semicontinuity properties of the weak limit ) . it remains to show that @xmath0 is @xmath42-good . fix @xmath77 and define @xmath84 and @xmath152 as before . note that @xmath153 is a lipschitz function in @xmath85 , and has support contained in @xmath154 . let @xmath155 be an open set with @xmath156 . the ( complex valued ) stone - weierstrass theorem for a locally compact space tells us that that the algebra of finite linear combinations of functions in @xmath157 is dense in @xmath158 ( with respect to the uniform norm in @xmath159 ) . let @xmath48 . there are functions @xmath160 and @xmath161 , all belonging to @xmath162 , such that @xmath163 for any @xmath164 . for each @xmath165 , we have@xmath166 it therefore follows that @xmath167 on the other hand , @xmath124 is @xmath42 nice , and so @xmath168 . since @xmath48 was arbitrary , we conclude that @xmath169 as @xmath170 . as a result of this convergence , we have that @xmath171 but since both @xmath172 and @xmath173 are in @xmath20 , the right hand side of this inequality equals @xmath174 . we now wish to appeal to the density of @xmath19 in @xmath1 to extend this inequality to all @xmath175 . let @xmath176 . as @xmath0 is @xmath42-nice , we saw in section [ statementsec ] that @xmath177 . but then , since the space of lipschitz function compactly supported in @xmath178 is dense in @xmath179 , we conclude that @xmath180 . finally , taking the limit as @xmath181 , the monotone convergence theorem guarantees that @xmath182 , and hence @xmath0 is @xmath42-good . the proof of the next lemma is left as an exercise . [ adsupp ] suppose that @xmath124 is a sequence of @xmath132 @xmath131-regular measures converging weakly to a measure @xmath0 . if @xmath183 with @xmath184 as @xmath170 , then @xmath62 . our last task is to check that the bilinear form @xmath185 has nice weak convergence properties . for this , let @xmath77 . for @xmath38 , we write @xmath186 the first and third terms are bounded by @xmath187 . the second term converges to @xmath188 as @xmath170 . therefore @xmath189 but @xmath38 was arbitrary , and so @xmath190 converges to @xmath191 as @xmath170 . throughout this section we fix a @xmath42-nice measure @xmath0 . a system of functions @xmath192 @xmath193 is called a @xmath30-riesz system if @xmath194 for each @xmath9 , and @xmath195 for every sequence @xmath196 . by a simple duality argument , we see that if @xmath192 is a @xmath30-riesz system , then @xmath197 suppose now that with each square @xmath65 , we associate a set @xmath198 of @xmath1 functions . we say that @xmath198 @xmath193 is a @xmath30-riesz family if , for any choice of functions @xmath199 , the system @xmath192 forms a @xmath30-riesz system . we now introduce a particularly useful riesz family . suppose that @xmath0 is a @xmath42-nice measure . fix @xmath200 , and define @xmath201 for any @xmath200 , @xmath202 is a @xmath30-riesz family , with constant @xmath203 . for each @xmath65 , pick a function @xmath204 . then we have @xmath205 and @xmath206 now , if @xmath207 with @xmath208 , then @xmath209 provided that @xmath210 if @xmath211 intersects @xmath212 , we instead have the bound @xmath213 indeed , note that @xmath214 , while the oscillation of @xmath215 on the set @xmath211 ( which contains the support of @xmath216 ) is no greater than @xmath217 . by multiplying these two estimates we arrive at the desired bound on the absolute value of the inner product . consider a sequence @xmath218 . then @xmath219 inserting our bounds on the inner products into this sum , we see that we need to bound the sum @xmath220 ( since all sums involving squares will be taken over the lattice @xmath12 , we will not write this explicitly from now on . ) using cauchy s inequality , we estimate @xmath221 it therefore suffices to estimate two double sums : @xmath222 and @xmath223 for each dyadic length @xmath15 greater than @xmath224 , there are at most @xmath225 squares @xmath226 of side length @xmath15 for which @xmath227 has non - empty intersection with @xmath211 . hence @xmath228 concerning @xmath229 , all the relevant squares @xmath230 in the inner sum are contained in the disc @xmath231 . therefore , at scale @xmath15 there are at most @xmath232 squares @xmath230 of side length @xmath15 that can contribute to the inner sum . as a result , @xmath233 combining our bounds , we see that the @xmath202 is a riesz family , with riesz constant @xmath234 . in this section we identify a local property of the support of a measure , which ensures that the measure is uniformly rectifiable . the mathematics in this section is largely due to david , jones , and semmes , see @xcite chapter 2.1 , and is simpler than jones geometric traveling salesman theory @xcite , which was used in @xcite . fix a @xmath42-nice measure @xmath0 , which is ad - regular with regularity constant @xmath132 . set @xmath71 . we will begin by constructing a certain graph . for our purposes , a _ graph _ @xmath235 is a set of points @xmath236 ( the vertices ) , endowed with a collection of line segments @xmath237 ( the edges ) where each segment has its end - points at vertices . a _ connected component _ of the graph is a maximal subset of vertices that can be connected through the edges . for example , the graph depicted in figure 1 below has two connected components . the _ distance between connected components _ of a graph is measured as the distance between the relevant sets of vertices . therefore , the distance between the components of the graph depicted in figure 1 is the distance between the vertices labeled @xmath238 and @xmath239 . [ comppic ] for a graph @xmath235 , and a square @xmath9 , we define @xmath240 to be the subgraph with vertex set @xmath241 , endowed with the edges from @xmath237 connecting those vertices in @xmath242 . let @xmath243 . fix @xmath244 ( this square is to be considered as the viewing window in the definition of uniform rectifiability ) . choose a ( small ) dyadic fraction @xmath245 with @xmath246 . we shall construct a graph adapted to @xmath247 inductively . set @xmath236 to be a maximal @xmath248 separated subset of @xmath23 . note that @xmath236 forms a @xmath249 net of @xmath23 . * _ the base step . _ * for each square @xmath65 with @xmath250 and @xmath251 , fix a point which lies in @xmath252 . then join together every point of @xmath253 to this fixed point by line segments , as illustrated in the figure below . in @xmath254 , there are at most @xmath255 points of @xmath236 , and so the total length of the line segments in @xmath254 is @xmath256 . [ basesteppic ] .,title="fig : " ] we thereby form the graph @xmath257 comprised of the vertex set @xmath236 , and the set of line segments @xmath258 obtained by carrying out the above procedure for all squares @xmath65 with @xmath250 . this is the base step of the construction . * _ the inductive step . _ * let @xmath15 be a dyadic fraction no smaller than @xmath245 . suppose that we have constructed the graph @xmath259 . the graph @xmath260 is set to be the pair @xmath261 , where @xmath262 is obtained by taking the union of @xmath263 with the collection of line segments obtained by performing the following algorithm : _ for every square @xmath65 with @xmath264 , consider the graph @xmath265 . if @xmath266 has at least two components that intersect @xmath254 , then for each such component , choose a vertex that lies in its intersection with @xmath267 . fix a point in @xmath268 , and join each of the chosen points to this fixed point with an edge_. [ comppic ] . the grey edges indicate the edges of @xmath269 not included in the subgraph @xmath270 . the dashed lines indicate the edges added by applying the algorithm . note that in this case the graph @xmath266 has seven components , four of which intersect @xmath254 . the fixed point in @xmath268 is denoted by @xmath271.,title="fig : " ] we carry out the inductive procedure for @xmath272 , and thereby obtain the graph @xmath273 . to continue our analysis , first note the following elementary fact : [ dyadicfact ] let @xmath65 with @xmath264 . for any two points @xmath274 with @xmath275 , there is a dyadic square @xmath230 of sidelength @xmath15 , such that @xmath276 , and @xmath277 . pick the square @xmath230 to be the dyadic square of side length @xmath15 containing @xmath278 . then @xmath279 , so @xmath280 . since @xmath264 , we have that @xmath281 is a union of dyadic squares of side - length @xmath15 . therefore @xmath230 is contained in @xmath281 . as the square annulus @xmath282 is of width @xmath283 , we conclude that @xmath276 . we shall use this lemma ( or rather a weaker statement with @xmath281 replaced by @xmath254 ) to deduce the following statement . [ compsep ] for each @xmath284 , and @xmath65 with @xmath264 , any two connected components of @xmath285 which intersect @xmath254 are @xmath15-separated in @xmath254 . first suppose @xmath286 . let @xmath287 with @xmath288 . choose @xmath230 as in lemma [ dyadicfact ] . we have that @xmath277 , and @xmath289 . but then @xmath278 and @xmath290 must have been joined when the base step rule was applied to the square @xmath230 , and so they lie in the same component of @xmath285 . now suppose that @xmath291 , and @xmath292 with @xmath275 . again , let @xmath230 be the square of lemma [ dyadicfact ] . the induction step applied at level @xmath293 to the square @xmath230 ensures that @xmath278 and @xmath290 are joined by edges in @xmath263 that are contained in @xmath276 . therefore , @xmath278 and @xmath290 lie in the same component of @xmath285 . [ countcla ] there exists a constant @xmath294 , such that for each @xmath284 , and for every @xmath65 with @xmath264 , the iterative procedure applied to @xmath9 increases the length of @xmath260 by at most @xmath295 . from claim [ compsep ] , we see that the graph @xmath285 can have at most @xmath30 components which have non - empty intersection with @xmath254 . consequently , the application of the inductive procedure can generate at most @xmath30 new edges , each of which having length no greater than @xmath296 . the claim follows . * _ adapting the graph to @xmath247 . _ * we begin with another observation about the induction algorithm . note that any two vertices in @xmath297 can be joined by edges in @xmath298 that are contained in @xmath299 . thus , the graph @xmath300 has only one connected component which intersects @xmath301 , and we denote this component by @xmath266 . let us denote by @xmath302 the total length of @xmath266 . by euler s theorem , there is a walk through @xmath266 which visits each vertex of @xmath266 at least once , and travels along each edge at most twice . by a suitable parametrization of this walk , we arrive at the following lemma : [ walklem ] there exists @xmath66\rightarrow \mathbb{c}$ ] , with @xmath303 , and such that @xmath304)\supset \mathcal{n}\cap 3p$ ] . if we have a suitable control over @xmath305 independently of @xmath245 , then @xmath306 is contained in the image of a lipschitz graph : [ lipfun ] suppose that there exists @xmath64 such that @xmath307 for every @xmath308 . then there exists @xmath66\rightarrow \mathbb{c}$ ] , such that @xmath309 , and @xmath310)$ ] . let @xmath311 . let @xmath312 denote the function of lemma [ walklem ] . then @xmath313)$ ] is a @xmath314-net of @xmath306 . by appealing to the arzela - ascoli theorem , we see that there is a subsequence of the @xmath312 ( which we again denote by @xmath312 ) , converging uniformly to some limit function @xmath315 . the function @xmath315 is lipschitz continuous with lipschitz norm no greater than @xmath316 . now , for any @xmath317 , there exists a sequence @xmath318 where @xmath319)$ ] , and @xmath320 . take @xmath321 $ ] with @xmath322 . there is a convergent subsequence of @xmath323 which converges to some @xmath324 $ ] . but then @xmath325 , and the proof is complete . we shall now estimate @xmath305 in terms of the total side length of squares where the induction step has been carried out . note that only the base and inductive steps applied to the dyadic squares @xmath9 contained in @xmath299 can contribute to the length . we shall first estimate the contribution to the length by the base step . [ basesteplength ] the contribution to the length of @xmath266 from the base step is no greater than @xmath326 . let @xmath327 denote the number of dyadic sub - squares of @xmath299 with side length @xmath245 where the base step has been carried out . for any such square @xmath9 , we must have @xmath328 . from the ad - regularity of @xmath0 , it follows that @xmath329 . hence @xmath330 and therefore @xmath331 . consequently , the contribution to the length of @xmath266 from the base step is no greater than @xmath332 , as required . we now denote by @xmath333 the collection of dyadic squares @xmath65 such that @xmath334 $ ] , @xmath335 , and the inductive step has been carried out non - vacuously in @xmath9 at scale @xmath336 . claim [ countcla ] guarantees that for each @xmath337 , an application of the inductive procedure increases the length @xmath305 by no more than @xmath338 . combining this observation with claim [ basesteplength ] , we infer the following bound : @xmath339 given the construction above , we would like to find a convenient way of identifying whether a square has been used in the inductive procedure at some scale . since we do nt want these squares to occur very often , we call them _ bad squares_. we say that @xmath65 is a @xmath340-bad square if there exist @xmath341 , such that @xmath342 , and there exists @xmath343 $ ] such that @xmath344 . we now justify the use of this definition : suppose that @xmath345 . suppose that the inductive algorithm has been applied to @xmath65 . then @xmath9 is a bad square . if the inductive algorithm has been applied , then there is a graph @xmath346 , for some @xmath347 . ] , with the following properties : 1 . the set @xmath236 forms a @xmath348 net of @xmath23 . 2 . for every dyadic square @xmath230 with @xmath349 and @xmath276 , we have that if @xmath350 , then @xmath278 and @xmath290 lie in the same component of @xmath240 . 3 . the connected components of @xmath240 that intersect @xmath254 are at least @xmath336 separated in @xmath254 . ( in fact , property ( 2 ) implies property ( 3 ) , as was seen in claim [ compsep ] ) . by assumption , there exist two points @xmath351 and @xmath352 in @xmath268 that lie in different components of @xmath240 . then @xmath353 . consider the line segment @xmath354 $ ] . cover this segment with overlapping discs of radius @xmath355 , such that the centre of each disc lies in the line segment @xmath356 $ ] ( see figure 4 ) . [ badsquarepic ] , along with their concentric doubles . the cloud of points represents those points of @xmath236.,title="fig:",width=453 ] suppose that every disc has positive @xmath0 measure . if @xmath345 , the concentric double of each disc is contained in @xmath281 . furthermore , in the concentric double of each disc , there must be a point from @xmath236 . we therefore form a chain of points in @xmath357 , with every consecutive pair of points in the chain are separated by a distance of at most @xmath358 . furthermore , the first point in the chain is within a distance of @xmath359 of @xmath351 , and the last point in the chain is no further than @xmath359 from @xmath352 . therefore , lemma [ dyadicfact ] ensures that each consecutive pair of points in the chain are contained in the concentric triple of some dyadic square @xmath230 with @xmath349 and @xmath276 . but then property ( 2 ) yields that each such pair lies in the same component of @xmath240 . as a result , @xmath351 and @xmath352 also lie in the same component of @xmath240 . from this contradiction , we see that one of the discs of radius @xmath355 has zero measure , which implies that @xmath9 is a bad square . now let @xmath360 denote the set of those squares @xmath65 that are bad . to prove theorem [ thm1 ] , it suffices to prove the following proposition : [ carlreduction ] suppose that @xmath0 is a @xmath42-good measure with ad regularity constant @xmath132 . there is a constant @xmath361 such that for each @xmath244 , @xmath362 let us see how theorem [ thm1 ] follows from this proposition . fix @xmath244 , and construct @xmath273 for @xmath246 . from proposition [ carlreduction ] , the bound ( [ lbound ] ) for the length @xmath305 is no more than @xmath363 , where @xmath364 can be chosen to depend on @xmath365 , and @xmath366 ( in particular , @xmath364 can be chosen independently of @xmath247 ) . but now lemma [ lipfun ] yields the existence of a function @xmath66\rightarrow \mathbb{c}$ ] with lipschitz norm no greater than @xmath363 , such that @xmath367)$ ] . this is the required uniform rectifiability . the condition ( [ carl ] ) is very well known in harmonic analysis , and a family of squares @xmath360 satisfying ( [ carl ] ) is often referred to as a _ carleson family_. the best constant @xmath294 such that ( [ carl ] ) holds for all @xmath244 is called the _ carleson norm _ of @xmath360 . fix a @xmath42-good measure @xmath0 with ad - regularity constant @xmath59 . choose @xmath369 , with @xmath370 . recall the riesz family @xmath202 introduced in section 5 . for each @xmath65 , we define @xmath371 consider a fixed @xmath244 . then for each @xmath372 there exists a function @xmath373 such that @xmath374 ( note here that @xmath375 whenever @xmath372 ) . hence @xmath376 since @xmath192 ( @xmath65 ) forms a @xmath377-riesz system , the right hand side of this inequality is bounded by @xmath378 . as @xmath0 is @xmath42-good , this quantity is in turn bounded by @xmath379 , which is at most @xmath380 . therefore @xmath381 as an immediate corollary of this discussion , we arrive at the following result : let @xmath382 . consider the set @xmath383 of dyadic squares @xmath9 satisfying @xmath384 . then @xmath383 is a carleson family , with carleson norm bounded by @xmath385 . in order to prove proposition [ carlreduction ] ( from which theorem [ thm1 ] follows ) , it therefore suffices to prove the following proposition : [ badlower ] suppose @xmath0 is a @xmath42-good measure with ad regularity constant @xmath59 . there exist constants @xmath386 , and @xmath382 , such that for any square @xmath387 , @xmath388 we end this section with a simple remark about scaling . [ scalerem ] fix a square @xmath9 , a function @xmath389 , and a compact set @xmath390 . for @xmath391 , set @xmath392 , @xmath393 and @xmath394 . then @xmath395 , @xmath396 , and @xmath397 in this section , we explore what happens if proposition [ badlower ] fails . to do this , we shall need a workable definition of the cauchy transform operator of a good measure acting on the constant function @xmath32 . suppose that @xmath35 is a @xmath42-good measure with @xmath398 . let us begin with an elementary lemma . [ l1awayfromsupp ] suppose that @xmath400 is a @xmath42-nice measure with @xmath401 . let @xmath402 with @xmath403 . set @xmath404 . then @xmath405 note the estimate @xmath406 the first term on the right hand side has size no greater than @xmath407 . since the domain of integration in the second term can be replaced by @xmath408 , lemma [ tailest ] guarantees that the second integral is bounded by @xmath409 . for @xmath410 , define @xmath411d\mu(\xi ) = \int_{\mathbb{c}}[k(z-\xi)-k(-\xi)]d\nu(\xi).\ ] ] lemma [ l1awayfromsupp ] guarantees that this integral converges absolutely . to extend the definition to the support of @xmath35 , we shall follow a rather standard path . we shall initially define @xmath399 as a distribution , before showing it is a well defined function @xmath35-almost everywhere . recall from section [ weaklims ] how we interpret @xmath412 as a bounded operator in @xmath413 . fix @xmath414 . choose @xmath415 satisfying @xmath416 on a neighbourhood of the support of @xmath417 . then define @xmath418 d\nu(\xi)d\nu(z ) . \end{split}\ ] ] note that lemma [ l1awayfromsupp ] , applied with @xmath419 , yields that @xmath420 therefore the inner product in ( [ c1defn ] ) is well defined . we now claim that this inner product is independent of the choice of @xmath421 . to see this , let @xmath422 and @xmath423 be two compactly supported lipschitz continuous functions , that are both identically equal to @xmath32 on some neighbourhood of @xmath424 . if @xmath425 , then @xmath426d\nu(\xi)-\int_{\mathbb{c}}(1-\varphi_2(\xi))\bigl[\frac{1}{z-\xi}+\frac{1}{\xi}\bigl]d\nu(\xi)\\ & = \int_{\mathbb{c}}(\varphi_2(\xi)-\varphi_1(\xi))\bigl[\frac{1}{z-\xi}+\frac{1}{\xi}\bigl]d\nu(\xi)\\ & = \int_{\mathbb{c}}(\varphi_2(\xi)-\varphi_1(\xi))k(z-\xi)d\nu(\xi)+\cau_{\nu}(\varphi_1)(0 ) - \cau_{\nu}(\varphi_2)(0 ) . \end{split}\ ] ] ( all integrals in this chain of equalities converge absolutely . ) now consider @xmath427 as a result of the anti - symmetry of @xmath428 , this equals @xmath429- \psi(\xi)[\varphi_2(z)-\varphi_1(z)]\bigl]d\nu(\xi)d\nu(z).\ ] ] however , as we saw in section [ weaklims ] , @xmath430 for each @xmath431 . hence , by using the linearity of the integral , and applying fubini s theorem , we see that the last line equals @xmath432 . by definition , this is equal to @xmath433 . the claim follows . we have seen that @xmath399 is well - defined as a distribution . for any bounded open set @xmath434 , if we choose @xmath421 to be identically equal to @xmath32 on a neighbourhood of @xmath435 , then @xmath436 . since lipschitz functions with compact support are dense in @xmath437 , we find that @xmath399 is well - defined @xmath35-almost everywhere . finally , we note that the smoothness of the function @xmath421 is not essential . if @xmath414 , let @xmath435 be a bounded open set containing @xmath424 . then it is readily seen that @xmath438 equals @xmath439d\nu(\xi),\psi\bigl\rangle_{\nu}. \end{split}\ ] ] we shall introduce a couple more sets of functions . @xmath440 will denote those functions @xmath417 with @xmath441 , that satisfy @xmath442 and @xmath443 . we define @xmath444 to be the set of compactly supported functions @xmath417 with @xmath441 , satisfying @xmath442 . we start with another standard estimate . [ l1meanzero ] suppose that @xmath35 is @xmath42-nice measure . for @xmath176 , suppose that @xmath445 . then @xmath446 . simply note that @xmath447 this quantity is no greater than @xmath448 , which is less than or equal to @xmath449 . our next lemma concerns a weak continuity property of @xmath399 . [ weakcontcau1 ] let @xmath450 be a sequence of @xmath42-good measures , with @xmath451 . suppose that @xmath450 converge weakly to @xmath35 ( and so @xmath35 is @xmath42-good ) , with @xmath398 . fix non - negative sequences @xmath452 and @xmath453 , satisfying @xmath454 , and @xmath455 $ ] . if @xmath456 for all @xmath457 , then @xmath458 ( here @xmath459 if @xmath460 . ) if @xmath461 , then there is nothing to prove , so assume that @xmath462 . let @xmath48 . pick @xmath463 . then there exists @xmath464 such that @xmath465 for all sufficiently large @xmath139 , and @xmath466 . fix @xmath467 with @xmath468 , such that @xmath469 . define @xmath470 note that @xmath471 is supported in @xmath178 , and has @xmath124-mean zero . since @xmath472 , we have that @xmath473 for all sufficiently large @xmath139 . therefore , for these @xmath139 , we have @xmath474 . now pick @xmath475 with @xmath148 on @xmath476 and @xmath477 on @xmath478 , such that both @xmath479 for all sufficiently large @xmath139 , and @xmath480 to see that such a choice is possible , note that if @xmath148 on @xmath481 for @xmath482 , then @xmath483 is bounded by @xmath484 ( recall here that @xmath471 has @xmath450 mean zero ) . for any @xmath485 , note that @xmath486 , and so by applying lemma [ l1awayfromsupp ] , we see that the above quantity is no greater than @xmath487 . applying lemma [ l1meanzero ] , we see that @xmath488 , which can be made smaller than @xmath116 with a reasonable choice of @xmath489 . the same reasoning shows that @xmath490 provided @xmath489 is chosen suitably . on the other hand , as @xmath450 is @xmath42-good , we have @xmath491 . since @xmath421 and @xmath492 are compactly supported lipschitz functions , the right hand side of this inequality converges to @xmath493 , and so it is bounded independently of @xmath139 . bringing together these observations , we see that @xmath494 converges to @xmath495 as @xmath170 . but since @xmath474 for @xmath139 large enough , we deduce from the triangle inequality that @xmath496 . let us now suppose that proposition [ badlower ] is false . fix @xmath497 . for each @xmath498 , @xmath499 , there is a @xmath42-good measure @xmath124 with ad - regularity constant @xmath59 , a square @xmath500 , and a set @xmath501 such that @xmath502 in addition , by the scale invariance of the condition ( [ gammaksmall ] ) ( see remark [ scalerem ] ) , we may dilate and translate the square @xmath503 so that it has side length @xmath32 , and so that there are @xmath504 with @xmath505 , such that @xmath506 $ ] and @xmath507 . note that the translated and dilated square is not necessarily dyadic . by passing to a subsequence if necessary , we may assume that @xmath124 converge weakly to a measure @xmath508 ( using the uniform niceness of the @xmath124 ) . this limit measure is @xmath42-good , with ad - regularity constant @xmath132 , and @xmath509 . furthermore , it is routine to check that @xmath508 satisfies the following property ( recall lemma [ adsupp ] ) : @xmath510\text { and } b(0,\tau)\cap \supp(\mu^{(a)})=\varnothing.\end{split}\ ] ] now , for each @xmath139 we have that @xmath511 and @xmath512 . we claim that @xmath513 to see this , note that for any @xmath514 , @xmath515 is equal to @xmath516 the first term is smaller than @xmath517 in absolute value . to bound the second term , note that for any @xmath518 , @xmath519 so lemma [ l1awayfromsupp ] yields that this second term is no larger than @xmath520 , and applying lemma [ l1meanzero ] yields the required estimate . we now apply lemma [ weakcontcau1 ] with @xmath521 , @xmath522 , and @xmath523 . our conclusion is that @xmath524 for all @xmath525 . we now set @xmath526 , for @xmath527 . the above argument yields a measure @xmath528 satisfying @xmath529 for all @xmath530 . we now pass to a subsequence of @xmath531 so that @xmath532 weakly as @xmath170 . the measure @xmath0 is @xmath42-good with ad - regularity constant @xmath132 , and satisfies the property ( [ badmua ] ) with @xmath0 replacing @xmath508 . by applying lemma [ weakcontcau1 ] with @xmath533 , @xmath534 , @xmath535 , and @xmath536 , we arrive at the following result : [ absurd ] suppose that proposition [ badlower ] fails . then there exists a @xmath42-good measure @xmath0 with ad - regularity constant @xmath132 , such that @xmath537 and there exist @xmath538 , with @xmath539 such that @xmath540 $ ] and @xmath541 we call any measure @xmath0 that satisfies ( [ reflectionless ] ) a _ reflectionless measure_. it turns out that there are nt too many good ad - regular reflectionless measures . [ reflectioncharac ] suppose that @xmath0 is a non - trivial reflectionless good ad - regular measure . then @xmath542 for a line @xmath26 , and a positive constant @xmath543 . note that proposition [ reflectioncharac ] contradicts the existence of the measure @xmath0 in lemma [ absurd ] . therefore , once proposition [ reflectioncharac ] is proved , we will have asserted proposition [ badlower ] , and theorem [ thm1 ] will follow . hence it remains to prove the proposition . it is at this stage where the precise structure of the cauchy transform is used . our goal is now to prove proposition [ reflectioncharac ] . suppose that @xmath0 is a reflectionless @xmath42-good measure . we may assume that @xmath545 . all constants in this section may depend on @xmath42 without explicit mention . since @xmath546 is a well defined @xmath0-almost everywhere function and satisfies ( [ reflectionless ] ) , we conclude that it is a constant @xmath0-almost everywhere in @xmath478 , say with value @xmath547 . [ refconstsupp ] suppose that @xmath0 is a @xmath42-good reflectionless measure , and @xmath545 . then there exists @xmath547 such that @xmath548 @xmath0-almost everywhere . our considerations up to now have been quite general , but now our hand is forced to use the magic of the complex plane . the main difficulty is to obtain some information about the values of @xmath546 away from the support of @xmath0 in terms of the constant value @xmath549 . [ resolvlem ] for every @xmath550 , @xmath551 ^ 2 = 2\kap\cdot \widetilde{\cau}_{\mu}(1)(z).\ ] ] an immediate consequence of lemma [ resolvlem ] is that either @xmath552 or @xmath553 for any @xmath554 . since @xmath546 is a continuous function away from @xmath25 , it follows that @xmath546 is constant in each connected component of @xmath544 . a variant of lemma [ resolvlem ] , where the cauchy transform is considered in the sense of principal value , has previously appeared in work of melnikov , poltoratski , and volberg , see theorem 2.2 of @xcite . we shall modify the proof from @xcite in order to prove lemma [ resolvlem ] . we shall first provide an incorrect proof of this lemma . indeed , note the following regularized version of the resolvent identity : for any three distinct points @xmath555 , @xmath556\cdot\bigl[\frac{1}{\xi-\omega}+\frac{1}{\omega}\bigl ] + \bigl [ \frac{1}{z-\omega}+\frac{1}{\omega}\bigl]\cdot\bigl[\frac{1}{\omega-\xi}+\frac{1}{\xi}\bigl ] \\ & = \bigl [ \frac{1}{z-\xi}+\frac{1}{\xi}\bigl]\cdot\bigl [ \frac{1}{z-\omega}+\frac{1}{\omega}\bigl ] . \end{split}\ ] ] integrating both sides of this equality with respect to @xmath557 , we ( only formally ! ) arrive at @xmath558 ^ 2 $ ] . once this is established , lemma [ refconstsupp ] completes the proof . the proof that follows is a careful justification of this integration . we shall define @xmath559 \varphi(\xi)d\mu(\xi)$ ] for @xmath415 . in particular , as @xmath40 tends to @xmath188 , @xmath560 converges to @xmath561 weakly in @xmath1 . set @xmath562 . snce @xmath563 , lemma [ l1awayfromsupp ] tells us that @xmath564\in l^1(\mu)$ ] . for @xmath565 , define a bump function @xmath566 , satisfying @xmath567 on @xmath568 , and @xmath569 . consider the identity ( [ resolve ] ) , and multiply both sides by @xmath570 . after integration against @xmath557 , the right hand side of this equality becomes @xmath571\varphi_n(\xi)d\mu(\xi)\int_{\mathbb{c}}\bigl[\frac{1}{z-\omega } + \frac{1}{\omega}\bigl]\varphi_n(\omega)d\mu(\omega).\ ] ] but since @xmath572\in l^1(\mu)$ ] , the dominated convergence theorem ensures that as @xmath573 , this expression converges to @xmath574 ^ 2 $ ] . now , let @xmath38 , and note that @xmath575.\ ] ] consider the integral @xmath576\cdot\bigl[\frac{1}{z-\xi}+\frac{1}{\xi}-\frac{1}{z-\omega } - \frac{1}{\omega}\bigl]\\ & \varphi_n(\xi)\varphi_n(\omega)d\mu(\xi)d\mu(\omega ) . \end{split}\ ] ] note that @xmath577 for @xmath578 , and so this integral is bounded in absolute value by a constant multiple of @xmath579 , which is bounded by @xmath580 this converges to zero as @xmath74 . making reference to ( [ resolve ] ) , we have thus far shown that @xmath581\bigl[k_{\delta}(\xi-\omega)+\frac{1}{\omega}\bigl ] \\ & + \bigl [ \frac{1}{z-\omega}+\frac{1}{\omega}\bigl]\bigl[k_{\delta}(\omega-\xi)+\frac{1}{\xi}\bigl]\bigl\}d\mu_n(\xi)d\mu_n(\omega)= [ \widetilde{\cau}_{\mu}(1)(z)]^2 . \end{split}\ ] ] by fubini s theorem , and the weak convergence of @xmath560 to @xmath582 , the left hand side of this equality is equal to twice the following limit @xmath583\widetilde{\cau}_{\mu}(\varphi_n)(\xi ) d\mu(\xi)\bigl ] . \end{split}\ ] ] therefore , to prove the lemma , it suffices to show that this limit equals @xmath584 . to do this , let @xmath585 . first consider @xmath586 note that , for @xmath587 , we have @xmath588 applying lemma [ l1awayfromsupp ] yields an upper bound for the right hand side of @xmath589 but as @xmath564\in l^1(\mu)$ ] , we conclude that @xmath590 as @xmath573 . next , note that @xmath591\widetilde{\cau}_{\mu}(1)(\xi ) d\mu(\xi)\ ! = \!\kap\ ! \int_{b(0 , n^{\alpha } ) } \!\!\varphi_n(\xi)\bigl[\frac{1}{z-\xi } + \frac{1}{\xi}\bigl ] d\mu(\xi),\ ] ] which converges to @xmath592 as @xmath573 . to complete the proof the lemma , it now remains to show that @xmath593 to do this , first note that @xmath594 ( this merely uses the @xmath42-niceness of @xmath0 ) . on the other hand , for sufficiently large @xmath327 , @xmath595 for @xmath596 . therefore , there is a constant @xmath597 such that @xmath598 finally , since @xmath599 , and @xmath600 , we estimate the right hand side here by a constant multiple of @xmath601 , which tends to zero as @xmath573 . in this section we conclude our analysis by proving proposition [ reflectioncharac ] . to do this , we shall use the notion of a tangent measure , which was developed by preiss @xcite . suppose that @xmath35 is a borel measure on @xmath602 . the measure @xmath603 is called a @xmath604-blowup of @xmath35 at @xmath7 . a tangent measure of @xmath35 at @xmath7 is any measure that can be obtained as a weak limit of a sequence of @xmath604-blowups of @xmath35 at @xmath7 with @xmath605 . now suppose that @xmath0 is a nontrivial @xmath42-good with ad regularity constant @xmath132 . then any @xmath604-blowup measure of @xmath0 at @xmath62 will again have these properties ( @xmath42-goodness , and @xmath132-ad regularity ) . therefore , both properties are inherited by any tangent measure of @xmath0 . in particular , every tangent measure of @xmath0 at @xmath62 is non - trivial , provided that @xmath0 is non - trivial . lastly , we remark that if @xmath0 is reflectionless , then any tangent measure of @xmath0 is also reflectionless . this follows from a simple application of lemma [ weakcontcau1 ] . in what follows , it will often be notationally convenient to translate a point on @xmath25 to the origin . whenever this is the case , the definition of @xmath546 in ( [ c1offsupp ] ) is translated with the support of the measure , and becomes @xmath606d\mu(\xi),\ ] ] for some @xmath607 . if @xmath0 is reflectionless , then @xmath546 is constant in each component of @xmath544 , and takes one of two values in @xmath544 . suppose that @xmath608 , for some line @xmath26 . then by translation and rotation we may as well assume that @xmath28 . if the support is not the whole line , then there exists an interval @xmath609 disjoint from the support of @xmath0 , with either @xmath610 or @xmath611 in the support of @xmath0 . by rotating the support if necessary , we may assume that @xmath612 . denote by @xmath613 a non - zero tangent measure of @xmath0 at @xmath611 . then @xmath613 has support contained in the segment @xmath614 , and @xmath615 . since @xmath613 is reflectionless , we may apply lemma [ resolvlem ] to deduce that @xmath616 is constant for all @xmath617 . differentiating this function with respect to @xmath618 , we arrive at @xmath619 this integral is strictly positive as @xmath613 is not identically zero . from this contradiction we see that @xmath620 . consequently , we have that @xmath621 , where @xmath622 . now let @xmath623 and consider , for @xmath624 , @xmath625 the expression on the left hand side is constant in @xmath624 and @xmath623 . on the other hand , the integral on the right hand side is a constant multiple of the harmonic extension of @xmath626 to @xmath627 . the poisson kernel is an approximate identity , and so by letting @xmath628 we conclude that @xmath626 is a constant . therefore @xmath629 with @xmath543 . we now turn to the general case . we first introduce some notation . for @xmath402 and a unit vector @xmath630 , @xmath631 denotes the ( closed ) half space containing @xmath7 on the boundary , with inner unit normal @xmath630 . with @xmath632 , we denote @xmath633 , where @xmath634 is the standard inner product in @xmath635 . [ nothingotherside ] suppose that @xmath554 . let @xmath636 be a closest point in @xmath25 to @xmath7 , and set @xmath637 . for each @xmath632 , there is a radius @xmath638 such that @xmath639 is disjoint from @xmath25 . we may suppose that @xmath640 for some @xmath8 and @xmath641 , ( and so @xmath642 ) . we shall examine the imaginary part of the cauchy transform evaluated at @xmath643 for @xmath644 : @xmath645 = \int_{\mathbb{c}}\bigl[\frac{\im(\xi)+t}{|\xi+it|^2}-\frac{\im(\xi - z_0)}{|\xi - z_0|^2}\bigl]d\mu(\xi).\ ] ] next , note that if @xmath656 , then @xmath657 , provided that @xmath658 and @xmath659 . to see this , note that @xmath660 , and so by elementary geometry , @xmath661 . this is at least @xmath662 under our assumptions on @xmath352 and @xmath618 . therefore , if @xmath659 , then @xmath663 both terms on the right hand side are bounded in absolute value by @xmath664 ( recall that @xmath665 ) . note that the integral on the left hand side is at most zero . our conclusion thus far is that there is a constant @xmath666 , depending on @xmath667 and @xmath668 , such that for any @xmath659 , @xmath669 note that the integrand in this integral is positive for any @xmath670 . suppose now that the statement of the lemma is false . then there exists @xmath671 , along with a sequence @xmath672 with @xmath673 as @xmath674 . by passing to a subsequence , we may assume that @xmath675 for each @xmath676 , and also that the balls @xmath677 are pairwise disjoint . each ball @xmath678 , and provided that @xmath679 , we have @xmath680 as a result , we see that @xmath681 but @xmath682 , and so the previous integral over @xmath683 has size at least @xmath684 . however , if @xmath618 is sufficiently small , then this quantity may be made larger than the constant @xmath666 appearing in ( [ iiibound ] ) . this is absurd . [ outsideconv]suppose that @xmath450 is a sequence of @xmath42-nice measures with ad - regularity constant @xmath132 that converges to @xmath35 weakly as @xmath170 ( and so @xmath35 is @xmath42-nice with ad - regularity constant @xmath132 ) . if @xmath685 , then for any @xmath410 , @xmath686 is well defined ( as in ( [ z0defn ] ) ) for sufficiently large @xmath139 , and @xmath687 as @xmath170 . first note that there exists @xmath8 such that @xmath688 . but then , by the ad regularity of each @xmath450 , we must have that @xmath689 for sufficiently large @xmath139 , and hence @xmath686 is well defined . let @xmath565 , and choose @xmath690 satisfying @xmath567 on @xmath691 and @xmath692 in @xmath478 . for large enough @xmath139 , @xmath693 is no greater than the sum of @xmath694\varphi_n(\xi ) d(\nu-\nu_k)(\xi)|$ ] and @xmath694[1-\varphi_n(\xi ) ] d(\nu-\nu_k)(\xi)|$ ] . the first of these two terms tends to zero as @xmath695 , while the second has size at most @xmath696 ( for sufficiently large @xmath327 ) due to lemma [ l1awayfromsupp ] . this establishes the required convergence . fix @xmath703 small enough to ensure that @xmath704 for any @xmath705 . the existence of @xmath706 is guaranteed by lemma [ nothingotherside ] . now set @xmath707 . notice that the value of @xmath400 is independent of the choice of @xmath708 in ( [ z0defn ] ) , so we shall fix @xmath709 . now , let @xmath710 denote a tangent measure to @xmath0 at @xmath188 . on account of lemma [ nothingotherside ] , the support of @xmath710 is contained in the line @xmath26 through @xmath188 perpendicular to @xmath630 . by step 1 , @xmath711 with @xmath712 $ ] . as a result , for any @xmath623 , we have that @xmath713 we claim that @xmath714 . to see this , note that for @xmath715 small enough so that @xmath716 , we have @xmath717 . but this equals @xmath400 because @xmath718 and @xmath719 lie in the same connected components of @xmath720 as @xmath7 and @xmath721 respectively . since @xmath710 is a weak limit of measures @xmath722 for some sequence @xmath723 , applying lemma [ outsideconv ] proves the claim . consequently , we have that @xmath400 determines the direction of tangency from @xmath7 to @xmath25 ( the angle @xmath724 ) . the right hand side of ( [ secondjump ] ) is non - zero , and so @xmath725 lies in a different component of @xmath720 to @xmath7 . as there are only two possible values that @xmath546 can take in @xmath720 , @xmath400 is determined by @xmath726 . since @xmath546 is constant in each connected component of @xmath544 , the direction of tangency from any point in the connected component of @xmath544 containing @xmath7 to @xmath25 is the same . finally , set @xmath727\}&\text { lies in the same connected component}\\ & \text { of } \mathbb{c}\backslash\supp(\mu)\text { as } z\bigl\}.\end{split}\ ] ] we claim that if @xmath728 , then @xmath701 . indeed , otherwise there is a point @xmath729 which is a closest point in @xmath25 to @xmath730 . but then it follows that @xmath731 . given that @xmath732\}\cap \supp(\mu)=\varnothing$ ] , this is a contradiction . from this claim , we see that if @xmath733 , then @xmath734 . since @xmath735 , it follows that @xmath736 , so @xmath701 for any @xmath702 . now , suppose that there are three points @xmath737 which are not collinear . then they form a triangle . since @xmath0 is ad - regular , there is a point in the interior of this triangle outside of the support of @xmath0 . let s call this point @xmath738 . but then there is a half space , with @xmath738 on its boundary , which is disjoint from @xmath25 . this half space must contain at least one of the points @xmath7 , @xmath352 or @xmath351 . this is absurd .

we use new wfc3 observations of the nearby grand design spiral galaxy m83 to develop two independent methods for estimating the ages of young star clusters . the first method uses the physical extent and morphology of h emission to estimate the ages of clusters younger than myr . it is based on the simple premise that the gas in very young ( few myr ) clusters is largely coincident with the cluster stars , is in a small , ring - like structure surrounding the stars in slightly older clusters since massive star winds and supernovae have had time to push out the natal gas ( e.g. , myr ) , and is in a larger ring - like bubble for still older clusters ( i.e. , myr ) . if no h is associated with a cluster it is older than myr . the second method is based on an observed relation between pixel - to - pixel flux variations within clusters and their ages . this method relies on the fact that the brightest individual stars in a cluster are most prominent at ages around 10 myr , and fall below the detection limit ( i.e. , m ) for ages older than about 100 myr . older clusters therefore have a smoother appearance and smaller pixel - to - pixel variations . the youngest clusters also have lower flux variations , hence the relationship is double valued . this degeneracy in age can be broken using other age indicators such as h morphology . these two methods are the basis for a new morphological classification system which can be used to estimate the ages of star clusters based on their appearance . we compare previous age estimates of clusters in m83 determined from fitting _ _ ubvi__h measurements using predictions from stellar evolutionary models with our new morphological categories and find good agreement at the% level . the scatter within categories is dex in for young clusters ( myr ) and dex for older ( myr ) clusters . a by - product of this study is the identification of 22 `` single - star '' hii regions in m83 , with central stars having ages myr .

1103.4026

age estimates are required for studying the evolutionary history of star clusters . based on such age estimates , a general framework has been developed starting from formation within the dense cores of giant molecular clouds ( gmcs ) ; a stage where the young stars are completely obscured by their dusty gas cocoons ; an emerging stage where the clusters become visible in the infrared ( ir ) and then visible in the optical as stellar winds and supernovae blow away the gas and dust ; a stage where an expanding bubble of ionized gas is blown ; and later stages with no evidence of ionized gas ( e.g. , see the lada & lada 2003 review article ) . in the past , two general methods for estimating the ages of unresolved extragalactic star clusters have been used . the first requires obtaining high quality spectroscopic observations covering wavelength regions of lines which change as a function of time . examples of this approach include bica & alloin ( 1986 ) , schweizer & seitzer ( 1993 ) , whitmore ( 1999 ) , bastian et al . ( 2009 ) and wofford et al . while these spectroscopic observations typically provide high quality age determinations , they are limited to age - dating relatively small numbers of bright clusters due to constraints on the brightness of clusters that can be observed spectroscopically in a reasonable amount of time . the standard method for age - dating large numbers of extra - galactic clusters compares photometry in several broadband filters ( e.g. , _ ubvi _ ) with predictions from population synthesis models ( e.g. , chandar et al . 2010 ) . comparisons of age estimates using spectroscopic and photometric observations of the same clusters show good agreement in most cases . an early example of this is the famous searle et al . ( 1980 ) paper where they compare integrated four - filter _ ugvr _ photometry of 61 star clusters in the magellanic clouds with the strength of different spectral features ( e.g. , balmer lines , @xmath14 band , etc . ) a recent example is wofford et al . ( 2010 ) who compared ages estimated from spectral energy distribution ( sed ) fitting with those derived from uv spectroscopy for 14 young ( @xmath15 myr ) clusters in the nuclear starburst region of m83 and found that the photometric ages are within a factor of 1.4 of the spectroscopic ones . whitmore et al . ( 2010 ) find similar agreement between their photometric age estimates and the spectroscopic estimates from bastian et al . ( 2009 ) for clusters in the antennae galaxies . while generally satisfactory , both the spectroscopic and photometric techniques have limitations . it is therefore important to develop independent methods for estimating the ages of clusters , especially for cases where spectroscopic or multi - band photometric observations are not available and it is necessary to estimate ages based on morphological appearance alone . here , we develop two new methods for estimating the ages of young star clusters in nearby galaxies based on high - resolution images at optical wavelengths . the first method uses the morphology of the ionized gas and its position relative to the cluster stars , as measured from narrowband h@xmath0 emission , to estimate ages ( @xmath16 ) for clusters younger than @xmath17 myr . this method relies on the general premise that the distribution of h@xmath0 will be largely coincident with the distribution of optical light in the youngest clusters ( i.e. , @xmath10 few myr ) , will be in a small ring - like structure around the optical stellar emission in slightly older clusters where massive star winds and supernovae have had time to blow a bubble ( i.e. , @xmath18 myr ) , and will be in a larger ring - like bubble for still older clusters ( i.e. , 510 myr ) . many past observational and theoretical studies of hii regions and `` supershells '' in the milky way and nearby galaxies have laid the groundwork for this method . for example , walborn ( 2002 ; see also walborn & parker 1992 and especially walborn 2010 for related discussions ) outlined an evolutionary cluster sequence based on observed properties of several well known clusters and ob associations in the milky way with ages ranging from @xmath19 myr to @xmath5 myr . this sequence was then used to illustrate observed changes with age in the visually brightest stars , in the ionized gas and dust content , and in the existence of red supergiants . much of this sequence was based on spectroscopy of individual stars . while we are unable to make similarly detailed observations at the distance of m83 , many of the same basic correlations and underlying physical processes are relevant for the age sequence outlined in this paper . theoretically , several works have made predictions for the size evolution of an expanding hii region over time . oey & clark ( 1997 , 1998 ) model the size evolution due to mass loss and supernova - injected energy from cluster stars , and assume that expanding bubbles `` stall '' when their internal pressure equals the ambient pressure in the ism . these simulations predict a strong dependence of bubble size on both the age and mass of the central cluster . ( e.g. , weaver et al . 1977 , oey & massey 1995 , oey & garcia - segura 2004 , dopita et al . 2006a , 2006b ) . our observations of clusters in m83 reveal a strong dependence of bubble size on cluster age , and possibly a weak dependence on cluster mass , as discussed in section 6 . the second method developed in this paper uses the surface brightness fluctuations of cluster stars to estimate their ages . young clusters have strong pixel - to - pixel flux variations , due to the presence of massive , luminous stars . as a cluster ages , the bright , short - lived massive stars disappear , and these fluctuations fade in strength . this technique is especially useful in the range @xmath20100 myr , ages that can be somewhat difficult to deal with using the sed method , because the predicted integrated colors loop back on themselves . our primary target for this study is the spiral galaxy m83 . at a distance of 4.5 mpc ( corresponding to a distance modulus of @xmath21 ; thim et al . 2003 , and a pixel scale of 0.876 pc pixel@xmath22 ) , m83 , nicknamed the `` southern pinwheel , '' is the nearest massive grand - design spiral galaxy . it is a slightly barred galaxy , with a hubble type sab(s)c ( 3rc ) . in this work , we make use of observations taken with the wide - field camera 3 ( wfc3 ) , which were described in detail in dopita et al . ( 2010 ) and chandar et al . briefly , the observations are part of the early release science project 1 ( ers1 ) program 11360 ( pi : oconnell ) and were taken in august , 2009 . observations were taken in several broadband ( `` uv '' - f225w , `` u '' - f336w , `` b '' - f438w , `` v '' - f555w , `` i '' - f814w , `` j '' - f110w , and `` h '' - f160w ) , and narrowband filters ( [ oiii ] - f373n , h@xmath23 - f487n , [ oii ] f502n , h@xmath0 - f657n , [ sii ] - f673n , paschen@xmath23 - f128n , and [ feii ] - f164n ) . in the appendix we briefly investigate clusters in m51 and find that the age versus morphological category relationship derived for m83 is appropriate for this galaxy as well . this paper is organized as follows : in section 2 we outline our working scheme for an evolutionary cluster classification system based on observables at optical wavelengths . in section 3 we investigate the correlation between h@xmath0 morphology and sed age estimates while section 4 examines the correlation between the strength of surface brightness fluctuations within clusters and their ages . a catalog of smaller hii regions , apparently ionized by `` single stars , '' is presented in section 5 . we examine the effect of cluster mass on h@xmath0 bubble size in section 6 , and summarize our primary results in section 7 . the early formative stages of star cluster evolution are best studied in the ir or microwave part of the spectrum , since the stars remain embedded in their placental dust cocoons for the first million years or so ( lada & lada 2003 ) . for example , in nearby groups and clusters in the milky way , individual young stellar objects ( ysos ) can be age - dated using a classification scheme ranging from category 0 to iii , as developed by wilking et al . ( 1989 ) and andre et al . ( 1993 ) , based on near - ir observations . in this paper we outline a working classification scheme for cluster evolution in external galaxies . we focus on the later stages based on the optical portion of the spectrum . our primary goals are to test how well we can use h@xmath0 morphology and surface brightness fluctuations to age date star clusters . the basic categories are defined below . in all cases we assume that the clusters have radial profiles that are broader than the point - spread function ( psf ) , hence confusion with individual stars is not a major issue ( see chandar et al . 2010 ) . examples of clusters in categories 3 through 6 are shown in figure [ fig:18plot ] . m83 clusters in categories 1 and 2 will be discussed in a later paper which presents our wfc3 observations in the @xmath24 and @xmath25 bands . _ category 1 : _ * opaque dust cloud * - core of a gmc - dark region on optical image with no associated ir / uvis source - often studied using millimeter observations of the hcn line ( e.g. , gao & solomon 2004 ) 0.1 in _ category 2a : _ - * embedded cluster * - weak ir source with no optical ( i band ) counterpart ( i.e. , av @xmath12 10 ) - no ionized h@xmath0 gas visible 0.1 in _ category 2b : _ - * obscured cluster * - strong ir and weak optical source ( 3 @xmath10 av @xmath10 10 ) 0.1 in _ category 3 : _ - * emerging cluster * - ionized gas spatially coincident with cluster stars - cluster has a reddish color due to dust - surface brightness fluctuations from individual stars are relatively small since the brightest , evolved stars have not yet appeared 0.1 in _ category 4a : _ - * very young cluster * - ionized gas in a small bubble surrounding the cluster - cluster has a bluish color since most of the dust has been expelled - surface brightness fluctuations from individual stars are strong 0.1 in _ category 4b : _ - * young cluster * - similar to category 4a but the ionized gas is now in a large bubble surrounding the cluster ( i.e. , radii larger than approximately 20 pc ) - surface brightness fluctuations from individual stars reach maximum 0.1 in _ category 5a : _ - * young / intermediate - age cluster * - no ionized gas is observed - surface brightness fluctuations are still present , but weaker 0.1 in _ category 5b : _ - * intermediate - age cluster * - ionized gas and surface brightness fluctuations among cluster stars are not observed - cluster has a slightly redder color due to aging of stars 0.1 in _ category 6 : _ - * old cluster * - no ionized gas or surface brightness fluctuations among cluster stars are observed , - cluster appears yellow / red with no evidence of dust in vicinity in principle , it should be possible to approximately estimate the ages of young ( @xmath17 myr ) star clusters from the size of the ionized gas ( h@xmath0 ) bubble that surrounds them , as outlined in sections 1 and 2 . in this section we test this idea by selecting a representative catalog of young clusters primarily based on their apprearance in the narrowband h@xmath0 image ( i.e. , the red color in figure [ fig : large_image ] ) , and then visually classify each source based on the scheme outlined in section 2 . each region is then matched with the apparent central source of ionization from the chandar et al . ( 2010 ) catalog , and compared with the ages from that paper based on sed fitting of the _ _ ubvi__h@xmath0 bands . we do not , at this point , separate the clusters by mass ( the effect of cluster mass on bubble size is discussed in section 6 ) . basic properties of the selected clusters are presented in table 1 , including position ( ra and decl . ) and photometric measurements ( @xmath26 , @xmath27 , and @xmath28 ) for the central cluster . the top row of figure [ fig:18plot ] shows examples of clusters in categories 3 - 6 , with different h@xmath0 morphologies as outlined in section 2 . the bottom set of panels shows the location of each cluster in the category in a @xmath27 versus @xmath28 two - color diagram , where the photometry is drawn from the catalog described in chandar et al . these panels show how well our morphological categories group the clusters in color - color space . predictions from single stellar population g. bruzual & s. charlot ( 2009 ; private communications , hereafter bc09 ) models are superposed for comparison . as expected , clusters in the two youngest categories ( 3 and 4a ) have somewhat higher extinction than those in later categories , which results in their colors spreading out along the reddening vector . the middle set of panels present histograms of ages determined from the photometric dating method used in chandar et al . ( 2010 ) , for all clusters within a given category . these age estimates are included in table 1 . there is a clear trend for clusters in later categories to have older ages . for clusters with associated h@xmath0 emission , i.e. , categories 3 , 4a , 4b , we find an rms scatter of approximately 0.1 in @xmath29 within each category . while our sed dating method includes photometry in the narrowband filter directly in the fit , the photometry is measured in a small ( 3 pixel radius ) aperture that misses most or all of the h@xmath0 emission after category 3 . hence the age estimates from the sed fitting and from the h@xmath0 morphology are largely independent of one another . the rms scatter in photometric ages for clusters identified within morphological categories 5 and 6 increases to @xmath11 in @xmath29 , primarily because morphology becomes a cruder estimate in this age regime . the good overall correlation between morphological classification and photometric age estimates from chandar et al . ( 2010 ) is also apparent in figure [ fig : large_image ] . this figure includes the sed ages for the clusters where morphological classifications have been determined using the prescripts defined in section 2 . the region covers a @xmath30 kpc@xmath31 area of m83 . we note in particular the intermediate - age and old clusters in the upper central part of the image which have a fuzzy appearance due to the absence of bright young stars . the appendix shows similar figures for selected regions in m51 . figure [ fig : ha_vs_sed ] plots the morphological category assigned here versus @xmath32yr ) based on the sed age estimates from chandar et al . the top panel shows the full range of cluster ages and categories . there is a clear , albeit nonlinear , correlation , with a steeper correlation between morphological category and age for clusters younger than @xmath33 yr . the nonlinearity is to be expected since h@xmath0 morphology evolves rapidly between the ages of 110 myr . note that in figure [ fig : ha_vs_sed ] we have subdivided the morphological types into a finer grid of categories ( based on subjective estimates of the h@xmath0 bubble size ) than outlined in section 2 . the bottom panel of figure [ fig : ha_vs_sed ] shows one of the main results of this work , _ that between 2 and 10 myr , the clusters in m83 show a strong correlation between their h@xmath0 morphology and age_. this suggests that the morphology of ionized gas alone can be used as an age indicator . the best linear fit , shown in the figure , is given by : @xmath34 ( i.e. , a 9 @xmath35 correlation ) for @xmath36 , where @xmath37 is the morphological category defined in section 2 . this relationship is only relevant when h@xmath0 is present ( i.e. , categories 3 and 4 ) . a similar , flatter , and less well - defined relationship exists for category 5 and higher , as shown in the top figure : @xmath38 ( i.e. , a 5 @xmath35 correlation ) for @xmath39 . we note that a similar correlation exists between our morphological categories and ages estimated by comparing color magnitude diagrams of individual stars with stellar tracks for stars around some of our objects , as will be discussed in a study of 50 regions in m83 by h. kim et al . ( 2011 , in preparation ) . while the dispersion in the estimated sed ages within each morphological category is remarkably small ( see figure [ fig:18plot ] and table 2 ) , a close inspection reveals four outliers in category 5b . three of these clusters ( 44034 , 18032 and 74692 ) have no associated h@xmath0 emission , have relatively small pixel - to - pixel flux variations ( discussed in section 4 ) , and have optical colors consistent with those predicted for extinction - free intermediate age clusters , and therefore almost certainly have ages @xmath40 . the sed fitting , however , erroneously assigned them lower ages ( @xmath36 ) , and higher extinctions ( i.e. , @xmath41 , 0.36 , 0.50 mag ) . the fourth cluster ( 10114 ) has significantly bluer colors than the other clusters in category 5b . in retrospect , this cluster probably belongs in morphological category 4a , since there does appear to be a small amount of h@xmath0 emission associated with it . in this case it is the morphological classification which appears to be in error rather than the sed age estimate ( the latter is @xmath42 ) . overall , we find good agreement between the previously determined sed age and currently determined morphological category for 95% of the clusters ( i.e. , 84 out of 88 cases ) , demonstrating that _ both _ techniques are quite reliable . while the good correlation between the morphological category and sed age found in figure 3 is encouraging , this approach has several limitations . in particular , placing the clusters into the different categories is subjective and hence not easily automated . an additional complication is that an isolated , full 360 degree ring with a single dominant central cluster is very rare . partial rings , multiple loops , and multiple clusters in the region are more typical situations . here , we attempt to better quantify the relationship between h@xmath0 morphology and cluster age by measuring the radii of the most prominent `` coherent '' ring or partial shell of ionized gas associated with the cluster stars . while the measurement of h@xmath0 bubble size is also difficult to automate , this more quantitative approach provides the opportunity for improving the correlation still further , and for exploring the impact that other physical parameters ( e.g. , cluster mass ) have on this relationship . figure [ fig : ha_vs_r ] plots the measured h@xmath0 bubble size versus photometric age for clusters younger than @xmath5 myr ( top panel ) , and bubble size versus morphological category ( bottom panel ) . there is clearly a strong correlation which suggests that h@xmath0 morphology alone can be used as an age indicator for clusters younger than @xmath5 myr . this figure hints at other intriguing results as well . the correlation between age and bubble size is strongest for clusters younger than @xmath43 , with typical bubbles growing from a few pc to @xmath44 pc in size . after this time , between ages @xmath436.9 , there is a large range observed in bubble size , from @xmath44 pc to nearly 80 pc , and no strong dependence on age . these trends are illustrated by the solid lines , which represent simple linear fits to observations in the two different regimes . the best linear fits for the top panel are given by : @xmath45 for r @xmath10 20 pc , and @xmath46 for r @xmath12 20 pc . similar fits are shown in the bottom panel , but numerical values are not included here since the use of morphological category makes them more qualitative . the large , open circles in the top panel of figure [ fig : ha_vs_r ] show the clusters with @xmath47 and @xmath48 . these will be discussed further in section 6 . as discussed in section 2 , bright individual stars are expected to be visible within m83 clusters at ages younger than @xmath49 myr , giving the clusters a mottled appearance with relatively large pixel - to - pixel flux variations . after this , the flux variations have mostly disappeared , leading to a more uniform appearance among cluster stars . compare , for example , the images in categories 4b and 5a in the top panel of figure [ fig:18plot ] , where a number of bright individual cluster stars are clearly visible , with the images for categories 5b and 6 , where individual cluster stars are no longer observed . figure 2 also shows a variety of cases where older clusters appear as fuzzy objects while single individual stars are clearly observed within the young clusters . here , we use the strength of these surface brightness fluctuations to develop a new method for estimating the ages of clusters in m83 . this is reminiscent of the surface brightness fluctuation method used to estimate distances to early - type galaxies ( e.g. , tonry & schneider 1988 ) , but in our case the clusters are all at the same distance and the surface brightness fluctuations are caused by differences in age . one must exercise caution when using this technique for other galaxies , since the relationships derived below for m83 would need to be renormalized for galaxies at different distances . we first need to remove the overall radial gradient in the cluster luminosity profile , since this gradient typically dominates over the peak - to - peak variations in pixel flux within the cluster . we accomplish this by creating a median divided image ( with a @xmath50 smoothing box size ) , which effectively flattens the radial profile of the cluster , leaving behind primarily the smaller scale pixel - to - pixel variations that we are interested in here . we then measure the rms scatter in the fluxes of pixels located within a 10 pixel radius . a more sophisticated method is being developed by c. kaleida et al . ( 2011 , in preparation ) where the rms variations are measured within the half - light or effective radius r@xmath51 of the cluster , hence accounting for clusters of different sizes in a more systematic manner . figure [ fig : sbf ] shows a comparison of the f555w image and the median divided f555w image in a region including clusters younger than @xmath52 ( blue circles and labels ) and older than this ( red circles and labels ) . the rms measurements are indicated in the bottom panel . one can easily see that old clusters ( e.g. , nos . 65 and 70 ) nearly disappear from the median divided image , resulting in low values of the rms , while bright , individual stars are observed in the younger clusters ( e.g. , nos . 32 and 49 ) , resulting in a higher rms . small , barely visible dust lanes also contribute to larger values of rms for the young clusters . in a few very concentrated clusters there are small artifacts at their centers , for example in cluster no . 64 . however , the rms is dominated by pixels outside of this central region , hence this is a relatively small effect . we note that for four clusters , identified in table 1 , we have used a slightly smaller radius to avoid single bright stars that are almost certainly unrelated to the cluster ( e.g. , nos . 66 , 69 , 70 and 79 ) . in the upper panel of figure [ fig : m83_rms_2plot ] we plot the rms in the pixel - to - pixel flux measured in the @xmath53 band versus the morphological category . the resolution is better here for clusters younger than @xmath54 yr ( i.e. , category 3 and 4 objects ) . in the lower panel of figure [ fig : m83_rms_2plot ] we plot the rms vs. the sed age , which gives better resolution for clusters older than @xmath33 yr . these figures show the second main result of our work , _ that older clusters have smaller rms values of the pixel - to - pixel flux variations . _ however , we also note , especially in the top panel , that the rms is degenerate , and increases in strength over the age range 27 myr ( i.e. , morphological categories 3 and 4 ) , peaks in category 4 , corresponding to an age of approximately 510 myr , and decreases in strength for clusters with older ages and in later categories . there are several possible explanations for this behavior . the first , and probably most important effect , is that the brightest stars ( i.e. , red and blue super giants ) do not appear until about 5 myr , and are largely gone by about 20 myr . two other evolutionary effects may potentially contribute to the low rms measured for the very young clusters . ( 1 ) the central part of the cluster , where it is difficult to detect individual stars due to the high background , might emerge from its dust cocoon before the outskirts . ( 2 ) clusters may start out very compact and expand with age . hence crowding may make it impossible to detect individual stars at very early ages . there is both theoretical and observational support for both of these conjectures , as discussed below . these trends are illustrated by the lines in figure [ fig : m83_rms_2plot ] , which represent linear fits to observations . the best fits for the top panel are given by : @xmath55 ( i.e. , a 5 @xmath35 correlation ) for category @xmath104.7 , and @xmath56 ( i.e. , a 6 @xmath35 correlation ) for category @xmath574.7 . for the bottom panel , only the trend for older clusters is shown , with @xmath58 for @xmath39 . we will better quantify the correlation between cluster age and the rms in pixel - to - pixel fluxes in a future paper using a significantly larger number of clusters . we note that these relationships will be different for galaxies at different distances , and in different filters . for example , fluctuations will generally be larger for older clusters ( 1001000 myr ) in the near - ir due to the presence of asymptotic giant branch ( agb ) stars . values of rms pixel - to - pixel flux variations for each cluster are included in table 1 , and the mean values for clusters in each morphological category are included in table 2 . the rms method is particularly promising for estimating the ages of clusters between 10 and 100 myr , a range which can be somewhat difficult for the sed method because the integrated colors of clusters loop back on themselves during this period . clusters in this age range can be identified initially by their lack of ionized gas . it is instructive to look at the outliers in figure [ fig : m83_rms_2plot ] , just as we did for the correlation between sed age and morphological category in section 3.1 . in the bottom panel the cluster with rms = 0.03 and @xmath59 ( i d = 44034 ) is an apparent outlier . it is also one of the sources which we believe , based on the discussion in section 3.1 , has an incorrect sed age estimate . based on its measured colors and assuming @xmath60 instead of 0.28 , this cluster has a likely age of @xmath61 , more appropriate for it s low rms value . similarly , many of the data points just to the right of this one are the same category 5b outliers discussed in section 3.1 that have suspect ages . other data points in the same region of the @xmath29 versus rms plot are located in the large bubble just below the nucleus of m83 , where it is difficult to measure the rms accurately due to the very high background . we note that the four clusters with rms @xmath62 and @xmath63 are all in morphological category 3 , and are largely responsible for the decline in the measured rms for the youngest clusters . many of the trends discussed above are seen in table 2 , which lists the mean values of several parameters as a function of morphological category . in particular , the trends in @xmath64 , @xmath29 , and h@xmath0 shell radius ( when present ) are clearly evident . in addition , the double - valued nature of pixel - to - pixel rms values is evident , with a peak at category 4a . one correlation that was mentioned briefly is the trend of increasing cluster size with age , quantified by the increase in the concentration index ( ci ) from values around 2.5 for categories 3 and 4a to @xmath65 for categories 5b and 6 . we find the same trend in our m51 data , which is briefly discussed in the appendix . this effect appears to be real , and is likely related to the rapid expansion of the clusters ( see , for example , mackey & gilmore 2003 ; bastian et al . 2009 ; pfalzner 2009 ) . several different physical mechanisms may be responsible for this expansion , including the expulsion of leftover ism due to feedback from massive stars ( e.g. , goodwin & bastian 2006 ; baumgardt & kroupa 2007 ) , and heating by binary stars and stellar mass black holes ( e.g. , mackey et al . 2008 ; see portegies zwart et al . 2010 for a review of this subject ) . the early expansion of star clusters will be discussed in more detail in a future paper ( r. chandar et al . 2011 , in preparation ) . while the dominant sources of ionizing flux responsible for the h@xmath0 emission in m83 are massive , young star clusters , which were studied in the previous sections , there is also a population of compact hii regions which are ionized by what appear to be single stars . in this section we identify and study a sample of 22 hii regions with very small h@xmath0 radii and an unresolved central point source ( based on their ci , i.e. , the magnitude difference between 0.5 and 3 pixel radii ; see chandar et al . color images of the selected sources are shown in figure [ fig : target_mosaic ] . only the brighter candidates have been retained for this first exploratory study ; the sample could be increased by a factor of two or more if fainter , less distinct objects were included . the measured colors of these sources , shown in figure [ fig : mini_cmd ] , either coincide with the colors predicted for the bluest ( youngest ) stars in the upper left portion of the diagram , or are found downstream along the reddening vector . basic properties of these sources , including positions , @xmath26 , ci , colors and measured radii of the h@xmath0 bubble , are compiled in table 3 . we note that it is unlikely that all of these objects are actually individual massive stars , hence the use of quotes around `` single star . '' some , or even most of these may be a dominant star among a close grouping of stars which are either too close to the primary source , or too faint , to be detected . what we can say is that a single ( or very close binary ) star dominates the light profile , resulting in a ci that is indistinguishable from a single star . if we assume that all of these objects have a similar age , and that the distribution in the two - color diagram is primarily due to reddening , we can correct for the effects of reddening and extinction . we show the corrected photometry in a color magnitude diagram in the left panel of figure [ fig : mini_cmd ] . here , we have assumed an intrinsic color of @xmath66 , the color of the bluest object , and solve for the @xmath27 color excess of each hii region . we find that this procedure moves most of the objects to a very young isochrone ( 4 myr is shown ) , although three sources scatter to the left of the models . these may be especially young stars , or the offset may result from observational uncertainties , since these are the three most reddened sources , and hence have the largest corrections . while many of these single - star hii ( i.e. , sshii ) regions are found near large regions of recent star formation , several of them are quite isolated ( see figure 3 from whitmore 2010 ) , raising the possibility that these massive stars formed in the field rather than in clusters or associations . for example , five of the 22 sshii regions are @xmath671 kpc away from any region of active star formation . if these are stars that have been dynamically ejected from their birthsites , i.e. , runaways , they must have velocities @xmath67200 km / s . however , they do not appear to have the prominent bow - front morphologies typical of many runaway stars in the galaxy and magellanic clouds ( e.g. , gvaramadze et al . velocity measurements are needed to establish if these massive stars formed in the field or are runaways from larger star - forming regions . as summarized in the introduction , the spatial relationship between young @xmath68 yr star clusters and their hii regions has been previously studied both observationally and theoretically . to our knowledge however , no study has yet systematically measured the sizes of h@xmath0 bubbles in a nearby galaxy and correlated these sizes with the properties of their ionizing star clusters , as we have done here . we have demonstrated that in m83 there is a good correlation between bubble size and age for clusters younger than @xmath43 , and that clusters with ages @xmath436.9 have a larger range in bubble size . this suggests that some of the expanding bubbles at these older ages may have effectively `` stalled . '' theoretical work suggests that bubble size should depend not only on the ambient pressure in the ism and the age of the cluster , but also on cluster mass . to look for evidence of a mass - dependence on bubble size , we compare the sizes measured for h@xmath0 bubbles associated with @xmath696.9 clusters that are more and less massive than @xmath70 . at these ages the bubbles are presumably approaching their stall radius , and the effect of cluster mass should be more apparent . the large , open circles in the top panel of figure [ fig : ha_vs_r ] show the sizes of clusters with @xmath47 and @xmath48 . while clusters at high and low masses have an overlapping range in cluster size , there is a tendency for more massive clusters to have larger bubbles , with median sizes of 18 pc and 41 pc for @xmath71 and @xmath72 , respectively . a formal fit of h@xmath0 radius versus @xmath73 for these clusters gives a slope of @xmath74 , suggesting that there may be a weak correlation between bubble size and mass . future studies that include a larger number of sources , selected in a more systematic way , are needed to confirm this result . however , other effects may also be important . for example , stochasticity in the number of massive stars formed in lower mass clusters can strongly affect their integrtated colors and hence their estimated ages . in this case , a lower mass cluster can have an intrinsically redder color than a higher mass cluster of the same age ( e.g. , fouesneau & lancon 2010 ) , and hence an older estimate for the age of the cluster . in addition to the correlation between h@xmath0 bubble size and cluster age ( and possibly mass ) , we also found that the rms variations in pixel - to - pixel brightness correlates with cluster age . this result has applications beyond star clusters , and can be used to constrain the ages of stellar populations in general . for example , portions of m83 itself that contain young stars and star clusters , such as along the spiral arms , have larger pixel - to - pixel flux variations than portions of m83 dominated by older stars , such as between the arms ( see h. kim et al . 2011 , in preparation for a discussion ) . this is also true in large portions of m82 and in portions of the tidal tails of galaxy mergers , where regions dominated by intermediate age star clusters ( @xmath67100 myr ) have small fluctuations in the surface brightness of the field stars . we have used observations taken with the newly installed wfc3 camera on - board the _ hubble space telescope _ to develop two independent methods for age - dating young star clusters in the nearby spiral galaxy m83 . our primary results are summarized below . a working classification system , largely based on h@xmath0 morphology and pixel - to - pixel flux variations , was developed to map an observed age sequence onto a proposed sequence of cluster evolution . the underlying evolutionary picture includes the formation of dense cores in giant molecular clouds ( gmcs ) ; a stage where the young stars are completely obscured by their dust cocoon ; an emerging stage where the clusters become visible in the ir , and then in the optical as stellar winds and supernovae blow away the dust ; a stage where an expanding h@xmath0 bubble is blown and the existence of very bright young stars leads to large pixel - to - pixel flux variations ; and later stages with no evidence of h@xmath0 and diminishing pixel - to - pixel flux variations . \2 . we found that h@xmath0 morphology , i.e. , the size of the ionized gas bubble , provides a viable method for age - dating clusters in the range 110 myr . this method is based on the simple premise that the gas in very young ( @xmath2 few myr ) clusters is largely coincident with the cluster stars , is in a small , ring - like structure surrounding the stars in slightly older clusters since the winds from massive stars have had time to push out the natal gas ( e.g. , @xmath75 myr ) , and is in a larger ring - like bubble for still older clusters ( i.e. , @xmath67510 myr ) . if no h@xmath0 is associated with a cluster it is generally older than @xmath6710 myr . we first made _ qualitative _ estimates based on the classification scheme outlined above , and find that the ages of the clusters , as determined from the sed method described in chandar et al . ( 2010 ) , correlate well with the morphological categories , with a scatter of @xmath670.1 in @xmath29 within each category for the clusters with h@xmath0 emission , and a scatter of @xmath670.5 in @xmath29 for the older clusters . we then _ quantified _ this technique by correlating the measured radii of the most conspicuous h@xmath0-emitting ring or shell which appears to be physically related to the cluster with the sed ages determined in chandar et al . we found tentative evidence for a weak correlation between bubble size and cluster mass , but a larger , more objectively selected sample will be required to confirm this . we then used pixel - to - pixel flux variations to age - date clusters . this technique is based on the fact that individual stars are bright enough to be visible within clusters when they are young ( e.g. , m@xmath6 , the approximate detection level , for the brightest stars with ages @xmath10100 myr ) , leading to relatively large pixel - to - pixel variations in flux . the strength of the fluctuations peaks in clusters with ages of @xmath67510 myr , presumably because this is when the brightest stars ( e.g. , red and blue super giants ) appear . the number of luminous , evolved stars falls off for both younger and older clusters . this degeneracy in age can be broken using other age indicators such as the h@xmath0 morphology . the technique is especially useful for identifying clusters older than 100 myr . a by - product of this study was the identification of 22 `` single - star '' hii regions in m83 . by assuming that all of these objects have a similar age , and that the distribution in the two - color diagram is primarily due to reddening , we corrected for the effects of reddening and extinction . we found that this procedure moves most of the objects to a very young isochrone with an age approximately 4 myr . some of these massive stars are located far from any star - forming region , indicating that they either formed in the field or were dynamically ejected from their birthsites at very high velocities . these sshii regioins will be studied in more detail in h. kim et al . ( 2011 , in preparation ) . in the future , we will extend the classification system into the near - ir ( i.e. , categories 1 and 2 ) using our j and h observations . we will also calculate the energy budget of cluster stars and compare with physical properties of the clusters and the ism . finally , we will extend this analysis to other galaxies in the ers1 sample ( including `` low pressure '' systems such as the dwarf starburst galaxy ngc 4214 ) in order to determine whether the relationships are universal or are strongly dependent on environment . we thank zolt levay for making the color images used in figures [ fig:18plot ] and [ fig : large_image ] . this paper is based on observations taken with the nasa / esa _ hubble space telescope _ obtained at the space telescope science institute , which is operated by aura , inc . , under nasa contract nas5 - 26555 . the paper makes use of early release science observations made by the wfc3 science oversight committee . we are grateful to the director of stsci for awarding director s discretionary time for this program . r. c. is grateful for support from the nsf through career award 0847467 . 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 nasa . the spiral galaxy m51 was originally used to perform a pilot study to test how well h@xmath0 morphology can be used to estimate cluster ages . examples of the results are included in figures [ fig : m51_ha_morph_1 ] and [ fig : m51_ha_morph_2 ] . these are similar to figure [ fig : large_image ] for m83 and show that even at twice the distance of m83 the quality of the images are comparable , and the morphological categories match sed age estimates at a similar level . a more complete analysis will be included in a future paper ( r. chandar et al . 2011 ; in preparation ) . 1&65815&204.244637&@xmath7629.8507043&@xmath766.10 & 2.57&@xmath761.35&@xmath760.12&0.24&6.52&1.1e3 & 3&2&0.082 + 2&63611&204.2426354&@xmath7629.8515924&@xmath766.42 & 2.35&@xmath761.27&0.07&0.34&6.54&2.1e3 & 3&1&0.091 + 3&41736&204.251687&@xmath7629.8640556&@xmath768.70 & 2.68&@xmath760.21&0.80&0.50&6.7&2.6e4 & 3&7&0.064 + 4&39744&204.2524132&@xmath7629.8651911&@xmath7611.03 & 2.58&@xmath761.33&0.23&0.30&6.56&1.2e5 & 3&7&0.119 + 5&49651&204.2564416&@xmath7629.8586912&@xmath767.17 & 2.50&@xmath760.38&0.68&0.50&6.48&7.1e3 & 3&3&0.063 + 6&32461&204.2880034&@xmath7629.8692326&@xmath767.67 & 2.62&@xmath761.36&0.30&0.44&6.52&8.6e3 & 3&5&0.073 + 7&53072&204.2889299&@xmath7629.8569125&@xmath766.64 & 2.61&@xmath760.93&0.56&0.50&6.48&4.2e3 & 3&4&0.076 + 8&93801&204.2795357&@xmath7629.8392001&@xmath767.10 & 2.50&@xmath761.37&0.08&0.32&6.54&3.6e3 & 3&1&0.106 + 9&50287&204.2865616&@xmath7629.8582915&@xmath767.35 & 2.66&@xmath760.54&0.69&0.50&6.70&7.4e3 & 3&6&0.072 + 10&59214&204.2818977&@xmath7629.8537395&@xmath767.56 & 2.53&@xmath761.08&0.36&0.50&6.54&9.3e3 & 3&6&0.105 + 11&24767&204.2493318&@xmath7629.8729735&@xmath766.89 & 2.32&@xmath760.46&0.70&0.50&6.56&5.1e3 & 3&7&0.096 + 12&58013&204.2475174&@xmath7629.8543366&@xmath766.49 & 2.43&@xmath761.29&0.15&0.42&6.46&5.4e3 & 3&7&0.104 + 13&52726&204.2458956&@xmath7629.8570619&@xmath768.13 & 2.78&@xmath761.50&@xmath760.06&0.26&6.42&1.2e4 & 3&7&0.130 + 14&12721&204.2908632&@xmath7629.8785688&@xmath765.98 & 2.43&@xmath761.30&0.32&0.50&6.46&4.3e3 & 3&5&0.070 + 15&19354&204.2525635&@xmath7629.8755537&@xmath766.06 & 2.83&@xmath761.39&0.27&0.44&6.44&2.5e3 & 3&7&0.065 + 16&51616&204.2799185&@xmath7629.8575701&@xmath769.31 & 2.43&@xmath761.33&@xmath760.02&0.20&6.58&1.8e4 & 4a&6&0.131 + 17&49790&204.2546965&@xmath7629.8585957&@xmath767.88 & 2.27&@xmath760.85&0.27&0.42&6.68&1.1e4 & 4a&10&0.091 + 18&17035&204.2906967&@xmath7629.8766101&@xmath767.23 & 2.24&@xmath761.48&@xmath760.16&0.00&6.70&1.8e3 & 4a&15&0.119 + 19&66897&204.2654298&@xmath7629.8502774&@xmath768.60 & 2.96&@xmath761.50&@xmath760.12&0.00&6.72&4.4e3 & 4a&20&0.134 + 20&39842&204.2520397&@xmath7629.8651314&@xmath7610.94 & 2.71&@xmath761.51&@xmath760.15&0.08&6.58&5.2e4 & 4a&13&0.086 + 21&31457&204.2855284&@xmath7629.8698283&@xmath768.83 & 2.19&@xmath761.31&@xmath760.03&0.04&6.78&8.9e3 & 4a&17&0.164 + 22&74860&204.262148&@xmath7629.8474511&@xmath768.59 & 2.67&@xmath761.43&@xmath760.14&0.00&6.78&5.5e3 & 4a&21&0.141 + 23&89929&204.2581863&@xmath7629.8412261&@xmath766.79 & 2.23&@xmath761.35&@xmath760.05&0.04&6.72&1.3e3 & 4a&17&0.118 + 24&25607&204.2492229&@xmath7629.8725982&@xmath767.85 & 2.61&@xmath760.66&0.42&0.50&6.70&1.2e4 & 4a&21&0.120 + 25&75041&204.2775285&@xmath7629.847385&@xmath767.31 & 2.24&@xmath761.61&@xmath760.15&0.12&6.52&2.7e3 & 4a&7&0.137 + 26&64462&204.2811447&@xmath7629.8512236&@xmath768.19 & 2.30&@xmath760.81&0.34&0.46&6.68&1.7e4 & 4a&7&0.121 + 27&61923&204.263904&@xmath7629.8523354&@xmath766.02 & 2.52&@xmath761.28&0.06&0.36&6.54&1.5e3 & 4a&9&0.105 + 28&21601&204.2861725&@xmath7629.8744932&@xmath769.59 & 2.27&@xmath761.48&@xmath760.12&0.00&6.74&1.5e4 & 4a&15&0.158 + 29&62771&204.2811659&@xmath7629.8519608&@xmath767.25 & 2.27&@xmath760.96&0.28&0.38&6.70&5.6e3 & 4a&16&0.130 + 30&13265&204.2547344&@xmath7629.8783208&@xmath767.64 & 2.44&@xmath761.63&@xmath760.27&0.00&6.64&2.3e3 & 4a&13&0.148 + 31&5990&204.2551409&@xmath7629.8813282&@xmath768.17 & 2.47&@xmath761.53&@xmath760.26&0.00&6.72&3.9e3 & 4a&46&0.124 + 32&67958&204.2887307&@xmath7629.8499286&@xmath769.84 & 3.20&@xmath761.35&@xmath760.04&0.02&6.78&1.1e4 & 4a&19&0.161 + 33&66054&204.2644967&@xmath7629.8506016&@xmath769.97 & 2.63&@xmath761.45&0.39&0.00&6.88&2.8e4 & 4a&49&0.127 + 34&69793&204.2694643&@xmath7629.8493238&@xmath7611.41 & 2.63&@xmath761.48&@xmath760.16&0.00&6.78&7.5e4 & 4a&25&0.136 + 35&7716&204.2816067&@xmath7629.8807039&@xmath769.17 & 2.47&@xmath761.43&0.41&0.00&6.88&1.4e4 & 4b&68&0.151 + 36&36651&204.2524088&@xmath7629.8668398&@xmath7612.17 & 3.09&@xmath761.58&@xmath760.22&0.00&6.78&1.1e5 & 4b&&0.070 + 37&35937&204.2537197&@xmath7629.8672182&@xmath7610.31 & 2.79&@xmath761.24&0.45&0.02&6.90&4.2e4 & 4b&&0.052 + 38&54268&204.2932226&@xmath7629.8563319&@xmath768.57 & 2.87&@xmath760.91&0.79&0.34&6.86&1.7e4 & 4b&32&0.108 + 39&23366&204.2529648&@xmath7629.8736403&@xmath769.39 & 2.43&@xmath761.29&@xmath760.08&0.02&6.78&1.3e4 & 4b&41&0.116 + 40&27423&204.2896066&@xmath7629.8718058&@xmath768.38 & 2.34&@xmath760.99&0.50&0.36&6.78&1.4e4 & 4b&71&0.170 + 41&78807&204.2629122&@xmath7629.8459862&@xmath769.71 & 2.46&@xmath761.55&@xmath760.23&0.00&6.78&1.6e4 & 4b&45&0.169 + 42&36729&204.2525635&@xmath7629.8667944&@xmath7612.15 & 3.08&@xmath761.45&0.00&0.00&6.74&1.0e5 & 4b&&0.073 + 43&70777&204.2607456&@xmath7629.8489709&@xmath768.18 & 2.68&@xmath761.23&0.49&0.08&6.88&6.7e3 & 4b&43&0.087 + 44&70769&204.2650559&@xmath7629.8489734&@xmath767.77 & 2.92&@xmath761.34&@xmath760.08&0.00&6.76&2.1e3 & 4b&18&0.108 + 45&37589&204.2519048&@xmath7629.8663287&@xmath7611.96 & 3.20&@xmath761.13&0.46&0.24&6.74&1.4e5 & 4b&&0.064 + 46&49893&204.2897929&@xmath7629.8585381&@xmath769.76 & 2.74&@xmath761.49&@xmath760.24&0.00&6.78&1.6e4 & 4b&48&0.189 + 47&37095&204.2523893&@xmath7629.8665896&@xmath7611.18 & 2.79&@xmath761.32&0.35&0.00&6.88&8.0e4 & 4b&&0.059 + 48&78154&204.257905&@xmath7629.8462475&@xmath768.28 & 2.76&@xmath760.86&0.18&0.00&7.86&3.4e4 & 5a&&0.080 + 49&66216&204.2897363&@xmath7629.8505371&@xmath768.87 & 2.56&@xmath760.98&0.70&0.34&6.84&2.5e4 & 5a&&0.145 + 50&76156&204.2556622&@xmath7629.8469956&@xmath7610.21 & 3.15&@xmath761.08&0.62&0.06&7.48&1.0e5 & 5a&&0.089 + 51&66123&204.2903301&@xmath7629.8505755&@xmath769.39 & 2.50&@xmath761.11&0.59&0.22&6.86&3.0e4 & 5a&&0.125 + 52&66069&204.2844254&@xmath7629.8505955&@xmath768.88 & 3.13&@xmath760.78&0.47&0.06&7.81&4.5e4 & 5a&&0.067 + 53&85836&204.2694279&@xmath7629.8431197&@xmath769.43 & 3.25&@xmath761.12&0.18&0.10&6.78&8.7e3 & 5a&&0.093 + 54&40779&204.2926898&@xmath7629.8645926&@xmath7610.17 & 3.06&@xmath760.65&0.49&0.08&7.96&2.1e5 & 5a&&0.044 + 55&30950&204.2578713&@xmath7629.8701108&@xmath7610.85 & 2.94&@xmath761.08&0.60&0.08&7.49&2.5e5 & 5a&&0.085 + 56&83925&204.2643496&@xmath7629.8439131&@xmath768.71 & 2.79&@xmath760.97&0.77&0.14&7.44&4.3e4 & 5a&&0.129 + 57&44034&204.2499199&@xmath7629.8626794&@xmath769.56 & 3.10&@xmath760.82&0.44&0.28&6.78&2.1e4 & 5b&&0.030 + 58&17159&204.2616672&@xmath7629.8765474&@xmath768.22 & 3.18&@xmath760.42&0.57&0.16&8.06&4.3e4 & 5b&&0.052 + 59&25716&204.2827697&@xmath7629.8725452&@xmath767.28 & 2.76&@xmath760.14&0.42&0.00&8.41&2.6e4 & 5b&&0.054 + 60&18032&204.2849161&@xmath7629.876142&@xmath767.72 & 2.98&@xmath760.72&0.61&0.36&6.78&5.7e3 & 5b&&0.064 + 61&74692&204.2760088&@xmath7629.8475135&@xmath766.99 & 3.17&@xmath760.04&0.70&0.50&6.74&3.3e3 & 5b&&0.075 + 62&85964&204.2565326&@xmath7629.843069&@xmath767.59 & 2.99&@xmath760.33&0.44&0.00&8.36&2.7e4 & 5b&&0.049 + 63&14748&204.2752747&@xmath7629.8776532&@xmath767.14 & 3.29&@xmath760.11&0.40&0.00&8.36&1.2e4 & 5b&&0.038 + 64&65733&204.2851579&@xmath7629.8507302&@xmath769.82 & 2.73&@xmath760.94&0.61&0.10&7.59&1.3e5 & 5b&&0.089 + 65&65479&204.2862739&@xmath7629.8508316&@xmath769.06 & 3.19&@xmath760.37&0.42&0.00&8.31&7.8e4 & 5b&&0.050 + 66&65304&204.2849235&@xmath7629.8508999&@xmath768.51 & 3.39&0.07&0.80&0.18&8.41&5.8e4 & 5b&&0.039 + 67&10114&204.2664853&@xmath7629.8796816&@xmath767.17 & 2.64&@xmath761.53&@xmath760.33&0.00&6.64&1.4e3 & 5b&&0.105 + 68&58911&204.2807662&@xmath7629.8539005&@xmath767.69 & 2.88&0.15&1.55&0.32&9.11&2.9e5 & 5b&&0.050 + 69&58363&204.2831825&@xmath7629.8541717&@xmath767.23 & 3.27&0.27&0.54&0.00&8.86&2.9e4 & 5b&&0.034 + 70&66553&204.2887134&@xmath7629.8504088&@xmath768.06 & 3.23&0.64&1.08&0.20&9.01&1.5e5 & 5b&&0.035 + 71&55591&204.2710795&@xmath7629.8556304&@xmath767.27 & 3.14&0.34&0.74&0.04&8.86&4.1e4 & 5b&&0.036 + 72&94866&204.2790178&@xmath7629.8386224&@xmath768.34 & 3.28&0.06&0.39&0.00&8.46&4.0e4 & 5b&&0.031 + 73&3113&204.2651882&@xmath7629.8828492&@xmath766.93 & 3.28&0.04&0.89&0.00&9.16&3.8e4 & 6&&0.031 + 74&12505&204.2762954&@xmath7629.8786591&@xmath766.67 & 3.26&0.44&0.97&0.00&9.06&2.2e4 & 6&&0.029 + 75&24528&204.2890323&@xmath7629.8730866&@xmath769.15 & 3.13&@xmath760.14&0.80&0.22&8.36&1.8e5 & 6&&0.040 + 76&55985&204.2695705&@xmath7629.8553997&@xmath767.97 & 2.91&0.02&1.01&0.00&9.26&2.1e5 & 6&&0.040 + 77&18044&204.2956504&@xmath7629.8761376&@xmath768.29 & 3.24&0.12&0.54&0.00&8.61&5.3e4 & 6&&0.034 + 78&54416&204.2898224&@xmath7629.8562536&@xmath768.01 & 3.15&@xmath760.37&0.91&0.38&7.96&6.1e4 & 6&&0.042 + 79&46572&204.2634717&@xmath7629.8608741&@xmath768.54 & 3.05&0.18&0.81&0.04&8.91&1.7e5 & 6&&0.058 + cccccccc 3 & @xmath77 & @xmath78 & @xmath79 & @xmath80 & @xmath81 & @xmath82 & @xmath83 + 4a & @xmath84 & @xmath85 & @xmath86 & @xmath87 & @xmath88 & @xmath89 & @xmath90 + 4b & @xmath91 & @xmath92 & @xmath93 & @xmath94 & @xmath95 & @xmath96 & @xmath97 + 5a & @xmath98 & @xmath99 & @xmath100 & @xmath101 & @xmath102 & & @xmath103 + 5b & @xmath104 & @xmath105 & @xmath106 & @xmath107 & @xmath108 & & @xmath109 + 6 & @xmath110 & @xmath111 & @xmath112 & @xmath113 & @xmath114 & & @xmath115 + 1&3550&204.2548844&@xmath7629.8825632&@xmath765.51 & 2.13&@xmath761.63&@xmath760.22&9 + 2&3916&204.2554865&@xmath7629.8823689&@xmath765.19 & 2.30&@xmath761.60&@xmath760.28&8 + 3&10373&204.2549063&@xmath7629.8795771&@xmath765.24 & 2.28&@xmath761.49&@xmath760.16&7 + 4&11098&204.2672170&@xmath7629.8792752&@xmath766.89 & 2.02&@xmath761.64&@xmath760.33&4 + 5&29250&204.2745636&@xmath7629.8709998&@xmath764.41 & 2.09&@xmath761.21&0.02&8 + 6&35101&204.2745919&@xmath7629.8676963&@xmath765.23 & 2.11&@xmath761.49&@xmath760.39&6 + 7&37778&204.2667959&@xmath7629.8662227&@xmath765.93 & 2.21&@xmath761.13&0.08&9 + 8&47544&204.2823703&@xmath7629.8602099&@xmath765.86 & 2.10&@xmath761.51&@xmath760.14&7 + 9&47772&204.2689918&@xmath7629.8600630&@xmath765.09 & 2.12&@xmath761.31&0.27&7 + 10&48660&204.2894733&@xmath7629.8594131&@xmath764.26 & 2.25&@xmath761.62&@xmath760.48&2 + 11&49679&204.2535423&@xmath7629.8586761&@xmath765.78 & 2.18&@xmath760.49&0.32&9 + 12&50923&204.2552793&@xmath7629.8579399&@xmath766.13 & 2.23&@xmath761.15&0.25&5 + 13&51301&204.2887045&@xmath7629.8577311&@xmath765.19 & 2.18&@xmath761.27&0.15&2 + 14&52904&204.2462139&@xmath7629.8569822&@xmath764.69 & 2.23&@xmath761.59&0.14&1 + 15&56031&204.2828693&@xmath7629.8553764&@xmath766.15 & 2.18&@xmath761.36&0.02&6 + 16&56485&204.2849019&@xmath7629.8551313&@xmath766.19 & 2.04&@xmath761.42&0.06&8 + 17&56492&204.2829518&@xmath7629.8551272&@xmath766.24 & 2.24&@xmath760.12&1.07&9 + 18&56937&204.2841803&@xmath7629.8548844&@xmath768.06 & 2.13&@xmath760.21&0.90&6 + 19&60624&204.2516243&@xmath7629.8530186&@xmath764.99 & 2.20&@xmath761.28&@xmath760.02&7 + 20&61345&204.2496746&@xmath7629.8526196&@xmath764.79 & 2.23&@xmath761.25&@xmath760.32&4 + 21&72114&204.2810230&@xmath7629.8485123&@xmath766.34 & 2.08&@xmath760.77&0.33&2 + 22&88339&204.2804302&@xmath7629.8420518&@xmath764.49 & 2.30&@xmath761.08&@xmath760.13&7 +

this paper is the first in a series that aims to understand the long - term evolution of neutron star magnetic fields . we model the stellar matter as an electrically neutral and lightly - ionized plasma composed of three moving particle species : neutrons , protons , and electrons ; these species can be converted into each other by weak interactions ( beta decays ) , suffer binary collisions , and be affected by each other s macroscopic electromagnetic fields . since the evolution of the magnetic field occurs over thousands of years or more , compared to dynamical timescales ( sound and alfvn ) of milliseconds to seconds , we use a slow - motion approximation in which we neglect the inertial terms in the equations of motion for the particles . this approximation leads to three nonlinear partial - differential equations describing the evolution of the magnetic field , as well as the movement of two fluids : the charged particles ( protons and electrons ) and the neutrons . these equations are first rather than second order in time ( involving the velocities of the three species but not their accelerations ) . in this paper , we restrict ourselves to a one - dimensional geometry in which the magnetic field points in one cartesian direction , but varies only along an orthogonal direction . we study the evolution of the system in three different ways : ( i ) estimating timescales directly from the equations , guided by physical intuition ; ( ii ) a normal - mode analysis in the limit of a nearly uniform system ; and ( iii ) a finite - difference numerical integration of the full set of nonlinear partial - differential equations . we find good agreement between our analytical normal - mode solutions and the numerical simulations . we show that the magnetic field and the particles evolve through successive quasi - equilibrium states , on timescales that can be understood by physical arguments . depending on parameter values , the magnetic field can evolve by ohmic diffusion or by ambipolar diffusion , the latter being limited either by interparticle collisions or by relaxation to chemical quasi - equilibrium through beta decays . the numerical simulations are further validated by verifying that they satisfy the known conservation laws in highly nonlinear situations .

0801.4372

observations of the surface magnetic fields of neutron stars have shown a correlation between the age of the star and the magnetic - field strength . for instance , young radio pulsars and high - mass x - ray binaries have surface magnetic fields of the order of @xmath0 g , while millisecond pulsars and low - mass x - ray binaries , which are older objects , have magnetic fields @xmath1 g. these observations suggest that the magnetic field is decaying , although this might be a by - product of accretion ( @xcite ; @xcite ; @xcite ; @xcite ) . on the other hand , it is thought that the spontaneous decay of the ultrastrong @xmath2 g magnetic field in _ magnetars _ is the main source of their x - ray luminosity since these objects appear to radiate substantially more power than that available from their rotational energy loss ( @xcite ; @xcite ; @xcite ) . since these objects appear to be isolated , the field decay must be attributed to processes intrinsic to the neutron stars . ref.@xcite ; ( hereafter gr-92 ) identified three long - term mechanisms that can promote the spontaneous decay of the magnetic field in the interior of neutron stars : 1 . _ ambipolar diffusion _ , i.e. the drift of the magnetic field and the charged particles relative to the neutrons . @xcite ; ( hereafter td-96 ) , using magnetar parameters , estimated the timescale of the magnetic field decay by means of ambipolar diffusion , finding that was consistent with the timescale of @xmath3 years observed for these objects . 2 . _ hall drift _ , a non - dissipative advection of the magnetic field by the associated electrical current ( @xcite ; @xcite ; @xcite ; @xcite , @xcite ; @xcite ; @xcite ; @xcite , @xcite ; @xcite ) . 3 . _ ohmic diffusion _ , a dissipative process caused by electrical resistivity ( see also baym et al . @xcite ) . all of these processes occur over timescales of thousands of years or more , compared to typical dynamical timescales ( sound and alfvn ) of miliseconds to seconds . the work of gr-92 was analytical , and therefore useful to identify general processes and relevant timescales , but it did not address the action of the identified processes in their full nonlinear development and their interactions with each other . the full evolution of the magnetic field can only be addressed by numerical simulations . regarding the geometrical configuration of the magnetic field , three - dimensional magnetohydrodynamics ( mhd ) simulations showed that in a stable stratified , non - rotating star , the initial field evolves on a short , alfvn - like timescale to a large - scale equilibrium configuration that can be axisymmetric ( @xcite , @xcite ; @xcite ) or non - axisymmetric ( @xcite ) , depending on the radial dependence of the initial magnetic - field strength . these simulations , being focused on the evolution of the magnetic field over short dynamical timescales , were carried out within the framework of the mhd theory , which considers the stellar matter to be a single fluid . the evolution of these configurations has not yet been studied over longer timescales , on which relative motions of different particle species are present and the processes studied by gr-92 become important . the description of these long - term processes requires a multi - fluid theory . in this paper we extend the model of gr-92 by allowing both neutron and charged - particle movements , rather than considering the neutrons as a fixed background ( see sect . [ phys ] ) . for simplicity , we restrict ourselves to a plane - parallel system where the magnetic field points in one cartesian direction but varies only along an orthogonal direction . although this one - dimensional model is unrealistic , it allows us to study the basic evolutionary processes in an analytical and numerical fashion with no serious complications before considering a realistic stellar geometry in more dimensions . gr-92 considered all the species in the star as non - interacting fermions ; in contrast , we use equations of state including interactions among protons and neutrons . since in neutron stars the magnetic pressure is much less than the degeneracy pressure of the particles , we treat the particle densities in terms of small perturbations around a non - magnetized background in full equilibrium . thus , we obtain a system of equations that is nonlinear with respect to the magnetic field but linear with respect to the particle densities ( see sect . [ 1dmodel ] ) . we will show that the magnetic field and particle species evolve through successive quasi - equilibrium states . we describe the basic physical processes of this evolution and estimate analytically the characteristic timescales required to reach these quasi - equilibrium states . we find that these timescales depend on the values of physical parameters in the system , such as the magnetic - field intensity , the collision rate between the charged particles and the neutrons , the weak interaction rate , and the ohmic resistivity ( sect . [ carsca ] ) . to obtain an analytic solution of the system of partial differential equations , we linearize the equations with respect to the magnetic field and find normal - mode solutions ( sect . [ normmodes ] ) and obtain analytical approximations for the decay times in appendix [ apa ] . we develop numerical simulations ( whose basic algorithm is described in appendix [ apb ] ) in the linear regime and compare with the normal - mode results ( sect . [ finline ] ) . in sect . [ finnoli ] , we carry out numerical simulations in the nonlinear regime and verify the conservation laws of our model . finally in sect . [ conc ] , we summarize the main results of this paper and provide our conclusions . we model the stellar interior as an electrically - neutral and slightly - ionized plasma composed of three moving - particle species : neutrons @xmath4 , protons @xmath5 , and electrons @xmath6 . this is an extension of the model of gr-92 , since these authors considered the neutrons to form a motionless background . furthermore , we account for strong interactions between neutrons and protons by writing each of their chemical potentials as a function of both of their number densities : @xmath7 $ ] . we consider the electrons to be an ideal , relativistic fermi gas , and write their chemical potential as a function only of their own density , @xmath8^{2/3}}$ ] , with @xmath9 the electron rest mass . the plasma species are described as three fluids coupled by collisions and electromagnetic forces , satisfying the equations of motion ( gr-92 ; reisenegger et al . @xcite ) : @xmath10 where @xmath11 is the mean velocity ; @xmath12 , @xmath13 is the effective mass of each species ( which includes corrections due to interactions and relativistic effects ; @xcite ) ; @xmath14 is the degeneracy - pressure gradient of the species @xmath15 ; @xmath16 is the gravitational potential ; @xmath17 and @xmath18 are the electric and magnetic fields ; and the last term represents the drag forces due to elastic binary collisions , which damp the relative motions of the different species . the collisional coupling strengths are parametrized by the symmetric matrix @xmath19 that depends generally on the local density and temperature . we ignore the effects of superfluidity or superconductivity , as well as additional particles that might be present in neutron stars . in the early stages of the neutron - star formation , particles are locked by collisions , which behave as a single fluid . alfvn waves propagate across the star , allowing it to reach a _ magnetohydrostatic quasi - equilibrium _ state in which all forces acting on a fluid element are closely in balance . since in this paper we are interested in simulating the evolution over far longer timescales characteristic of the processes of gr-92 , we use a slow - motion approximation in which we neglect the acceleration terms on the left - hand side ( lhs ) of eq . ( [ movecond ] ) ( i.e. , @xmath20 ) . we assume implicitly that all the forces acting on a given particle species balance each other at all times . to some extent , this will happen , in the sense that the relative velocities will take such values that the drag forces balance all the other forces , and these drag forces will decay as the other forces come into balance . however , if eq . ( [ movecond ] ) is summed over all three particle species , the drag forces cancel and no mechanism remains to ensure that the entire fluid reaches magnetohydrostatic quasi - equilibrium . to create such mechanism , we introduce an artificial friction - like force acting on the neutrons ( the most abundant species ) of the form @xmath21 such that the neutron equation of motion becomes @xmath22 as a result , we obtain the equation of magnetohydrostatic quasi - equilibrium for a fluid element , @xmath23 where @xmath24 is the electrical current density . thus , the system approaches magnetohydrostatic quasi - equilibrium on a timescale controlled by the parameter @xmath25 ( see sect . [ tmhdcar ] ) . this parameter is chosen in such a way that this timescale is sufficiently long for the numerical code to be able to deal with it ( and therefore much longer than the true dynamical timescales [ sound and alfvn ] ) , but shorter than the long timescales of interest in this paper . if we add eq . ( [ movecond ] ) for electrons and protons , assuming charge neutrality , @xmath26 , we obtain the diffusion equation for the combined fluid of charged particles , @xmath27 where @xmath28 , @xmath29 , and we define the _ ambipolar diffusion velocity _ by : @xmath30 if the charged particles are not in _ diffusive quasi - equilibrium _ , that is the forces on the right - hand side ( rhs ) of eq . ( [ chargeddiff ] ) are not in balance , the latter will drive them to move relative to the neutrons at the ambipolar diffusion velocity @xmath31 . on the other hand , if the system is in magnetohydrostatic quasi - equilibrium ( see eq . ( [ mhdmod ] ) and the charged particles have reached their diffusive quasi - equilibrium ( @xmath32 ) , this implies that neutrons are also in a diffusive quasi - equilibrium state given by @xmath33 to derive the equation that governs the evolution of the magnetic field , we combine eq . ( [ movecond ] ) , for electrons and protons without the inertial terms , and the induction equation @xmath34 , to obtain @xmath35-\vec{\nabla}\times \left(\frac{c^2\vec{\nabla}\times{\vec{b}}}{4\pi\sigma}\right)\nonumber\\ & & -\frac{c}{2e}\vec{\nabla}\left(\frac{\gamma_{en}-\gamma_{pn}}{\gamma_{cn}}\right)\times\vec{\nabla}\mu_c\nonumber\\ & & -{1\over ec}\vec{\nabla}\left({\gamma_{en}\mu_p-\gamma_{pn}\mu_e}\over{\gamma_{cn}}\right)\times\vec{\nabla}\psi,\end{aligned}\ ] ] where we define the _ hall drift velocity _ , which is proportional to the electrical current density , by @xmath36 the first term on the rhs of eq . ( [ magn ] ) shows that the magnetic field is transported by the sum of the neutron velocity @xmath37 , the ambipolar diffusion velocity @xmath31 , and the hall drift velocity @xmath38 . the second term represents the ohmic diffusion , where the electrical conductivity is @xmath39^{-1}.\ ] ] finally , the last two terms on the rhs of eq . ( [ magn ] ) represent battery effects . to complete a set of equations to describe the system evolution , we need equations for the evolution of the particle densities . weak interactions between the particles tend to reduce chemical - potential imbalances between the charged particles and neutrons . we define the differences between the rates , per unit volume , of the reactions @xmath40 and @xmath41 as @xmath42 , where @xmath43 is the chemical imbalance , and the parameter @xmath44 depends generally on density and temperature ( gr-92 ) . , where @xmath45 is the boltzmann constant and @xmath46 is the temperature , @xmath44 must also be allowed to depend on @xmath47 ( reisenegger @xcite ; fernndez & reisenegger @xcite ) . ] the _ chemical quasi - equilibrium _ between the charged particles and neutrons is reached when @xmath48 . the continuity equations for the particle densities are then given by @xmath49 in eq . ( [ densn ] ) , the @xmath50 sign corresponds to the neutrons and the @xmath51 sign to the electrons and protons . defining the _ baryon _ number density as @xmath52 and adding eq . ( [ densn ] ) for charged particles and neutrons , we obtain the conservation law for @xmath53 , @xmath54 the set of differential eqs . ( [ magn ] ) , ( [ densn ] ) , and ( [ densb ] ) are first , rather than second order in time , that is they involve the velocities of the three fluid species , but not their accelerations . we consider a one - dimensional geometry in which the magnetic field points in one cartesian direction @xmath55 , but varies only along an orthogonal direction @xmath56 as @xmath57 . we also assume that all of the other physical variables vary only along @xmath56 , and therefore that the gradient operator is @xmath58 . from ampere s law , @xmath59 , thus , @xmath60 . using eqs . ( [ ambi ] ) , ( [ magn ] ) , ( [ hall ] ) , ( [ densn ] ) , and ( [ densb ] ) , and defining @xmath61 , we obtain the following non - linear set of equations for the evolution of the magnetic field and particle densities : @xmath62 @xmath63 @xmath64 where @xmath65,\end{aligned}\ ] ] and @xmath66.\ ] ] in this one - dimensional geometry @xmath67 , and , because we are considering only variations along @xmath56 , the gradients that appear in the battery terms of eq . ( [ magn ] ) are parallel . thus , there is no contribution from the hall drift and the battery terms to the evolution of the magnetic field . we note that eqs . ( [ 1dmagne ] ) and ( [ nb1d ] ) can be written as flux - conserving equations of the form @xmath68 where @xmath69 represents either @xmath70 or @xmath53 , and @xmath71 is defined in each case by @xmath72 and @xmath73 on the other hand , if we define @xmath74 , we obtain the conservation law @xmath75 we note that @xmath76 is the magnetic flux , while @xmath77 is the baryon number . we see from eq . ( [ conserv ] ) that @xmath78 is a conserved quantity if @xmath79 , which depends on the boundary conditions imposed on the variables contained in @xmath71 . to conserve both the magnetic flux and the baryon number during evolution , we impose the boundary conditions @xmath80 @xmath81 @xmath82 of course , this restriction does not provide a realistic description of a neutron star , in which magnetic flux can be lost through the boundaries ; however , it allows us to control the precision of the calculation in this one - dimensional case . although this geometry is quite unrealistic , it enables us to study the timescales involved in an analytical and numerical fashion . in the future , when we study this system in higher numbers of dimensions , we will use more appropiate boundary conditions for the magnetic field . since in neutron star conditions , the ratio between the magnetic pressure @xmath83 and the degeneracy pressure of the charged particles is very small , we consider the magnetic field to be a small perturbation on a non - magnetized background in full hydrostatic , chemical , and diffusive quasi - equilibrium . if we label the physical variables characterizing the background by the sub - index @xmath84 , the hydrostatic , diffusive , and chemical equilibria are given by @xmath85 @xmath86 @xmath87 the magnetic field can not force significant displacements of the particles ; we therefore write @xmath88 , where @xmath89 . the chemical potentials are @xmath90 , where @xmath91 , @xmath92 , and @xmath93 , with @xmath94 and the remaining coefficients are calculated from the equation of state . we also use the cowling approximation , neglecting the perturbations of the gravitational potential with respect to the background value , i.e. , @xmath95 . by neglecting terms of second order in @xmath96 and using that @xmath97 , @xmath98 , and the definitions @xmath99 , @xmath100 , @xmath101 , and @xmath102 , we derive the following set of equations , which are nonlinear with respect to the magnetic field @xmath103 , but linear with respect to the density perturbations @xmath104 and @xmath105 , namely , @xmath106 @xmath107 @xmath108 where @xmath109 , \end{aligned}\ ] ] and @xmath110 . \end{aligned}\ ] ] we proceed in writing the set of eqs . ( [ bz ] ) , ( [ nb ] ) , ( [ nc ] ) , ( [ vn ] ) , and ( [ va ] ) in terms of dimensionless variables . we start by defining dimensionless variables as @xmath111 , with the sub - index @xmath112 that represents a characteristic value of the corresponding dimensional variable . in the following , we explain the meaning of each of the characteristic values . we normalize the space variable @xmath56 with respect to the length of the system , that is the length of our computational domain @xmath113 , thus @xmath114 . we note , however , that there is another characteristic length @xmath115 , which is the length over which the functions vary spatially ; we therefore use this scale when dealing with order - of - magnitude estimates of spatial derivatives . the time variable @xmath116 is normalized with respect to @xmath117 . the meaning of this quantity will become clearer in the following section ; it is related to the shortest relevant timescale of the system . we assume that @xmath118 is the maximum value of the initial magnetic field , @xmath119 is the maximum neutron background density , @xmath120 is the maximum of the charged background density , @xmath121 and @xmath122 are properties of the background , @xmath123 is the maximum resistivity in the background . from the third term on the rhs of eq . ( [ vn ] , we estimate a characteristic order - of - magnitude value for the neutron velocity induced by the magnetic stress to be @xmath124 . if we compare the first and third parentheses on the rhs of eq . ( [ vn ] ) , and use that close to magnetohydrostatic quasi - equilibrium @xmath125 , @xmath126 , and @xmath127 , then we can estimate a characteristic neutron - density perturbation to be @xmath128 . finally , defining @xmath129 and @xmath130 where @xmath45 is any of the corresponding parameters defined in the last section , we obtain the following dimensionless set of equations , @xmath131 @xmath132 @xmath133 where @xmath134 , \end{aligned}\ ] ] @xmath135 . \end{aligned}\ ] ] in the above set of equations , the parameter @xmath136 controls the coupling strength between the magnetic field and the particle species , @xmath137 controls the importance of the ohmic diffusion , @xmath138 controls the weak interaction rate , and finally @xmath139 controls the importance of the ambipolar diffusion process . in diffusive quasi - equilibrium , ( [ vn ] ) with @xmath140 , and eq . ( [ va ] ) with @xmath141 , imply that @xmath142 where we used that the fraction @xmath143 . for instance , in the non - interacting particle limit of gr-92 this fraction is approximately equal to @xmath144 since the chemical potential of each particle species @xmath15 is a function only of its number density @xmath145^{2/3}}$ ] , which implies that @xmath146 , @xmath147 , @xmath148 , and @xmath149 . on the other hand , for the background numerical values that we use in this paper this fraction is @xmath150 . these background parameters are extracted from the equation of state provided by akmal et al . ( @xcite ) , evaluated at a neutron number density @xmath151 which corresponds to a charged - particle number density of @xmath152 and to a total mass density ( including neutrons , protons and electrons ) of @xmath153 for these values , @xmath154 , @xmath155 , @xmath156 , @xmath157 , and @xmath158 . in addition , the linearization with respect to the variable @xmath159 requires that @xmath160 , i.e. , @xmath161 and therefore @xmath162 we consider a dynamical system of three independent variables controlled by three differential equations that are first order in time . thus , in the linear limit , the system will have three exponentially decaying `` modes '' of different timescales . although , this is not strictly true in the general nonlinear case , one can always identify three characteristic timescales on which the system approaches successive quasi - equilibrium states . real neutron stars approach a magnetohydrostatic quasi - equilibrium on a timescale not much longer than the alfvn time , which is of the order of seconds . here , as described in sect . [ phys ] , this early evolution is mimicked by a artificial friction force term proportional to a parameter @xmath25 ; this parameter is chosen so that this timescale is sufficiently long enough for the numerical code to be able to deal with it , but shorter than the timescales of modeled processes , such as ambipolar diffusion and weak interactions . in what follows , we make analytic estimates of the characteristic evolutionary timescales and evaluate them for typical magnetar core parameters ( see e.g. , arras et al . @xcite ) . in doing this , we can make sense of our numerical results and estimate the order of magnitude of the timescales involved in a real system , under the limitations of our one - dimensional model . we assume that our non - magnetized background star is in hydrostatic quasi - equilibrium [ see eqs . ( [ hydro1 ] ) and ( [ hydro2 ] ) ] , in other words , the net force on a fluid element containing all species is zero . when a magnetic field is present in the system , the magnetic force applies pressure to the charged particles ( electrons and protons ) , inducing density perturbations that create an imbalance between the different forces acting on a fluid element [ see eq . ( [ vn ] ) ] . the particles must move until a magnetohydrodystatic ( mhs ) quasi - equilibrium is reached . the neutron velocity necessary to achieve this balance is expressed in eq . ( [ vn ] ) . during the first stages of evolution , the collisional coupling between the charged particles and neutrons compels them move with about the same velocity , @xmath163 . neglecting weak interactions , from eqs . [ nbnorm ] and [ ncnorm ] we obtain the consistency condition @xmath164 the induced charged - particle and neutron - pressure gradients tend to choke the magnetic force . the magnetohydrostatic quasi - equilibrium state is reached when there is a close balance between these opposing forces [ see rhs of eq . ( [ vn ] ) ] , that is , @xmath165 where we neglected the second term in the rhs of eq . ( [ vn ] ) since @xmath166 and used the eq . ( [ mhdc ] ) . the velocity induced by the initially - unbalanced magnetic - pressure gradient is @xmath167 on the other hand , from eq . ( [ nb ] ) , we obtain the time required to create ( or destroy ) a perturbation @xmath168 as @xmath169 using eqs . ( [ condi1 ] ) , ( [ condi2 ] ) , and ( [ nn ] ) , and @xmath170 we estimate the timescale to reach the magnetohydrostatic quasi - equilibrum as @xmath171 thus , @xmath172 . in eq . ( [ tnn ] ) we used that @xmath173 . in fact , in the non - interacting particle limit of gr-92 we have @xmath147 , thus , this fraction is of the order of @xmath174 . for the numerical background values that we use in this paper , this fraction is @xmath175 . we note that the scale @xmath176 is the shortest relevant timescale in the system , and is controlled by the artificial @xmath25 parameter that was introduced in our slow - motion approximation ( sect . [ phys ] ) . real neutron stars evolve to the magnetohydrostatic quasi - equilibrium on a short timescale not much longer than the alfvn time , which for typical magnetar core parameters scales as @xmath177 where @xmath178 denotes the radius of the star in units of @xmath179 and @xmath180 the magnetic field in units of @xmath181 ( this timescale as well as the following ones are evaluated at the typical mass density @xmath182 ) . this timescale is far shorter than the timescales of the processes that promote the long - term evolution of the magnetic field ( see sects . [ tcardiff]-[tmagnf ] ) , which are of the order of years or much longer . a numerical code simulating the evolution on the alfvn timescale would require a time step many orders of magnitude shorter than that required to simulate the long - term evolution in a computational time that is not prohibitively long . we overcome this difficulty by replacing the short - term dynamics by the artificial friction term proportional to the parameter @xmath25 . this parameter is chosen so that the timescale in eq . ( [ tnn ] ) is long enough for the numerical code to be able to deal with it ( and therefore much longer than the alfvn time ) , but shorter than the timescales of the long - term processes that we discuss in the following sections . we now assume that the magnetohydrostatic quasi - equilibrium , discussed in the last section , has been reached . however , the charged particles continue to move relative to the neutrons due to ambipolar diffusion ( subject to collisional drag among different species ) , and they reach diffusive quasi - equilibrium when there is a close balance between the magnetic force and the charged - particle pressure gradients ( see rhs of eq . ( [ va ] ) . by an argument analogous to that in the previous section and using the temperature dependence of the collisional frequencies from ref . @xcite , we obtain the timescale @xmath183 for charged particles to reach diffusive quasi - equilibrium as @xmath184 where @xmath185 and @xmath186 we can write this time in units of our normalization time @xmath187 as @xmath188 we require @xmath189 , therefore @xmath190 . the @xmath191 parameter in eq . ( [ bartnc ] ) , which is inversely proportional to the collisional parameter @xmath192 , controls the timescale on which the charged particles reach the diffusive quasi - equilibrium . if there is a perturbation of the chemical quasi - equilibrium ( @xmath193 ; see rhs of eq . ( [ nc ] ) , the characteristic timescale on which the chemical quasi - equilibrium is restored through weak interactions ( charged particles decaying into neutrons and viceversa ) can be estimated from eq . ( [ nc ] ) . neglecting the first term on the rhs , which takes into account the ambipolar diffusion , comparing the terms with @xmath159 , and using the temperature dependence of the @xmath44 parameter from ref.@xcite , which assumes modified urca reactions ( e.g. , ref . @xcite ) , we obtain @xmath194 if we write eq . ( [ dmu1 ] ) in units of our normalization time , we derive @xmath195 using eq . ( [ bartdmu ] ) , we obtain @xmath196 ; we therefore require @xmath197 , so that @xmath198 . the @xmath199 parameter in eq . ( [ bartdmu ] ) is directly proportional to the weak interaction rate parameter @xmath44 , which controls the timescale on which the chemical quasi - equilibrium is achieved . both weak interactions and ambipolar diffusion processes contribute in general to the decay of the charged - particle density perturbations . the more rapidly - acting processes determine the evolutionary timescale of the charged particles . we propose an approximate general interpolation formula : @xmath200^{-1}\sim \left[\left(\frac{d}{l}\right)^2 \overline { \epsilon}+ \overline{\theta}\right]^{-1}.\ ] ] the timescale on which the magnetic field decays by ohmic diffusion can be estimated from eq . ( [ bz ] ) , neglecting the first term on the rhs and using the temperature dependence of the electrical conductivity from ref . @xcite , as @xmath201 on the other hand , using eq . ( [ tnn ] ) , we obtain @xmath202 the @xmath203 parameter is directly proportional to the parameter @xmath204 and controls the ohmic diffusion timescale . we require @xmath205 , so that @xmath206 . the two terms on the rhs of eq . ( [ magbar ] ) set the two basic timescales on which the magnetic field decays . the first of these terms couples the magnetic - field evolution with the particle dynamics , while the second one gives the magnetic - field evolution due to ohmic diffusion . for the remainder of this section , we estimate the magnetic - field evolution timescale by neglecting the second term ( formally we set @xmath207 or @xmath208 ) . again , it is useful to distinguish between the two opposite extreme regimes discussed in sects . [ tcardiff ] and [ tchemcar ] . in the first regime , the ambipolar diffusion occurs more rapidly than the weak - interaction processes , that is @xmath209 . first , the system reaches the magnetohydrostatic quasi - equilibrium in the short timescale @xmath176 . in the second stage , all particles reach the diffusive quasi - equilibrium in the timescale @xmath183 and there is a close balance between the lorentz force and the charged particle pressure gradient [ see eq . ( [ va ] ) ] . since the magnetic field generates density perturbations in the charged particles but not in the neutrons , it prevents the system from reaching chemical quasi - equilibrium . weak interactions tend to restore the local chemical quasi - equilibrium on a characteristic timescale @xmath210 . this tendency causes a slight reduction of the pressure gradient in the charged particles with respect to the lorentz force , compelling the charged particles and the magnetic flux to move together at a small ambipolar velocity @xmath211 . this movement maintains the charged - particle pressure gradients , which tend to be erased by weak interactions that continue to operate . this interplay of weak interactions and ambipolar diffusion stops only once the pressure and magnetic field gradients disappear . to estimate the timescale of this process , we first note that the ambipolar velocity needed to keep the charged - particle density perturbations stationary with respect to time ( @xmath212 ) can be estimated from eqs . ( [ nc ] ) and ( [ fluctuch ] ) as @xmath213 since we neglect ohmic diffusion , the magnetic - field evolution is governed by the coupling with the charged - particle movement at the ambipolar velocity given by eq . ( [ vaequi ] ) . from eq . ( [ bz ] ) , and using eq . ( [ vaequi ] ) , we can estimate the timescale on which the magnetic field evolves in this case to be @xmath214 using eqs . ( [ fluctuch ] ) and ( [ tb ] ) , we write this timescale in dimensionless form as @xmath215 in this regime , the weak interactions operate very slowly , converting particles of one species into the other ( beta decays ) in a tendency to erase the chemical imbalance . the beta decays perturb the diffusive quasi - equilibrium , promoting magnetic - field evolution by means of ambipolar diffusion in a timescale that is limited by how rapidly the weak interactions operate ( @xmath210 ) . we note that the timescale given by eq . ( [ tb ] ) is the timescale to reach chemical quasi - equilibrium , @xmath216 amplified by the factor @xmath217 ( which is of the order of the ratio of the charged - fluid pressure to the magnetic pressure ) . the dependence of the timescale on this factor is expected since the lorentz force drives the ambipolar diffusion , which maintains the chemical imbalance as long as there is a magnetic - field gradient . in the opposite regime , the weak interactions occur much faster than the ambipolar diffusion process , i.e. @xmath218 . as before , during the first stage of this regime , the system reaches the magnetohydrostatic quasi - equilibrium in the short timescale @xmath176 . at the end of this stage , the gradient of the magnetic pressure is balanced by the combined degeneracy pressure of all particles , with the neutrons providing the main contribution due to the much higher density . during the second stage , chemical quasi - equilibrium is established at every point in the system , coupling the neutron and charged - particle density perturbations . since the charged - particle pressure is far smaller than that of the neutrons , it can not by itself balance the magnetic - pressure gradient , which , according to eq . ( [ va ] ) , causes an ambipolar drift of velocity @xmath219 on the other hand , from eq . ( [ bz ] ) we have @xmath220 , thus , we can estimate the magnetic - field evolutionary timescale in this case to be @xmath221 if we use eq . ( [ bartnc ] ) , we obtain @xmath222 in this case , the beta decays restore the chemical quasi - equilibrium quickly but the collisions between particles are frequent . this prevents the particles from moving easily relative to each other to achieve diffusive quasi - equilibrium . in this regime , the main mechanism promoting magnetic - field evolution is ambipolar diffusion . this process carries the magnetic field with a velocity that is limited by the collision rate between particles and depends on the magnetic - field strength . equation . ( [ tambi22 ] ) gives the magnetic - field evolutionary timescale in this regime , which is that required for the particles to reach diffusive quasi - equilibrium , @xmath223 but , as above , amplified by the factor @xmath217 . we conclude that , in general , where both weak interations and ambipolar diffusion contribute to the magnetic - field evolution , the slower of these processes determines the evolutionary timescale of the magnetic field . for a general estimate , we use the approximate interpolation formula that recovers the timescales of the regimes discussed above to the corresponding limits @xmath224 equation ( [ gre92 ] ) corresponds to the estimate of eq . ( 35 ) in gr-92 in the limit of non - interacting particles . for a case that includes ohmic diffusion , the shorter of the timescales given in , eqs . ( [ tresis ] ) and ( [ gre92 ] ) provides the timescale for magnetic - field evolution , which we estimate using the interpolation formula @xmath225^{-1},\ ] ] that can be written in dimensionless form as @xmath226\right\}^{-1}+ \left(\frac{d}{l}\right)^2\overline \omega \right]^{-1}.\ ] ] the ordering between @xmath227 and @xmath228 depends on the relative size of @xmath203 and the other dimensionless parameters . we can place the relevant timescales in increasing order as @xmath229 @xmath230 @xmath231 where @xmath232 . it is easy to show that , for @xmath207 , the ordering between the timescales is @xmath233 , @xmath234 , and @xmath235 . the set of eqs . ( [ magbar ] ) , ( [ nbnorm ] ) , and ( [ ncnorm ] ) is nonlinear with respect to the magnetic field variable @xmath70 . to find a linear solution of this set , we assume that the properties of the background star are homogeneous , i.e. , quantities with sub - index zero do not depend on position , and we linearize the magnetic field as @xmath236 where @xmath237 is a constant and @xmath238 is a small magnetic perturbation . if we choose @xmath239 as the magnetic - field normalization unit ( i.e. , @xmath240 ) , we obtain @xmath241 we find normal - mode solutions for the dynamical variables in this linear system as @xmath242 @xmath243 @xmath244 where @xmath245 is an integer labeling each mode , @xmath246 is the decay time of each mode , @xmath247 is an integer measuring the mode wave - number , and @xmath248 is the amplitude of the corresponding variable @xmath249 in each mode . we note that the function @xmath250 satisfies the boundary conditions imposed in the previous section and that the characteristic length for spatial variations in this normal - mode solution is @xmath251 . we note that the linear decay times @xmath252 correspond to the estimated characteristic timescales @xmath253 in the general nonlinear regime [ see eqs . ( [ torder1 ] ) , ( [ torder2 ] ) , and ( [ torder3 ] ) ] , and we use the same notation in the remainder of this paper . the normal - mode solution translates into an eigenvalue problem that can be solved either numerically or using analytical approximations . in figs . [ t1]-[t3 ] , we compare the decay times of the normal modes obtained from an analytical approximation to the eigenvalue problem ( open triangles ; see appendix [ apa ] ) with those obtained from a numerical solution of the eigenvalue problem ( black diamonds ) . we observe that in most cases there is good agreement between these two methods . figures . [ t1](b ) and [ t1](d ) show a discrepancy between these two methods , which is more noticeable for control parameters ( @xmath254 , @xmath191 , @xmath199 and @xmath203 ) of order of 1 . this is expected since the analytical approximation to the eigenvalue problem should be accurate for control parameters significantly less than 1 ( see appendix [ apa ] ) . in figs . [ t1]-[t3 ] we also plot the evolutionary timescales obtained from the order - of - magnitude estimates in sect . [ carsca ] ( open circles ) [ see eqs . ( [ torder1 ] ) , ( [ torder2 ] ) and ( [ torder3 ] ) ] with numerical coefficients inserted by hand to ensure reasonable agreement with the other , more precise determinations , @xmath255 @xmath256^{-1},\end{aligned}\ ] ] @xmath257^{-1}\right.\nonumber\\ & & \left.+\left[\left\{\frac{1}{7.5\overline \upsilon}\left(\frac{1}{(l\pi)^2 \overline \epsilon}+\frac{1}{2.5\overline \theta}\right)\right\}^{-1}+0.81(l\pi)^2 \overline \omega \right]^{-1}\right].\end{aligned}\ ] ] the fact that this agreement can be reached using coefficients that do not differ significantly from unity ( see eqs . ( [ torder1lin ] ) , ( [ torder2lin ] ) and ( [ torder3lin ] ) ) corroborates the order - of - magnitude estimates in the framework of the linear - mode regime . we developed a numerical code based on a finite - difference algorithm that solves the nonlinear set of eqs . ( [ magbar ] ) , ( [ nbnorm ] ) , and ( [ ncnorm ] ) ( see appendix [ apb ] ) . in this section , we use that code to simulate the evolution of a linear magnetic - field perturbation of the form given by eq . ( [ bpert1 ] ) , as well as the associated particle - density evolution . in this task , we aim to achieve two main goals : 1 . to verify the quasi - equilibrium states that occur during the evolution . 2 . to compare the linear normal - mode solution given by eqs . ( [ nm1 ] ) , ( [ nm2 ] ) , and([nm3 ] ) with the solution obtained by solving numerically the set of eqs . ( [ magbar ] ) , ( [ nbnorm])and ( [ ncnorm])(see appendix [ apb ] ) , to measure the precision of the numerical code . in our numerical simulation , we set the initial condition to correspond to a magnetic field of the form given by eq . ( [ bpert1 ] ) , with @xmath258 , @xmath259 and @xmath260 . we set the initial particle density perturbations to be @xmath261 . in what follows , we neglect the ohmic diffusion , therefore , we expect that @xmath233 , @xmath234 , and @xmath235 . we use the parameters : @xmath207 , @xmath262 , @xmath263 , @xmath264 . we note we are in a regime in which ambipolar diffusion occurs more rapidly than the weak interaction processes ( @xmath265 and @xmath266 ) . from a numerical solution of the eigen - value problem [ see eq . ( [ eigen ] ) ] , we obtain @xmath267 , @xmath268 , and @xmath269 . in fig . [ tmhd ] , we show the system s evolution during the first timescale @xmath270 , which is the timescale over which the system can reaches magnetohydrostatic quasi - equilibrium . we label the different instants of this evolution with progressive numbers as described in the figure . this magnetic perturbation perturbs the hydrostatic quasi - equilibrium of the homogeneous background star ; thus particles are therefore compelled to move in order to compensate for this imbalance , and reach magnetohydrostatic quasi - equilibrium in a short timescale of the order of @xmath270 . in fig . [ tmhd ] ( a ) we observe the evolution of the magnetic - field perturbation during this short timescale , and that it has not evolved significantly . this is expected since the magnetic pressure is small compared with the background - degeneracy pressure , and can therefore induce only a small relative density perturbation in the particles ( see figs . [ tmhd ] ( b ) and [ tmhd ] ( c ) ) . we can also see in figs . [ tmhd ] ( b ) and [ tmhd ] ( c ) that during this short timescale @xmath271 , as expected from eq . ( [ mhdc ] ) . in fig . [ tdif ] , we show the system s evolution during the second timescale @xmath272 , on which the diffusive quasi - equilibrium of the charged particles is attained . at the end of this timescale , @xmath273 , and @xmath274 has grown significantly to balance the magnetic pressure . towards the end of this timescale , the separation between the lines is narrower , showing that the charged particles have reached their diffusive quasi - equilibrium . in fig . [ tchem ] , we show the system s evolution during the third timescale @xmath275 , on which the chemical quasi - equilibrium is restored and the magnetic perturbation decays . at the end of this timescale , the density perturbations as well as the magnetic - field perturbation have decreased substantially . on the other hand , we note that a linear magnetic - field perturbation can be written in general , as a superposition of the linear normal modes given by eq . ( [ nm3 ] ) as @xmath276 where @xmath277 and the coefficients @xmath278 are calculated from the initial conditions @xmath279 , @xmath280 , and @xmath281 . for the initial conditions of this simulation and from the numerical solution of eq . ( [ eigen ] ) , we derive that @xmath282 , @xmath283 , @xmath284 , @xmath285 , @xmath286 and @xmath287 . we note that , for @xmath288 , @xmath289 on the other hand , the total magnetic energy associated with this perturbation in physical units is @xmath290 thus , in the asymptotic limit of eq . ( [ asymmm ] ) , @xmath291 where @xmath292 is the initial energy contained in the third mode . to measure the precision of our numerical code , in fig . [ linen ] we compare the evolution of the magnetic - energy perturbation obtained from the simulation with the asymptotic limit of eq . ( [ asymagneenergl ] ) . in fig . [ linen ] , we see that there is good agreement between our numerical result and the asymptotic limit of eq . ( [ asymagneenergl ] ) . using our numerical code , we simulate the evolution of a non - linear magnetic field , by applying the set of eqs . ( [ magbar ] ) , ( [ nbnorm ] ) , and ( [ ncnorm ] ) without any linearization of @xmath293 . the main aims of this section are : 1 . to compare the estimate obtained using eq . ( [ tbch ] ) with the result of the numerical code in the non - linear regime . 2 . to verify the conservation laws given by eq . ( [ conserv ] ) for the magnetic flux and baryon number , which provide a measure of the right performance of our numerical code . defining the initial conditions of our simulation , we set the particle densities , @xmath261 , for an initial gaussian magnetic field profile given by @xmath294 we normalize the magnetic field in eq . ( [ nonlinerab ] ) to its maximum value @xmath295 . the parameter @xmath112 in eq . ( [ nonlinerab ] ) is related to its standard deviation @xmath296 by @xmath297 , and @xmath298 is the center of the gaussian function . the characteristic spatial variation in the gaussian function can be estimated by the width @xmath299 of the gaussian function at its half - height , which is given by @xmath300 ; we can therefore approximate the characteristic length over which the magnetic force varies to be @xmath301 . in our analysis , we again neglect ohmic diffusion . to estimate the characteristic decay time of the magnetic field in eq . ( [ nonlinerab ] ) , we use as a first approximation the characteristic length of the magnetic force in eq . ( [ tbch ] ) as the width of the initial gaussian , therefore , @xmath302 in our numerical analysis , we assume that @xmath303 , @xmath304 , @xmath207 , @xmath305 , @xmath306 , and @xmath307 . if we substitute these parameters in eq . ( [ tbchnol ] ) , we obtain a characteristic decay time @xmath308 . in fig . [ gauss ] ( a ) , we see the evolution of the initial magnetic field given by eq . ( [ nonlinerab ] ) , and that at times close to @xmath228 the maximum of the magnetic field has decayed to @xmath309 of its initial amplitude . on the other hand , fig . [ gauss ] ( b ) shows the evolution of the charged - particle density perturbations , and fig . [ gauss ] ( c ) shows the evolution of the neutron density perturbations . we verified the conservation law for the dimensionless baryon number @xmath310 , where @xmath311 was the backgound baryon number . according to the initial conditions @xmath312 which implies that the conservation law requires @xmath313 the calculated values of the baryon number in the simulation time steps show a maximum percentage error of @xmath314 with respect to @xmath315 . regarding the magnetic flux , we found exact conservation with respect to the initial magnetic flux @xmath316 down to 7-digit precision . we have studied the long - term evolution of the magnetic field and the densities of neutrons and charged particles ( protons and electrons in the interior of a neutron star , using a multi - fluid model with a simplified geometry in which the magnetic field points in one cartesian direction and varies along an orthogonal direction . we found a set of three non - linear partial differential equations of first order in time describing this evolution , and we analyzed them in three different ways : ( i ) estimating evolutionary timescales directly from the equations , guided by physical intuition ; ( ii ) a normal - mode analysis of the equations in the limit of a nearly uniform system ; and ( iii ) a finite - difference numerical integration of the equations . in this section , we summarize the main results of each one of these approaches and present the main conclusions of this work . the three partial differential equations of our model constitute a dynamical system of three independent variables ( magnetic - field , charged - particle , and neutron density perturbations ) . we identified three characteristic evolutionary timescales on which the system approaches successive quasi - equilibrium states . in the following , we summarize the basic physical processes governing the evolution and the estimates of each timescale . during the early stages of a neutron star s life , the lorentz force moves the bulk stellar fluid , inducing perturbations in its pressure . in this short time , all particles move together as a single fluid with the same bulk velocity . there is also not enough time for weak interactions between particles ( beta decays ) to operate , so the composition is frozen . the system evolves until it reaches a magnetohydrodystatic quasi - equilibrium state , in which the lorentz and the fluid forces are close to balancing . neutron stars with no rotation or convection reach this quasi - equilibrium state in a short time , not much longer than the alfvn time , which for typical magnetar core parameters scales as @xmath317 where @xmath178 denotes the radius of the star in units of @xmath318 and @xmath180 the magnetic field in units of @xmath319 on timescales far longer than the alfvn time , relative movements between the different species of particles in the star as well as weak interactions between them become relevant ( gr-92 ) . therefore , a description of the system during these stages requires a multi - fluid theory . in this paper , we develop a model to study the decay of the magnetic field induced by these long - term mechanisms . the alfvn time is far shorter than the timescales of these long - term processes ( see sects . [ conc12 ] and [ conc13 ] ) , which are of the order of years or much more . a numerical code simulating the evolution during the alfvn timescale would require a time step many orders of magnitude shorter than that required to simulate the long - term evolution in a computational time not prohibitively long . in our model , we overcome this difficulty by replacing the short - term dynamics by a ficticious friction force term acting on the neutrons ( the most abundant species ) . its strength is controlled by an artificial parameter @xmath25 , chosen so that the timescale to reach magnetohydrostatic quasi - equilibrium is long enough for the numerical code to be able to deal with it ( and therefore much longer than the alfvn time ) , but shorter than the timescales of the long - term processes that we are interested in modeling . on timescales far longer than required to reach magnetohydrostatic quasi - equilibrium , the neutrons and charged particles move relative to each other with different velocities and are affected by collisional drag . also , weak interactions ( beta decays ) convert particles from one species into another , erasing departures from chemical quasi - equilibrium among neutrons , protons , and electrons caused by the induced particle density perturbations . both processes contribute to the evolution of the particle densities . in estimating the timescale for this evolution , it is useful to distinguish the regimes in which each one of these processes is dominant . if the collisions between the charged particles and the neutrons are rare and the weak interactions are slow , the charged particles move easily relative to the neutrons ( ambipolar diffusion ) allowing the particles to reach a diffusive quasi - equilibrium state ( in which gravitational , electromagnetic , and pressure forces are closely balanced ) but staying avoid chemical quasi - equilibrium . the diffusive quasi - equilibrium is reached in a timescale controlled by the collision rate between neutrons and charged particles given by @xmath320 where @xmath185 and @xmath186 in the opposite regime , when the collisions between charged particles and neutrons are frequent , and the weak interaction rate is high , the relative diffusion of particles will be impeded by the collisional drag , but the system reaches the chemical quasi - equilibrium in a timescale controlled by weak interactions of magnitude ( assuming modified urca reactions ) @xmath321 the ohmic dissipation promotes the decay of the magnetic field in a timescale @xmath322 this extremely long time implies that ohmic dissipation is not very effective in producing magnetic - field decay in neutron stars . a more rapid decay of the magnetic field can occur through ambipolar diffusion , i.e. , a joint drift of the magnetic field and the charged particles relative to the neutrons , in which magnetic energy becomes dissipated by collisions . this diffusion process induces local density perturbations that deviate from chemical quasi - equilibrium . weak interactions tend to erase the chemical imbalance by converting particles of one species into the other . during these conversions , neutrinos , and antineutrinos remove part of the magnetic energy . in estimating the timescale of magnetic field decay induced by these processes , it is useful to distinguish between the two regimes we discussed above . if ambipolar diffusion occurs far more rapidly than the weak interaction process , i.e. , @xmath323 diffusive quasi - equilibrium is quickly reached , but chemical quasi - equilibrium is not . thus , the rate at which chemical quasi - equilibrium is restored by weak interactions determines the characteristic timescale over which the magnetic field evolves , @xmath324 this timescale is the corresponding one to reach chemical quasi - equilibrium @xmath210 but amplified by a factor of the order of the ratio of the charged fluid pressure to the magnetic pressure . this is expected since the lorentz force drives the ambipolar diffusion that prevents weak interactions from restoring chemical quasi - equilibrium . in the opposite regime , collisions are frequent and weak interactions are fast , i.e. @xmath325 so the system reaches chemical quasi - equilibrium long before diffusive quasi - equilibrium . therefore , the evolution of the magnetic field is limited by the collisions that control the ambipolar diffusion , on a timescale of @xmath326 since the magnetic field sustains the diffusive imbalance and drives the diffusion , the timescale given in eq . ( [ tbb2 ] ) is similar to that taken by particles to achieve diffusive quasi - equilibrium but , as above , amplified by a factor that depends on the magnetic field strength . in the limit of non - interacting protons and neutrons studied by gr-92 , eqs . ( [ tbb1 ] ) and ( [ tbb2 ] ) correspond to the estimate of the timescale for magnetic field evolution due to the ambipolar diffusion process of eq . ( 35 ) in gr-92 . we note that gr-92 considered the neutrons to be a static species in diffusive quasi - equilibrium . from this correspondence , we confirm that the motion of the neutrons , which was included in our model , plays no important role in the one - dimensional evolution . in the spirit of finding an analytical solution to the set of non - linear partial differential equations that describes the evolution of the system , we completed a linear perturbation analysis of the equations in the limit of a nearly uniform system and tried a normal mode solution ( sect . [ normmodes ] ) . we found three exponentially - decaying modes that could be interpreted as the episodes of successive relaxation to the quasi - equilibrium states identified in the previous physical analysis ( see sects . [ carsca ] and [ conc1 ] ) . we found analytical and numerical solutions for the decay times of these modes ( see appendix [ apa ] ) that agree with the estimated formulae for the evolutionary timescales of the general non - linear system ( see sects . [ carsca ] , [ conc1 ] and figs . [ t1],[t2 ] , and [ t3 ] ) . we constructed a finite - difference numerical code to solve the full system of non - linear partial differential equations . this was tested against the normal - mode solutions ( fig . [ linen ] ) and by verifying the known conservation laws of magnetic flux and baryon number in highly non - linear situations ( fig . [ gauss ] ) . applying this numerical code , we propose to consider , in the future , conditions in which the background is non - homogeneous , as well as non - linear situations . we have established a general multifluid formalism in which it is possible to study the magnetic field decay processes in neutron stars identified by gr-92 , which are likely to represent the basis of the magnetar phenomenon . as a first step , we have focused on a simplified geometry in which the magnetic field points in one cartesian direction and varies in an orthogonal direction to it . we estimated the timescales of the relevant processes , and using numerical simulations we followed the temporal evolution of some simple magnetic - field configurations . the present work is far from exhausting the possibilities of this formalism , which should be applied to more realistic situations , including neutron stars with spherical symmetry , non - uniform background density and three - dimensional magnetic fields with components that depend on two or three spatial coordinates . further steps might consider convective motions and additional species of particles as well as the effects of superfluidity and superconductivity . the set of eqs . ( [ magbar ] ) , ( [ nbnorm ] ) , and ( [ ncnorm ] ) is , in general , non - linear with respect to the magnetic field variable @xmath70 . to find a linear solution to this set , we linearized the equations respect to the magnetic field , as described in sec . [ normmodes ] . assuming that the properties of the background star are homogeneous , that is quantities with sub - index zero do not depend on position , we obtain the following linear set of equations : where @xmath353 is normalized with respect to @xmath239 , while @xmath354 are both normalized with respect to @xmath355 . if we substitute eq . ( [ norm1 ] ) into the linear matrix system eq . ( [ setlin ] ) , we obtain the eigenvalue problem we also note that all @xmath370 whereas @xmath371 . in the limit @xmath372 , the eigenvalues are @xmath373 , @xmath374 . in the general case , we can write @xmath375+\varphi , \end{aligned}\ ] ] and treat @xmath376 as small quantities , which we derive below . we calculate the characteristic polynomial of the matrix eq . ( [ matrsm ] ) and compare terms of the same order in this polynomial with those in the equation equations . ( [ val1 ] ) and ( [ val2val3 ] ) give approximate analytical expressions for the decay times of the normal modes [ see eq . ( [ eigentimes ] ) ] as functions of the dimensionless parameters , @xmath392 . equations . ( [ val1 ] ) and ( [ val2val3 ] ) are useful because from them we can understand in an analytical way the relative importance of the different physical processes for the behavior of the decay times . the linear decay times @xmath252 correspond to the estimated characteristic timescales @xmath253 in the general non - linear regime ( see eqs . ( [ torder1 ] ) , ( [ torder2 ] ) , and ( [ torder3 ] ) ) . to solve the dimensionless system of eqs . ( [ magbar ] ) , ( [ nbnorm ] ) , and ( [ ncnorm ] ) numerically we discretize them using an ftcs ( forward time centered space ) first order scheme . we define the spatial and temporal grid by @xmath393 , @xmath394 where @xmath395 and @xmath396 . in this method , we evaluate the spatial derivatives of a quantity @xmath397 at a half point between consecutive grid points , namely , we thank fondecyt for financial support through the postdoctoral project 3060103 and regular projects 1060644 and 1070854 . we also thank the gemini project 32070014 , the eso - chile mixed committee , and the fondap center for astrophysics ( 15010003 ) . j.h . and a.r . are grateful for the hospitality of the max planck institute for astrophysics and the max planck institute for extraterrestrial physics , in garching , germany , where part of this research was developed . we thank the editor of _ astronomy_&_astrophysics _ steven shore and an anonymous referee , whose suggestions and comments were useful to improve the manuscript . 32 bhattacharya d. , 1995 , in x - ray binaries , ed . w. h. g. lewin , j. van paradijs , & e. p. j. van den heuvel ( cambridge : cambridge university press ) , 233 cumming , a. , zweibel , e. , bildsten , l. 2001 , apj , 557 , 958 cumming , a. 2002 , mnras , 333 , 589 payne , d. j. b. & melatos , a. 2004 , mnras , 351 , 569 duncan , r.c . , & thompson , c. 1992 , apj , 392 , l9 thompson , c. & duncan , r.c . , 1996 apj , 473 , 322 arras , p. , cumming , a. , & thompson , c. 2004 , apj , 608 , l49 goldreich , p. , & reisenegger , a. 1992 , apj , 395 , 250 jones , p. b. 1988 , mnras , 233 , 875 rheinhardt , m. , & geppert , u. 2002 , phys . lett . , 88 , 101103 geppert , u. , & rheinhardt , m. 2002 , a&a , 392 , 1015 hollerbach , r. , & rdiger , g. 2002 , mnras , 337 , 216 hollerbach , r. , & rdiger , g. 2004 , mnras , 347 , 1273 rheinhardt , m. , konenkov , d. , & geppert , u. 2004 , a&a , 420 , l33 cumming , a. , arras , p. , & zweibel , e.g. 2004 , apj , 609 , 999 reisenegger , a. , prieto , j.p . , benguria , r. , lai , d. , & araya , p.a : 2005 , in magnetic fields in the universe : from laboratory and stars to primordial structures , aip conference proceedings , vol . 784 , eds . de gouveia dal pino , g. lugones , & a. lazarian , p. 263 reisenegger , a. , benguria , r. , prieto , j.p . , araya , p.a . , & lai , d. 2007 , a&a , 472 , 233 reisenegger , a. , 2007 , an , 328 , 1173 baym , g. , pethick , c. , & pines , d. 1969 , nature , 224 , 872 braithwaite , j. , & spruit , h.c . 2004 , nature , 431 , 819 braithwaite , j. , & spruit , h.c . 2006 , a&a , 450 , 1097 braithwaite , j. , & nordlund , a. 2006 , a&a , 450 , 1077 braithwaite , j. 2008 , mnras , 386 , 1947 akmal , a. , pandharipande , v.r . & ravenhall , d.g . 1998 , phys . . c , 58 , 1804 reisenegger , a. , 1995 , apj , 442 , 749 fernndez , r. , & reisenegger , a. 2005 , apj , 625 , 291 yakovlev , d.g . , & shalybkov , d.a . , 1990 , soviet astron.lett . , 16 , 86 sawyer , r.f . 1989 , phys . d. , 39 , 3804 shapiro , s.l . , & teukolsky , s.a . 1983 , black holes , white dwarfs , and neutron stars ( new york : wiley ) haensel , p. , urpin , v.a . , & yakovlev , d.g . 1990 , a&a , 229 , 133

we conduct a systematic examination of the properties of models for chemically homogeneous , differentially rotating , main - sequence stars of mass . the models were constructed using a code based on a reformulation of the self - consistent field method of computing the equilibrium stellar structure for a specified conservative internal rotation law . the code has recently been upgraded with the addition of new opacity , equation of state , and energy generation routines , and a mixing - length treatment of convection in the outer layers of the stellar interior . relative to nonrotating stars of the same mass , these models all have reduced luminosities and effective temperatures , and flattened photospheric shapes ( i.e. , decreased polar radii ) with equatorial radii that can be larger or smaller , depending on the degree of differential rotation . for a fixed ratio of the axial rotation rate to the surface equatorial rotation rate , increasingly rapid rotation generally deepens convective envelopes , shrinks convective cores , and can lead to the presence of a convective core ( envelope ) in a ( ) model , a feature that is absent in a nonrotating star of the same mass . the positions of differentially rotating models for a given mass in the h - r diagram can be shifted in such a way as to approximate the nonrotating zams over ranges in luminosity and effective temperature that correspond to a mass interval between and about 0.7 . we briefly note a few of the implications of these results , including ( i ) possible ambiguities arising from similarities between the properties of rotating and nonrotating models of different masses , ( ii ) a reduced radiative luminosity for a young , rapidly rotating sun , ( iii ) the nuclear destruction of lithium and other light metallic species in the layers beneath an outer convective envelope , and ( iv ) , the excitation of solar - like oscillations and the operation of a solar - like hydromagnetic dynamo in some stars .

0704.1275

rotation is a universal stellar physical attribute . this conclusion is supported by an enormous body of observational data , accrued over a period of nearly 100 years , from which information about rotational speeds and periods , the dependence of these quantities on parameters such as mass and age ( i.e. , evolutionary state ) , and the effects of rotation on the shape , effective temperature , chemical homogeneity , and other basic properties of stars have been inferred . despite increasing evidence that rotation can have a significant impact on a variety of stellar characteristics , it is generally not included as a component of the structural / evolutionary models which are the primary tools for the interpretation of observations . this omission is , in part , a consequence of the increased complexity of the problem of determining the structure and evolution of a rapidly , differentially rotating star : a one - dimensional model necessarily becomes two - dimensional , the gravitational potential must be derived by solution of poisson s equation , and the uncertain physics of convective and circulatory flows , along with other rotation - dependent mechanisms that contribute to angular momentum redistribution and chemical mixing , needs to be addressed . some progress toward the development of a straightforward yet robust technique for computing the internal structure of a rotating star has recently been made with the implementation of a new version of the self - consistent field ( scf ) method ( jackson , macgregor , & skumanich 2005 , hereafter paper i ) . in its original form ( ostriker & mark 1968 ; see also jackson 1970 ) , the scf approach to computing a model for an assumed conservative law of rotation consisted of two separate steps : ( i ) determination of the gravitational potential for a given distribution of the mass density @xmath5 , followed by ( ii ) solution of the equations of stellar structure with the potential from ( i ) to update the equilibrium distributions of @xmath5 and other quantities . the process was initiated with a trial distribution for @xmath5 , and the two steps were executed sequentially and iterated upon until the densities from ( i ) and ( ii ) agreed to within a specified tolerance . while the scf method was successfully used to construct detailed models of rotating upper - main - sequence stars ( bodenheimer 1971 ) , it failed to converge for objects less massive than about 9 @xmath6 , behavior which precluded its application to intermediate- and low - mass stars ( see , e.g. , chambers 1976 ; clement 1978 , 1979 ; paper i ) . as described in paper i ( see also 2 below ) , the reformulated scf method circumvents the difficulties responsible for nonconvergence in lower - mass , more centrally condensed stars with an approach that entails the specification of trial functions for the pressure @xmath7 , the temperature @xmath8 , and the shape of constant - density surfaces , together with iterative adjustment of both the profiles and central values of @xmath7 and @xmath8 . with these modifications , the method is capable of producing self - consistent models for rotating main - sequence stars of all masses . it has been validated through detailed comparisons with stellar models ( both rotating and nonrotating ) for masses @xmath9 computed by other investigators using alternative techniques ( paper i ) , and has been applied in an examination of the photospheric shape of the be star achernar , an object revealed by interferometric observations as highly flattened by rapid rotation ( jackson , macgregor , & skumanich 2004 , and references therein ) . more recently , the structure code based on the method has undergone considerable renovation , with the replacement of routines for the equation of state , nuclear energy generation rates , and opacities , and the addition of a mixing length treatment of convection . it is now equipped for use in an investigation of the effects of rotation on the structure and properties of stars with masses @xmath10 . in the present paper , we use the formalism described above to conduct a detailed survey of the structural characteristics of differentially rotating stars having masses @xmath3 in the range @xmath11 . in the absence of rotation , this mass interval encompasses considerable variation in internal properties , with the gross structure of objects at the lower limit consisting of an inner radiative core and an outer convective envelope , changing to a convective core and radiative envelope for objects at the upper limit . the influence of rotation on the basic morphology of stellar interiors for these masses has not received much attention , with rotation - related effects usually treated as perturbations to the nonrotating structure , if at all ( see , e.g. , thompson et al . however , observations indicate the occurrence of surface rotation speeds rapid enough to imply non - negligible modifications to many stellar properties , particularly if the rotation is differential : the projected rotation speeds of 2 @xmath6 main - sequence stars are typically in excess of 100 km s@xmath12 , and comparable values have been measured for near zero - age main - sequence ( zams ) 1 @xmath6 stars in young clusters ( stauffer 1991 ; wolff & simon 1997 ; tassoul 2000 , and references therein ) . motivated by these considerations , we have constructed an extensive set of self - consistent models for @xmath13 zams stars , to systematically study the dependence of stellar characteristics on the rate and degree of differential rotation . our scf models extend previous computational results for stars in this mass range , some of which were obtained for uniformly rotating configurations ( faulkner , roxburgh , & stritmatter 1968 ; sackmann 1970 ; kippenhahn & thomas 1970 ; papaloizou & whelan 1973 ; roxburgh 2004 ) , while others were obtained using either approximate methods for nonconservative differential rotation ( endal & sofia 1981 ; pinsonneault et al . 1989 ; eggenberger , maeder , & meynet 2005 ) or a non - scf , finite - difference technique in the case of conservative rotation ( clement 1979 ) . in 2 we provide a synopsis of the new scf method , briefly describing its implementation in a code for computing the structure of chemically homogeneous , differentially rotating main sequence stars , the improvements and extensions to the input physics that have been made since paper i , and the results of tests of code reliability through comparisons with extant models for nonrotating zams stars with masses @xmath14 . an examination of the properties of models spanning a wide range of internal rotation characteristics is presented in 3 , with particular attention paid to the behavior of such quantitites as luminosity and effective temperature , central thermodynamic properties , the location and extent of convective regions , and the size and shape of the stellar photosphere . among the consequences of these rotation - induced changes in stellar properties is a shift in the position of objects in the classical h - r diagram ( hrd ) ; indeed , for some adopted rotation laws , the resulting structural modifications can enable a differentially rotating star to have the same @xmath15 and @xmath16 values as a non - rotating star of significantly lower mass and to thereby occupy the same position in the hrd . in the concluding section of the paper ( 4 ) , we use the model results to consider how the changes brought on by rapid differential rotation might affect other features of main - sequence stars in this mass range , such as the radiative luminosity of the young sun and associated effects on the planets , the abundance of lithium in the surface layers of mid - f dwarf stars , and the presence of strong , large - scale magnetic fields in some a stars . the new version of the scf method ( described in detail in paper i ) is an iterative scheme that is initialized by specifying ( i ) a pair of one - dimensional trial functions , one for the temperature distribution and one for the pressure distribution ( each normalized by its central value and defined over a spatial range that is normalized by the equatorial radius of the star ) , and ( ii ) a two - dimensional normalized function describing the shape of the equidensity surfaces . the normalized two - dimensional trial density distribution , which is used as the source term in poisson s equation for the gravitational potential , follows from the equation of state and the trial functions for @xmath7 and @xmath8 . for a conservative law of rotation , in which the angular velocity depends only on the perpendicular distance from the axis of rotation , it is possible to define an effective potential from which both the gravitational and centrifugal forces can be derived . the surfaces of constant effective potential ( i.e. , level surfaces ) can then be identified and used to solve a set of ordinary differential equations analogous to the usual equations of stellar structure for nonrotating stars . this solution step yields updated temperature and pressure distributions , allowing the iterative cycle to be repeated and the process continued until convergence is achieved . when the input and output functions representing the normalized temperature and pressure profiles are in agreement , the two parameters corresponding to the central temperature ( @xmath17 ) and central pressure ( @xmath18 ) are adjusted by a newton - raphson technique to bring them closer to the actual physical conditions at the center of the final equilibrium model . the entire procedure ( consisting of an scf loop nested inside a newton - raphson loop ) is repeated until an acceptable level of agreement between the input and output values of the two central parameters is attained . when used to construct models for differentially rotating main - sequence stars , this reformulation of the scf method has been shown to converge for all masses in the range @xmath19 , and for values of the dimensionless rotational kinetic energy @xmath20 ( the ratio of the rotational kinetic energy to the absolute value of the gravitational potential energy of the configuration ) as high as 0.100.12 for intermediate- and high - mass models , and up to nearly 0.26 for some fully convective , highly flattened , disk - like , @xmath1 models . by comparison , the largest @xmath20 values among the non - scf models computed by clement ( 1978 , 1979 ) were @xmath21 for 30 @xmath6 and @xmath22 for models in the range @xmath23 , whereas bodenheimer ( 1971 ) obtained a 60 @xmath6 scf model with @xmath24 . as in paper i ( see also jackson , macgregor , & skumanich 2004 ) , the internal rotation of each of the stellar models discussed in 3 is given by an angular velocity distribution of the form ( ) = _ 0 1 + ( / r_e ) ^2 , where @xmath25 , and @xmath26 is the equatorial radius of the star . for a given model , the constants @xmath27 and @xmath28 are prescribed parameters that characterize , respectively , the axial rotation rate and the ratio of the surface equatorial rotation rate to the axial rate ( a measure of the degree of differential rotation ) , @xmath29 . in practice , the value of the first of these quantities is specified through the parameter @xmath30 , with @xmath31 the equatorial angular velocity for which the magnitudes of the gravitational and centrifugal forces at @xmath26 are equal . rigorous answers to questions about the existence and uniqueness of solutions to complicated systems of integro - differential equations are generally very difficult to obtain . although specifying the two rotational parameters @xmath28 and @xmath32 for chemically homogeneous models of fixed mass and composition does not always lead to a converged model , when it does , our experience with the code suggests that the model is unique . for a given set of @xmath33 values , using the scf code to converge models of the same mass and chemical composition from two or more _ different _ trial models always seems to lead to the _ same _ final converged model . on the other hand , it can be demonstrated that specifying the parameters @xmath28 and @xmath27 does not , in general , lead to a unique solution . a minor drawback to use of the parameters ( @xmath28 , @xmath32 ) in presenting the results is that several important global properties , including the luminosity @xmath16 , the total angular momentum @xmath34 , and the dimensionless rotational kinetic energy @xmath20 , are not monotonic functions of @xmath32 when @xmath28 is held fixed ( see also 3.1 ) . despite this shortcoming , we feel that the apparent uniqueness of the models corresponding to particular @xmath33 values justifies the use of these quantities in presenting and discussing results . we emphasize that sequences of models obtained by holding @xmath28 fixed and varying @xmath32 should not be interpreted as any sort of evolutionary sequence . the significance of models characterized by the same value of @xmath28 is that they have the same degree of differential rotation , that is , the same @xmath35 . the `` half - width '' of the rotation profile , @xmath36 , does , however , change from model to model for constant @xmath28 . since the publication of paper i , the input physics for the scf code described therein has been updated considerably , with the installation of software components that at various times were parts of the stellar - evolution code developed by don vandenberg at the university of victoria . all of the models presented in this paper were computed for the following abundances by weight of h , he , and heavy elements : @xmath37 , @xmath38 , and @xmath39 . the opacities were obtained , as in vandenberg et al . ( 2000 ) , from tables of opal opacities calculated by rogers & iglesias ( 1992 ) and from tables of low - temperature opacities calculated by alexander & ferguson ( 1994 ) , using interpolation subroutines written by vandenberg ( 1983 ) . other subroutines written by vandenberg ( 1992 ) were utilized for the following : ( i ) the equation of state formulated by eggleton , faulkner , & flannery ( 1973 , eff ) ; ( ii ) nuclear energy generation rates for hydrogen burning from caughlan & fowler ( 1988 ) , including the effect of electron screening as treated by graboske et al . ( 1973 ) for the case of equilibrium abundances of cno isotopes ; and ( iii ) a standard mixing - length treatment of surface convective zones ( see , e.g. , baker & temesvary 1966 ; kippenhahn , weigert , & hofmeister 1967 ) . we have made no attempt to incorporate any of the direct effects of rotation into the adopted convection model ; instead , we have simply modified the nonrotating mixing - length description of convection by replacing the local gravitational acceleration , @xmath40 , with the effective gravitational acceleration , @xmath41 ( i.e. , @xmath40 as reduced by the local centrifugal acceleration ) , averaged over equipotential surfaces . for all of the models , a value of 1.9 for the ratio of the mixing length to the pressure scale height has been adopted . in view of the revisions and updates that have been made to the scf code of paper i , it seems worthwhile to make a careful comparison of our nonrotating models for stars on the lower main sequence with models for the same mass obtained from a standard ( nonrotating ) stellar evolution code for this purpose , we have used the current version of the evolutionary code of christensen - dalsgaard ( 1982 , hereafter referred to as the jcd code ) to generate two evenly spaced sequences of seven models each , spanning the mass range @xmath42 along the zams . these models have chemical composition @xmath43 , @xmath44 , quite close to the abundances adopted for our scf models , and were computed using the same value ( 1.9 ) of the mixing length parameter . the basic eff equation of state was used to construct one of the sequences , while the other was derived with the so - called ceff equation of state , a modification of the eff treatment that includes coulomb corrections ( see , e.g. , christensen - dalsgaard & dappen 1992 ) . aside from the inclusion or omission of coulomb effects in the equation of state , the package of input - physics routines used to generate the jcd models is very similar to that installed in the scf code ( see , e.g. , christensen - dalsgaard , proffitt , & thompson 1993 ; di mauro & christensen - dalsgaard 2001 ) . in addition to the intrinsic distinctions between the mathematical techniques used to obtain the two sets of models , the current version of the scf code differs from the jcd code in the following ways : ( i ) the equation of state does not include treatment of coulomb corrections ; ( ii ) the photospheric pressure in the scf code follows from the application of a different , simplified boundary condition ( see paper i ) ; and ( iii ) , there is a slight difference in the vintage of the nuclear energy generation data . we have conducted a quantitative comparison of some of the important properties ( @xmath45 , @xmath16 , @xmath15 , @xmath18 , @xmath46 , @xmath47 ) of scf , jcd / eff , and jcd / ceff models for nonrotating zams stars of the same mass . a theoretical hrd indicating the positions of the various models is displayed in figure 1 . the eff and ceff main sequences ( the dashed and dotted lines , respectively ) are very nearly coincident , with the former models shifted shifted along the locus relative to the latter models , toward somewhat lower luminosities and effective temperatures . on the basis of the preceding discussion , we expect the scf models to agree better with eff models than with ceff models , and inspection of the results plotted in figure 1 indicate that this is generally the case . for the luminosity , the most sensitive of the stellar properties , the largest discrepancies occur for the @xmath48 models , with the scf @xmath16 about @xmath49 lower than that of the eff model of the same mass ; this latter value is , in turn , about @xmath50 lower than @xmath16 for the corresponding ceff model . the luminosities of the scf and eff models are essentially equal for @xmath2 , while a comparison of the eff and ceff models for that mass reveals that the ceff model is about @xmath51 more luminous . the effective temperatures and radii of the scf models deviate from those of the corresponding eff models by less than @xmath52 , except for those models between 1.2 and 1.4 @xmath6 where the the magnitudes of the discrepancies are somewhat larger , @xmath53 . the scf values for @xmath18 and @xmath47 differ from the eff values of those quantities by about @xmath54 , with the relative difference in the @xmath17 values @xmath55 . there is also good qualitative agreement between the scf and jcd models with respect to the appearance or absence of convective cores and envelopes , and there is good quantitative agreement with respect to the radial extent and the enclosed mass of the principal radiative - convective interfaces . the properties of scf models for rotating zams stars with masses @xmath56 are summarized in figures 2 - 8 and table 1 . to facilitate discussion of the computed stellar characteristics , we adopt the convention , established in paper i , of identifying each model by its mass , @xmath3 , and the two rotational parameters , @xmath28 and @xmath32 , for the reasons discussed in 2 . regions of the ( @xmath57 ) parameter space in which converged @xmath58 scf models can and can not be obtained have been illustrated in figure 3 of paper i. while the ( @xmath28 , @xmath32 ) planes for the lower - mass scf models considered here closely resemble that for the @xmath58 models , there are some important differences . of particular relevance to the present paper are the regions corresponding to region ii in paper i , regions of relatively high angular momentum wherein the scf method is incapable of producing converged models . these _ forbidden zones _ are considerably expanded for the 1 and 2 @xmath6 models : for the former models , the values ( @xmath59 ) corresponding to the lower tip of this region are @xmath60 , @xmath61 , while for the latter , @xmath62 and @xmath63 . along each constant-@xmath28 sequence for @xmath64 , quantities such as the axial angular velocity @xmath27 and the total angular momentum @xmath34 are non - monotonic functions of @xmath32 , increasing to a maximum for @xmath65 and decreasing thereafter . as @xmath66 , convergence in the vicinity of this maximum becomes significantly slower , and fails completely when @xmath28 is large enough ( i.e. , @xmath67 ) to place the model within the forbidden zone . for @xmath68 ( the largest value of @xmath28 we have considered ) , converged models can be readily obtained on the low-@xmath32 sides of these regions . on the high-@xmath32 sides , converged models can be obtained only within the ranges @xmath69 for the 1 @xmath6 models , and @xmath70 for the 2 @xmath6 models . the physical , mathematical , and computational reasons for the lack of convergence of models on either side of the forbidden zone are discussed in considerable detail in paper i. in this paper , we present results for complete constant-@xmath28 sequences of models having @xmath71 , but confine our attention to just the low-@xmath32 sides of forbidden zones for models with @xmath72 . models on the high-@xmath32 sides of these regions have highly rotationally flattened , disk - like structures that will be the focus of a subsequent paper . ccccc ccccc ccccc ccccc cc [ tab : mod ] @xmath3 & 0.8 & 1 & 1 & 1 & 1 & 1.2 & 1.2 & 1.6 & 2 & 2 & 2 + @xmath28 & 0 & 0 & 1.5 & 3.75 & 5 & 0 & 4 & 0 & 0 & 3 & 4.75 + @xmath32 & 0 & 0 & 1.55 & 3.58 & 4.15 & 0 & 3.74 & 0 & 0 & 5.64 & 5.9 + @xmath20 & 0 & 0 & 0.044 & 0.079 & 0.052 & 0 & 0.065 & 0 & 0 & 0.047 & 0.078 + @xmath34 & 0 & 0 & 6.44 & 7.27 & 5.49 & 0 & 8.93 & 0 & 0 & 23.71 & 23.73 + @xmath73 & 0 & 0 & 219 & 123 & 81 & 0 & 112 & 0 & 0 & 205 & 127 + @xmath16 & 0.224 & 0.653 & 0.332 & 0.224 & 0.336 & 1.591 & 0.653 & 6.075 & 15.47 & 7.560 & 6.083 + @xmath74 & 4710 & 5540 & 4710 & 4700 & 5080 & 6110 & 5550 & 7490 & 9090 & 5890 & 7480 + @xmath75 & 4.635 & 4.551 & 4.478 & 4.709 & 4.679 & 4.412 & 4.616 & 4.310 & 4.337 & 3.855 & 4.404 + @xmath26 & 0.714 & 0.878 & 0.934 & 0.744 & 0.767 & 1.129 & 0.904 & 1.466 & 1.590 & 2.927 & 1.512 + @xmath76 & 1 & 1 & 0.697 & 0.638 & 0.712 & 1 & 0.668 & 1 & 1 & 0.371 & 0.549 + @xmath77 & 1 & 1 & 0.697 & 0.764 & 0.837 & 1 & 0.791 & 1 & 1 & 0.530 & 0.770 + @xmath78 & 17.078 & 17.178 & 17.100 & 16.967 & 17.019 & 17.259 & 17.101 & 17.326 & 17.287 & 17.317 & 17.282 + @xmath79 & 7.046 & 7.121 & 7.069 & 7.004 & 7.039 & 7.187 & 7.094 & 7.280 & 7.326 & 7.289 & 7.255 + @xmath80 & 1.892 & 1.918 & 1.893 & 1.823 & 1.841 & 1.935 & 1.869 & 1.911 & 1.827 & 1.893 & 1.892 + @xmath81 & 25 & 40 & 35 & 17 & 20 & 73 & 29 & 114 & 96 & 446 & 83 + @xmath82 & & & & 0.138 & 0.138 & 0.046 & 0.090 & 0.094 & 0.122 & 0.054 & 0.110 + @xmath83 & & & & 0.023 & 0.031 & 0.007 & 0.014 & 0.082 & 0.141 & 0.085 & 0.074 + @xmath84 & 0.678 & 0.718 & 0.618 & 0.718 & 0.722 & 0.822 & 0.746 & 0.990 & 0.990 & 0.718 & + @xmath85 & 0.921 & 0.968 & 0.905 & 0.935 & 0.950 & 0.997 & 0.972 & 1.000 & 1.000 & 0.995 & + @xmath86 & 6.458 & 6.392 & 6.467 & 6.452 & 6.440 & 6.104 & 6.388 & 4.780 & 4.771 & 5.988 & + fig . & & & 3@xmath87 & 3@xmath88 & 3@xmath89 & & 3@xmath90 & & & 3@xmath91 & 3@xmath92 + in figure 2 , we show how some of the characteristics of the @xmath1 models depend on the dimensionless rotation parameter @xmath32 for @xmath94 . the relation between @xmath32 and the physical quantity @xmath73 , the equatorial rotation speed at the stellar surface , is given in panel ( a ) , from which it can be seen that for each of the @xmath28 sequences depicted , @xmath73 increases monotonically with @xmath32 . along the curves with @xmath95 in figure 2 , the plotted models span the range from @xmath96 ( i.e. , nonrotating ) to the value @xmath97 for which @xmath98 . along the curves with @xmath72 , the last plotted model is located adjacent to the boundary of the forbidden region in the @xmath99 plane ; for these differentially rotating models , the centrifugal and gravitational forces have nearly equal magnitudes in the core of the star . we note that the degree of differential rotation increases with @xmath28 , in the sense that the configurations corresponding to larger values of @xmath28 have a greater difference between the axial and surface equatorial rates of rotation ( see 2.1 ) . as panel ( b ) of figure 2 makes evident , the radiative luminosities of these models are diminished relative to the luminosity @xmath100 of a nonrotating @xmath1 star . this is a well - known consequence of including rotation in the determination of the equilibrium stellar structure ( see , e.g. , clement 1979 ; bodenheimer 1971 ) . in the results shown in figure 2 , the reduction in @xmath16 is larger for differentially rotating models than it is for models that are uniformly or nearly uniformly rotating . a model with @xmath101 rotating at the break - up rate ( @xmath102 ) has @xmath103 , while an @xmath104 model with @xmath105 has @xmath106 , a reduction of more than a factor of 6 from the nonrotating value . much of the reason for this behavior lies in the effect of rotation on the thermodynamic conditions in the deep , energy - producing regions of the stellar interior . for these @xmath1 models , the contribution of the centrifugal force to supporting material against gravity enables the star to emulate an object of lower mass with correspondingly reduced values of @xmath18 , @xmath17 , and @xmath47 ( e.g. , sackmann 1970 ) . the results presented in panel ( c ) illustrate the dependence of @xmath17 on model rotational properties ; similar variations are found for both @xmath18 and @xmath47 . for rigidly rotating configurations , this centrifugal support is largest in the outermost layers of the interior , which contain only a small fraction of the stellar mass ; in this case , @xmath18 , @xmath17 , and @xmath47 are little changed from the values appropriate to a nonrotating star of the same mass . for the @xmath101 , @xmath102 model noted previously , @xmath107 , @xmath108 , and @xmath109 , where the subscript 0 indicates the nonrotating value . alternatively , in models for higher values of @xmath28 , the effects of rotation are increasingly concentrated toward the central regions of the star , with the result that the perturbations to the central thermodynamic quantities can be more substantial ; for @xmath110 , @xmath105 , @xmath111 , @xmath112 , and @xmath113 . panels ( b ) and ( c ) also indicate that the magnitudes of the changes in @xmath16 , @xmath17 , and other quantities depend on the assumed profile of internal differential rotation . the model for @xmath114 , @xmath115 has @xmath116 , with @xmath117 , @xmath118 , and @xmath119 , smaller reductions relative to the nonrotating model than those for @xmath104 , @xmath120 . this behavior is an outgrowth of the structural modifications arising from the centrifugal force distributions associated with the different rotation profiles . in the @xmath114 model , the ratio of the centrifugal to gravitational force in the equatorial plane , @xmath121 ( @xmath122 is the radial coordinate in the equatorial plane ) , is sharply peaked in the innermost portion of the stellar core , with maximum value 0.83 at the center , decreasing to @xmath123 at @xmath124 . for the shallower angular velocity profile of the @xmath104 model , the force ratio decreases from a smaller central value of 0.42 to 0.23 at the stellar surface , 10 times the value found throughout the outer 50% of the interior of the @xmath114 model . in this connection , we note that bodenheimer ( 1971 ) found that the luminosities of models for rotating 30 @xmath6 stars , computed assuming a variety of internal rotation laws , depended primarily on the total angular momentum content of a given stellar model and not on the details of its distribution within the interior . specifically , his results indicated that the luminosities of models corresponding to four different prescribed distributions of the angular momentum per unit mass decreased with increasing total angular momentum @xmath34 , with the relation between @xmath16 and @xmath34 nearly the same for each model sequence . if the luminosities of our 1 @xmath6 models are plotted versus their respective @xmath34 values , an analogous reduction in @xmath16 for increasing @xmath34 can be discerned . however , the relation between @xmath16 and @xmath34 is roughly independent of @xmath28 only for small @xmath34 , and exhibits a clear dependence on @xmath28 that becomes increasingly pronounced as @xmath34 is made larger . that the luminosity can be even approximately expressed as a function of @xmath34 necessarily reflects the modifications produced by rotation to the thermodynamic conditions in the energy - producing core of the star , as described previously ( see also mark 1968 ) . the nature and origin of the relation between @xmath16 and @xmath34 for stars with masses @xmath125 and higher , including its dependence on the structural characteristics of the models , will be addressed in detail in a subsequent paper in this series . as in paper i , we define an average effective temperature through the relation @xmath126 , where @xmath127 is the stefan - boltzmann constant and @xmath128 the area of the stellar surface . the results for @xmath15 depicted in panel ( d ) exhibit dependences on @xmath28 and @xmath32 that are similar to those seen for the quantities plotted in the preceding panels . reductions by more than @xmath129 from the nonrotating value ( @xmath130 ) are possible as , for example , in the case @xmath131 , @xmath132 , for which @xmath133 . the rotation - induced variations in @xmath15 represent the combined effects of changes in both @xmath16 and @xmath128 . an indication as to the behavior of @xmath128 can be gleaned from examination of the influence of rotation on the stellar size and shape . this information is presented in panels ( e ) and ( f ) , where we show , respectively , the equatorial radius @xmath26 ( in units of the radius @xmath134 of the nonrotating , spherical @xmath1 model ) , and the polar radius @xmath135 together with @xmath136 , the maximum perpendicular distance from the equatorial plane to the stellar surface . these latter two quantities are both given as fractions of @xmath26 ; values of @xmath137 that are @xmath138 reflect a rotational flattening of the configuration , while values of @xmath139 that are @xmath140 are indicative of a deviation from a convex spheroidal shape through the development of a concavity in each of the two polar regions of the star . note that the response of the equatorial radius to increasing @xmath32 differs greatly depending upon whether @xmath28 is @xmath141 or @xmath142 . in the former case , the centrifugal force attains its largest value relative to the gravitational force at the stellar surface , consequently producing a distension of the outer , equatorial layers of the stellar interior and an overall increase in @xmath26 . although @xmath135 decreases somewhat relative to the radius of the corresponding nonrotating model , the net effect of the changes in @xmath26 and @xmath135 is usually an increase in the volume of the rotating star . for some models with @xmath143 , the decrease in @xmath135 can more than compensate for the the increase in @xmath26 , and the volume of the rotating star is reduced . in the case where @xmath144 , both @xmath26 and the stellar volume shrink with increasing @xmath32 along a constant-@xmath28 sequence . for these models , the centrifugal force is largest in comparison to the gravitational force in the central region of the stellar core , causing the central thermodynamic conditions , luminosity , and size ( i.e , radius , surface area , volume ) to assume values that are characteristic of nonrotating stars of lower mass . this raises the possibility that a differentially rotating star can imitate a less massive nonrotating star in radius and effective temperature , as well as in luminosity . in panel ( f ) , it can be seen that for @xmath145 , the curves for @xmath139 ( solid lines ) and @xmath137 ( dotted lines ) are coincident and @xmath146 , implying that the photospheric shape of these models is oblate spheroidal . such a model ( @xmath131 , @xmath147 ) is depicted in panel ( a ) of figure 3 . for @xmath148 , however , @xmath149 , symptomatic of the development of a `` dimple '' or indentation at either pole , as in the cases of the models shown in panels ( b ) ( @xmath150 , @xmath151 ) and ( c ) ( @xmath114 , @xmath152 ) of figure 3 . panels ( g ) and ( h ) of figure 2 contain results pertaining to the location , extent , and properties of convective regions in the models . in the absence of rotation , the internal structure of a @xmath1 star consists of an inner , radiative core that encompasses @xmath153% of the stellar radius and contains @xmath154% of the stellar mass , surrounded by an outer , convective envelope . as is apparent in panel ( g ) , the inclusion of rotational effects can modify this basic morphological picture in two ways : either by increasing the size of the convective envelope or by promoting the formation of a convective core . for lower values of @xmath28 ( i.e. , @xmath155 in panel [ g ] ) , the rotationally induced changes in internal structure cause the outer convection zone to deepen as @xmath32 increases . for example , in the model with @xmath131 , @xmath147 ( see panel [ a ] in figure 3 ) , the radius in the equatorial plane of the base of the convective envelope is @xmath156 , significantly deeper than the the base radius @xmath157 in the nonrotating model . associated with this reduction in @xmath158 is a decrease in the mass of the core ( @xmath159 ) , and an increase in the temperature @xmath160 at the bottom of the envelope , as can been seen in panel ( h ) ; for the @xmath131 , @xmath147 model , @xmath161 , as compared with @xmath162 for the nonrotating model . since some chemical species ( e.g. , li , be , and b ) can be destroyed by thermonuclear reactions at temperatures @xmath163 , enhancements of @xmath160 of this magnitude are likely to have consequences for the surface abundances of these elements . in the case of the strongly differentially rotating models with @xmath164 , the largest centrifugal effects are concentrated in the innermost portion of the core , so that the fractional thickness of the outer envelope is little affected . however , as the magnitude of the centrifugal - to - gravitational force ratio in the core increases for larger @xmath32 , the decreasing pressure gradient implied by the requirement of hydrostatic equilibrium forces the otherwise stably stratified central regions of the interior to become convective . under these conditions , with the presence of a convective core , the structure of the deep interior resembles that of a higher - mass star . the models shown in panels ( b ) and ( c ) of figure 3 are examples of @xmath1 stars with convective cores ; in each of these models , the radial extent of this region in the equatorial plane is about 14% of @xmath26 and contains @xmath165% of the stellar mass . for comparison , the convective core in a model for a nonrotating @xmath2 star has a radius that is about 12% of @xmath26 and contains @xmath166% of the stellar mass . to illustrate the structural differences between differentially and near - uniformly rotating stars , in figure 4 we show the profiles of several physical quantities for models with @xmath167 , and @xmath168 , along with the corresponding results for the nonrotating model . the luminosities and total angular momenta of the two rotating models are @xmath169 and @xmath170 ( @xmath171 ) , and @xmath172 and @xmath173 ( @xmath174 ) , respectively , where @xmath34 is meaasured in units of @xmath175 g @xmath176 s@xmath12 . in panels ( a ) and ( b ) , the temperature @xmath8 and mass density @xmath5 are depicted as functions of the radial position @xmath122 ( measured in units of the present - day solar radius @xmath177 ) in the equatorial plane of the star . the central values of both quantities exhibit rotation - induced reductions relative to the nonrotating case , the magnitudes of these modifications being larger for @xmath171 ( @xmath178 ) than for @xmath179 ( @xmath180 ) . as noted previously , this behavior is a consequence of differences in the magnitudes and distributions of the centrifugal force in the two rotating models . these distinctions can be clearly seen in the profiles of the centrifugal - to - gravitational force ratio shown in panel ( d ) . for @xmath181 , the force ratio is largest in the deep interior , attaining a maximum value of 0.96 at the stellar center and decreasing outwards to a magnitude @xmath123 in the photosphere . for @xmath179 , the ratio increases monotonically throughout the interior , rising from a central value of just 0.026 to a maximum of 0.36 at @xmath26 . the substantial contribution of the centrifugal force to the support of the innermost regions of the @xmath182 model is responsible for the considerable enhancement of the density scale height there , evident in panel ( b ) ; @xmath5 declines by only @xmath183% over the inner 20% of the stellar interior . the resulting changes in the internal mass distribution ( panel [ c ] ) lead to a star with a smaller radius , @xmath184 as opposed to @xmath185 for the nonrotating model . alternatively , for the @xmath179 model , the lack of centrifugal support in the core of the star leads to temperature , density and mass distributions therein that closely resemble those of the nonrotating model . closer to the surface , however , the density distribution becomes extended , a product of the increasing centrifugal reduction of gravity in the outer layers of the interior ; as a result , the stellar radius is larger than that of the nonrotating model , @xmath186 . figure 5 is a theoretical hrd for @xmath1 models that span a broad range of internal rotational states , from uniform rotation to extreme differential rotation ( up to @xmath187 for @xmath188 ) . the nonrotating zams is delineated by a dotted line , with the positions of several specific models for @xmath189 indicated . clearly , as already implied by panels ( b ) and ( c ) of figure 2 , the locations of rotating models in such a diagram are displaced to the right of and below the position they would occupy in the absence of rotation , toward lower values of both the luminosity and the effective temperature . models with uniform or near - uniform internal rotation ( i.e. , @xmath190 ) lie well above the nonrotating zams , while those for which the degree of differential rotation is substantial ( @xmath191 ) have positions in the hrd that collectively approximate the nonrotating zams over virtually the entire mass range depicted . for example , the rotating @xmath1 model ( @xmath150 , @xmath151 ) shown in panel ( b ) of figure 3 is characterized by @xmath192 and @xmath193 , which is indistinguishable from the values @xmath194 , @xmath195 for a nonrotating , @xmath48 model . such a coincidence of the positions of stars of different mass in the classical hrd is made possible by the fact that the largest structural changes take place in the cores of strongly differentially rotating models . specifically , the decrease in @xmath16 brought about by the smaller values of @xmath18 , @xmath17 , and @xmath47 , together with the corresponding reduction in @xmath26 , enable such a model to effectively mimic a lower mass stellar model in which rotational effects are not included . figure 6 illustrates some of the internal properties of three so - called _ solar look - alike _ models , that is , models for differentially rotating stars with masses @xmath196 that have many physical attributes in common with the model for a nonrotating , @xmath1 star . inspection of panels ( a ) and ( b ) reveals that the profiles and the central values of the temperature and density for the 1.1 , 1.2 , and 1.3 @xmath6 models depicted therein are quite close to those of the solar - mass model with @xmath197 . a consequence of these structural similarities is that many of the general characteristics of the models are nearly identical . for example , the @xmath198 model which , in the absence of rotation , would have @xmath199 , @xmath200 , and @xmath201 instead has @xmath202 , @xmath203 , and @xmath204 , compared with @xmath205 , @xmath206 , and @xmath207 for the nonrotating , @xmath1 model . a cross - section in the meridional plane of the rotating @xmath208 interior is shown in panel ( d ) of figure 3 , from which it can be seen that a solar - like convection zone having @xmath209 is present in the layers beneath the photosphere . with a surface equatorial rotation speed @xmath210 km s@xmath12 , such an object could be mistakenly identified as a rapidly rotating @xmath1 star , if observations were analyzed through comparison with nonrotating stellar models . _ that the values of measureable or inferrable stellar properties can be the same in rotating and nonrotating models for different masses represents a potential source of ambiguity in the interpretation of a variety of observations . _ panels ( c ) and ( d ) reveal subtle differences in the internal mass distributions and in the variation of the ratios of the centrifugal to gravitational force in the equatorial planes of these solar look - alike models . future space - based asteroseismological observations may be capable of exploiting such differences in internal structure to distinguish slowly rotating stars from more rapidly , differentially rotating higher - mass stars that happen to have the same values of @xmath16 and @xmath15 and thus occupy the same position in the hrd ( see , e.g. , lochard et al . 2005 ) . in figure 7 , we present a summary of the rotational dependence of @xmath211 model characteristics , using the same format as adopted for figure 2 . the model for a non - rotating star of this mass has @xmath212 , @xmath213 , and @xmath214 , and a convective core with @xmath215 , @xmath216 ; such a model has a thin , subsurface convective layer , @xmath217 , that contains a negligible fraction of the stellar mass . the results shown in figure 7 , which span the rotational parameter ranges @xmath218 , @xmath219 , exhibit many of the same behaviors as noted previously in connection with the @xmath1 models of figure 2 . in particular , along each constant-@xmath28 sequence , the luminosity ( panel [ b ] ) , central temperature , ( panel [ c ] ) and effective temperature ( panel [ d ] ) all initially decrease as @xmath32 is increased , with the largest reductions in these quantities occurring in the cases of differentially rotating models with @xmath191 . for @xmath220 1 , 2 , both @xmath221 and @xmath222 pass through minima and then increase slightly with @xmath32 beyond those points , whereas @xmath15 continues to decrease with @xmath32 for all @xmath28 sequences . the central pressure @xmath18 ( panel [ e ] ) , on the other hand , is an increasing function of @xmath32 for @xmath223 , and displays a non - montonic dependence on @xmath32 for the model sequences corresponding to @xmath224 . the origin of this variation is the mass dependence of @xmath18 for stars on the non - rotating zams . in our scf models without rotation , @xmath18 has a maximum value for @xmath3 near @xmath225 , and is decreasing for both larger and smaller @xmath3 values . the maximum occurs for a mass close to the value marking the internal structural transition between stars with radiative cores and convective envelopes and stars with convective cores and radiative envelopes . hence , insofar as the properties of a rotating star are like those of a non - rotating star of lower mass , we expect the @xmath18 values of these @xmath211 models to increase as the effects of rotation become more pronounced . for cases in which the central thermodynamic conditions are significantly perturbed by rapid , differential rotation in the core of the star , this mass - lowering effect can be large enough to shift @xmath18 to values characteristic of nonrotating stars with @xmath226 ( i.e. , on the other side of the central pressure peak ) , as in the @xmath2 models for @xmath220 4 , 5 . panels ( f ) and ( g ) of figure 7 convey information pertaining to the photospheric sizes and shapes of the models . those that rotate nearly rigidly are equatorially distended , and , in the direction perpendicular to the equatorial plane , have their largest dimension along the rotation axis ( @xmath227 ) ; since @xmath228 , these models ( like their @xmath1 counterparts ) have a flattened , spheroidal shape . models with increasing degrees of differential rotation ( i.e. , with @xmath229 ) develop polar concavities ( as indicated by @xmath149 ) , and ultimately ( i.e. , for @xmath224 and sufficiently large @xmath32 ) become more compact , with @xmath230 . panels ( e ) and ( f ) of figure 3 give cross - sectional representations of the photospheric shapes of @xmath2 models for @xmath231 and @xmath232 , repectively ; in the former model , @xmath233 , while for the latter , @xmath234 . the influence of rotation on the occurrence and extent of convective regions within the models is explored in panel ( h ) of figure 7 . the convective core , a salient feature of the non - rotating stellar interior , generally decreases in size as the value of @xmath32 is increased . this contraction stems from a reduction in the magnitude of the radiative gradient @xmath235 in the core region of the star , a result of the way in which the central thermodynamic conditions are modified by rotation ( see , e.g. , macgregor & gilliland 1986 ) ; with @xmath235 smaller , the size of the region wherein it exceeds the adiabatic gradient @xmath236 shrinks accordingly . only for the most rapidly , differentially rotating models does @xmath237 show a modest increase compared to the non - rotating model ; in panel [ h ] , the model for @xmath238 has @xmath239 . in this case , as was seen for the @xmath1 models having convective cores in figure 2 , the growth of @xmath237 can be traced to a _ larger _ value of @xmath235 , produced by a centrifugal - to - gravitational force ratio that is nearly unity at the stellar center . note also that for some @xmath2 models , the convection zone underlying the stellar photosphere can extend into the stellar interior by more than the few @xmath240 that is the thickness of this region in the absence of rotation . as can be seen , for example , in panel ( e ) of figure 3 , the base of the convective envelope in the equatorial plane of the model with @xmath241 is located at @xmath242 , a fractional depth which is the same as in the non - rotating @xmath1 model . this leads to the intriguing possibility that solar - like oscillations , driven by turbulent convection , may be excited in stars that would normally be too massive to generate them . the results shown in panel ( h ) suggest that such a solar - like convective envelope is most likely to occur in @xmath2 models with intermediate differential rotation ( say . , @xmath243 ) . for a profile of this kind , the inner and outer portions of the interior can each rotate rapidly enough to both significantly perturb the thermodynamic conditions in the core and increase @xmath26 by extending the stellar envelope . in figure 8 , we show the positions of models for differentially rotating , @xmath2 stars in a theoretical hrd . as in figure 5 , the location of the non - rotating zams is indicated by a dotted line , with the positions of models for stars with masses in the range @xmath244 indicated along it . this hrd for rotating @xmath2 models has a number of features in common with the corresponding representation of @xmath1 properties . without exception , models are shifted from the non - rotating location to new positions characterized by lower @xmath15 and @xmath16 values . for @xmath245 , these new positions lie to the right - hand side of ( i.e. , above ) the non - rotating zams ; for rotation that is increasingly differential , however , the model locations approach the non - rotating zams , moving just to the left - hand side of ( i.e. , below ) it for @xmath114 . for @xmath246 , the model positions are distributed along the non - rotating zams , delineating its track in the hrd for masses greater than about 1.5 @xmath6 . this coincidence again raises the possibility that a rapidly , differentially rotating @xmath2 star could effectively masquerade as a non - rotating star of lower mass . as an example , with @xmath247 , @xmath248 , and @xmath249 the model for @xmath250 shown in panel ( f ) of figure 3 closely resembles a non - rotating @xmath251 star , for which @xmath252 , @xmath253 , and @xmath254 . the scf approach to computing the structure of rotating stars is robust and efficient , capable of yielding converged models of stars of all masses , with internal angular - velocity distributions covering the range from uniform to extreme differential rotation . the code described in paper i has been upgraded throught the introduction of opacities , energy generation rates , and an equation of state that are close to state - of - the - art , and with the introduction of a mixing - length treatment of convective energy transport . we have used this revised code to conduct a systematic study of the properties of models for rotating zams stars with masses @xmath255 , assuming that the rotation of the stellar interior can be adequately represented by the parameterized angular - velocity profile given in equation ( 1 ) . our results suggest that rotation affects virtually every characteristic of the intermediate- and low - mass stars considered herein , with the magnitude , and sometimes even the sense , of the changes in fundamental attributes depending on both the rate and degree of the differential rotation in the stellar interior . for models that are uniformly or nearly uniformly rotating , the centrifugal contribution to the hydrostatic support of the star is largest in the outer layers of the interior ; for models that are strongly differentially rotating , the centrifugal force is largest compared to gravity in the core of the star . luminosities and effective temperatures are always diminished relative to nonrotating stars of the same mass , while equatorial radii increase in uniformly rotating or moderately differentially rotating models , and decrease in models where the difference between the angular velocity at the center and that at the equator is sufficiently large . as the degree of differential rotation is increased , the photospheric shape of the star changes from a convex surface closely approximating a spheroid flattened along the rotation axis to a roughly oblate surface with deepening polar concavities ; in such cases , the greatest perpendicular distance from the equatorial plane to the surface occurs away from the axis of rotation , and is always less than the equatorial radius of the star . in @xmath1 models , the effects of rotation can either increase the thickness of the outer convective envelope , or contribute to the formation of a convective core . in rotating @xmath2 models , the size of the convective core is diminished relative to that found in the absence of rotation , and an extensive , solar - like convection zone can be present in the outer layers of what would otherwise be a stable radiative envelope . the extent to which the results of these computations accurately depict the stellar structural modifications arising from rotation depends on the validity of a number of simplifying assumptions and approximations made in the course of developing the basic model , as well as on the inherent limitations of the scf method itself . here , we briefly address the most salient of the factors that might restrict the applicability of these results , noting where improvements and extensions can ( or can not ) be made ( see also paper i ) . 1 ) . the modified scf iterative scheme utilized in the present investigation necessarily requires the angular velocity distribution within the star to be conservative , and thus expressible as a function of just the perpendicular distance from the rotation axis , @xmath256 ( see eq . [ 1 ] ) . although this specification facilitates the construction of models through the considerable mathematical simplification it introduces , its effect on some computed stellar structural characteristics , such as the photospheric shapes and core physical properties of very rapidly , differentially rotating models , is likely to differ from that produced by a rotation law of ( say ) the form @xmath257 . there is some evidence , from both simulations and observations , for the occurrence of differential rotation of the type given by equation ( 1 ) ( see , e.g. , dobler , stix , & brandenburg 2006 , and references therein ) . yet analyses of solar acoustic oscillations yield a picture of the large - scale internal dynamics of the sun which is not in accord with the angular velocity being constant on cylindrical surfaces ( see thompson et al . we note that use of the scf method precludes consideration of non - conservative rotation laws ( e.g. , @xmath258 ) ; however , detailed treatment of the hydrodynamical and magnetohydrodynamical processes affecting the internal rotational states of the sun and stars is presently beyond the scope of @xmath259 of the extant structural / evolutionary models . 2 ) . the model treats only the prescribed rotational motion of the stellar interior , omitting any meridional circulatory flow and its consequent effects on the internal angular momentum distribution . for models with radiative envelopes , rotationally driven circulation will cause the rotation profile to deviate over time from the state given by equation ( 1 ) , unless such evolution is mitigated by additional angular momentum transport mechansims ( see , e.g. , maeder & meynet 2000 , and references therein ) . for models with convective envelopes , the rotation profile in the outer layers of the interior is the product of the complex interplay between meridional circulation and turbulent heat and angular momentum transport ( rempel 2005 ; miesch , brun , & toomre 2006 ) ; whether @xmath260 is solar - like or constant on cylindrical surfaces depends on the latitudinal entropy distribution in the subadiabatically stratified layers below the convection zone . the model employs a simplified treatment of convection , locating unstable regions by application of the schwarzschild criterion and utilizing an averaged , rotationally modified mixing - length description ( see 2.2 ) to determine the structure of an outer convective envelope , if present . use of the solberg - hiland condition ( e.g. , ledoux 1965 ; kippenhahn & weigert 1990 ) to ascertain the onset of convective instability would account for the direct influence of the centrifugal force . since models computed for the rotation law of equation ( 1 ) have @xmath261 , where @xmath262 is the specific angular momentum ( see paper i ) , this change would likely decrease the equatorial - plane thickness of an outer convection zone , and produce a latitudinal variation in @xmath158 . however , for rapid rotation , there is considerable uncertainty regardless of the convection criterion / model adopted , because of both the rudimentary nature of extant treatments of convective energy transport and the universal assumption of axisymmetry among rotating stellar models . we reiterate that the present model describes a chemically homogeneous , zams star , and neglects effects associated with the structural and compositional evolution of the stellar interior . we are presently developing a mean - field hydrodynamics - based treatment of turbulent chemical and angular momentum transport which maintains conservative rotation profiles , thus making it possible to investigate the main - sequence evolution of these scf models . despite the shortcomings enumerated above , we believe that the basic model and method described herein compare favorably with alternative approaches to determining the structure and evolution of rotating stars . the most widely used of these ( e.g. , meynet & maeder 1997 ) relies on the use of approximate , roche - like equipotentials to represent the internal gravity of the star , thereby strictly limiting it to describing slowly rotating stars to ensure the accuracy of the computed models . whatever its drawbacks , the scf method yields two - dimensional , axisymmetric configurations that represent fully consistent solutions to both the set of stellar structure equations _ and _ poisson s equation for the gravitational potential . the results presented herein have a number of implications for young stars with masses between 1 and @xmath2 . measurements of projected equatorial rotation speeds in excess of 100 km s@xmath12 for some solar - type stars in young open clusters ( see , e.g. , stauffer 1991 ) raise the possibility that the structure and properties of these objects could be significantly altered from those of nonrotating stars of the same mass . such stars , provided that their interior rotation is sufficiently differential , could in actuality be somewhat more massive objects for which surface rotation as rapid as that indicated by observations is the normally expected zams state . the question of whether or not the kind of strong differential rotation required to produce such ambiguity is present within the interiors of some low- and intermediate - mass stars may ultimately be resolved through space - based asteroseismological observations , which should allow low - resolution inversions of the rotation profiles in the inner @xmath263 of the stellar radius ( see gough & kosovichev 1993 ) . asteroseismology from space may also afford the means for identifying individual stars whose rotation enables them to pose as lower - mass objects , since , when effects associated with asphericity are not too large , the average mode frequency spacing is sensitive to the mean density , a quantity which table 1 reveals to be different for _ look - alike _ models . if rapid differential rotation is a possibility for objects in the mass range spanned by the models of 3 , then the associated changes in structure and properties could have consequences for a variety of important astrophysical processes that take place within and around such stars . the reduced radiative luminosity of a rapidly rotating young sun would likely influence the evolution of the solar nebula , and further exacerbate the discrepancies between the properties of nonrotating models and observational inferences indicating a higher zams luminosity ( see , e.g. , sackmann & boothroyd 2003 ) . the rotationally induced deepening of a sub - photospheric convection zone , together with the increase in temperature of the material at the base , could contribute to the depletion of lithium in the stellar surface layers by reducing the thickness of the region through which chemical species must be transported in order to be destroyed by nuclear processes . the formation of a solar - like convective envelope in a young , differentially rotating , @xmath2 star could excite global oscillations , and be accompanied by the operation of a solar - like hydromagnetic dynamo . dynamo - generated fields that diffuse into and are retained by the radiative interior ( see , e.g. , dikpati , gilman , & macgregor 2006 ) could enable the star to remain magnetic long after spin - down and the elimination of nonuniform rotation have led to the disappearance of both the surface convective layer and the dynamo . each of these possibilities will be addressed in forthcoming papers . we wish to thank d. a. vandenberg for providing software from his stellar - evolution code that was used to handle the input physics in our scf code , and we wish to thank j. christensen - dalsgaard for the use of his stellar - evolution code to help us validate our models . this work was supported in part by an astronomy & astrophysics postdoctoral fellowship under award ast-0401441 ( to t. s. m. ) from the national science foundation . alexander , d. r. , & ferguson , j. w. 1994 , , 437 , 879 baker , n. h. , & temesvary , s. 1966 , studies of convective stellar envelopes ( nasa goddard inst . for space studies ) bodenheimer , p. 1971 , , 167 , 153 caughlan , g. r. & fowler , w. a. 1988 , atomic data and nuclear data tables , 40 , 283 chambers , r. h. 1976 , ph.d . thesis , university of toronto christensen - dalsgaard , j. 1982 , , 199 , 735 christensen - dalsgaard , j. , & dppen , w. 1992 , a&a review , 4 , 267 christensen - dalsgaard , j. , proffitt , c. r. , & thompson , m. j. 1993 , , 403 , l75 clement , m. 1978 , , 222 , 967 clement , m. 1979 , , 230 , 230 dikpati , m. , gilman , p. a. , & macgregor , k. b. 2006 , , 638 , 564 di mauro , m. p. , & christensen - dalsgaard , j. 2001 , in helio- and asteroseismology at the dawn of the millenium , proc . of soho 10/gong 2000 workshop , ed . a. wilson ( noordwijk : esa ) , 373 dobler , w. , stix , m. , & brandenburg , a. 2006 , , 638 , 336 eggenberger , p. , maeder , a. , & meynet , g. 2005 , , 440 , l9 eggleton , p. p. , faulkner , j. , & flannery , b. p. 1973 , , 23 , 325 endal , a. s. , & sofia , s. 1981 , , 243 , 625 faulkner , j. , roxburgh , i. w. , & strittmatter , p. a. 1968 , , 151 , 203 gough , d. o. , & kosovichev , a. g. 1993 , asp conf . 40 : iau colloq . 137 : inside the stars , 40 , 541 graboske , h. , dewitt , h. , grossman , a. , & cooper , m. 1973 , , 181 , 457 jackson , s. 1970 , , 161 , 579 jackson , s. , macgregor , k. b. , & skumanich , a. 2004 , , 606 , 1196 jackson , s. , macgregor , k. b. , & skumanich , a. 2005 , , 156 , 245 kippenhahn , r. , weigert , a. , & hofmeister , e. 1967 , meth . comp . phys . , 7 , 129 kippenhahn , r. , & thomas , h. -c 1970 , in stellar rotation , ed . a. slettebak ( new york : gordon & breach pub . ) , 20 kippenhahn , r. , & weigert , a. 1990 , stellar structure and evolution ( berlin : springer - verlag ) ledoux , p. 1965 , in stellar structure , eds . l. h. aller & d. b. mclaughlin ( chicago : univ . of chicago press ) , 499 lochard , j. , samadi , r. , & goupil , m. j. 2005 , , 438 , 939 macgregor , k. b. , & gilliland , r. l. 1986 , 310 , 273 maeder , a. , & meynet , g. 2000 , , 38 , 143 mark , j. w. -k . 1968 , , 154 , 627 meynet , g. , & maeder , a. 1997 , , 321 , 465 miesch , m. s. , brun , a. s. , & toomre , j. 2006 , , 641 , 618 ostriker , j. p. , & mark , j. w. -k . 1968 , , 151 , 1075 papaloizou , j. c. b. , & whelan , j. a. j. 1973 , , 164 , 1 pinsonneault , m. h. , kawaler , s. d. , sofia , s. , & demarque , p. 1989 , , 338 , 424 rempel , m. 2005 , , 622 , 1320 rogers , f. j. , & iglesias , c. a. 1992 , , 79 , 507 roxburgh , i. w. 2004 , , 428 , 171 sackmann , i. -j . 1970 , , 8 , 76 sackmann , i. -j . , & boothroyd , a. i. 2003 , , 583 , 1024 stauffer , j. r. 1991 , in angular momentum evolution of young stars , ed . s. catalano & j. r. stauffer ( dordrecht : kluwer academic pub . ) , 117 tassoul , j. -l . 2000 , stellar rotation ( cambridge : cambridge univ . press ) thompson , m. j. , christensen - dalsgaard , j. , miesch , m. s. , & toomre , j. 2003 , , 41 , 599 vandenberg , d. a. 1983 , , 51 , 29 vandenberg , d. a. 1992 , , 391 , 685 vandenberg , d. a. , swenson , f. j. , rogers , f. j. , iglesias , c. a. , & alexander , d. r. 2000 , , 532 , 430 wolff , s. c. , & simon , t. 1997 , , 109 , 759

power - counting arguments are used to organize the interactions contributing to the reactions near threshold . we estimate the contributions from the three formally leading mechanisms : the weinberg - tomozawa ( wt ) term , the impulse term , and the-excitation mechanism . sub - leading but potentially large mechanisms , including-wave pion - rescattering , the galilean correction to the wt term , and short - ranged contributions are also examined . the wt term is shown to be numerically the largest , and the other contributions are found to approximately cancel . similarly to the reaction , the computed cross sections are considerably smaller than the data . we discuss possible origins of this discrepancy .

nucl-th9904031

computations of pion production in nucleon - nucleon ( @xmath4 ) collisions near threshold , allow a confrontation of our understanding of @xmath4 interactions with data in a kinematic region for which chiral symmetry , and therefore quantum chromodynamics qcd , could be very important . the reaction @xmath3 near threshold has attracted much attention in the years since the first iucf data appeared @xcite and has exposed serious disagreements with earlier theoretical calculations @xcite . the existence of many conflicting models claiming to explain this discrepancy @xcite calls for a principle to organize the several potentially significant mechanisms of pion production . chiral perturbation theory ( @xmath5pt ) has been applied to mesonic @xcite , one - baryon @xcite , and nuclear @xcite processes where typical momenta of the order of the pion mass , @xmath6 , allow a systematic expansion of observables in powers of @xmath7 , where @xmath8 gev . et al _ @xcite have adapted the power counting and applied @xmath5pt to near - threshold pion production , where momenta are of order @xmath9 , @xmath10 being the nucleon mass . they estimated leading and next - to - leading contributions , the latter including important short - range contributions , related to the isoscalar components of the potential ( and possibly described by @xmath11 and @xmath12 meson exchanges ) @xcite . subsequently , van kolck _ et al _ @xcite showed that next - to - next - to - leading contributions ( _ e.g. _ , from @xmath13 meson exchange ) are also relevant ; data could then be explained within the very large theoretical uncertainties associated with @xmath2-wave pion rescattering and the short - range structure of the nuclear force . other @xmath5pt - inspired calculations have also stressed the importance of understanding rescattering @xcite and the effect of loops @xcite . since these @xmath5pt - inspired calculations include a larger number of contributions than other , model calculations , one concludes that a large theoretical uncertainty plagues all calculations performed to date . given such uncertainties , it is natural to examine other channels using the same techniques . here we are going to discuss @xmath14 and @xmath15 , which traditionally have been considered better understood than @xmath3 . we consider energies near threshold where the calculation simplifies , because the pion emerges mostly in an @xmath2 wave . we will show that an understanding of these channels is still in the future . we adopt here the conventional nuclear approach of grouping all @xmath4 interactions generated by mesons of small momentum in a potential , while the contributions associated with energies comparable to @xmath6 are accounted for in a kernel to be evaluated between wavefunctions generated by the potential . splitting the problem this way , one should still strive to calculate wavefunctions and kernels from the same theory or model , otherwise ambiguities arise from off - shell extrapolations ( or equivalently , from nucleon - field redefinitions ) . @xmath5pt is the only known tool for performing this task , and at the same time is consistent with qcd , because its symmetries are treated correctly . @xmath5pt is also unique in that it offers the possibility of doing systematic calculations : an expansion in momenta provides a power counting to organize the calculation even though coupling constants are not small . @xmath5pt separates interactions in long - range effects calculated explicitly with pion exchange and short - range effects accounted for by contact interactions with an increasing number of derivatives . parameters not constrained by chiral symmetry depend on details of qcd dynamics , and are at the present unknown functions of qcd parameters . predictive power is not lost , however , because at any given order in the power counting only a finite number of unknown parameters appear ; after they are fitted to a finite set of data , all else can be predicted at that order . these predictions are called `` low - energy theorems '' . since the lagrangian of @xmath5pt is the most general one consistent with qcd symmetries , @xmath5pt is a generalization of current algebra . unfortunately , the only @xmath4 potential derived in @xmath5pt @xcite and fitted to low - energy phase shifts @xcite produces poor results near the pion production threshold . attempts to remedy the situation are in progress @xcite . for the time being we will rely on modern , `` realistic '' phenomenological potentials which fit @xmath4 data very well . by considering more than one of those , we can estimate the otherwise uncontrolled error stemming from our inconsistent use of potential and kernel . we are going to see that , in contrast to neutral pion production , the error is small in the channel considered here . because such realistic potentials reproduce low - energy phase shifts with identical long - range tails , they must contain the equivalent to the leading order in the chiral expansion . although our approach should be considered phenomenological , our leading - order result will be an approximation to a low - energy theorem . at sub - leading orders in the expansion , an apparent ambiguity arises , concerning the correct treatment of the energy transferred in pion rescattering . this can be seen in the conflicting results of estimates of the same kernel @xcite . this issue is under study @xcite ; here we will limit ourselves to the most natural prescription that the transferred energy is @xmath16 . we will concentrate most of our efforts on the kernel . we will discuss a reasonable power counting for pion production that generalizes for @xmath17 and @xmath18 channels the discussion of ref . @xcite . in leading order , the new ingredient here is pion rescattering using the weinberg - tomozawa ( wt ) term that dominates isospin - dependent @xmath19 scattering . this term does not contribute to the reaction @xmath20 . in the latter , the leading non - vanishing order consists only of an impulse term ( ia ) , in which a single pion is emitted from a nucleon , and a similar contribution from the delta ( @xmath1 ) . this leading order underpredicts the experimental data by a factor of approximately 5 , due to two cancellations not incorporated in our power counting : _ ( i ) _ among different regions in coordinate space for each term evaluated between initial and final wavefunctions ; and _ ( ii ) _ between the total impulse and delta contributions . ( i ) _ results from our present inability to treat the potential and the kernel on the same footing . oversight _ ( ii ) _ is somewhat accidental , but actually expected from the fact that , in energy , the pion threshold sits midway between the elastic threshold and the delta pole . the sensitivity of the delta contribution to the realistic potential used is the main source of dependence on the @xmath4 potential of the final result . as a consequence of the accidentally small leading order , effects which are usually negligible acquire prominence in neutral pion production in the @xmath21 reaction . one effect is isospin - independent pion rescattering . @xmath5pt is critically necessary to assess the size of this contribution . first , in principle @xmath5pt allows one to determine from @xmath19 data not only the momentum- and energy - independent @xmath22 vertex , but also terms which are quadratic in energy and momentum . this is important in view of the fact that in pion production the virtual pion has energy of order @xmath23 and the combination of parameters of relevance is thus different from the combination that appears at the @xmath19 scattering threshold . there is by now a number of consistent estimates of the relevant parameters to third order @xcite . an estimate of the uncertainty of this contribution to pion production can be made by using also a lower - order determination @xcite . second , @xmath5pt is the only way to account for rescattering as a component of a feynman diagram without destroying chiral symmetry . there have been attempts to estimate rescattering by simply connecting a nucleon line to a @xmath19 amplitude with an arbitrary off - shell extension @xcite . by considering field redefinitions in the most general chiral lagrangian , it is easy to show that the off - shell ambiguity in the pion leg is equivalent to a set of short - range @xmath24 interactions , and that an inconsistent treatment of both effects leads to violation of chiral symmetry in a way that is contradictory with qcd @xcite . using @xmath5pt , it was found @xcite that rescattering interferes destructively with the leading - order effects in @xmath20 . this further interference makes agreement with data more difficult . although the magnitude of the effect is still being assessed @xcite , it is certain that the uncertainty stemming from different determinations of the @xmath5pt parameters is large @xcite . this disagreement between theoretical evaluations and cross - section data for @xmath20 can be largely removed if @xmath11 , @xmath25 and other heavy mesons are included @xcite . when first suggested @xcite , it looked as if @xmath20 was a clear signal of these otherwise elusive mechanisms . among the first corrections in our power counting , one finds two - pion - exchange loop graphs and @xmath24 counterterms ( that behave properly under chiral symmetry ) . a full @xmath5pt calculation requires the calculation of these loops , and although some steps have been taken in this direction @xcite , it is a herculean task that remains to be completed . even then , it will still require that the counterterms be fitted to some pion production data ( say , right at threshold ) so that other data ( say , the energy dependence close to threshold , or other channels ) be predicted . in the case of @xmath20 , even the most sophisticated phenomenological models have to recourse to such a fit of a short - range counterterm @xcite . in any case , after all this one would then be interested in determining whether such counterterms are of natural size , and whether they can be further understood as the result of heavy - meson exchange . in view of all other uncertainties , the authors of ref @xcite took the point of view that an estimate of this class of sub - leading contributions could more easily be made by modeling them with meson exchange in z - graphs , following ref . it turns out that the counterterms so produced are of natural size , which lends them credence , but that they are not sufficient to achieve agreement with data . in order to study convergence , further meson exchanges ( chiefly @xmath13 ) that contribute to higher - order counterterms were considered , and shown to be smaller but still relevant @xcite . the conclusion of ref.@xcite was that it is possible to describe the @xmath20 reaction consistently with qcd , reasonable meson exchanges , and realistic potentials , but only within a very large theoretical uncertainty . it is our intention to assess here our theoretical understanding of the @xmath26 reactions by analyzing the effect of the same microscopic mechanisms . these reactions have been studied for some time . conservation of parity , angular momentum , and isospin constrains the possible channels for these reactions . in the case of unbound final nucleons , sufficiently close to threshold the strong @xmath2-wave two - nucleon interaction implies that the most important channels should be those in which the final nucleons have relative orbital angular momentum @xmath27 . some of these partial waves are listed in table [ tab.01 ] . .partial - wave amplitudes for the @xmath28 reactions that can contribute to the isotropic cross section right at threshold with an @xmath27 final state . @xmath29 denotes the pion angular momentum . [ cols="<,^,^,^ " , ] [ tab.01 ] near threshold we expect the pion , too , to be in an @xmath2-wave , and the cross sections to be nearly isotropic , with @xcite @xmath30 @xmath31 @xmath32 here @xmath33 and @xmath34 have quite different magnitudes due to the different wavefunctions ; in the case of the deuteron final state the effect of the @xmath35 wave is also important , and it is incorporated in the following . the cross section for @xmath36 is related to that of @xmath37 by isospin considerations . there is a relative factor 2 from @xmath21 being pure isospin @xmath38 and @xmath39 having @xmath38 and @xmath40 in equal probabilities ; and there are coulomb effects . the total cross - section for the deuteron channels is function of @xmath41 where @xmath42 is the maximum pion momentum . it is well - known @xcite that sufficiently close to threshold the cross - section for @xmath43 is @xmath44 while for @xmath45 it is @xmath46 here @xmath47 is a constant which depends on the strong interactions that produce the pion and @xmath48 is a coulomb factor arising mostly from the pion - deuteron electromagnetic interaction . if the experimental data for @xmath45 are corrected for these electromagnetic effects , then both reactions near threshold can be viewed as a determination of the single constant @xmath47 . the threshold behavior of the reaction @xmath49 was estimated by gell - mann and watson @xcite , who predicted an energy dependence @xmath50 under the approximation of zero - range nuclear forces . early theoretical analyses of the reaction @xmath51 , like in ours , split it into a kernel where the two - nucleon system emits the pion and effects of initial- and final - state interactions taken into account by using deuteron and scattering wave functions that satisfy the schrdinger equation with a given @xmath4 potential . the numerical results of the early analyses @xcite were that this reaction is dominated by the weinberg - tomozawa ( wt ) term ( fig . [ fig.1]b ) . the main competitor process was thought to be the impulse term ( ia ) ( fig . [ fig.1]a ) . the early estimate of this ia gave a small contribution @xcite due to a cancellation in the matrix elements between the @xmath2 and @xmath35 waves of the deuteron final state . it was believed that the wt term alone could account for the essential features of the data . = 13.0 cm = 4.0 cm with the advent of accelerators capable of producing intense high - quality beams of protons with gev energies and very precise detecting systems , good and accurate data for @xmath36 @xcite , @xmath51 @xcite , and @xmath52 @xcite near threshold became available . for the deuteron channels , the new data confirm the threshold behavior ( [ snp ] ) @xcite , and show that the theoretical calculations started by koltun and reitan were in the right direction . more recent calculations of the wt and ia mechanisms @xcite , however , indicated that their strength might not be sufficient to explain the new data . as for the unbound final state , the recent measurements in the indiana cooler @xcite give @xmath53 a calculation @xcite using ia , on - shell pion rescattering , and @xmath54 z - graphs finds good agreement with data in all @xmath55 , @xmath56 , and @xmath57 channels with a soft @xmath19 form factor . our motivation here is to examine what a chiral power counting suggests for the pion production in the @xmath58 and @xmath49 reactions near threshold . for the @xmath59 reaction , we will compute @xmath60 but will disregard the coulomb interaction in both the initial state which is not so important due to the not - so - small initial energy and final state . this means we do not calculate @xmath48 , and can only compare with coulomb - corrected @xmath21 data . we will see that @xmath60 is indeed approximately linear with @xmath61 at threshold , so our result can be considered a calculation of @xmath47 in eq . ( [ spp ] ) . alternatively , it is a calculation of @xmath62 . for the @xmath56 final state , we will also disregard the coulomb interaction and see that our cross section has an energy dependence similar to the data . moreover , we will show that the wt term is indeed the dominant one , not only for the @xmath63 @xmath56 final state but also for the deuteron channel , which is consistent with the early analyses for this final state . the ia term is smaller than wt because of cancellations between different regions in coordinate space , and between the @xmath2 and @xmath35 waves of the deuteron final state . the same cancellations affect the contribution from an explicit @xmath1 in the intermediate state ( fig . [ fig.1]d ) , more so in the @xmath59 channel than in the @xmath64 channel . the relative smallness of this term here reduces the effects of using different @xmath4 potentials compared to the @xmath20 reaction . however the delta contribution is affected by a number of uncertainties . we include also the same sub - leading terms that proved important in the @xmath20 reaction : isospin - independent rescattering ( st , fig . [ fig.1]b ) , and heavy meson exchange simulating short - range mechanisms ( @xmath54 and @xmath13 , fig . [ fig.1]c , e ) . the sub - leading galilean correction to the wt term ( gc ) is included as well . these contributions are all relatively small . we will see that , if we neglect the uncertain @xmath1 contributions , each of the the cross - sections is underestimated by a factor of @xmath65 . in section ii we discuss the power counting and the chiral lagrangian . in section iii the kernel is obtained . in section iv the calculation of the cross - sections is outlined ; some technical details are relegated to an appendix . section v describes our input and discusses our results . an outlook is presented in section vi . near threshold for pion production the total energy of the two colliding nucleons is of order @xmath66 , so that the center - of - mass initial kinetic energy of each nucleon is @xmath23 . this energy is smaller than the nucleon mass , so we can use a non - relativistic framework . the nonrelativistic kinetic energy formula holds ; the mass @xmath67 plays no dynamical role , and the typical momentum of real and virtual particles involved in this process is @xmath68 . this requires some adaptation of the usual effective theory ideas which have been developed for momenta typical of most nuclear systems , @xmath69 . because the mass difference between the delta isobar and the nucleon , @xmath70 , is numerically of order of the typical excitation energy we are interested in , @xmath6 , the @xmath1 must be included explicitly as a degree of freedom in the lagrangian . on the other hand , @xmath71 is smaller than the characteristic mass scale of qcd , @xmath8gev , at least in the chiral limit @xmath72 , so that the contribution of other states ( the roper , the @xmath25 meson , _ etc . _ ) can be buried in short - range interactions . we thus seek a theory of non - relativistic nucleons and deltas interacting with pions that is consistent with the symmetries of qcd . @xmath5pt is implemented via the most general lagrangian involving these degrees of freedom aided by power - counting arguments . the seminal idea was contained in a paper by weinberg @xcite . this idea was developed systematically for interactions of mesons @xcite and for interactions of mesons with a baryon @xcite . the generalization of these techniques to describe properties of more than one baryon was also due to weinberg @xcite and was carried out in detail in ref . @xcite ; see also refs . @xcite . when dealing with typical momenta of order @xmath69 , the usual power counting suggests that we order terms in the chiral lagrangian according to the `` index '' @xmath73 , where @xmath74 is the sum of the number of derivatives , the number of powers of @xmath75 , and the number of powers of @xmath76 ; and @xmath77 is the number of fermion field operators . in the following we will consider interactions with @xmath1 up to 4 . the lagrangian with @xmath78 for each interaction @xcite is @xmath79n + \frac{g_{a}}{2 f_{\pi } } n^{\dagger}({\mbox{\boldmath $ \tau$}}\cdot\vec{\sigma}\cdot\vec{\nabla}{\mbox{\boldmath $ \pi$}})n \nonumber \\ & & + \delta^{\dagger}[i\partial_{0}- \delta]\delta + \frac{h_{a}}{2 f_{\pi}}[n^{\dagger}({\mbox{\boldmath $ t$}}\cdot \vec{s}\cdot\vec{\nabla}{\mbox{\boldmath $ \pi$}})\delta + h.c . ] + \cdots \ , , \label{la0}\end{aligned}\ ] ] where @xmath80 mev is the pion decay constant , @xmath81 is the isobar - nucleon mass difference , @xmath82 is the axial - vector coupling of the nucleon , @xmath83 is the @xmath84 coupling , and @xmath85 and @xmath86 are the transition spin and isospin matrices , normalized such that @xmath87 notice that we defined the fields @xmath88 and @xmath1 in such a way that there is no factor of @xmath89 in their time evolution . hence @xmath10 does not appear explicitly at this index , corresponding to static baryons . we also wrote @xmath90 in the rest frame of the baryons , which is the natural choice . ( galilean invariance will be ensured by including terms with additional derivatives . ) chiral symmetry determines the coefficient of the so - called weinberg - tomozawa term ( @xmath91 ) but not of the single - pion interactions ( @xmath92 ) . the lagrangian with @xmath93 is @xcite , @xmath94 \nonumber \\ & & + \frac{1}{f_{\pi}^{2}}n^{\dagger}[(c_2 + c_3 - \frac{g_a ^2}{8 m_{n } } ) \dot{{\mbox{\boldmath $ \pi$}}}^{2 } -c_3 ( \vec{\nabla}{\mbox{\boldmath $ \pi$}})^{2 } -2c_1 m_{\pi}^{2 } { \mbox{\boldmath $ \pi$}}^{2 } - \frac{1}{2 } ( c_4 + \frac{1}{4m_{n } } ) \varepsilon_{ijk } \varepsilon_{abc } \sigma_{k } \tau_{c } \partial_{i}\pi_{a}\partial_{j}\pi_{b}]n \nonumber \\ & & + \frac{\delta m_{n}}{2 } n^{\dagger}[\tau_{3}-\frac{1}{2 f_{\pi}^{2 } } \pi_3 { \mbox{\boldmath $ \pi$}}\cdot{\mbox{\boldmath $ \tau$}}]n + \frac{1}{2m_{n}}\delta^{\dagger}[\vec{\nabla}^{2 } + \cdots]\delta \nonumber \\ & & -\frac{g_{a}}{4 m_{n } f_{\pi}}[in^{\dagger}{\mbox{\boldmath $ \tau$}}\cdot\dot{{\mbox{\boldmath $ \pi$ } } } \vec{\sigma}\cdot\vec{\nabla}n + h.c . ] -\frac{h_{a } } { 2 m_{n } f_{\pi } } [ in^{\dagger}{\mbox{\boldmath $ t$}}\cdot\dot{{\mbox{\boldmath $ \pi$}}}\vec{s}\cdot\vec{\nabla } \delta + h.c . ] \nonumber \\ & & -\frac{d_1}{f_{\pi } } n^{\dagger}({\mbox{\boldmath $ \tau$}}\cdot\vec{\sigma}\cdot\vec{\nabla}{\mbox{\boldmath $ \pi$}})n\ , n^{\dagger}n -\frac{d_2}{2 f_{\pi } } \varepsilon_{ijk } \varepsilon_{abc } \partial_{i}\pi_{a } n^{\dagger}\sigma_{j}\tau_{b}n\ , n^{\dagger}\sigma_{k}\tau_{c}n + \cdots \ , , \label{la1}\end{aligned}\ ] ] where the @xmath95 s are coefficients of @xmath96 , @xmath97 is the quark mass difference contribution to the neutron - proton mass difference , and the @xmath98 s are coefficients of @xmath99 . these seven numbers are not fixed by chiral symmetry , but it is important to point out that galilean invariance requires that the other coefficients explicitly shown above be related to those appearing in @xmath90 . this in particular fixes the strength of the single - pion interactions in terms of the lowest - order coefficients @xmath100 and @xmath101 , and of the common mass @xmath102 . the lagrangian with @xmath103 is @xmath104 \nonumber \\ & & -\frac{e_1}{2 m_n f_{\pi } } [ in^{\dagger}{\mbox{\boldmath $ \tau$}}\cdot\dot{{\mbox{\boldmath $ \pi$}}}\vec{\sigma}n\ , \cdot n^{\dagger}\vec{\nabla}n + h.c . ] \nonumber \\ & & + \frac{e_2}{2 m_n f_{\pi } } [ n^{\dagger}{\mbox{\boldmath $ \tau$}}\cdot\dot{{\mbox{\boldmath $ \pi$}}}\vec{\sigma}\times \vec{\nabla}n\ , \cdot n^{\dagger}\vec{\sigma}n + h.c . ] + \cdots \ , , \label{la2}\end{aligned}\ ] ] where the @xmath105 s are other coefficients of @xmath99 . among the lagrangians with higher indices , we find @xmath106 + \cdots \ , , \label{la3}\ ] ] where @xmath107 is a coefficient of @xmath108 . the two nucleons in the @xmath109 reaction can interact repeatedly by the exchange of mesons of momenta @xmath110 before and after the emission of the pion . we account for this through the iteration of a potential , which produces initial and final wavefunctions that differ from the free ones . the emission of the pion , on the other hand , involves the larger momentum @xmath68 . the sub - diagrams that involve such typical momentum form the kernel of `` irreducible diagrams '' , which is evaluated between wavefunctions . the unusually high momentum in the kernel requires modification in the usual power counting . while in the usual power counting energy and momenta are counted as equal , here energies are @xmath111 but momenta @xmath112 . this changes the usual correspondence between index and order . the leading nucleon propagator , for example , includes both the static term in ( [ la0 ] ) and the kinetic term in ( [ la1 ] ) at the same order in power counting for pion production , contrary to the situation in @xmath5pt applied to one - baryon systems . the nucleon propagator then is @xmath113 . relativistic corrections @xmath114 are relatively smaller by @xmath115 and can be considered higher - order insertions . note that this is completely consistent with our decomposition of the full amplitude into a kernel and wavefunctions obtained from a schrdinger equation , and , contrary to what is stated in ref . @xcite , does _ not _ imply the need of a relativistic framework . the delta propagator differs from the nucleon by the presence of the mass difference @xmath116 . as pointed out first in ref . @xcite , the delta propagator is then actually @xmath117 and tends to interfere destructively with the nucleon contributions to the kernel . the pion propagator , on the other hand , is @xmath118 . in interactions , each time derivative is associated with a factor of @xmath6 , while a space derivative a factor of @xmath71 . finally , a loop brings a @xmath119 . a detailed analyses of various contributions can found in ref . the new elements here are those associated with the isospin - dependent wt pion rescattering . the order of the wt contribution is evaluated as follows . the term proportional to @xmath120 of eq . ( [ la0 ] ) yields an explicit factor of @xmath121 ; the pion - nucleon interaction provides a factor of @xmath122 , and the pion propagator @xmath123 . the total net result is of order @xmath124 , which is the same order of the impulse approximation and the delta contribution @xcite . therefore due to power counting arguments , we expect that contribution of wt , ia , and @xmath1 terms to have the same importance ; they constitute our leading order . we will also include other terms which are of order @xmath125 or higher relative to the leading one : the isospin - independent pion rescattering ( sg ) , the galilean correction to the wt term ( gc ) , and contact terms , modeled by the heavy meson exchange ( @xmath11 , @xmath12 and @xmath13 ) . we now obtain the explicit forms of the various contributions by evaluating the most important irreducible diagrams in momentum space . our notation is as follows : @xmath126 is the energy of the ( on - shell ) pion produced with momentum @xmath127 in the center of mass ; @xmath128 ( @xmath129 ) is the center - of - mass momentum of the incoming ( outgoing ) nucleon labeled `` 1 '' ( those of nucleon `` 2 '' are opposite ) ; @xmath130 ( @xmath131 ) is the momentum ( energy ) transferred ; @xmath132 ; @xmath133 ; @xmath134 is the spin of proton @xmath135 ; @xmath136 where @xmath137 is the isospin of emitted pion ; and @xmath138 . we define the @xmath139-matrix in terms of the @xmath2-matrix via @xmath140 . according to the previous discussion , we expect the leading contributions to arise from the diagrams in figs . [ fig.1]a , b , and d. in the case of the weinberg - tomozawa diagram ( fig . [ fig.1]b ) , we get @xmath141 the galilean correction to wt contribution is smaller by a factor of @xmath142 and is given by @xmath143 the impulse term ( fig . [ fig.1]a ) is discussed in detail in @xcite . the principle of irreducibility says that we have to redraw this diagram . since the outgoing pion carries an energy of the order of the pion mass , the energies of the @xmath4 intermediate state before and after pion emission differ by @xmath111 . therefore both of the intermediate states can not simultaneously be within @xmath144 of being on - shell : at least one intermediate state , before or after emission , is off shell by @xmath111 . this single , relatively - high - momentum ( @xmath145 ) pion exchange must therefore be included in the irreducible class of operators for our process ( unlike the usual case ) . all other initial- and final - state interactions will be considered reducible and included in the wave functions . therefore , in the case of pion exchange with a nucleon in the intermediate state we get @xmath146\ , , \label{ia}\ ] ] which is listed for comparative purposes only . we will actually calculate the impulse approximation directly from eq.(6 ) in ref . @xcite , in the same fashion as done by koltun and reitan @xcite . recoil corrections are expected to be smaller by a factor of @xmath147 . since the impulse contribution will prove to be small , we can ignore the recoil contributions in this first approach . the @xmath1 contribution ( fig . [ fig.1]d ) to the kernel is given by @xmath148\nonumber \\ & & + i\epsilon_{abc}\tau_b^{(1)}\tau_c^{(2)}\left[\left(\omega_q\,\vec{k } \cdot\vec{p}-\delta\,\vec{k}^2\right)\vec{s}\cdot\vec{k}- \frac{i}{4}\delta\left(\vec{\sigma}^{(1 ) } \cdot\vec{k}\;\vec{\sigma}^{(2)}\cdot(\vec{p}\times\vec{k } ) + \vec{\sigma}^{(1)}\cdot(\vec{p}\times\vec{k } ) \;\vec{\sigma}^{(2)}\cdot\vec{k}\right)\right ] . \label{delta}\end{aligned}\ ] ] results similar to eqs . ( [ ia ] ) and ( [ delta ] ) follow for shorter - range terms where the two nucleons exchange a heavier meson rather than a pion . in the case of a nucleon intermediate state , such a contribution is automatically included in the potential . in any reasonable model , the contributions from a delta intermediate state turn out to be smaller than those in diagram of fig . [ fig.1]d . for example , they could arise from @xmath33 exchange , but then the relatively high @xmath33 mass suppresses this contribution ; the contribution from the @xmath25 which is formally of higher - order is likely to be more important . since , as we are going to see , the delta contribution from pion exchange is not large , we will not go into such detailed analysis for the purpose of estimating the effect of the @xmath1 : we use eq.([delta ] ) . we will further discuss the uncertainties related to the delta contribution below . there are other corrections of order @xmath147 compared to the leading terms . [ fig.1]b represents also isospin - independent rescattering : @xmath149\vec{\sigma}_a\cdot\vec{k } \right .. \nonumber \\ & - & \left.\frac{\delta m_n}{8}\left[\delta_{3a}\vec{\tau}^{(1)}\cdot \vec{\tau}^{(2)}\,\left(\vec{\sigma}^{(1)}-\vec{\sigma}^{(2)}\right ) + \vec{\sigma}^{(1)}\tau_3^{(1)}\tau_a^{(2)}- \vec{\sigma}^{(2)}\tau_a^{(1)}\tau_3^{(2)}\right]\cdot\vec{k}\right\}. \label{st}\end{aligned}\ ] ] the short - range mechanisms provided by the @xmath150 lagrangians involve several unknown constants . chiral symmetry tells us nothing about the strength of these coefficients . we can use data to determine some of them . alternatively , we can use a model to determine these coefficients and then try to explain the experimental results . here we use the mechanism first proposed by lee and riska @xcite and by horowitz _ et al _ @xcite , where the short - range interaction is supposed to originate from z - graphs with @xmath11 and @xmath12 exchanges , as shown in fig . [ fig.1]c . in this case , @xmath151 , \label{sigom}\end{aligned}\ ] ] where @xmath152 ( @xmath153 ) and @xmath154 ( @xmath155 ) are the mass and the vector coupling to nucleons of the @xmath11 ( @xmath12 ) meson , and @xmath156 denotes the ratio of tensor to vector coupling for the @xmath12 meson . in the case of @xmath157 , the contribution due to @xmath13 exchange ( fig . [ fig.1]e ) is not negligible @xcite , so we also include it here in order to get an estimate of the convergence of our expansion . this contribution leads to @xmath158 where @xmath159 is the @xmath160 coupling @xcite and the other coefficients have the same meaning as in the previous equation . at momenta much smaller than the meson masses , these short - range contributions are indeed contact interactions . we are concerned with evaluating the matrix elements of the above operators between the initial @xmath161 and the deuteron final wave functions . to evaluate the influence of the potential in the amplitudes , we use reid93 @xcite and argonne v18 @xcite potentials which , for a given @xmath21 channel , are local potentials . thus we evaluate the operators between coordinate space initial ( @xmath135 ) and final ( @xmath77 ) wave functions expressed by @xmath162 where @xmath163 ( @xmath164 ) is for third - component of the angular momentum @xmath165 ( @xmath166 ) , and @xmath167\;,\ ] ] where the deuteron wave functions are normalized as @xmath168=1\;.\ ] ] here and in the following the spin angular part is expressed , as usual , as @xmath169 where @xmath170 is the spherical harmonic function . we convert the operators of eqs . ( [ wt]-[rhoom ] ) to configuration space by inverting the fourier transforms . the resulting operators can then be used in configuration - space matrix elements . we define the matrix elements of the operators of eqs . ( [ wt]-[rhoom ] ) as @xmath171\,u_{1,1}(r ) , \label{mdef}\end{aligned}\ ] ] where @xmath172 represents @xmath173 , @xmath174 , _ etc . _ @xmath175 is the corresponding operator , obtained using the matrix elements given in the appendix . to compare results before the cross - section evaluation , we define dimensionless amplitudes @xmath176 , where @xmath177 _ etc . _ via @xmath178 we will plot @xmath179 in terms of @xmath61 to evaluate the energy dependence of the amplitude , and @xmath180 in terms of @xmath181 to study the @xmath181 dependence of the integrand . the contribution from the weinberg - tomozawa term is given by @xmath182 where @xmath183 . the galilean correction to the wt term is given by @xmath184 , \label{gcs}\\ h^{gc}_d(r)&=&{g_a \over 4 f^3_\pi}\frac{3 m_\pi}{4\pi } \sqrt{\frac{2}{3}}\ ; \frac{m_\pi}{4m_n}\left[a_{gc}(r)+b_{gc}(r)+d_{gc}(r ) \right ] , \label{gcd}\end{aligned}\ ] ] where @xmath185 the impulse approximation is given by @xmath186 , \label{ias}\\ h^{ia}_d(r)&=&{g_a \over f_\pi}\frac{m_\pi}{m_n}\sqrt{\frac{1}{3}}\ ; \left[-\frac{2}{r}+\frac{\partial}{\partial r}\right ] . \label{iad}\end{aligned}\ ] ] the @xmath1 resonance contribution be written as @xmath187 , \label{dels}\\ h^{\delta}_d(r)&=&\frac{1}{9m_n}\left({g_a \over f_\pi}\right)^3\left ( \frac{h_a}{g_a}\right)^2\frac { m_\pi \tilde{m}_\pi^2}{(m_\pi-\delta ) } \sqrt{\frac{2}{3}}\;\frac{1}{4\pi}\left[a_{\delta}(r)+b_{\delta}(r)+ d_{\delta}(r)\right ] , \label{deld}\end{aligned}\ ] ] where @xmath188 the isospin - independent seagull term is given by @xmath189 where @xmath190 we also include the effects of the @xmath11 and @xmath12 exchange and @xmath13 term . the result for the @xmath11 and @xmath12 exchange is @xmath191 \left({\partial\over\partial r}+{1\over r}\right ) + \frac{1}{2}\frac{\partial}{\partial r}\left [ f_\sigma(r)+ f_\omega(r)\right]\right\}\ , , \label{sig - oms}\ ] ] and @xmath192 \left({\partial\over\partial r}-{2\over r}\right ) + \frac{1}{2}\frac{\partial}{\partial r}\left [ f_\sigma(r)+ f_\omega(r)\right]\right\}\ , , \label{sig - omd}\ ] ] where the function @xmath193 accounts for exchange of the meson @xmath194 between nucleons , @xmath195 we follow ref.@xcite by including the effects of form factors in these heavy - meson contributions , as given in the bonn potential @xcite . for completeness , we repeat that monopole form factors are used at each meson vertex according to the replacement @xmath196 where @xmath197 is the transferred momentum and @xmath198 is the cutoff mass . with @xmath199 denoting the common @xmath13 mass ( 780 mev ) , @xmath13 exchange yields @xmath200 \left({\partial\over\partial r}+{1\over r}\right)\right . \nonumber \\ & & \left . \;\ ; -[b_{\rho-\omega}(r ; m , \lambda ) -b_{\rho-\omega}(r ; \lambda , m ) ] \left({\partial\over\partial r}-{1\over r}\right)\right\ } , \label{ros}\end{aligned}\ ] ] where @xmath201,\nonumber \\ b_{\rho-\omega}(r ; m_1 , m_2 ) & = & m_1 \ , \frac{e^{-m_1r}}{m_1r}\left [ 4m_1 ^ 2\left(1+\frac{3}{m_1r}+ \frac{3}{(m_1r)^2}\right)-(m_2 ^ 2-m_1 ^ 2)(1+m_1r)\right].\end{aligned}\ ] ] @xmath202 can be obtained making the replacement @xmath203 above . the final steps consist of computing the total matrix element @xmath204 , @xmath205 squaring it , and integrating over the available phase space . we find @xmath206 where @xmath207 is the magnitude of the center - of - mass 3-momentum , @xmath208 and @xmath209 is energy of the produced deuteron of mass @xmath210 , @xmath211 . as we did in the deuteron case , we evaluate the matrix elements for the unbound final state using the same operators of eqs . ( [ wt])-([rhoom ] ) . we will consider here just the absolute threshold limit where @xmath212 . according to selection rules , we have 2 channels , as stated in tab . [ tab.01 ] : @xmath213 ( @xmath214 ) and @xmath215 ( @xmath216 ) . again , to evaluate the influence of the potential in the amplitudes , we use reid93 @xcite and argonne v18 @xcite potentials which , for a given @xmath21 or @xmath217 channel , are local potentials . thus we evaluate the operators between coordinate space initial ( @xmath135 ) and final ( @xmath77 ) wave functions for the two channels , expressed by * @xmath218 @xmath63 channel : + @xmath219 * @xmath220 @xmath221 channel : + @xmath222 where @xmath163 ( @xmath164 ) is for third - component of the angular momentum @xmath165 ( @xmath166 ) . we convert the operators of eqs . ( [ wt]-[rhoom ] ) to configuration space by inverting the fourier transforms . the resulting operators can then be used in configuration - space matrix elements . we define the matrix elements of the operators of eqs . ( [ wt]-[rhoom ] ) as * @xmath218 @xmath63 channel + @xmath223 * @xmath220 @xmath221 channel + @xmath224 with @xmath225 j^y&=&\frac{m_\pi}{p\,p'}\int_0^\infty dr\,u_{0,0}\,h^y(r)\,u_{1,0}(r)\end{aligned}\ ] ] where @xmath172 and @xmath226 represents @xmath1 , @xmath174 , etc . , @xmath176 and @xmath227 are dimensionless integrals , @xmath228 and @xmath229 are the corresponding operators , obtained using the matrix elements given in the appendix , which are the same used in the deuteron final state ( for the @xmath218 @xmath63 channel ) and in the @xmath20 @xcite ( for the @xmath220 @xmath221 channel ) . we will plot @xmath179 and @xmath227 in terms of @xmath230 to evaluate the energy dependence of the amplitude , and @xmath180 and @xmath231 in terms of @xmath181 to study the @xmath181 dependence of the integrand . for the @xmath218 @xmath63 channel , the coordinate space expressions for the amplitudes are pretty much the same of the @xmath2-wave deuteron amplitudes . for instance , the weinberg - tomozawa term gives @xmath232 where @xmath183 . for the @xmath233 channel , the expressions follow closely the work of cohen _ et al _ @xcite , the isospin matrix elements being the only difference . as we did in the deuteron case , we include the effects of form factors when dealing with heavy meson contributions . the final steps consist of computing the total matrix element @xmath204 for the two channels , using eq . ( [ calm ] ) for the @xmath63 final state and the same equation but without wt and the gc terms for the @xmath221 final state , since in this state the isovector contributions are zero . the cross section is obtained by squaring the total amplitude , and integrating over the available phase space . we find @xmath234 where @xmath235 is the laboratory velocity of the incident proton , @xmath42 is the pion 3-momentum , @xmath236 , @xmath207 is the magnitude of the center - of - mass initial 3-momentum and @xmath237 . the @xmath238 indicates that ( a ) a sum over final spin states and ( b ) an average over initial spin states must be made , which result in factors of @xmath239 for the @xmath63 and @xmath240 for the @xmath221 final states . the various amplitudes considered in the last section depend on several parameters that we can determine from other processes . the wt , gc to wt , and the impulse - approximation operators depend on the pion mass , @xmath241 mev @xcite , and on @xmath242 we use the value of @xmath243 appropriate for each potential . the @xmath1 operator of eq.([delta ] ) further depends on the @xmath244 mass splitting @xmath245 mev @xcite and on the @xmath246 coupling constant , @xmath101 . this has been fixed from @xmath247-wave @xmath19 scattering ( see , e.g. , ref.@xcite ) , @xmath248 the seagull operator of eq.([st ] ) depends on four parameters @xmath249 and @xmath250 . the @xmath95 s can be obtained by fitting @xmath2-wave @xmath19 scattering . in ref.@xcite they were found to be @xmath251=-\frac{2.31}{2m_n}\ ; , \ ] ] from the @xmath11-term , the isospin - even scattering length , and the axial polarizability , to @xmath252 . we refer to this as `` seai '' . a different determination from an @xmath253 fit to @xmath19 sub - threshold parameters @xcite gives @xmath254 instead . we refer to this as `` seaii '' . newer determinations @xcite give values closer to the more negative value , but we use both values in order to estimate the importance of this contribution . note that the analysis of ref.@xcite does not include the isobar explicitly . since the inclusion of the @xmath246 interaction only affects @xmath2 waves at one order higher than the @xmath95 s , the above values can still be used to estimate the effect of @xmath2-wave rescattering . the parameter , @xmath250 , can in principle also be determined from s - wave @xmath19 scattering , but would require a careful analysis of other isospin - violating effects . chiral symmetry relates it to the strong interaction contribution to the nucleon mass splitting , which is also difficult to determine directly . estimates of the electromagnetic contribution @xmath255 are more reliable , @xmath256 mev @xcite , and give @xmath257 mev . to be definite , we use @xmath258 finally , the @xmath259 and the @xmath13 operators involve @xmath260 , @xmath261 , @xmath262 , and @xmath156 , parameters listed in table a.3 of ref . they also involve @xmath159 , discussed in ref . these heavy - meson contributions correspond to chiral lagrangian coefficients of natural size : @xmath263 , @xmath264 , @xmath265 , and @xmath266 . the relative sizes of the various contributions to the matrix element @xmath267 of this reaction as function of @xmath61 are shown in fig . [ fig.2 ] for the reid93 potential and in fig . [ fig.3 ] for the av18 potential . there is a very strong similarity between these two sets of results . = 13.0 cm = 13.0 cm our leading order comprises wt , ia and @xmath1 . the first noticeable observation is that wt is by far the largest contribution . this is a consequence of cancellations in ia and @xmath1 which were not anticipated by the power counting . in figs . [ fig.4 ] , [ fig.5 ] , and [ fig.6 ] we can see typical integrands for the three contributions at @xmath268 , in the case of the argonne v18 potential . while for wt the contributions from the @xmath2 and @xmath35 deuteron waves add and are dominated by the region around @xmath269 fm , for ia and @xmath1 the @xmath2 and @xmath35 waves tend to interfere destructively , and contributions from different @xmath181 regions approximately cancel . = 13.0 cm = 13.0 cm = 13.0 cm our formally sub - leading contributions are all considerably smaller than wt . this suggests that the theoretical control of this reaction is greater than for @xmath20 . moreover , sub - leading contributions come with different signs , partially cancelling . in fact , the @xmath11,@xmath12 and gc terms come out with similar size as predicted by the power counting but with opposite signs ; there is an almost complete accidental cancellation between them . @xmath13 exchange is of higher order and indeed very small . as a result , when the small contribution from the combination of seagull parameters `` seaii '' is considered , the sum of all sub - leading contributions is small . the sum is somewhat larger and of opposite sign to leading order when `` seai '' is used . in fig . [ fig.10 ] we compare our leading and sub - leading results for the @xmath21 cross - section ( without coulomb ) using the reid93 potential with data from refs . we see that our curves are approximately constant , as expected . recent data points for @xmath270 are not all consistent , but they cluster around an average @xmath47 of about 180 to 190 @xmath271b . the value of @xmath47 in leading order ( wt+ia+@xmath1 ) is a factor of about 1.5 below data . the destructive interference with next - order contributions increases the disagreement by an amount depending on the value of @xmath19 isoscalar rescattering term . the change is bigger when `` seai '' is used . this can be seen in fig . [ fig.10 ] as `` sumi '' and `` sumii '' . because of the cancellations among sub - leading terms , the result exhibits comparable dependence on the short - range contributions , as can be seen in fig . [ fig.10 ] as `` sumii without @xmath11 , @xmath25 , @xmath12 '' . we summarize our results in fig . [ fig.11 ] where we show the sum of all the contributions we considered for the two sets of seagull parameters , and for the two potentials . = 13.0 cm = 13.0 cm the relative sizes of the various contributions to the matrix element @xmath267 of this reaction as function of @xmath230 are shown for the @xmath63 final state in fig . [ fig.12 ] for the reid93 potential and in fig . [ fig.13 ] for the av18 potential ; and for the @xmath221 final state in fig . [ fig.14 ] for the reid93 potential and in fig . [ fig.15 ] for the av18 potential . = 13.0 cm = 13.0 cm = 13.0 cm = 13.0 cm here again the wt contribution is the largest ; since it contributes only to the @xmath63 final state , this channel is dominant . most of the other contributions are much smaller and tend to cancel to some extent . the exception is the @xmath1 , which has a significant destructive interference with wt in the @xmath63 state . the ia contribution is small due to the same type of cancellation observed before among different regions in coordinate space . in fig . [ fig.19 ] we compare our results for the different final states using the reid93 potential with data from ref . @xcite . in fig . [ fig.20 ] we summarize our results for the two potentials considered and compare them to the same data . we see that the theory produces a correct shape for the @xmath61 dependence but fails in magnitude by a factor of @xmath272 . use of `` seai '' further worsens the results . once again the differences between the two potentials are minimal . = 13.0 cm = 13.0 cm the @xmath57 final state has been considered before in the literature . @xcite has found that the impulse term was very small due to the strong cancellation between the @xmath2 and @xmath35 waves in the matrix element shown above . mostly due to the wt , @xmath47 of 146 to 160 @xmath271b was obtained using the older , higher value of the @xmath273 coupling constant ; using the more recent value , we get 124 to 136 @xmath271b . a similar analysis @xcite included also form factors at the @xmath274 vertices , which led to a decrease of approximately 20% in the cross section ; with the more modern coupling constants , the overall result was near 100 @xmath271b . recently , ref . @xcite re - analyzed this reaction using a covariant approach . the result for @xmath47 using only the wt term was again near 100 @xmath271b , but the amplitude contains what we refer to as the galilean correction to the wt term , which has an opposite sign . our results for the wt term alone are in numerical agreement with these works , since we find that it gives an @xmath47 of about 100 @xmath271b . other contributions tend to worsen the description of the data . particularly damaging are the contributions of the rescattering type . here , as for @xmath20 , the @xmath2-wave seagulls tend to interfere destructively with the leading mechanism . as in the case of the @xmath20 reaction , the @xmath1 contribution shows appreciable dependence on the potential used . but unlike that reaction , here it generates relatively small dependence on the potential in the final result , as consequence of the large relative size of wt . the @xmath1 contribution is the one that presents the largest change when we compare @xmath217 and @xmath74 final states . in fig . [ fig.21 ] we compare the radial distributions of the delta contribution for the @xmath63 @xmath217 final state at @xmath275 @xmath276 and for the deuteron final state ( same as in fig . [ fig.6 ] ) . from this we can see how this contribution has different signs in the two channels , being small and constructive with wt for @xmath74 ( @xmath277 and @xmath278 ; @xmath268 ) and larger and destructive for @xmath217 ( @xmath279 and @xmath280 ; @xmath268 ; @xmath275 ) = 13.0 cm note , however , that the @xmath1 contribution is subject to large uncertainties . first , the @xmath281 coupling constant is not well determined and appears squared ; although we do not know whether this coupling should be bigger or smaller than the value used here , it could contribute to a decrease of the @xmath1 effect . second , the @xmath281 form factor seems to be much softer than the corresponding @xmath273 form factor ; our neglecting both enhances the @xmath1 contribution relative to the wt term . third , we have neglected the @xmath244 mass difference in energies ; since the @xmath1 mass would appear in the denominator , the delta contribution would decrease by a factor of @xmath282 . fourth , we have neglected the kinetic energy of the delta , account of which would further decrease its amplitude . fifth , it is known that there can be some cancellation between the pion- and rho - exchange delta terms ; including the latter would also diminish the delta contribution . assuming each of these gives a 10% decrease , we could very well be overestimating the @xmath1 contribution by 50% or more . if we neglect this contribution altogether , we obtain the results in fig . [ fig.22 ] for the cross section in the @xmath74 channel and in fig . [ fig.23 ] for the cross section in the @xmath217 channel . this brings theory to underestimate both sets of data by a common factor of @xmath283 . this implies that there must be further corrections to the amplitude of about 50% . this is not unlike the @xmath20 reaction considered in ref . @xcite , where theory tends to fail by a similar factor . the case for failure of theory there is less clear - cut , however , because of the lack of a large contribution as for the wt term here . as a consequence , the usually small effect of other mechanisms is enhanced and the result is dominated by shorter - range dynamics ; more sensitivity to the potential and seagull terms surfaces , and it is possible to find a combination of parameters that includes data @xcite . no such gimmicks work here . for example , we find that heavy - meson exchange hailed as solution in the @xmath20 reaction does not help much in @xmath284 and @xmath285 , in agreement with the findings in refs . @xcite . the dependence on the @xmath1 contribution ( which is a particular type of rescattering ) suggests that we need better control over longer - range contributions , such as @xmath19 rescattering and two - pion exchange . = 13.0 cm = 13.0 cm we have calculated the cross section near threshold for the reactions @xmath286 , using a chiral power counting to order interactions . the interactions included contain most of the interactions found in the literature . in particular , they include mechanisms used in various models of the @xmath20 reaction . in contrast to the latter , results here depend relatively little on the wavefunction employed and on short - range interactions . the weinberg - tomozawa term dominates , but the delta contribution can be important , with large uncertainty . our computed results typically fall a factor of about two below the data . our kernels vary as the cube of a generic meson - nucleon coupling constant , so such discrepancy can be parametrized as a 12% deficiency in the coupling constants . the pion - nucleon coupling constant is known to higher precision than that , but not to much higher precision . thus the discrepancy we find might not be a very serious problem . our main conclusion is that a relatively long - range mechanism such as @xmath19 rescattering and/or two - pion exchange is needed for the description of these reactions . accordingly , we suggest that advance in understanding pion production in @xmath4 collisions must follow not from the study of @xmath20 by itself as has been the trend of theoretical study to date but from focus on an understanding of long - range effects that afflict all channels . we thank c. hanhart for a helpful discussion . this research was supported in part by the u.s . doe under grant de - fg03 - 97er41014 and nsf under grant phy 94 - 20470 . the work of c.a.dr . was supported by the brazilian fapesp under contract numbers 97/05817 - 0 and 97/6209 - 4 . he thanks the nuclear theory group at the university of washington for hospitality during the initial stages of this work . where @xmath288 and @xmath289 are spin operators or scalar products . the index @xmath137 represents the isospin of the emerging pion , @xmath290 is the isospin state of the deuteron and @xmath291 the @xmath21 isospin initial state . u. van kolck , in _ proceedings of the workshop on chiral dynamics 1997 , theory and experiment _ , ed . a. bernstein , d. drechsel , and t. walcher ( springer - verlag , 1998 ) , hep - ph/9711222 ; nucl . * a645 * ( 1999 ) 273 . cohen , j.l . friar , g.a . miller and u. van kolck , phys . c * 53 * , 2661 ( 1996 ) ; bull . . soc . * 40 * , 1629 ( 1995 ) , presented at the dnp / aps fall meeting in bloomington , indiana , october 1995 .

we determine the wavefunctions of electrons bound to a positively charged mesoscopic metallic cluster covered by an insulating surface layer . the radius of the metal core and the thickness of the insulating surface layer are of the order of a couple of ngstrm . we study in particular the electromagnetic decay of externally located electrons into unoccupied internally located states which exhibits a resonance behaviour . this resonance structure has the consequence that the lifetime of the `` mesoscopic atoms '' may vary by up to 6 orders of magnitude depending on the values of the parameters ( from sec to years ) .

cond-mat0311377

clusters consisting of some 10@xmath0 to 10@xmath1 atoms , which can be produced for instance in the course of the condensation of vapour , have attracted great interest during the last 20 years . these mesoscopic clusters were found to exhibit classical as well as quantum mechanical features . in the special case of metallic clusters @xcite , the conduction electrons , which move in the average potential produced by the positively charged ions , exhibit shell effects . as a consequence , the binding energy of the cluster becomes a function of the number of free electrons ( and thus also of the number of ions ) . in turn , this implies measurable variations of the formation probability of a cluster as a function of its mass in a thermal vapour - gas equilibrium . other interesting types of mesoscopic systems are the so - called fullerenes which represent a crystal of atoms covering a closed , usually spherical surface @xcite . a number of review articles exist on this exciting new field of physics ( see for instance ref . @xcite and ref . @xcite and books mentioned at the end of ref . @xcite and ref . @xcite ) . in the present paper , we consider a mesoscopic system which consists of a spherical metallic core of radius @xmath2 carrying @xmath3 positive charges and surrounded by an insulating layer of external radius @xmath4 ( see fig . 1 ) . we determine analytically the bound states of a single electron in the potential @xmath5 which is produced by the positive surface charge density of the metallic core and the insulating dielectric surface layer . as in the case of an ordinary atom , these single electron wave functions describe also approximately the case that several electrons are bound in this potential . for @xmath3 bound electrons , the system has total charge 0 . because of the obvious similarity of this system with an atom or ion , we refer to it as a mesoscopic atom or mesoscopic ion . as one can surmise , there are two different types of bound electronic states : electrons localized in the metallic core inside of the insulating layer ( region ii , fig . 1 ) and electrons localized in the vacuum outside of the insulating layer ( region 0 , fig . 1 ) . the bound states localized outside are reminiscent of the bound states of an electron in the coulomb potential of the atomic nucleus . we are mainly interested in these externally localized electronic states ( class 1-states ) and their decay into lower - lying bound states predominantly localized within the metallic core ( class 2-states ) . the paper is organized as follows : in sec . 2 , we formulate the single - particle hamiltonian and determine the eigenfunctions and eigenenergies of bound electrons . in doing so , we take into account the polarization of the insulating surface layer which is produced by the uniform positive charge distribution on the surface @xmath6 of the metal , but we neglect the additional polarization effects produced by the externally localized electrons . in sec . 3 we determine these additional polarization effects within the classical maxwellian electrodynamics . they lead to an additional potential acting on the electrons which is of importance close to the surface @xmath7 of the insulator . so far , we have not yet studied the effect of this additional potential on the externally localized electrons . the sec . 4 is dedicated to a presentation of the results . in subsec . 4.1 , we describe the procedure which we used in order to obtain the eigenvalues and eigenfunctions and we present typical examples of spectra and wavefunctions . in subsec . 4.2 , we discuss the electromagnetic decay of a predominantly externally localized bound electron into a lower - lying internally localized state . we show that the transition rate and the corresponding life - time may vary by several orders of magnitude depending on the parameters of the system . in general , the transition is hindered by the small overlap between externally and internally localized states ( class 1-states and class 2-states ) if the insulating surface layer is sufficiently thick . in particular , we discuss the special role in this decay of class 3-states which exhibit non - negligible amplitudes both in the external region 0 and the internal region ii . finally , in sec . 5 , we summarize the results and discuss some related open questions . [ fig1 ] , surface @xmath6 ) covered by an insulating layer ( outer radius @xmath4 , surface @xmath7 ) . states of bound electrons predominantly localized in the surrounding vacuum ( @xmath8 , region 0 ) are called class 1-states in distinction from states localized in the metallic core ( class 2-states).,width=264 ] in the single - particle hamiltonian for an electron bound to a positively charged and insulated metallic core ( @xmath9 = mass of the electron ) @xmath10 the potential @xmath5 contains the interaction of the electron with the positive surplus charge @xmath11 ( @xmath12 = elementary charge ) and with the ionic background ( jellium ) of the metal . furthermore , it contains the effects of the insulating surface layer . the radius @xmath2 of the metal and the thickness ( @xmath13 ) of the surface layer are assumed to be of the order of 5 to 20 (1 = 10@xmath14 m ) and the number @xmath3 of positive charges of the metal is considered to be 1 to 5 . for these dimensions of the system we may expect that , on the one hand , the dynamics of the electron is fully quantum - mechanical and , on the other hand , we may use macroscopic concepts in formulating the potential @xmath5 , at least as a first approximation . so we assume that the positive surplus charge is uniformly distributed over the surface @xmath6 of the metal @xmath15 the coulomb potential produced by this simple charge distribution acting upon an electron ( charge @xmath16 ) is given by @xmath17 where @xmath18 is the heaviside function @xmath19 inside of the metal , the conduction electrons feel not only the constant potential produced by the positive surface charge but also the attractive interaction with the ionic background which we represent by a constant negative potential @xmath20 @xmath21 ) . on the other hand , the average energy gap above the fermi level in the insulator may act upon the electron like a repulsive barrier of a height @xmath22 . last not least , there are polarization effects produced by the positive surplus charge and also by the charge distribution @xmath23 of the externally ( i.e. predominantly in region 0 ) localized electrons . the polarization effects are investigated in sec . 3 on the basis of maxwell s equations for the electric field @xmath24 and for the displacement field @xmath25 in the presence of homogeneous dielectric media . in the case that we only take into account the polarization effects produced by the positive surplus charge ( 2 ) , the following modifications of the coulomb potential ( 3 ) are to be considered : within region i , i.e. within the insulating layer , the coulomb potential @xmath26 is to be replaced by @xmath27 and within region ii , i.e. within the metallic core , the constant potential @xmath28 is to be supplanted by @xmath29 within region 0 , i.e. in the vacuum surrounding the insulating layer , the value @xmath30 of the coulomb potential ( 3 ) is unchanged . one should note that the potential in region i assumes the value @xmath31 at the outer surface @xmath32 of the insulating layer . let us denote the sum of the constant parts of the potentials in region i and ii by @xmath33 and @xmath34 if @xmath3 is equal to 1 or 2 and the dielectric constant @xmath35 exceeds 1 only slightly , and if the radii are in the range of values treated in sec . 4 , the 2@xmath36 terms on the right - hand sides of eqs . ( [ eq2.4 ] ) and ( [ eq2.5 ] ) are much smaller than the 1@xmath37 terms . so far , it is only these cases which we treated numerically . altogether , the potential @xmath5 felt by the electron has the form @xmath38 \theta_0(r_2 - r ) + \left[b - { ze^2_0\over \varepsilon_{\rm i}r}\right ] \theta_0(r_1 - r ) \theta_0(r - r_2 ) - { ze^2_0\over r } \theta_0(r - r_1)\,\ , . \label{eq2.5.1}\ ] ] the constants @xmath39 , @xmath35 and @xmath40 are empirical parameters . the potential @xmath5 is shown in fig . 2 for a specific choice of the parameters . [ fig2 ] for the following choice of parameters : @xmath41 ; @xmath42 ; @xmath43 ; @xmath44 ev ; @xmath45 ev ; @xmath46 ev ; @xmath47 ; @xmath48,width=264 ] the eigenfunctions @xmath49 of the schrdinger equation @xmath50 separate @xmath51 and the radial functions @xmath52 and @xmath53 have to satisfy the differential equations @xmath54 @xmath55 respectively . , the potential @xmath5 is constant and thus the eq . ( [ eq2.8 ] ) represents the differential equation for spherical bessel functions . of the 2 linearly independent solutions , only the regular one , @xmath56 , qualifies as a physical solution . up to a normalization factor which will be determined later on , the radial function @xmath52 in region ii is simply given by @xmath57 where @xmath58 is defined by @xmath59 and @xmath60 is the spherical bessel function of the 1@xmath37 kind ( see chapt . 9 of ref . @xcite ) . as bound states have an energy between the well depth @xmath61 and 0 , the parameter @xmath58 is assured to be real . let us postpone the treatment of the general case and first turn to the simpler situation when the @xmath62-dependent term @xmath63 is negligible in comparison to the constant @xmath64 in the entire region i. this is so , if the inequality @xmath65 holds . for the case shown in fig . 2 , the condition is fulfilled . generally speaking , the barrier heights @xmath40 range from 2 to 5 ev , and the dielectric constant @xmath66 may have values between slightly more than 1 up to more than 100 ( examples : @xmath35 ( caoutchouc ) = 4 ; @xmath35 ( ti0@xmath67-crystal ) = 115 ) . let us now investigate the case that the potential ( [ eq2.5.1 ] ) can be approximated by @xmath68 \theta_0(r_2 - r ) + b\theta_0 ( r_1 - r ) \theta_0(r - r_2 ) - { ze^2_0\over r}\theta_0(r - r_1)\,\ , . \label{eq2.13}\ ] ] the radial schrdinger equation ( [ eq2.8 ] ) in region i then assumes again the form of bessel s differential equation . but as the energy @xmath69 is lower than the barrier height @xmath64 , the solution in region i is given by a linear combination of two independent modified spherical bessel functions ( see chapt . 10.2 of ref . @xcite ) . we choose the pair @xmath70 , where @xmath71 and @xmath72 are the spherical bessel - function and spherical hankel - function of the 1@xmath37 kind , resp . , with imaginary arguments . the dimensionless real variable @xmath73 is related to the radial coordinate @xmath62 by @xmath74 @xmath75 the two factors @xmath76 and @xmath77 are chosen so that the pair of solutions are simply related to the modified bessel function of integer order @xmath78 and @xmath79 for which a number of rapidly converging series expansions exist ( see ref . @xcite , chapt . 10.2 ) @xmath80 the radial function @xmath52 in region i thus has the general form @xmath81 , it is more convenient to deal with the differential eq . ( [ eq2.9 ] ) for @xmath53 which has the form @xmath82 we have to find a normalizable solution of ( [ eq2.18 ] ) . it turns out that the only normalizable solution of ( [ eq2.18 ] ) has the form @xmath83 where the dimensionless variable @xmath84 is related to @xmath62 by @xmath85 with @xmath86 and where @xmath87 is a particular solution kummer s differential equation @xmath88 the parameters @xmath89 and @xmath90 in ( [ eq2.22 ] ) are related to the parameters in eq . ( [ eq2.18 ] ) by @xmath91 @xmath92 @xmath93 and the solution @xmath94 of ( [ eq2.22 ] ) is defined as follows : + we write @xmath90 in the form @xmath95 which implies @xmath96 then the function @xmath97 , where @xmath98 reads ( see ref . @xcite , eq . ( 13.1.6 ) or ref . @xcite , eq . ( 13 ) of chapt . 6.7.1 ) @xmath99\right\}\nonumber \\ & + & { ( n-1)!\over \gamma(a_0 ) } \sum^{n-1}_{\nu=0 } { ( a_0-n)_\nu\over ( 1 - n)_\nu } \cdot { z^{\nu - n}_0\over \nu ! } \label{eq2.25}\end{aligned}\ ] ] in ( [ eq2.25 ] ) , the function @xmath100 is the logarithmic derivative of the @xmath101-function @xmath102 and @xmath103 is pochhammer s symbol @xmath104 furthermore , @xmath105 is the confluent hypergeometric function @xmath106 we note in passing that the complicated form of the solution ( [ eq2.25 ] ) is due to the fact that the parameter @xmath90 in @xmath94 is an integer . for non - integer @xmath90 , the function @xmath94 depends in a simple way on two confluent hypergeometric functions ( see ref . @xcite , eq . ( 13.1.3 ) ) . this form of @xmath94 becomes singular if @xmath90 tends to an integer value . the limiting process which leads to the form ( [ eq2.25 ] ) is discussed in ref . @xcite . for any value of @xmath90 , the asymptotic behaviour of @xmath107 for @xmath108 is given by @xmath109 \label{eq2.29}\ ] ] ( see ref . @xcite , eq . ( 13.1.8 ) ) . from ( [ eq2.29 ] ) and ( [ eq2.19 ] ) it is clear that @xmath53 is a normalizable function irrespectively of the value of the energy @xmath69 . however , it turns out that the continuity conditions can only be met for certain discrete values of @xmath69 . the radial function @xmath110 in region 0 is thus defined by @xmath111 before we proceed to the explicit formulation of the continuity conditions , we return to the form of the solution in region i in the general case that the @xmath62dependent term @xmath63 is neglected . then the differential equation for @xmath53 in region i has the form @xmath112 eq . ( [ eq2.31 ] ) resembles eq . ( [ eq2.18 ] ) . in fact , with the replacements @xmath113 @xmath114 eq . ( [ eq2.18 ] ) becomes eq . ( [ eq2.31 ] ) . + with the ansatz @xmath115 we obtain again kummer s differential equation for @xmath116 @xmath117 where @xmath90 is again given by ( [ eq2.23.2 ] ) or ( [ eq2.24.1 ] ) , but @xmath118 is obtained from @xmath89 by the replacements ( [ eq2.32.1 ] ) , ( [ eq2.32.2 ] ) : @xmath119 @xmath120 and the dimensionless variable @xmath121 is related to @xmath62 by @xmath122 the parameter @xmath123 defined in ( [ eq2.15 ] ) agrees with @xmath124 because @xmath125 and @xmath126 . the general solution @xmath127 of eq . ( [ eq2.34 ] ) can be written as a linear combination of the function @xmath128 ) and , as linearly independent 2@xmath36 solution , the confluent hypergeometric function @xmath129 : @xmath130 we note that we had to suppress the solution @xmath131 in region 0 because it would lead to a non - normalizable solution . the radial solution @xmath52 in region i thus assumes the form @xmath132 @xmath133 \,\ , . \label{eq2.37p}\ ] ] we now turn to the continuity conditions which enable us to determine the amplitudes @xmath134 , @xmath135 ( or @xmath136 , @xmath137 ) , @xmath138 , and the energy eigenvalues @xmath69 . once these parameters will be obtained , we shall normalize the entire radial function to 1 . the radial wave function @xmath52 and its 1@xmath37 derivatives must be continuous at the limits @xmath139 and @xmath32 of the regions ii and i : @xmath140 @xmath141 @xmath142 @xmath143 we proceed as follows : + eqs . ( [ eq2.38 ] ) and ( [ eq2.39 ] ) are 2 linear inhomogeneous equations for the amplitudes @xmath134 , @xmath135 ( or @xmath136 , @xmath137 ) . we solve them for these amplitudes and substitute the result into the equation @xmath144 i.e. the ratio of the eqs . ( [ eq2.41 ] ) and ( [ eq2.40 ] ) . ( [ eq2.42 ] ) represents a transcendental equation for the energy . it is solved numerically . the amplitude @xmath138 can subsequently be obtained from ( [ eq2.40 ] ) . the resulting wavefunction @xmath52 is not yet normalized to 1 . this can be easily achieved by the replacement @xmath145 with @xmath146 let us present this procedure in more detail : + we introduce the abbreviations : @xmath147_{x=\kappa_{\rm i}r } = \left[\sqrt{{\pi\over 2x } } i_{l+{1\over 2}}(x)\right]_{x=\kappa_{\rm i}r } \,\ , , \label{eq2.45.1}\ ] ] @xmath148_{x=\kappa_{\rm i}r } : = \left[\sqrt{{\pi\over 2x } } k_{l+{1\over 2}}(x)\right]_{x=\kappa_{\rm i}r } \,\ , , \label{eq2.45.2}\ ] ] @xmath149_{y=2\kappa_{\rm i}r } : = \left[y^l e^{-{y\over 2 } } m(a_{\rm i } , 2l+2,y)\right]_{y=2\kappa_{\rm i}r}\,\ , , \label{eq2.46.1}\ ] ] @xmath150_{y=2\kappa_{\rm i}r } : = \left[y^l e^{-{y\over 2 } } u(a_{\rm i } , 2l+2,y)\right]_{y=2\kappa_{\rm i}r}\,\ , . \label{eq2.46.2}\ ] ] furthermore , we use the prime to denote the derivative with respect to the variable given as argument : @xmath151_{x=\kappa_{\rm i}r_2 } \,\,,\ ] ] @xmath152_{y=2\kappa_{\rm i}r_2 } \,\,,\ ] ] @xmath153_{x=\kappa_{\rm ii } r_2 } \,\,,\ ] ] a.s.o . + from the continuity equations ( [ eq2.38 ] ) , ( [ eq2.39 ] ) we find the amplitudes @xmath134 , @xmath135 and @xmath136 , @xmath137 to be given by very similar relations @xmath154_{\kappa_{\rm i}r_2 } } \,\ , , \label{eq2.47.1}\ ] ] @xmath155_{x=\kappa_{\rm i}r_2 } } \,\ , . \label{eq2.47.2}\ ] ] the wronskian in the denominators is given by ( see ref . @xcite , eq . ( 10.2.8 ) ) @xmath156 @xmath157_{y=2\kappa_{\rm i}r_2 } } \,\ , , \label{eq2.48.1}\ ] ] @xmath158_{y=2\kappa_{\rm i}r_2 } } \,\ , . \label{eq2.48.2}\ ] ] the wronskian in the denominators of ( [ eq2.48.1 ] ) and ( [ eq2.48.2 ] ) can be related to the wronskian @xmath159 : @xmath160 using the differential eq . ( [ eq2.34 ] ) , one finds for @xmath161 ( see ref . @xcite , eq . ( 13.1.22 ) ) @xmath162 and thus for @xmath163 : @xmath164 substituting ( [ eq2.47.3 ] ) into ( [ eq2.47.1 ] ) and ( [ eq2.47.2 ] ) and , analogously , substituting ( [ eq2.48.5 ] ) into ( [ eq2.48.1 ] ) and ( [ eq2.48.2 ] ) we obtain the following forms of the amplitudes in region i : @xmath165 \,\ , , \label{eq2.49.1}\ ] ] @xmath166 \,\ , , \label{eq2.49.2}\ ] ] @xmath167 \,\ , , \label{eq2.50.1}\ ] ] @xmath168 \,\ , . \label{eq2.50.2}\ ] ] with ( [ eq2.49.1 ] ) , ( [ eq2.49.2 ] ) or ( [ eq2.50.1 ] ) , ( [ eq2.50.2 ] ) the amplitudes of the 2 linearly independent solutions in region i are represented as a function of the energy @xmath69 . the eigenvalues can now be obtained from the ratio of the two continuity conditions ( [ eq2.41 ] ) and ( [ eq2.40 ] ) , i.e. from the equation @xmath169 this is a transcendental equation for the eigenenergies . it has infinitely many solutions close to the energy @xmath170 due to the infinite range of the coulomb potential . in spite of the complicated nature of the wavefunctions it is not difficult to find the numerical solutions of ( [ eq2.51 ] ) . more explicitly , the logarithmic derivatives in ( [ eq2.51 ] ) look as follows : @xmath171_{r = r_1 } = \left[{{d\over dr}((2\lambda r)^l u(a_0 , 2l+2 , 2\lambda r ) e^{-\lambda r } ) \over ( 2\lambda r)^l u(a_0 , 2l+2 , 2\lambda r ) e^{-\lambda r}}\right]_{r = r_1 } \label{eq2.52}\ ] ] @xmath172_{r = r_1 } = \left[{{d\over dr } ( a_{\rm i}f^{(1)}_l(r ) + b_{\rm i } f^{(2))}_l(r ) ) \over a_{\rm i } f^{(1)}_l(r ) + b_{\rm i } f^{(2)}_l(r)}\right]_{r = r_1 } \label{eq2.53.1}\ ] ] @xmath172_{r = r_1 } = \left[{{d\over dr } ( \tilde a_{\rm i}\tilde f^{(1)}_l(r ) + \tilde b_{\rm i}\tilde f^{(2))}_l(r ) ) \over \tilde a_{\rm i } \tilde f^{(1)}_l(r ) + \tilde b_{\rm i } \tilde f^{(2)}_l(r)}\right]_{r = r_1 } \label{eq2.53.2}\ ] ] eq . ( [ eq2.53.2 ] ) and ( [ eq2.53.1 ] ) deal with the case of the exact potential and of the approximate constant potential in region i. in calculating the derivative @xmath173 or @xmath174 one can use properties of the different functions involved . having determined the eigenvalues from eq . ( [ eq2.51 ] ) one obtains the amplitude @xmath138 from eq . ( [ eq2.40 ] ) and thereafter the correct normalization of the entire function must be achieved by numerical calculation of the normalization constant ( [ eq2.44 ] ) . in sec . 2 , only the polarization of the insulator which is produced by a constant positive surface charge density @xmath175 on the surface @xmath6 of the metal was taken into account . however , the charge distribution @xmath23 of the externally localized electrons produces not only additional polarization effects in the insulating surface layer but also modifies the distribution of the positive surplus charge on the surface @xmath6 of the metal . in this section , we present a simple macroscopic description of the polarization effects which is based on maxwell s equations in the presence of homogeneous dielectric media . outside of the metallic core , i.e. in the region i ( @xmath176 ) of the insulator , and in the vacuum 0 ( @xmath177 ) surrounding the insulator the electric field @xmath24 and the displacement @xmath25 are related by : @xmath178 where @xmath35 and @xmath179 are the dielectric constant and the electric susceptibility of the insulator , respectively . the fields satisfy maxwell s equations @xmath180 @xmath181 for the sake of simplicity , we assume that the distribution @xmath23 of the externally located electrons and , as a consequence , also the positive surface charge distribution @xmath182 on @xmath6 , are , i.e. that they depend only on the polar angle @xmath183 and the radial coordinate @xmath62 . they can thus be written in the form @xmath184 integrating the surface charge density over the surface @xmath6 must yield the total positive surface charge @xmath11 . this implies the condition @xmath185 or , with ( [ eq77 ] ) @xmath186 the components @xmath187 of the electronic charge distribution are related to the total electronic charge distribution by @xmath188 the maxwell equations ( [ eq74 ] ) and ( [ eq75 ] ) are equivalent to the poisson equation @xmath189 in region i and to @xmath190 in region 0 . here , the potential @xmath191 is related to the electric field @xmath24 by @xmath192 the following boundary conditions must be fulfilled at the surfaces @xmath6 and @xmath7 : @xmath193 @xmath194 @xmath195 @xmath196 furthermore , the potential @xmath197 must satisfy the asymptotic condition @xmath198 the conditions ( [ eq86 ] ) and ( [ eq88 ] ) are obtained from the maxwell equation ( [ eq74 ] ) , and the conditions ( [ eq87 ] ) and ( [ eq89 ] ) result from ( [ eq75 ] ) and the requirement that the classical electric field should be everywhere finite . the solutions @xmath199 and @xmath200 of the poisson equations ( [ eq81 ] ) and ( [ eq82 ] ) in the regions i and 0 , respectively , can be written in the form : @xmath201 where @xmath202 is the coulomb potential produced by the electronic charge distribution @xmath203 @xmath204 the charge distribution @xmath23 itself is to be calculated from the wavefunctions of the bound and predominantly externally located ( i.e. in region 0 , @xmath177 ) electrons . conduction electrons which are located internally ( i.e. in region ii , @xmath205 ) do not contribute to @xmath23 because they are already taken into account by the number @xmath3 of positive surplus charges . the density @xmath23 is thus mainly located in region 0 and it is very small in region i , i.e. inside of the insulating surface layer , because the eigenenergies of the electrons are far below the positive barrier in region i. the function @xmath206 is the general azimuthally symmetric solution of the laplace equation in region i @xmath207 and @xmath208 the corresponding most general solution of the laplace equation in region 0 satisfying in addition the asymptotic condition ( [ eq90 ] ) : @xmath209 the coulomb potential @xmath202 can be written as @xmath210 where the functions @xmath211 and @xmath212 are related to the components @xmath187 of the electronic charge distribution by @xmath213 substituting the explicit forms ( [ eq94 ] ) to ( [ eq96 ] ) into the boundary conditions ( [ eq86 ] ) , ( [ eq88 ] ) , and ( [ eq89 ] ) , we obtain the relations @xmath214 furthermore , using the identity ( see for instance ref . @xcite ) for @xmath215 @xmath216 \label{eq102}\ ] ] together with ( [ eq91 ] ) , ( [ eq92 ] ) and ( [ eq94 ] ) , the condition ( [ eq87 ] ) can be shown to be equivalent to the relation @xmath217 where @xmath218 . ( [ eq103 ] ) is seen to imply that the potential @xmath219 is constant on the surface @xmath6 . for @xmath220 , we obtain from ( [ eq99 ] ) @xmath221 \label{eq104}\ ] ] and from ( [ eq100 ] ) and ( [ eq101 ] ) the relations @xmath222 hence , with ( [ eq104 ] ) , we find @xmath223 \label{eq107}\ ] ] and @xmath224 \,\ , . \label{eq108}\ ] ] for @xmath218 , we obtain from ( [ eq99 ] ) and ( [ eq103 ] ) : @xmath225 \,\ , , \label{eq109}\\ & & d^{\rm i}_l = { 4\pi\over ( 2l+1)\varepsilon_{\rm i } } \left[-b_l(r_2 ) + \rho^{s_2}_l r^{l+1}_2\right ] \,\ , . \label{eq110}\end{aligned}\ ] ] for the coefficients @xmath226 , we find the relations @xmath227 and @xmath228 \label{eq112}\ ] ] from the eqs . ( [ eq101 ] ) and ( [ eq100 ] ) , which hold for @xmath229 . the equality of the righthand sides of ( [ eq111 ] ) and ( [ eq112 ] ) together with ( [ eq109 ] ) and ( [ eq110 ] ) yields a condition for the surface charge components @xmath230 for @xmath218 . from this equation one finds the explicit form @xmath231 for @xmath218 . the component @xmath232 is determined by the conservation of the total positive charge on @xmath6 and is given by eq . ( [ eq79 ] ) . for @xmath218 , one obtains from ( [ eq109 ] ) , ( [ eq110 ] ) and ( [ eq112 ] ) @xmath233 as we have mentioned already , the charge distribution @xmath234 of the externally located electrons is vanishingly small for @xmath235 . consequently , we have @xmath236 and @xmath237 the potentials @xmath238 , @xmath239 , @xmath240 , and the charge distribution @xmath241 on the surface @xmath6 are thus fully determined for a given azimuthally symmetric distribution @xmath234 of the externally located electrons . the potential @xmath242 acting on an electron in our mesoscopic atom is related to the potential functions as follows : @xmath243 the potential @xmath244 represents the interaction of a conduction electron with the average ionic background charge in the metal and has to be added to the constant potential @xmath245 produced by the positive surplus charge on @xmath6 . the potential barrier @xmath40 to be added in the insulating surface layer is related to the average energy gap between the electronic bands . since it is of quantum mechanical origin , it is not contained in the classically determined term @xmath246 . the schrdinger equation with the potential @xmath247 can no longer be solved analytically . one could diagonalize the hamiltonian @xmath248 within the set of analytical eigenfunctions which we have determined in sec . 2 . this represents a selfconsistency problem , because the eigenfunctions of the hamiltonian ( [ eq120 ] ) depend on the charge distribution @xmath23 which is calculated from them . the solution thus requires an iteration process . let us consider two special cases for which the results of this section simplify considerably : * if we neglect the polarization effects produced by the external electrons , that is if we put @xmath249 the quantities @xmath211 , @xmath212 vanish for @xmath229 , and , as a consequence , so do the components @xmath250 . the positive surplus charge density on @xmath6 is then spherically symmetric and given by @xmath251 it is easily seen that , in this case , the potential @xmath252 becomes equal to the potential @xmath5 we considered in sec . 2 . * if we assume that the charge density @xmath23 of the external electrons is spherically symmetric @xmath253 only the component @xmath254 in ( [ eq76 ] ) is unequal 0 @xmath255 and , consequently , the quantities @xmath211 , @xmath212 , and @xmath230 vanish for @xmath218 . in this case , the potentials @xmath256 , @xmath199 and @xmath200 take the following form @xmath257 the case of a spherically symmetric distribution of the externally located electrons is realistic at finite temperature , because magnetic substates of a given electronic level are expected to be occupied with the same probability whenever the mesoscopic atom is in contact with a thermal reservoir of finite temperature . the functions @xmath258 and @xmath259 depend themselves on the electronic distribution . for @xmath260 , @xmath258 decreases to zero as one can see from ( [ eq97 ] ) . the quantity @xmath261 represents the total charge of the externally located electrons inside the sphere of radius @xmath62 @xmath262 using ( [ eq76 ] ) in ( [ eq128 ] ) we find @xmath263 for @xmath264 , @xmath265 tends to the total charge @xmath266 of the externally located electrons . if the mesoscopic atom is neutral , i.e. if there are as many externally bound electrons as there are positive surplus charges on the surface @xmath6 of the metallic core , the term @xmath267 in eq . ( [ eq127 ] ) cancels the potential @xmath268 for @xmath260 , as it should . let us study the asymptotic behaviour of the function @xmath258 and @xmath259 for @xmath269 : for simplicity , let us assume that only the lowest external electron state is occupied by two electrons , one with intrinsic spin up and the other one with spin down . the external state of lowest energy belongs necessarily to orbital angular momentum ( @xmath220 ) and to zero number of radial modes . we assume that the wavefunctions we determined in sec . 2 are realistic for @xmath270 . in region 0 , the electronic density distribution is then given by @xmath271 or , substituting the explicit wavefunction ( [ eq2.19 ] ) , by @xmath272 here , @xmath273 is the overall normalization constant introduced by eq . ( [ eq2.44 ] ) . for @xmath274 , the function @xmath275 can be replaced by ( see ref . @xcite , eq . ( 13.5.2 ) ) @xmath276 retaining only the 1@xmath37 term in ( [ eq132 ] ) , we obtain for the asymptotic density ( @xmath277 ) @xmath278 and for the function @xmath258 @xmath279 the integral in ( [ eq134 ] ) is the incomplete @xmath101-function ( see ref . @xcite , eq . ( 6.5.3 ) , @xmath280 ) @xmath281 with the asymptotic expansion ( ref . @xcite , eq . ( 6.5.32 ) @xmath282 from ( [ eq134])([eq136 ] ) , we obtain @xmath258 in the form @xmath283 for @xmath284 . of course , one could raise the accuracy of the result ( [ eq137 ] ) by incorporating one or two more terms in ( [ eq132 ] ) and ( [ eq136 ] ) . the important point , however , is that @xmath258 tends to zero exponentially for large values of @xmath260 . this justifies the assumption that the electronic charge density @xmath285 in the asymptotic regime @xmath260 is correctly described by the wavefunctions of sec . 2 , which do not contain the electronic polarization effects . the function @xmath259 can be written in the form ( see ( [ eq98 ] ) ) @xmath286 with @xmath287 the 2@xmath36 term one the r.h.s . of ( [ eq138 ] ) is easily seen to have the approximate form @xmath288 for @xmath284 . thus the function @xmath259 is asymptotically @xmath289 given by @xmath290 which means that it approaches exponentially the asymptotic value @xmath291 . in the general case of a @xmath183-dependent electronic charge density @xmath234 , the quantities @xmath292 and @xmath293 are different from zero and their asymptotic behaviour can be investigated in an analogous way . one finds that @xmath294 tends exponentially to zero and @xmath295 exponentially to constant values as @xmath62 goes to infinity . therefore , at large values of @xmath62 , the dominant part of the potential @xmath247 is the potential @xmath5 given by eq . ( [ eq2.5.1 ] ) . thereby , our assumption that the wavefunctions determined in sec . 2 are essentially correct at large values of @xmath296 is seen to be justified . for non - asymptotic values of @xmath62 , the electronic charge density and the resulting functions @xmath258 and @xmath259 must be obtained from the eigenfunctions of the hamiltonian ( [ eq119 ] ) which contains the potential @xmath247 defined in ( [ eq117 ] ) to ( [ eq119 ] ) . let us formulate the special case of a spherically symmetric charge density and the resulting spherically symmetric potential @xmath5 in somewhat more detail : the electronic charge density at a finite temperature @xmath297 is given by @xmath298 or , using the relation ( see for instance ref . @xcite , eq . ( 3.69 ) ) @xmath299 by @xmath300 here , @xmath301 is the @xmath302-th eigenvalue for given angular momentum @xmath303 and @xmath304 is the corresponding eigenfunction of the radial schrdinger equation . the parameter @xmath305 represents the chemical potential which corresponds to the number of externally located electrons . + the potential @xmath247 has the form @xmath306 where @xmath307 , @xmath308 , and @xmath309 are given by the eq . ( [ eq104 ] ) , ( [ eq107 ] ) , and ( [ eq108 ] ) , resp . the functions @xmath258 and @xmath259 are defined by ( [ eq97 ] ) and ( [ eq98 ] ) , where @xmath310 . the wavefunctions @xmath304 are solutions of the radial schrdinger equation with @xmath311 given by ( [ eq145 ] ) - ( [ 147 ] ) @xmath312 since @xmath311 depends itself on the functions @xmath313 , one has to solve ( [ eq148 ] ) by iteration . one way to find the solutions @xmath313 of ( [ eq148 ] ) is to expand the solutions @xmath304 in terms of the radial eigenfunctions @xmath314 defined in sec . 2 , where the potential @xmath5 is given by eq . ( [ eq2.5 ] ) and related to @xmath311 by @xmath315 @xmath316 @xmath317 are the eigenvalues obtained in sec . 2 . ( [ eq150 ] ) we left away the basis functions @xmath318 corresponding to continuous eigenvalues @xmath319 of eq . ( [ eq2.8 ] ) . substitution of ( [ eq150 ] ) into ( [ eq148 ] ) leads to a nonlinear system of equations for the expansion coefficients @xmath320 which must be solved by iteration . we present results on the spectrum of eigenenergies , the eigenfunctions , and on the electromagnetic transition rate between externally and internally localized states for a range of values of the parameters @xmath4 , @xmath2 , and @xmath321 . so far , we have not taken into account the effects presented in sec . 3 , i.e. the polarization of the insulating dielectric material by externally localized electrons . the additional potential produced by this polarization is expected to modify the wave functions in the vicinity of the surface @xmath7 , perhaps even rather substantially . at large distances ( say a couple of ngstrm outside of the surface @xmath7 ) , the additional polarization potential tends rapidly to zero . consequently , the wavefunctions presented in this article are expected to be trustworthy at distances of a couple of ngstrm from the surface @xmath7 . we studied only the case that the coulomb potential @xmath322 within the insulator ( @xmath323 ) is small compared to the barrier height @xmath64 which is related to the gap parameter @xmath40 by eq . ( [ eq2.4 ] ) . this means that we neglected the coulomb potential in region i. .the energies @xmath324 , @xmath325 , and the barrier height @xmath64 as a function of the radii @xmath4 and @xmath2 and of the dielectric constant @xmath35 . energies are given in mev and radii in . the parameter @xmath40 is chosen equal to 2500 mev or 4000 mev . [ tab1 ] [ cols="<,^,^,^,^,^,^,^,^",options="header " , ] augmenting @xmath4 for fixed value of @xmath2 enlarges the width of the barrier and , consequently , decreases the amplitude which an externally located electron has in the inner region ii . as the decay of a class 1-state into a class-2 state depends strongly on this tail , the transition probability must decrease and the lifetime must increase as a function of rising @xmath4 . the results presented in table 4 support this expectation . the lifetime @xmath326 is found to rise roughly exponentially with @xmath327 . this is so because the amplitude of the tail in region ii depends exponentially on @xmath327 as one can see from wkb . the dependence of @xmath326 on @xmath4 is also shown in fig . 6 in logarithmic scale . [ fig6 ] . apart from @xmath4 , all the other parameters are chosen as in fig . 3.,width=264 ] in fig . 7 , we show the dependence of the lifetime @xmath326 on the barrier height @xmath64 in logarithmic scale . the lifetime increases as a function of rising @xmath64 . the increase is somewhat smaller than exponential as can be expected from wkb . [ fig7 ] in sec and logarithmic scale as a function of the barrier height @xmath64 . apart from @xmath64 , all the other parameters are chosen as in fig . 3.,width=264 ] [ fig8 ] in sec and logarithmic scale as a function of @xmath2 . apart from @xmath2 , all the other parameters are chosen as in fig . 3.,width=264 ] if one changes @xmath2 with all the other parameters fixed , one modifies predominantly the internally located class 2 states and the corresponding energies . as one can see from fig . 8 , the lifetime @xmath326 as a function of rising @xmath2 decreases on the average as one would expect from the fact that the thickness @xmath327 of the surface layer decreases . however , beside the average decrease of the lifetime , one observes a kind of resonance structure : within a change of @xmath2 by 1 , the lifetime @xmath326 can change by 10 orders of magnitude . + how can we understand these huge fluctuations of the lifetime ? [ fig9 ] ) of lowest external electron to be found in the inner region ii ( @xmath328 ) as a function of @xmath2 . apart from @xmath2 , all the other parameters are chosen as in fig . 3.,width=264 ] in fig . 9 , we show the probability of an externally located electron to be found in the inner region ii . this localization probability in region ii , which is of crucial importance for the transition rate , is seen to show pronounced peaks at precisely those values of @xmath2 , where the lifetime has very small values as one can see by comparing fig . the large variations of the transition rate as a function of @xmath2 ( and of @xmath3 ) can be simply explained as follows : + as one changes @xmath2 ( or @xmath3 ) , the eigenvalues of class 2-states move whereas the ones of class 1-states remain essentially the same . more specifically , the eigenvalues @xmath329 of class-2 states are lowered , if the inner radius @xmath2 or the positive surplus charge @xmath3 is increased . thereby , it happens regularly that a class 2-state which has the same angular momentum @xmath303 as the decaying class 1-state , crosses the energy of this class 1-state . if one augments @xmath2 or @xmath3 on a sufficiently long scale , such crossings occur several times . close to the crossing , the external class 1-state develops a non - negligible amplitude within region ii and the class 2-state develops an appreciable amplitude in the external region 0 . we call such mixed states class 3-states . if the decaying class 1-state happens to have the nature of a class 3-state , its decay probability to lower - lying class 2-states is considerably enhanced as compared to the ordinary situation where no class 2-state of the same angular momentum is available . we note in passing that this enhancement of the transition probability to lower - lying class 2-states does not only occur for dipole transitions but also for transitions of higher multipolarity . so far we have not yet calculated the transition rates of higher multipolarity , because they would only matter if the decaying external state is a class 3-state . the phenomenon can be considered as an example of the mechanism which landau and zener had studied in famous papers @xcite . this is demonstrated in appendix . [ fig10 ] ( sec , log.scale ) as a function of radius @xmath4 and @xmath2 ( in ) in a 3-dimensional plot . except for @xmath4 and @xmath2 , all the parameters are chosen as in fig . 3.,width=377 ] in fig . 10 , the dependence of the lifetime @xmath326 on the radii @xmath4 and @xmath2 is shown in a 3-dimensional plot . again one notices that the dependence of @xmath326 on @xmath4 is smooth whereas the one on @xmath2 contains resonances . [ fig11 ] ( ) . energies are given in units of @xmath330 mev . except for @xmath2 , all the other parameters are chosen as in fig . 3.,width=377 ] in fig . 11 , the energy difference between the lowest - lying class 1 state and the closest lying class 2-state of the same angular momentum is shown . the plot supports the explanation of the resonance phenomenon in terms of a coupling between close - lying states . in the numerical results presented in this section , the values @xmath64 of the barrier in region i and the well depth @xmath39 of the average potential in region ii and the dielectric constant @xmath35 were chosen arbitrarily , and not determined for given choices of the material . in this paper , we determined the quantum states and eigenenergies of an electron in the average potential produced by a positively charged metal cluster which is covered by an insulating surface layer . because of the similarity of the externally localized electronic states with the bound states of an electron in the coulomb potential of an atom , we occasionally referred to the system as a mesoscopic atom or , in case of a non - zero total charge , as a mesoscopic ion . in particular , we investigated the electromagnetic decay of an externally located electron ( class 1-state ) to a lower - lying unoccupied electronic state located within the metal ( class 2-state ) . it turned out that the life - time depends sensitively on the radius @xmath2 of the metal core and on the number @xmath3 of positive surplus charges of the metal . in the model we studied , we found life - times varying between 10@xmath14 s and years depending on the value of @xmath2 or of @xmath3 . the origin of the large fluctuations was found to be the appearance of states with non - negligible amplitudes both outside inside of the insulating surface layer ( class 3-states ) . the mechanism leading to these class 3-states is described in appendix . the wavefunctions and eigenenergies were calculated neglecting the polarization of the insulating surface layer by externally located electrons . the potential produced by these electronic polarization effects was determined in sec . it was shown that this additional potential matters mainly in the vicinity of the outer surface @xmath7 of the insulator and decreases exponentially outside of this surface . the polarization potential could be taken into account by diagonalizing the total hamiltonian including the polarization potential in the basis of the wavefunctions determined in this paper . apart from this physical simplification of the problem we neglected the coulomb potential produced by the positive surplus charge within the insulator . this approximation is justified whenever the change of the coulomb potential within the insulating surface layer is much smaller than the barrier height @xmath40 . this condition was fulfilled in all the cases we studied numerically . the 1@xmath37 improvement we already mentioned is to incorporate the additional potential between the external electron and the polarization cloud it produces in the insulator . the form of this electronic polarization potential was derived in sec . a 2@xmath36 improvement would be to replace the constant potentials @xmath20 in the metal and @xmath331 in the insulator by periodic potentials defined by the crystal structure of these materials . of course , this is much more complicated and , to our knowledge , it has rarely been done so far for mesoscopic systems . furthermore , the crystal structure in mesoscopic aggregates is in general poorly known . as a result , the electronic eigenvalues of the total system are expected to exhibit a band structure similarly to a macroscopic solid , the insulating surface layer being characterized by a completely filled valence band and an empty conduction band . the empty conduction band in the insulator can be at negative or at positive energy if the energy of the neutral system plus one electron at infinity is put equal to 0 . describing the average potential in the insulator by a positive barrier @xmath22 is only meaningful if the conduction band of the insulator is at positive energy . as a further restrictive condition for our model to be a reasonable 1@xmath332 approximation there should be no localized traps for an additional electron in the uncharged insulator this implies that the individual atoms or molecules of the insulator should not have bound states with an additional electron as a monomer @xcite . in general , bound states of an electron with a neutral atom or molecule do not exist for atoms or molecules with completely filled shells . therefore , one should use insulators which exhibit this property . presumably , al@xmath670@xmath333 fulfills this condition . finally , we neglected effects of the electron - phonon coupling altogether . we presume that they tend to reduce the lifetime , but perhaps not drastically . as long as one can only produce a limited number of mesoscopic atoms , one could just investigate experimentally the effects which we studied in the present paper , especially the amusing strong dependence of the lifetime on the radius and the charge of the metallic core . although we would be highly pleased by an experimental scrutiny of our predictions , still more exciting open questions could be asked , if it were possible to produce longlived systems of the kind we presented in this paper , and in large quantities . in this case , one would have access to the study of molecules formed from mesoscopic atoms and to condensed matter consisting of mesoscopic atoms as building blocks . the presence of externally localized electrons would represent a fundamental difference from the case of a condensed system of uncharged mesoscopic clusters . on the one hand , one may expect that , in a condensed matter phase of mesoscopic atoms , the weakly bound external electrons would partly dissociate from their mother atoms and move freely in the average field produced by the charged insulated cores . we would thus surmise that such a system would be a ( good ) conductor . on the other hand , the fact that the building blocks of such a condensed matter carry charge , will have a profound influence on the equilibrium state of the material . probably , its density is lower than for uncharged constituents , where the cohesive forces are of van der waals type and , consequently , of much shorter range . as we learnt from this paper , a mesoscopic system of the kind we investigated can only be stable in a strict sense , if there is no empty internally located electronic state below the energy of the lowest externally located state which is at an energy slightly above @xmath335 . the positive surplus charge @xmath11 on the surface @xmath6 of the metal leads to a decrease of the well - depth @xmath336 as we have shown in sec . 3 : @xmath337 if , for the uncharged metal , the fermi level is at the energy @xmath338 ( with @xmath339 , @xmath340 ) , it lies at the modified energy @xmath341 in the charged metal . so the condition for strict stability is that there is no level of an internally located electron between @xmath342 and @xmath335 . as the separation energy @xmath343 of an electron from the uncharged metal and the fermi energy @xmath344 are related to @xmath339 by @xmath345 we can state that strict stability is achieved whenever there is no level between the energies @xmath346\right\}$ ] and @xmath335 . this means that , metal with small @xmath347 is favourable for stability . in an ordinary metal , the separation energy @xmath347 is of the order of a couple of ev ( examples : @xmath348 ev ; @xmath349 ev ) . in the case of a radius @xmath350 of the metal core , the distance between neighbouring electronic levels is of the order of 100 mev . this means that in the case we chose for our numerical investigations , there are many unoccupied levels below the energy of the lowest external electron . however , if we consider a system with an inner radius @xmath351 , the spacing of internally located electron levels of given @xmath303 approaches already 1 ev . it is therefore conceivable that one might succeed to model a strictly stable , , mesoscopic atom . a further trick to widen the spectrum of internally localized electrons would be to replace the metallic sphere with radius @xmath2 by a metallic bubble of outer radius @xmath2 and a thickness of 1 to 2 in this case the levels differing by the number of radial nodes would have a much larger spacing . this is perhaps possible , if one replaces the charged metal layer by a ( charged ) fullerene @xmath352 , which has a radius of about 3,5 . the separation energy @xmath347 of an electron from graphite is known to be @xmath353 ev . we surmise that the separation energy from the fullerene is about the same . of course , it would no longer be justified to describe free electrons in the fullerene as moving in a simple square well perpendicular to the sphere @xmath6 . but the spacing of electronic levels is expected to be in the order of ev . finally , one is not limited to the spherical geometry . we may also envisage to consider charged nanotubes consisting of a cylindrical surface crystal of @xmath354-atoms covered by an insulating surface layer . the charge could in this case be generated by applying an external voltage to the ends of the nanotubes . as there is a lot of activity spent on nanotubes presently , we hope that one could succeed to produce small macroscopic amounts of insulated charged nanotubes in a not too far future . of course , the problem of the stability of externally localized electrons with respect to electromagnetic decay would be similar for nanotubes as for the case of the system we investigated . so we hope that our investigation stimulates the interest of the experts in the field of mesoscopic physics . * acknowledgement : * + klaus dietrich gratefully acknowledges the hospitality and support on the occasion of his stay at the theoretical physics institute of the umcs in april 2003 . krzysztof pomorski is indebted to the physics department of the technical university of munich for granting of the honorarium of a visiting professor position in february 2002 . let us consider the simplified potential @xmath355 we decompose this potential as follows @xmath356 @xmath357 \,\ , , \label{eqa3}\ ] ] @xmath358 and define the two hamiltonians @xmath359 where @xmath360 is the kinetic energy operator of radial motion @xmath361 the eigenstates @xmath362 , @xmath363 of the hamiltonians ( [ eqa5 ] ) @xmath364 @xmath365 form a non - orthogonal set of basis functions . we consider an eigenstate @xmath366 of the hamiltonian @xmath367 with eigenenergy @xmath69 @xmath368 @xmath369 for the special case that the angular momentum @xmath303 is the same for all the states and that two specific eigenenergies @xmath370 and @xmath371 are very close to the energy @xmath69 in ( [ eqa10 ] ) . then , we may expect that the expansion of @xmath366 in terms of the basis states @xmath362 and @xmath363 may be limited to the two states with the energies @xmath370 , @xmath371 very close to @xmath372 : @xmath373 in principle , we would have to sum over all eigenstates @xmath362 and @xmath363 on the righthand side of ( [ eqa11 ] ) , but of all the expansion coefficients @xmath374 , @xmath375 only the ones corresponding to the energies @xmath370 , @xmath371 closest to @xmath69 are expected to be large and all the others to be much smaller . henceforth , we leave away the subscript @xmath376 for simplifying the notation . it is convenient to introduce the dual basis states @xmath377 , @xmath378 defined by the orthogonality conditions @xmath379 where @xmath380 and @xmath381 can be 1 or 2 . @xmath382 and @xmath383 can be obtained by diagonalizing the inverse of the overlap matrix @xmath384 it is easily seen that the schrdinger eq . ( [ eqa10 ] ) together with the assumption ( [ eqa11 ] ) leads to the two coupled linear equations @xmath385 where the matrix elements are defined as follows : @xmath386 introducing the abbreviations @xmath387 @xmath388 the solutions @xmath389 of the eigenvalue equation @xmath390 are given by @xmath391 writing the two states @xmath392 in the form @xmath393 where @xmath394 one obtains @xmath395 from the eq . ( [ eqa14 ] ) and the amplitudes @xmath396 from the normalisation of @xmath397 . the physically interesting quantities are the probabilities @xmath398 and @xmath399 to find the system described by @xmath397 in the substate @xmath400 or @xmath401 . these probabilities are given by the mean - values of the projection operators @xmath402 it is easily seen that the probabilities are given by the following equations @xmath403 @xmath404 @xmath405 @xmath406 where @xmath407 is the overlap of the two basis states @xmath408 and the quantities @xmath409 by @xmath410 where @xmath411 it is seen that for @xmath412 i.e. far away from the crossing of the energies @xmath413 and @xmath414 , one of the ratios @xmath409 is much larger than the other . for @xmath415 one finds @xmath416 @xmath417 \label{eqa33}\ ] ] and for @xmath418 the reversed result @xmath419 @xmath420 this shows that a predominantly externally localized state turns into a predominantly internally localized one as the energies @xmath413 and @xmath414 cross as a function of @xmath2 or @xmath3 . the coupled true states @xmath421 are thus seen to change their nature as a result of the crossing of the energies @xmath413 , @xmath414 of a class 1 and a class 2 state . close to the crossing point the states @xmath421 are seen to be delocalized , containing comparable contributions of the two basis states @xmath422 . w. d. knight , k. clemenger , w. a. de heer , w. a. saunders , m. y. chou , m. l. cohen : phys . lett . * 52 * , ( 1984 ) 2141 ; w. d. knight , w. a. de heer , k. clemenger , w. a. saunders : solid state commun . * 53 * , ( 1985 ) 44 ; hellmut haberland ( ed ) : clusters of atoms and molecules " , vol . 1 and 2 , springer verlag , 1994 . s. frank , n. malinowski , f. tast , m. heinebrodt , i.m.l . billas , t.p . martin : z. phys . * d40 * , ( 1997 ) 250 ; m. springborg , s. satpathy , n. malinowski , t.p . martin , u. zimmermann : phys . lett . * 77 * , ( 1996 ) 1127 ; h. kuzmany , j. fink , m. mehring , s. roth ( eds ) : electronic properties of fullerenes " , springer - verlag , 1993 . m. brack : rev . phys . * 65 * , 677 ( 1993 ) . w. a. de heer , rev . phys . * 65 * , 611 ( 1993 ) . m. abramowitz and j.a . stegun ( eds ) : `` handbook of mathematical functions '' , nat . bureau of standards , applied mathematics , series 55 , issued june 1964 . a. erdelyi , higher transcendental functions " , mcgraw - hill , 1953 . jackson : `` classical electrodynamics '' , john wiley , 1967 . bransden and c.j . joachain , `` physics of atoms and molecules '' , longman group limited , 1983 . m.e . rose , `` elementary theory of angular momentum '' , new york : dover , 1995 l.d . landau , phys . * 2 * ( 1932 ) 46 ; c. zener , proc . ( london ) * a137 * ( 1932 ) 696 . h. haberland and k.h . bowen : `` solvated electron clusters '' , chapt . 2.5 in `` cluster of atoms and molecules ii '' , vol . 56 of the springer series in chemical physics , 1994 .

for josephson junctions based on-wave superconductors , time - reversal symmetry is known to allow for powerful relations between the normal - state junction properties , the excitation spectrum , and the josephson current . here we provide analogous relations for josephson junctions involving one - dimensional time - reversal invariant topological superconductors supporting majorana - kramers pairs , considering both topological topological and-wave topological junctions . working in the regime where the junction is much shorter than the superconducting coherence length , we obtain a number of analytical and numerical results that hold for arbitrary normal - state conductance and the most general forms of spin - orbit coupling . the signatures of topological superconductivity we find include the fractional ac josephson effect , which arises in topological topological junctions provided that the energy relaxation is sufficiently slow . we also show , for both junction types , that robust signatures of topological superconductivity arise in the dc josephson effect in the form of switches in the josephson current due to zero - energy crossings of andreev levels . the junction spin - orbit coupling enters the josephson current only in the topological topological case and in a manner determined by the switch locations , thereby allowing quantitative predictions for experiments with the normal - state conductance , the induced gaps , and the switch locations as inputs .

1607.05730

josephson junctions involving majorana fermions@xcite are under intensive theoretical@xcite and experimental@xcite investigation for the promising routes they provide towards demonstrating topological superconductivity@xcite and as potential building blocks towards topological quantum computation.@xcite in the most frequently studied class of systems , majorana fermions are nondegenerate zero - energy end states in hybrid devices realising one - dimensional ( 1d ) effectively spinless @xmath1-wave superconductors.@xcite a number of concrete platforms exist to realise this time - reversal symmetry breaking ( so - called class d@xcite ) topological superconductivity , all of which use the proximity effect to combine @xmath0-wave superconductors , strong spin - orbit coupling ( e.g. , via nanowires,@xcite topological insulators,@xcite or the superconductor itself@xcite ) and magnetic fields ( e.g. , via zeeman coupling@xcite or ferromagnetism@xcite ) . the past years have seen a rapidly increasing interest in realising@xcite and detecting@xcite time - reversal invariant ( so - called class diii@xcite ) analogues of such majorana fermion systems . in 1d , class diii topological superconductors host a kramers pair of majorana fermions at each end which can be combined into a zero - energy fermion end mode with anomalous time - reversal properties.@xcite the proposed platforms for realisation again include hybrid devices based on spin - orbit coupling@xcite and the superconducting proximity effect,@xcite albeit now with unconventional ( but nontopological ) superconductors ( e.g. , iron - based superconductors with @xmath2-wave pairing@xcite ) . these hybrids realise spinful effectively @xmath1-wave pairing with the admixture of a smaller @xmath0-wave pairing component . across the junction . the superconductors ( dark grey slabs ) each have a 1d , spin - orbit coupled quantum wire ( light grey blocks ) deposited on top of them . by the proximity effect , superconductivity is induced in each quantum wire . the wires with majorana - kramers pairs ( red and blue hemispheres ) display induced topological superconductivity , arising from coupling to unconventional ( e.g. , @xmath2-wave ) superconductors . the normal region ( e.g. , tunnel barrier , point contact , disordered metal ) between the superconductors is indicated by cross - hatching . ] in this paper , we study josephson junctions involving 1d class diii topological superconductors . we obtain the energy spectrum and the consequent josephson current in terms of the normal - state scattering properties , requiring only that the junction be much shorter than the superconducting coherence length ( i.e. the short - junction limit ) . our results apply for a number of physically relevant situations including disordered junctions , or junctions of intermediate transparency that are away from both the tunnelling and the highly transparent limits considered in previous works.@xcite we obtain results for two types of setups ( fig . [ fig : setups ] ) : junctions between two topological superconductors [ fig . [ fig : setups](a ) ] and junctions between an @xmath0-wave superconductor and a topological superconductor [ fig . [ fig : setups](b ) ] . for the first type of setup [ fig . [ fig : setups](a ) ] , we study the conditions under which a time - reversal invariant generalisation@xcite of the so - called fractional josephson effect@xcite can arise . this hallmark signature in class d topological superconductors refers to the appearance of a @xmath3 periodic current - phase relation , replacing the more conventional @xmath4 periodic one as majorana fermions enable charge transfer in terms of electrons instead of cooper pairs . in addition to establishing the form in which such behaviour can appear in class diii topological josephson junctions with generic junction properties , we also consider the role of the characteristic timescale for energy relaxation . the latter aspect , to the best of our knowledge , has not so far been considered ; however , as we show , it presents an important channel for the loss of @xmath3 periodicity . for the second type of setup [ fig . [ fig : setups](b ) ] , we investigate how the anomalous time - reversal properties of the fermion @xmath5 , built from the two interface majoranas , appear in the features of the josephson current . here general considerations@xcite show that there is an anomalous , time - reversal protected contribution that gives nonzero current at @xmath6 ( where @xmath7 is integer ) with sign set by the parity of @xmath5 . to positively establish the origin of this contribution in an experimental system ( e.g. , that the current at @xmath6 is not due to broken time - reversal invariance@xcite ) , however , a theory for the current - phase relationship is needed that accounts for other , nonanomalous contributions and which holds for generic junctions . such a theory may also be useful in scenarios where access to changes in @xmath8 are more readily available than to @xmath8 itself ; for example , the ac josephson effect where the phase sweep speed is controlled by the voltage across the junction ( i.e. @xmath9 ) . in this paper we provide such a theory . the signatures mentioned above are for junctions with a conserved fermion parity . we also investigate , for both setups , the case where the fermion parity is allowed to relax ( i.e. the dc josephson effect regime ) . for all the regimes to be considered , we compare topological junctions to their nontopological counterparts ( i.e. without majorana fermions ) in order to search for unique signatures of topological superconductivity . the rest of this paper is organised as follows . we start , in sec . [ sec : hamiltonian ] , by describing 1d class diii topological superconductors in terms of the bogoliubov - de gennes hamiltonian near the fermi points . this will allow us , in sec . [ sec : smatrices ] , to obtain the andreev reflection matrices of normal - topological superconductor interfaces , and thus to formulate a scattering matrix description for our systems . in sec . [ sec : andreevbs ] we investigate the bound state spectrum supported by topological topological ( sec . [ sec : ps - ps ] ) and @xmath0-wave topological ( sec . [ sec : s - ps ] ) josephson junctions . then in sec . [ sec : josephsoncurrent ] we calculate the josephson currents for our junction setups and contrast the results to the corresponding nontopological cases . to test our scattering matrix predictions , in sec [ sec : numerics ] , we compare our results to numerical simulations of a lattice model of time - reversal invariant josephson junctions . finally , in sec . [ sec : concl ] we present our conclusions . in mean field theory , the excitations of superconductor systems can be described in an effectively single - particle picture in terms of the so - called bogoliubov - de gennes hamiltonian,@xcite which is a @xmath10 matrix in electron - hole space @xmath11 here @xmath12 and @xmath13 are the single - particle hamiltonians for electrons and holes , respectively , and @xmath14 is the superconducting pair potential . each of the blocks of eq . act on spin-@xmath15 electrons and we have @xmath16 where @xmath17 is the operator for time - reversal with pauli matrices @xmath18 acting in spin - space and complex conjugation operator @xmath19 . this choice of @xmath13 corresponds to a basis which makes spin - rotation properties manifest : in a spin - rotation symmetric system , all blocks of @xmath20 are proportional to the identity matrix in spin - space . a josephson junction is composed of two superconductors linked together by a normal region with a phase difference @xmath8 across the junction . we will describe such junctions , with normal regions @xmath21 , using the step - function model@xcite @xmath22 to describe topological superconductors in our setups , we focus on the case where @xmath12 describes a spinful system that is , at least in the superconductors and at the normal - superconductor interfaces , effectively 1d ( e.g. , the spin - orbit coupled nanowire in the hybrid proposals@xcite ) with a single conducting channel . the two spin components give rise to two right - moving and two left - moving fermi points . class diii topological superconductivity arises@xcite when the spectrum acquires superconducting gaps @xmath23 of opposite signs at the two right - moving fermi points . the gaps at the left - moving pair of fermi points follow by time - reversal symmetry . for our purposes , it is sufficient to focus on the physics in the vicinity of the fermi points . in this case , in addition to spin , @xmath12 acquires a further @xmath24 block structure for left- and right - movers . we have @xmath25 [ eq : hbdglin ] where the entries are ordered from the rightmost to the leftmost fermi point . in eq . , @xmath26 and @xmath27 are the fermi velocities where the upper sign refers to the electron hamiltonian and the lower to the hole hamiltonian . in eq . , @xmath28 refers to the left / right superconductor and , without loss of generality , we will take @xmath29 . moreover , by parameterising the pairings as @xmath30 , the system can be viewed as having a time - reversal invariant @xmath1-wave pairing @xmath31 with the admixture of a smaller @xmath0-wave pairing @xmath32 . in line with a number of concrete proposals describing systems conserving the @xmath33-component of spin , one can view @xmath34 as the right-/left - moving spin-@xmath35 velocities and @xmath23 as the pairing at the corresponding fermi points . ( the complementary spin-@xmath36 quantities follow via time - reversal . ) the hamiltonian is , however , more general and holds in the absence of a conserved spin component ; it can be used to describe the superconductors in both setups we consider . for the topological topological junction , we consider the gap - symmetric case @xmath37 , while for the @xmath0-wave topological junction we impose no such requirement . for nontopological , singlet superconductor based , josephson junctions in the short - junction limit , time - reversal invariance is known to allow for elegant relations to the normal - state scattering properties,@xcite such as beenakker s formula for @xmath0-wave junctions@xcite @xmath38 expressing the andreev ( i.e. subgap ) bound state energies in terms of the normal - state transmission probability @xmath39 at the fermi energy . as we show below , a number of analogous exact relations exist in the topological case . the starting point for establishing these are the andreev reflection matrices at normal ( topological ) superconductor interfaces . excitations with energies below the superconducting gap can be described in terms of andreev reflections at the superconductor / normal ( sn ) interfaces . these andreev reflections are described by the andreev reflection matrices @xmath40 and @xmath41 , which respectively contain amplitudes for a hole reflecting into an electron and vice versa . we now use eq . as a starting point to derive the andreev reflection matrices at sn interfaces , which will serve as key ingredients for formulating the scattering matrix description of our systems . the superconductors are 1d , forming single transport channels which carry electrons and holes of a particular spin , leading to four subchannels overall . the key assumption that we make is that the four electron - hole dirac cones at the four fermi points remain uncoupled for clean sn interfaces . this so - called andreev approximation@xcite holds when the superconducting coherence length is much longer than the inverse of the separation between neighbouring fermi points or , when the bulk superconducting system conserves a spin component , between the left- and right - moving fermi points . the approach to deriving the andreev reflection matrices is essentially the same as for josephson junction composed of @xmath0-wave superconductors , as treated by beenakker.@xcite the andreev reflection matrices take the form @xmath42 where we have introduced the subblocks @xmath43 with @xmath44 the scattering of electrons and holes about the normal region is described by the scattering matrices @xmath45 and @xmath46 , respectively . together these matrices must obey certain symmetry relations as a result of the structure of the hamiltonian . the first relation that must be satisfied as a consequence of particle - hole symmetry is @xmath47 the second relation that must be satisfied due to time - reversal invariance is @xmath48 the elements of @xmath49 change on the scale of @xmath50 , where @xmath51 is the dwell time in the junction . in terms of @xmath51 and @xmath14 , the short - junction limit requires @xmath52 . since we will consider energies on the scale of a few @xmath14 or less , the energy dependence of the scattering matrix @xmath49 may be neglected,@xcite allowing us to take the scattering matrix at the fermi level @xmath53 . the electron subblock of the scattering matrix has the general form @xmath54 and , upon using time - reversal symmetry [ eq . ] in the single channel case of interest , we can parameterise@xcite @xmath55 where @xmath56 and @xmath57 are complex numbers . the transmission probability @xmath39 encodes the normal - state conductance as @xmath58 . the spin - orbit scattering has been introduced through the @xmath24 unitary matrix @xmath59 , where @xmath60 is a real phase and @xmath61 is an su@xmath62 matrix which can be parameterised by euler angles ( @xmath63 ) via @xmath64 where the parameter @xmath65 is a measure of the degree of spin - flip scattering . furthermore , due to the unitarity of @xmath45 , we have the following identity @xmath66 at this point , we are now ready to consider specific junction setups and how the andreev energies may be obtained . the scattering processes described in the previous section lead to andreev bound states in the normal region of the josephson junction , where the bound states have a spectrum of energies dependent on @xmath8 . these energies are the roots of the secular equation@xcite @xmath67=0 , \label{eq : secular1}\end{aligned}\ ] ] where we have introduced @xmath68 as the @xmath69 identity matrix . we now turn to obtaining the andreev levels by solving the secular equation for the junction setups depicted in fig . [ fig : setups](a ) and fig . [ fig : setups](b ) . here we work towards obtaining the spectrum of andreev bound states in the topological topological junction depicted in fig . [ fig : setups](a ) . for simplicity , we take the pairing strengths to be identical on either side of the junction , allowing us to suppress the @xmath70 index for convenience . by substituting the scattering matrix and the andreev reflection matrices into the secular equation and also employing eq . , the particle - hole and time - reversal symmetries in eq . - , and the folding identity @xmath71 the secular equation may be recast into the form @xmath72=0 , \label{eq : toptopsecular0}\end{aligned}\ ] ] where we have used @xmath73 , @xmath74 , and @xmath75 . since @xmath76 , we may rewrite eq . as @xmath77=0 , \label{eq : toptopsecular1}\end{aligned}\ ] ] where @xmath78 is the diagonal matrix of eigenvalues of @xmath79 . we now demonstrate that the only relevant @xmath80 parameter in eq . is @xmath65 . directly substituting this parameterisation into @xmath79 , we find that @xmath81 so that the eigenvalues contained in @xmath78 are those of @xmath82 and hence only the parameter @xmath65 plays a role . since the determinant in the secular equation is of a diagonal matrix , the secular equation amounts to either one or both of two equations being satisfied . introducing @xmath83 and @xmath84 , where @xmath85 the two equations are @xmath86=0 \label{eq : toptopsecular2}\end{aligned}\ ] ] and these equations together can contribute up to four solutions to the andreev spectrum . if we consider the special case where @xmath87 ( i.e. setting the @xmath0-wave admixture to zero ) , a compact expression may be obtained for the bound state energies @xmath88 where the @xmath89 has a superscript prime in order to distinguish it from the @xmath90 . . already illustrates the effect of the spin - orbit scattering : as it is varied , it translates half of the andreev levels to the left and the other half to the right in @xmath91-@xmath8 space . furthermore , as a result of time - reversal symmetry , the branch crossings at @xmath6 , where @xmath7 is an integer , are protected by kramers theorem . -wave pairing @xmath92 corresponding to @xmath93 . the transmission probability @xmath39 and spin - orbit parameter @xmath65 vary throughout . the transmission probability above which bound states are lost to the continuum is @xmath94 [ see eq . ] . here and in what follows , we express results in terms of the gap parameter with smallest magnitude and also show data ( black crosses / dashed lines ) from a numerical lattice simulation ( see sec . [ sec : numerics ] ) of the corresponding junctions . the gaps in the simulations here are @xmath95 and @xmath96 . ] for the case where @xmath97 ( i.e. in the presence of an @xmath0-wave pairing admixture ) , the secular equation may be solved analytically , but such an expression is cumbersome due to its size . it , however , displays a number of salient and robust features shared with the purely @xmath1-wave case . the first of these is the crossing of the energy branches at @xmath6 ( with @xmath7 integer ) ; as these are time - reversal invariant phases , this is simply a consequence of kramers theorem . a less obvious finding regards the zero crossings @xmath98 of the energies , defined by @xmath99 . from the secular equation , we analytically extract the values of @xmath98 as @xmath100 the locations of these zero crossings are identical to the case of a junction without @xmath0-wave pairing , as in eq . , meaning that the inclusion of a weak @xmath0-wave pairing does not move or remove the zero crossings . as a result of this , the andreev branches will remain @xmath3 periodic in the presence of an @xmath0-wave pairing . ( strictly speaking , in the linear junctions we consider , the zero crossings and the consequent @xmath3 periodicity are only approximate , neglecting the hybridisation with the majorana - kramers pairs at the far ends of the wires given its exponential suppression in system size . exact zero crossings arise in a ring geometry when the only majorana fermions in the system are in the junction . ) the bound state energies for this junction display one of the electron - hole and time - reversal symmetry protected topological patterns proposed by zhang and kane@xcite in the context of anomalous topological pumps . examples of the subgap energies are depicted in fig . [ fig : p+sandreevenergies ] . ) corresponding to @xmath101 . the transmission probability is @xmath102 . the spin - orbit parameter @xmath65 is varied in ( a)-(f ) , highlighting its role in splitting the levels in @xmath8-space . levels are lost to the continuum around @xmath103 which is a generic feature when @xmath104 . the numerical lattice data ( black crosses / dashed lines ) has gap parameters @xmath105 and @xmath106 . ] an important feature of the finite @xmath0-wave pairing admixture is the existence of an intergap energy regime ( @xmath107 ) where the subgap energies generally can not exist . branches which would appear to extend past @xmath108 up towards @xmath109 vanish after hitting the @xmath108 threshold [ e.g. , see fig . [ fig : p+sandreevenergies](b ) ] . as these branches approach the threshold , their derivatives are found to smoothly tend to zero . the existence of an intergap regime has ramifications for the josephson current and is discussed in sec . [ sec : josephsoncurrent ] . whether there are subgap branches which reach the @xmath108 threshold is dependent only on the magnitude of @xmath39 and the pairings @xmath23 . we find a critical transmission probability @xmath110 above which branches of the andreev spectrum are lost to the intergap regime . it has the form @xmath111 an exception is the case of perfect transmission and a conserved @xmath33-component of spin ( i.e. @xmath112 and @xmath113 ) , where the @xmath109 and @xmath108 modes are not mixed and hence the subgap energies are not lost to the continuum . in the case where the @xmath0-wave admixture is stronger than the @xmath1-wave pairing ( i.e. @xmath114 ) , the superconductor is nontopological . from the secular equation , the andreev energies for a nontopological junction are given by the solutions to the two equations @xmath115=0 . \label{eq : nontopsecular}\end{aligned}\ ] ] from this equation it may be demonstrated that , other than the exceptional case of @xmath112 , there are no zero - energy modes for nontopological junctions . as a consequence of time - reversal symmetry , the andreev levels must cross at phases @xmath6 . we also note that the spin - orbit parameter plays a role in splitting up the andreev levels in @xmath8-space ; this is depicted for various values of @xmath65 in fig . [ fig : nontopp+sandreevenergies ] . these levels correspond to the trivial pump in terms of ref . . we now investigate the andreev bound state energies in the @xmath0-wave topological superconductor setup , depicted in fig . [ fig : setups](b ) . specifying that the @xmath0-wave superconductor is to the left of the topological superconductor , the andreev reflection matrices are as in eqs . - , where the gap parameters have the form @xmath116 and we also take @xmath117 and @xmath118 . now by employing the parameterisation of the scattering matrix and the folding identity , the secular equation may be brought to the form @xmath119 where we have introduced @xmath120 and @xmath121 and the branch choice of the square root is irrelevant as long as the same choice is made for both @xmath122 and @xmath123 . note here that the secular equation is already independent of the spin - orbit scattering . furthermore , it may be demonstrated from this equation that the zero crossings occur when @xmath6 , where @xmath7 is an integer . this is precisely the behaviour one expects as for these phases the junction realises a time - reversal invariant interface between a topological and a nontopological gapped system which must harbour a kramers pair of majorana zero modes . ( the zero crossings , again , are strictly speaking approximate , neglecting the hybridisation with the majorana - kramers pairs at the far ends of the wire . ) a number of interesting limiting cases exist where the subgap energies have a compact solution . if we first consider the gap - symmetric case ( i.e. @xmath124 ) , the andreev bound state energies have the analytical solution @xmath125 where two of the bound states are low energy ( corresponding to the negative sign of @xmath90 ) and the other two are high energy ( corresponding to the positive sign of @xmath90 ) . another interesting limit is when the conventional superconductor is strong compared to the topological superconductor ( i.e. @xmath126 ) , where the subgap energies are @xmath127 in the converse case ( i.e. @xmath128 ) , the subgap energies are independent of the transmission probability , taking the form @xmath129 provided that @xmath130 . the @xmath39-independence of the energies in this limit can be attributed to the fact that , for @xmath128 , on the effectively @xmath1-wave side , only @xmath131 is involved in the andreev reflections which therefore become resonant.@xcite in the absence of an @xmath0-wave admixture and also where the left and right superconducting gaps differ in magnitude ( i.e. @xmath132 ) , an intergap regime opens up where it is possible for andreev levels to escape into the continuum . this result is in agreement with ioselevich _ _ et al.__@xcite and also explains why the high - energy solutions are absent in eqs . - . upon the inclusion of a small @xmath0-wave pairing component in the topological superconductor , we find that a second intergap regime ( @xmath133 ) opens up where the andreev bound states are able to escape into the continuum . this second intergap regime is of the same type as in the case of topological topological superconductor junctions in sec . [ sec:3c ] . with both of these intergap regimes , there are three types of junction that are possible depending on the relative magnitudes of the superconducting gaps . in general , the intergap regime spans over the range of energies @xmath134 . we find that the feature of a pair of high - energy solutions and a pair of low - energy solutions that arose in the gap - symmetric case remains generally true in the gap - asymmetric case with one caveat : as the high - energy solutions would appear to extend up to the largest gap of the system , high - energy subgap states are generally lost to the continuum as the intergap regime always spans from the smallest gap parameter to the largest . illustrative examples of the bound state spectrum are depicted in fig . [ fig : sandp+sandreevenergies ] . -wave superconductor ( with gap @xmath135 ) and a topological superconductor ( with gaps @xmath23 ) with a finite @xmath0-wave pairing admixture . for ( a ) and ( b ) high - energy andreev levels are generally present , while for ( c ) and ( d ) they are lost to the continuum for all phase differences @xmath8 . the numerical lattice data ( black crosses / dashed lines ) uses gap parameters @xmath136 with ( a)-(b ) @xmath137 and ( c)-(d ) @xmath138 . ] the system becomes nontopological when the @xmath0-wave admixture becomes greater than the @xmath1-wave pairing ( i.e. @xmath114 in our convention ) . from the secular equation it follows that , in this case , andreev levels cross zero energy only in the exceptional case @xmath112 at phase differences @xmath139 with @xmath7 integer . the majorana - kramers pair at each @xmath6 are now absent . various nontopological subgap energies are depicted in fig . [ fig : nontopsandp+sandreevenergies ] . -wave superconductor with pairing @xmath135 and a nontopological superconductor with pairings @xmath109 and @xmath114 . the numerical lattice data ( black crosses / dashed lines ) uses gap parameters @xmath140 with ( a)-(b ) @xmath141 and ( c)-(d ) @xmath142 . ] the josephson current @xmath143 for our junctions comes from two contributions . one of these has been our focus so far : the andreev bound state spectrum states which are confined to the normal region of the josephson junction . the other current contribution originates from the continuous spectrum states with an energy larger than the smallest of the superconducting gaps , allowing them to extend into their respective superconductors . the total current is @xmath144 where @xmath145 is the current due to the andreev bound states and @xmath146 is the current due to the continuous spectrum . in what follows , for simplicity , we will consider the zero temperature limit and some related cases where the bogoliubov - de gennes energy levels have definite occupancies . ( we will comment on finite temperature considerations in sec . [ sec : concl ] ) . in this case , each of the contributions can be calculated as @xmath147 where @xmath148 is the contribution of the part @xmath149 of the spectrum to the total energy . we find , for both of our junction setups , that the contribution @xmath146 is zero for energies above the largest superconducting gap of the junction , while it is generally nonzero for energies that lie within an intergap regime . @xmath145 is also generally nonzero for both of our topological junction setups . the current that is measured depends on the speed at which @xmath8 changes relative to the energy and fermion parity relaxation times of the junction . in what follows , given that fermion parity relaxation naturally involves energy relaxation but not necessarily vice versa , we consider three complementary cases : ( i ) when both energy and fermion parity relaxation can be neglected , ( ii ) with fast energy relaxation but fermion parity conservation , and ( iii ) when both energy and fermion parity relaxation are fast . in all cases , we use a protocol where the intended sweep is preceded by some period of slow sweep in regime ( iii ) , which ensures definite bogoliubov - de gennes level occupancies . in addition to this , we assume the sweep speeds described in all cases are slow enough that unintended landau - zener tunnelling between branches or to the continuum is avoided . the contribution to the josephson current by the andreev levels is dependent on the fermion parity of the junction . as the andreev levels come in particle - hole pairs , only half of them may be occupied at a given moment ( due to the redundancy @xmath150 of the corresponding fermion operators ) . if we associate one pair to have energy @xmath151 and the other pair to have energy @xmath152 , where the andreev levels are labelled as in fig . [ fig : paritydemo](a ) , then the total energy from the andreev levels is@xcite @xmath153 , \label{eq : parityenergies}\end{aligned}\ ] ] where @xmath154 is the fermion parity of the junction . in using the quantum numbers @xmath155 , we neglect the hybridisation with the majorana - kramers pairs at the far ends of the wires as this is exponentially suppressed with system size . ( this approximation influences only the parity conserving cases discussed below , where it amounts to assuming that @xmath8 changes quickly enough so that landau - zener tunnelling occurs with probability unity across the exponentially small splittings that the zero crossings approximate . ) the concrete choice of @xmath155 , as @xmath8 is varied , will depend on the junction s interaction with its environment , specifically on whether the junction is able to relax to the ground state ( potentially subject to a parity constraint ) . we will separately discuss each of these cases in what follows . we consider the possible energies and currents in the absence of energy relaxation and , additionally for landau - zener tunnelling to the continuum to be in principle avoidable , with no levels escaping into the continuum . we have four branches of energy for the four values of @xmath156 , each of which is @xmath3 periodic in @xmath8 . this is illustrated in fig . [ fig : paritydemo](b ) . the consequent current contribution is also @xmath3 periodic . we now consider the subgap energy and its contribution to the current when the energy relaxation is much faster than the sweep speed of @xmath8 . this amounts to choosing the minimum energy @xmath157 within a given parity sector , resulting in branch switches at the locations @xmath158 . these locations correspond to those of finite energy andreev branch crossings which are where energy - minimising andreev branch occupancy switches can occur without a change in fermion parity . examples are depicted in fig . [ fig : paritydemo](c ) . the corresponding current contribution is @xmath4 periodic and has jumps ( current switches " ) at the branch switch locations . , while the transmission probability and spin - orbit parameter are @xmath102 and @xmath159 . in ( a ) the blue and red curves represent @xmath160 and @xmath161 , respectively [ as in eq . ] , while the dashed lines signify their particle - hole partners . in ( b)-(d ) , the blue lines are odd parity , the red lines are even parity energies . fermion parity is conserved in ( b ) and ( c ) , but not in ( d ) . energy relaxation is absent in ( b ) , but not in ( c ) and ( d ) . numerical lattice data has been omitted as the agreement has already been demonstrated to be excellent . ] finally we consider the case where the junction relaxes to its ground state without conserving fermion parity . in this case the junction takes the minimum energy regardless of fermion parity , resulting in energy branch switches at the locations specified by eq . . underlying the switches are now the zero - energy andreev branch crossings which , via changing the fermion parity , ensure that all of the negative energy andreev levels are occupied for all @xmath8 . such switches are indicative of the topological pattern of subgap branches of ref . . examples are shown in fig . [ fig : paritydemo](d ) . the energies and the corresponding current contribution are @xmath4 periodic . the branch switches , again , are accompanied by switches in the current contribution . these current switches are the class diii analogues of the switches in class d systems , discussed in ref . . in sec [ sec : ps - ps ] , we identified the energy regime @xmath162 as where the spectrum is continuous . the josephson current due to this spectrum will depend on the density of states @xmath56 , whose structure is dependent on whether the energy is in the intergap regime @xmath163 or the above gap regime @xmath164 . such a supercurrent may be calculated according to eq . and noting that the contribution of the filled negative energy bogoliubov - de gennes energy levels is @xmath165,@xcite @xmath166 where @xmath56 may be expressed in terms of the total scattering matrix of the junction @xmath167,@xcite @xmath168 + \mathrm{constant}. \label{eq : dos1}\end{aligned}\ ] ] the matrix @xmath167 has the general form @xmath169 with @xmath170 and @xmath171 describing reflection and transmission for modes incoming from within the superconductor at the sn interface , @xmath172 and @xmath173 describing reflection and transmission for modes incoming from the normal region at the sn interface , and @xmath174 describing scattering off the normal region . and ( e)-(h ) its nontopological counterpart with gap parameter @xmath101 . for topological junctions , the critical transmission above which subgap states escape into the continuum is @xmath94 [ see eq . ] . the numerical lattice data ( black crosses ) has gap parameters @xmath95 and @xmath96 for the topological junctions and @xmath105 and @xmath106 for the nontopological junctions . ] _ ( a ) above the gaps . _ in the above the gap regime ( @xmath164 ) , we demonstrate that there are no continuous contributions to the josephson current . the matrix @xmath172 has the general form @xmath175 where @xmath176 and @xmath177 are the @xmath178 matrices from eq . evaluated for @xmath164 . the other three scattering matrices are related via @xmath179 by using the relations in eq . in the expression for @xmath167 , it may be shown , with some algebra , that @xmath180 then , using that @xmath181 are real for energies above the gaps and substituting eq . into the density of states , we find that @xmath182 where @xmath183 . it follows that @xmath184 is real and hence the density of states is constant . therefore , there is no contribution to the josephson current in the above the gap regime . _ ( b ) between the gaps . _ for energies in the intergap regime ( @xmath163 ) we show that there are nonvanishing contributions to the josephson current . the matrix @xmath172 has the same form as in eq . ; however , the rest of the matrices which describe scattering at the sn interface have reduced dimension since the @xmath109 modes can not be transmitted into the superconducting leads . as a result , @xmath167 is a @xmath178 matrix . the continuous contributions to the josephson current are obtained numerically by substituting eq . into eq . . the currents are found to be @xmath4 periodic and their magnitude increases with the transmission probability . for small values of @xmath39 , the contributions resemble sinusoidal functions , becoming increasingly nonsinusoidal as @xmath39 is increased ; this must occur as the continuous contributions compensate for subgap levels which escape into the continuum above @xmath110 . examples of continuous contributions to the current ( measured in in terms of the flux quantum @xmath185 ) are depicted in fig . [ fig : intergapcurrents1](a)-(d ) for the topological case and in fig . [ fig : intergapcurrents1](e)-(h ) for the nontopological case . the total current is calculated by combining the subgap and continuous contributions , as in eq . . in the case without energy relaxation , as in the subgap case , we focus on junctions where andreev levels do not escape into the continuum . however , in the presence of relaxation our discussion includes also the case of escaping levels . examples are depicted in fig . [ fig : p+stotalcurrents ] for topological junctions and in fig . [ fig : nontopsandpandscurrentvaryingt](a ) for nontopological junctions . generally , we find the continuous contributions to be significant relative to the subgap current when the gap asymmetry between @xmath109 and @xmath186 is appreciable . the continuous contributions are typically most significant when there are subgap states which escape into the continuum , as then the continuous contributions must account for the missing subgap levels . there are some distinguishing features between topological and nontopological junctions . one such feature is the fractional josephson effect : when the phase is swept much faster than the relaxation time of the junction ( but not so fast that landau - zener tunnelling between branches or to the continuous spectrum occurs ) , the supercurrent of a topological topological junction will be @xmath3 periodic [ see fig . [ fig : p+stotalcurrents](a)-(b ) ] , twice that of the conventional josephson effect . this @xmath3 periodicity , however , is sensitive to energy relaxation , which motivates the need for a topological signature in junctions where energy relaxation takes place . a distinctive feature appears in the case with energy relaxation : in topological josephson junctions , the josephson current is @xmath4 periodic and displays a current switch at @xmath187 with @xmath188 the fermion parity of the junction and @xmath7 an integer [ see fig . [ fig : p+stotalcurrents](c)-(f ) ] . we have investigated the analogous situation in the nontopological case where , due to the absence of andreev level zero crossings and the use of the preparation protocol in sec . [ sec : josephsoncurrent ] , only the even parity ground state arises and the corresponding current exhibits switchlike behaviour at @xmath189 when @xmath39 is close to unity [ see fig . [ fig : nontopsandpandscurrentvaryingt](a ) ] . with regards to establishing switches as topological , it is most judicious to work with junctions having an intermediate value of @xmath39 in order to suppress nontopological switches , while still realising josephson currents of an appreciable magnitude . when both energy and fermion parity are allowed to relax , further qualitative distinctive features are found to emerge . a single pair of switches occurs in the interval @xmath190 , with the switch locations given by the zero crossings in eq . . as these switches occur symmetrically about @xmath6 , junctions with a value of @xmath65 sufficiently ( i.e. determined by the resolution of @xmath8 ) far enough away from @xmath191 have topological switches that are distinct compared to possible nontopological counterparts . in the case that @xmath65 is near @xmath191 , switches may be distinguished from their nontopological variants provided that @xmath39 is not too close to unity , as then potential nontopological switchlike false positives do not arise [ see fig . [ fig : nontopsandpandscurrentvaryingt](a ) ] . besides being a topological signature ( as explained in sec . [ sec : sxpsubgap ] ) , identifying the location of these switches provides a way to measure the value @xmath65 for the junction . lastly , an important characteristic of these topological switches is that the magnitude of the current is not necessarily the same after a switch . examples of these switches are depicted in fig . [ fig : p+stotalcurrents](g)-(j ) . for @xmath0-wave topological superconductor junctions , due to the high - energy and low - energy subgap branches being generically ( i.e. for @xmath39 away from unity ) well separated and using the protocol outlined in sec . [ sec : josephsoncurrent ] , the cases without relaxation and with energy relaxation only coincide : in terms of the subgap current , they both correspond to always taking the positive high - energy subgap level to be empty . therefore it is sufficient to consider the cases distinguished by the ( non)conservation of fermion parity . with a conserved fermion parity , we include the contribution of one of the low - energy branches where the choice of branch depends on the fermion parity of the junction . the resultant current has @xmath4 periodicity . the nature of the low - energy and high - energy contributions to the josephson current can be understood in terms of symmetries of the system . time - reversal symmetry relates an energy at @xmath8 to one of equal value at @xmath192 . as a result of this and the @xmath4 periodicity of the spectrum , the high - energy andreev branches are even in @xmath8 about the time - reversal invariant phases @xmath6 and the corresponding current contribution is odd , vanishing at these phases . for the low - energy branches , applying time - reversal brings the energy to the opposite branch related to the initial one by particle - hole symmetry . this implies that time - reversal flips fermion parity , signifying the anomalous time - reversal properties of majorana - kramers doublets.@xcite at the same time , this also means that the low - energy branches are odd in @xmath8 about @xmath6 and the corresponding contribution to the current is even with finite currents at @xmath6 . moreover , as a consequence of particle - hole symmetry , each parity sector of the low - energy andreev levels contributes oppositely to the current . -wave topological josephson junction and ( e)-(h ) its nontopological counterpart . the numerical lattice data ( black crosses ) has gap parameters @xmath140 with ( a)-(b ) @xmath137 , ( c)-(d ) @xmath138 , ( e)-(f ) @xmath141 , and ( g)-(h ) @xmath142 . ] for a nonconserved fermion parity , the energy of the junction is always minimised and branch switches occur at @xmath6 ( @xmath7 is an integer ) , where the low - energy branches cross each other at zero energy due to the emergence of a kramers pair of majorana zero modes . this leads to switches in the subgap josephson current contribution , in agreement with the work of chung _ _ et al.__@xcite in the tunnelling limit . as a result of the vanishing high - energy contribution at @xmath6 , these current switches are entirely due to the low - energy andreev levels . moreover , at the switches the sign of the current is flipped while its magnitude is preserved , which is a feature dictated by the time - reversal properties of the spectrum : the contributions just after the switch are the time - reverse of those just before the switch . the switch itself , in this sweep regime , is indicative of a flip in the fermion parity . taken together , these magnitude - preserving sign switches therefore indicate the anomalous time - reversal properties of majorana - kramers pairs . as we will see below , this feature is left intact when continuous contributions are taken into consideration . -wave topological junction . the blue and red curves represent odd and even parity currents . for ( a)-(d ) local fermion parity is conserved while for ( e)-(h ) it is violated via relaxation , resulting in current switches at phase differences @xmath6 with @xmath7 integer . numerical lattice data has been omitted here since the agreement with subgap energies and continuous contributions to the current has already been shown to be excellent . ] as the transmission probability @xmath39 approaches unity . displayed in ( a ) is the total supercurrent for a nontopological nontopological junction with spin - orbit parameter @xmath159 and superconducting gaps @xmath101 and in ( b ) the total supercurrent for an @xmath0-wave nontopological junction with superconducting gaps @xmath193 and @xmath194 . lattice numerics are omitted here as the agreement with subgap energies and continuous contributions to the current has already been shown to be excellent . ] the josephson current due to the continuous contributions is calculated by substituting the total scattering matrix @xmath167 , which has the same general form as in eq . , into the density of states , and finally into eq . . we find that the continuous contributions are generally nonzero , but only when the energies lie in an intergap regime . moreover , as with the subgap contributions , the continuous contributions are independent of the spin - orbit scattering . _ ( a ) above the gaps . _ the calculation of the contribution to the josephson current here is similar to the topological topological case . the andreev reflection matrix here assumes the same general form as in eq . and the relations @xmath195 also hold . as a result of this , @xmath167 takes the same form as in eq . and so an analogue to eq . can be derived ( i.e. involving a determinant which is real ) . therefore the contribution to the josephson current due to energies above the gaps is zero . _ ( b ) between the gaps . _ as mentioned in sec . [ sec : s - ps ] , there are a number of intergap energy regimes that arise . the specific form of @xmath167 will change depending on which intergap energy window one works in ( as this controls what is allowed to extend indefinitely into the superconductors ) and the relative magnitudes of the gap parameters . if we specify the relevant gap parameters @xmath196 , then the intergap contributions are calculated in an analogous fashion to eq . by the integral @xmath197 this integral is calculated numerically in the same way as with topological topological junctions . the continuous contributions for various gap parameters and transmission probabilities can be found for the topological case in fig . [ fig : sandpandscontcontribs](a)-(d ) and for the nontopological case in fig . [ fig : sandpandscontcontribs](e)-(h ) . the continuous contributions are zero at @xmath6 ( with @xmath7 integer ) as energy levels in the continuum must be even in terms of @xmath8 about the phases @xmath6 as a result of time - reversal symmetry . therefore , the contribution to the total current at these phases is entirely due to the low - energy andreev levels , implying that the highlighted key features in the preceding section are not obscured . following eq . , the total current is the combination of both the subgap and continuous contributions . we consider the two cases which differ by the conservation of fermion parity . examples are showcased in fig . [ fig : sandpandstotalcurrent ] for @xmath0-wave topological junctions and in fig . [ fig : nontopsandpandscurrentvaryingt](b ) for @xmath0-wave nontopological junctions . there are some key features in the josephson currents which can reveal whether or not a junction is topological . the most prominent of these arise in the case with both energy and fermion parity relaxation : topological junctions have magnitude - preserving current switches at phase differences @xmath6 ( with @xmath7 integer ) , where the current at these switch locations is solely due to the low - energy subgap levels . furthermore , provided that the transmission probability @xmath39 of the junction is not too close to unity , analogous magnitude - preserving switches are absent in the nontopological case [ see fig . [ fig : nontopsandpandscurrentvaryingt](b ) ] . however , as can be demonstrated using eq . , nontopological variants are only able to support switchlike features at @xmath139 , so the observation of current switches in phase increments of @xmath198 indicates the topological nature of a junction . in addition to these qualitative signatures , our results allow a quantitative fit to experimental data using the normal - state conductance and the induced gaps as inputs , which may facilitate progress in identifying topological josephson junctions . we now turn to testing our predictions on a lattice model@xcite for topological superconductivity in hybrid structures based on a rashba spin - orbit coupled nanowire in proximity to an @xmath2-wave superconductor . the hamiltonian consists of a left and a right subsystem , each with @xmath199 sites , @xmath200 where @xmath201 describes the left proximitised nanowire , @xmath202 describes the right proximitised nanowire , and @xmath203 , \\ h_{0\mathrm{r}}&=\sum_{j = n+1}^{2n-1 } \left[\left(c_{j+1}^{\dagger}v c_{j}+ \mathrm{h.c.}\right)-\mu c^\dagger_j c_j\right ] , \end{aligned}\ ] ] where @xmath204 is the chemical potential , @xmath205 , and @xmath206 with real hopping @xmath207 and spin - orbit energy @xmath208 . the pairing is @xmath209 and in terms of @xmath210 , this leads to @xmath211 the coupling between the two subsystems is via @xmath212 and we allow for arbitrary coupling consistent with time - reversal symmetry , i.e. we take @xmath213 with arbitrary @xmath214 and @xmath215 real . the normal - state transmission matrix @xmath216 , via wavefunction matching , can be shown to be @xmath217 where @xmath218 and @xmath219 ; the phase @xmath60 , explicitly expressible in terms of @xmath220 and @xmath221 , plays no role in what follows . this lattice model is used to numerically investigate the topological topological and @xmath0-wave topological josephson junctions and their nontopological counterparts . in general there is excellent agreement between the scattering matrix results and the numerical lattice model . this is despite considering finite size systems and without imposing the andreev approximation . parameter values that were used ( in units of @xmath207 ) are @xmath222 , @xmath223 , and @xmath224 . the corresponding value for @xmath215 to produce a desired value of @xmath39 may be derived from eq . . the values for gap parameters depend on the specific junction setup and may be found in the relevant figure captions . the numerical lattice data is plotted as black crosses and black dashed lines along with various subgap energies and continuous current contributions . for numerical simulations of topological junctions , the subgap energies display a constant line at zero energy which is due to the majorana fermions at the wire end points away from the junction . by taking a sufficiently long nanowire , the splitting of these modes is exponentially suppressed to zero . for the nontopological variant , there is no line at zero energy , reflecting the absence of majorana fermions in the system . andreev levels for topological topological junctions are depicted in fig . [ fig : p+sandreevenergies ] . the numerics fit closely to the scattering matrix results , correctly reproducing the effect of @xmath65 to move the zero crossings in @xmath8-space and the feature of subgap states escaping into the continuum . similar agreement between the lattice model and the scattering matrix results is also achieved for the nontopological variants in fig . [ fig : nontopp+sandreevenergies ] . the lattice model provides the energy levels of the occupied continuous spectrum and hence the continuous contributions to the current via @xmath225 . these contributions are seen for topological topological junctions in fig . [ fig : intergapcurrents1](a)-(d ) and for nontopological variants in fig . [ fig : intergapcurrents1](e)-(h ) , displaying excellent agreement between the lattice model and the scattering matrix calculation . similarly for @xmath0-wave topological junctions and the corresponding nontopological variants , the lattice model and scattering matrix calculation are in excellent agreement . the case where an @xmath0-wave admixture is included is depicted in fig . [ fig : sandp+sandreevenergies ] for the topological case and in fig . [ fig : nontopsandp+sandreevenergies ] for the nontopological case . the continuous contributions to the supercurrent are shown in fig . [ fig : sandpandscontcontribs](a)-(d ) for the topological case and in fig . [ fig : sandpandscontcontribs](e)-(h ) for the nontopological case . we have studied the josephson effect for time - reversal invariant topological topological and @xmath0-wave topological junctions in the short - junction limit . in terms of the subgap energy spectrum , we found topological generalisations of beenakker s formula [ eq . ] for both of the junction setups considered . for topological topological junctions , one such expression [ eq . ] holds in the case when the topological superconductors are characterised by @xmath226 . for @xmath0-wave topological junctions in the analogous @xmath227 regime , three expressions are found to hold for the cases @xmath228 in eq . , @xmath229 in eq . , and @xmath230 in eq . , where @xmath135 is the gap parameter for the conventional @xmath0-wave superconductor and @xmath32 and @xmath31 are the respective @xmath0-wave and @xmath1-wave pairings for the topological superconductor . for more general cases we are able to find simple analytical expressions to establish key features , such as the location of zero crossings and when subgap states are lost to the continuum . the supercurrent generally has contributions from both the subgap andreev levels and the continuous spectrum . the continuous contributions originate from states with an energy lying in an intergap regime ( arising when there is either an @xmath0-wave admixture present in the topological superconductor or a gap asymmetry across the junction ) . the continuous contributions are generally significant in magnitude relative to the subgap contributions provided that the junction is away from the tunnelling limit and that the intergap regime is not too small . for topological topological junctions , we found that the fractional josephson effect can occur , which has the hallmark of a @xmath3 periodic current . the establishment of such a current , however , is challenging due to energy relaxation . as with class d superconductors , the periodicity can be measured via the ac josephson effect : the phase sweep may be controlled by the voltage ( @xmath9 ) across the junction and the resulting periodicity of the supercurrent can be measured . the @xmath3 periodicity of the current will only remain intact provided that the phase sweep speed is much faster than @xmath231 ( @xmath232 ) , the typical timescale for energy relaxation ( but still slow enough so that landau - zener tunnelling between branches or to the continuum is avoided ) . for the case with energy relaxation ( @xmath233 ) , the current is @xmath4 periodic . this is in contrast to class d superconductors , where the conservation of local fermion parity at the junction suffices to ensure that the @xmath3 periodicity is robust against energy relaxation . due to the difficulties posed by energy relaxation in realising a @xmath3 periodic current , we have proposed signatures unique to topological junctions in the presence of energy relaxation . for topological topological junctions , the regime of fermion parity nonconservation is particularly useful . this is the dc josephson effect where the zero - energy crossings of the andreev levels [ eq . ] , directly linked to the topological symmetry protected energy - phase relation of ref . , lead to a pair of current switches in the interval @xmath234 . via eq . , the distance between the switches also provides a direct measure of the spin - orbit parameter @xmath65 . for the special values @xmath235 , the pair of switches degenerate and can be at the same location as the switchlike features of highly transmitting nontopological junctions . to be able to attribute the current switches to that of a topological origin , it is therefore useful to either move the junction away from such special points by changing the spin - orbit coupling or to work away from unit transparency . for @xmath0-wave topological superconductor junctions , we also found the dc josephson effect to be the regime with the clearest signatures of topological superconductivity . in this case , magnitude - preserving current switches occur at the time - reversal invariant phases @xmath6 ( with @xmath7 integer ) , reflecting the existence and anomalous time - reversal properties of majorana - kramers pairs at these phases . in contrast , nontopological analogues may only support switchlike features for @xmath39 close to unity at @xmath139 . therefore , the observation of magnitude - preserving switches after every phase increment of @xmath198 is an indicator of topology . we have worked in the zero temperature limit for both types of junction setup . in terms of the dc josephson effect , our work may be straightforwardly extended to finite temperature @xmath236.@xcite the current switches remain intact away from zero temperature provided that @xmath237 is satisfied , where @xmath238 is the magnitude of the current switch . as the temperature increases outside of this regime , the switchlike features are lost as they are smoothed out over @xmath8-space . our work on these junctions provides a direct relationship between the excitation spectrum , the josephson current , the normal - state conductance @xmath239 , the various superconducting gaps , and , in the case of topological topological junctions , the spin - orbit parameter @xmath65 . relationships of this type in the nontopological case have been known to provide a powerful way in quantitative experimental analyses of josephson junctions.@xcite our results allow the application of the same strategy to the topological case . 70ifxundefined [ 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://iopscience.iop.org/article/10.1070/1063-7869/44/10s/s29/meta [ * * , ( ) ] link:\doibase 10.1103/physrevlett.100.096407 [ * * , ( ) ] http://iopscience.iop.org/article/10.1088/0034-4885/75/7/076501/pdf [ * * , ( ) ] http://www.annualreviews.org/doi/abs/10.1146/annurev-conmatphys-030212-184337 [ * * , ( ) ] http://journals.aps.org/prb/abstract/10.1103/physrevb.79.161408 [ * * , ( ) ] http://journals.aps.org/prl/abstract/10.1103/physrevlett.106.077003 [ * * , ( ) ] http://journals.aps.org/prb/abstract/10.1103/physrevb.84.180502 [ * * , ( ) ] http://journals.aps.org/prl/abstract/10.1103/physrevlett.107.236401 [ * * , ( ) ] http://journals.aps.org/prl/abstract/10.1103/physrevlett.108.257001 [ * * , ( ) ] http://journals.aps.org/prb/abstract/10.1103/physrevb.86.140503 [ * * , ( ) ] http://journals.aps.org/prl/abstract/10.1103/physrevlett.111.056402 [ * * , ( ) ] http://journals.aps.org/prb/abstract/10.1103/physrevb.88.214514 [ * * , ( ) ] http://journals.aps.org/prx/abstract/10.1103/physrevx.4.021018 [ * * , ( ) ] link:\doibase 10.1103/physrevb.93.205430 [ * * , ( ) ] link:\doibase 10.1103/physrevb.93.125435 [ * * , ( ) ] link:\doibase 10.1103/physrevb.93.220507 [ * * , ( ) ] link:\doibase 10.1103/physrevb.94.014511 [ * * , ( ) ] @noop http://www.nature.com/nphys/journal/v8/n11/full/nphys2429.html [ * * , ( ) ] http://pubs.acs.org/doi/abs/10.1021/nl303758w [ * * , ( ) ] http://www.nature.com/nnano/journal/v10/n7/full/nnano.2015.86.html [ * * , ( ) ] link:\doibase 10.1038/ncomms10303 [ * * , ( ) ] @noop http://journals.aps.org/prb/abstract/10.1103/physrevb.78.195125 [ * * , ( ) ] http://journals.aps.org/prl/abstract/10.1103/physrevlett.102.187001 [ * * , ( ) ] link:\doibase 10.1088/1367 - 2630/12/12/125002 [ * * , ( ) ] link:\doibase 10.1103/physrevb.88.035121 [ * * , ( ) ] http://journals.aps.org/prl/abstract/10.1103/physrevlett.105.077001 [ * * , ( ) ] http://journals.aps.org/prl/abstract/10.1103/physrevlett.105.177002 [ * * , ( ) ] link:\doibase 10.1103/physrevb.83.054513 [ * * , ( ) ] link:\doibase 10.1103/physrevb.88.020407 [ * * , ( ) ] link:\doibase 10.1126/science.1259327 [ * * , ( ) ] http://journals.aps.org/prb/abstract/10.1103/physrevb.55.1142 [ * * , ( ) ] link:\doibase 10.1103/physrevb.81.125318 [ * * , ( ) ] http://journals.aps.org/prl/abstract/10.1103/physrevlett.108.147003 [ * * , ( ) ] http://journals.aps.org/prl/abstract/10.1103/physrevlett.110.117002 [ * * , ( ) ] http://journals.aps.org/prb/abstract/10.1103/physrevb.86.184516 [ * * , ( ) ] link:\doibase 10.1103/physrevb.89.220504 [ * * , ( ) ] link:\doibase 10.1103/physrevb.90.045118 [ * * , ( ) ] link:\doibase 10.1103/physrevb.90.155447 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.115.237001 [ * * , ( ) ] @noop @noop @noop , @noop http://journals.jps.jp/doi/abs/10.1143/jpsj.78.062001 [ * * , ( ) ] http://www.nature.com/nature/journal/v464/n7286/abs/nature08914.html [ * * , ( ) ] link:\doibase 10.1103/physrevlett.108.037002 [ * * , ( ) ] http://link.springer.com/article/10.1140%2fepjb%2fe2004-00066-4 [ * * , ( ) ] http://journals.aps.org/prb/abstract/10.1103/physrevb.84.081304 [ * * , ( ) ] @noop _ _ ( , ) link:\doibase 10.1103/revmodphys.51.101 [ * * , ( ) ] link:\doibase 10.1002/pssa.2210470266 [ * * , ( ) ] http://journals.aps.org/prl/abstract/10.1103/physrevlett.67.3836 [ * * , ( ) ] link:\doibase 10.1103/physrevb.53.r11957 [ * * , ( ) ] link:\doibase 10.1103/physrevb.56.892 [ * * , ( ) ] link:\doibase 10.1103/physrevb.57.6084 [ * * , ( ) ] http://stacks.iop.org/0034-4885/63/i=10/a=202 [ * * , ( ) ] @noop * * , ( ) , http://journals.aps.org/prb/abstract/10.1103/physrevb.77.045311 [ * * , ( ) ] http://link.aps.org/doi/10.1103/physrevb.90.020501 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.103.237001 [ * * , ( ) ] link:\doibase 10.1088/1367 - 2630/13/5/053016 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.90.226806 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.66.76 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.77.2542 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.85.170 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.99.127005 [ * * , ( ) ] link:\doibase 10.1038/nature12315 [ * * , ( ) ] http://journals.aps.org/prx/abstract/10.1103/physrevx.3.041034 [ * * , ( ) ]

photons that escape the interstellar medium of star - forming galaxies may be resonantly scattered by neutral hydrogen atoms in the circumgalactic and intergalactic media , thereby increasing the angular extent of the galaxy s emission . we present predictions of this extended , low surface brightness emission based on radiative transfer modeling in a cosmological reionization simulation . the extended emission can be detected from stacked narrowband images of emitters ( laes ) or of lyman break galaxies ( lbgs ) . its average surface brightness profile has a central cusp , then flattens to an approximate plateau beginning at an inner characteristic scale below.2 mpc ( comoving ) , then steepens again beyond an outer characteristic scale of mpc . the inner scale marks the transition from scattered light of the central source to emission from clustered sources , while the outer scale marks the spatial extent of scattered emission from these clustered sources . both scales tend to increase with halo mass , uv luminosity , and observed luminosity . the extended emission predicted by our simulation is already within reach of deep narrowband photometry using large ground - based telescopes . such observations would test radiative transfer models of emission from laes and lbgs , and they would open a new window on the circumgalactic environment of high - redshift star - forming galaxies .

1010.3017

high - redshift star - forming galaxies are becoming an important probe of galaxy formation , reionization , and cosmology . ionizing photons of young stars in star - forming galaxies ionize neutral hydrogen atoms in the interstellar medium ( ism ) , and each subsequent recombination has a probability of @xmath2 of ending up as a photon @xcite . after escaping the ism , these photons can be scattered by neutral hydrogen atoms in the circumgalactic and intergalactic media ( igm ) , which tends to make the emission extended . in this paper , we present predictions for the extended emission associated with high - redshift star - forming galaxies from a radiative transfer model of emission applied to a hydrodynamic cosmological simulation @xcite . as a result of reprocessed ionizing photons , prominent emission can be a characteristic of star - forming galaxies , which can be used to detect high - redshift galaxies . galaxies detected through the strong emission associated with them ( e.g. , from narrowband photometry ) are dubbed lyman - alpha emitters ( laes ) and an increasing number of such galaxies have been discovered ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? high - redshift star - forming galaxies can also be detected from broadband photometry with lyman break technique , which are termed lyman break galaxies ( lbgs ; e.g. , @xcite ) . as long as photons reprocessed from ionizing photons can escape from the ism , emission is also expected to be associated with lbgs . emission encodes useful information about star - forming galaxies and their environments , such as star formation rate ( e.g. , @xcite ) , kinematics of ism gas ( e.g. , @xcite ) , and neutral fraction of igm gas ( e.g. , @xcite ) . photons usually experience complex radiative transfer processing ( resonant scatterings ) in media with neutral hydrogen atoms , which complicates the interpretation of observed emission properties . @xcite ( hereafter paper i ) present a physical model of emission from laes by solving radiative transfer in the circumgalactic and intergalactic media . the radiative transfer calculation is performed in a cosmological volume ( @xmath3 on a side ) from a state - of - the - art radiation - hydrodynamic reionization simulation @xcite . the calculation uses a @xmath4 grid to sample the density , velocity , and temperature of the neutral hydrogen gas in the simulation and is applied to all the sources residing in halos above @xmath5 . resonant scatterings enable photons to probe the circumgalactic and intergalactic environments ( density and velocity structures ) around star - forming galaxies . this leads to a coupling between the observed emission properties and the environments . this simple physical model is able to explain an array of observed properties of @xmath65.7 laes the subaru/_xmm - newton _ deep survey ( sxds ; @xcite ) , including spectra , morphology , apparent luminosity function ( lf ) , shape of the ultraviolet ( uv ) lf , and the distribution of equivalent width . the selection imposed by the environment dependent radiative transfer also introduces interesting new features in the clustering of laes ( @xcite , hereafter paper ii ) . in the above radiative transfer model , while the number of photons is conserved after they escape the ism , the scatterings in the circumgalactic and intergalactic media cause the emission to spread spatially . therefore , one generic prediction of the model is an extended emission halo around a star - forming galaxy . observationally , only a fraction of photons can be detected for an individual source , those included in the central part of the extended emission with high enough surface brightness ( tip of the iceberg ) . the outskirts of the halo with low surface brightness is typically buried in the sky noise . in this paper , we show that it is possible to detect the bottom of the iceberg by stacking the narrowband images of a large number of sources to suppress the sky noise . we present the predictions of the extended emission in the stacked image from our radiative transfer model and discuss what we can learn from it . our radiative transfer modeling is performed for sources at @xmath7 . we study properties of surface brightness profile from stacked images in section 2 . in section 3 , we discuss the observational prospects . we summarize our results in section 4 . throughout the paper , we adopt a spatially flat @xmath8cdm cosmological model for our calculations , with a matter density parameter @xmath9 and a hubble constant @xmath10 in units of @xmath11 . distances are expressed in comoving units unless mentioned clearly otherwise . the radiative transfer calculation in paper i is performed in a simulation box of 100@xmath12 ( comoving ) on a side with the neutral hydrogen density , temperature , and peculiar velocity sampled with a 768@xmath13 grid . the cell size is therefore 130.2@xmath14 ( comoving ) , about the virial diameter of a @xmath15 halo . the star - forming region in a high - redshift galaxy , where photons in our study originate , is only a few kpc ( comoving ) across ( e.g. , * ? ? ? so the initial emission is treated as a point source at the halo center for the purpose of the radiative transfer calculation . the size of the pixel for collecting photons is 8 times finer , corresponding to 16.3@xmath14 ( comoving ) or 0.58 . with the cell resolution , the gas distribution is uniform inside the virial radius of small halos ( below a few times @xmath16 ) , which would have some effect on the surface brightness profile . however , since photons are initially emitted from a point source and the dynamics of the infall region around halos plays a significant role in determining the distribution of photons ( paper i ) , the resolution we use is sufficient for obtaining the generic features in the extended emission . we present a resolution test in 2.5 , and the main results presented in this paper would remain valid with improved resolution . our calculation does not address the effects of galactic wind and dust , and we discuss these model uncertainties in 2.4 and 3 . we start from the stacked image for sources residing in halos of fixed mass , @xmath17 . in our model , the intrinsic luminosity , which is the the total amount of emission from the reprocessed ionizing photons , is tightly correlated with the halo mass , therefore this stacking is basically for sources at fixed intrinsic luminosity or uv luminosity . our radiative transfer model produces a three - dimensional array , recording the spectra as a function of spatial position on the sky . we construct the narrowband image from this array , with a filter width similar to that in the @xmath7 sxds ( see paper i ) . our simulation box has about the same area as sxds with a redshift depth three times larger . so we are able to construct narrowband images for three sxds - like fields . each image corresponds to an ideal case with perfect continuum and sky subtraction . examples of images of individual sources can be found in paper i ( fig.4 and fig.5 ) . we stack all the source images together as would be done with the narrowband observation . for each source , the image includes the center and surrounding pixels ( up to a radius of @xmath0 10@xmath12 ) . the center pixel is chosen to be the one that contains the halo center . in other words , the center corresponds to that in the uv band . it is evident that the surrounding pixels can include photons from other neighboring sources . for all the sources in a narrow bin ( 0.16 dex ) of halo mass around @xmath17 ( about 440 sources for each of the sxds - like fields ) , we stack their images with their centers aligned . typically there are two algorithms to generate a stacked image , taking either the mean or the median . the stacked image from the mean algorithm is shown in figure [ fig : stack_img ] . the mean algorithm has the advantage of conserving the flux , even though it may be affected by outliers in the flux distribution , showing up as many small clumps in the stacked image . the median stacked image has the advantage of insensitivity to outliers . we find that at a fixed projected radius , the observed flux roughly follows a log - normal distribution among different sources , which causes the surface brightness from the median stacked image to be lower in amplitude than that from the mean stacked image ( see solid curves in figure [ fig : sb_m11 ] ) . the stacking cases mentioned above correspond to an ideal situation of no photon noise . when total photon noise ( source + sky ) is added , the surface brightness profile from the mean stacked image remains the same and that from the median stacked image would change . in the limit of large noise ( compared to the signal distribution ) , the flux distribution is completely modified by the noise distribution , and the median value is driven to be close to the mean ( also see @xcite ) . in practice , the large noise limit is typical at large radii for the case considered here ( see below ) , so the surface brightness profile from the median stacked image approaches that from the mean stacked one . see the thin solid curve in figure [ fig : sb_m11 ] . in what follows , we focus on presenting our results from the mean stacked image . we compute the surface brightness profile by averaging the flux in annuli at different radii , therefore the effect of the grainy feature seen in the image is greatly reduced . the surface brightness profile from the stacked image shows a few interesting features . it has two characteristic scales that are associated with steep changes in the profile slope . the inner slope change occurs at @xmath18 ( possibly upper limit , see [ sec : resolution ] ) and the outer one at @xmath19 . inside the inner characteristic radius @xmath20 , the profile appears as a central cusp , roughly following a steep power - law . between the inner and outer characteristic radii , @xmath21 , the surface brightness decreases slowly with increasing radius , close to a plateau . beyond the outer characteristic radius @xmath22 , the surface brightness profile steepens and drops , forming an extended tail . as shown later , the two scales have clear physical meanings . with the mean surface brightness profile , we compute the cumulative luminosity as a function of projected radius ( bottom panel of figure [ fig : sb_m11 ] ) . it turns out that the luminosity from the stacked image does not converge at large radius . because of the plateau in the surface brightness profile , the luminosity profile approximately follows @xmath23 , a trend similar to that from the global mean surface brightness ( dot - dashed line ) . the global mean profile is computed by uniformly distributing the total flux in the original narrowband image across the whole image . the plateau in the stacked surface brightness profile has a much higher amplitude than the global mean . as we show later , the higher signal is caused by the clustered neighboring sources . in the top panel of figure [ fig : sb_m11 ] , we also show the level of sky noise ( dotted lines ) . the calculation is based on a 4-hour observation with the nb816 filter using the subaru telescope , which is roughly the setup for the survey of @xmath7 laes in sxds @xcite . in the nb816 band , the sky ( ab magnitude of 20.4mag / arcsec@xmath24 in a dark night ) has a surface brightness of @xmath25 . the mean sky photon count in a 2diameter aperture from a 4-hour exposure with the subaru telescope reaches @xmath26 . therefore , the 1@xmath27 sky noise in a 2diameter aperture corresponds to a surface brightness level of @xmath28 , which is higher than the signal of interest on large scales ( fig . [ fig : sb_m11 ] ) . the noise per pixel is expected to be suppressed by @xmath29 with @xmath30 sources stacked . the suppression works even better for detecting the signal at large radii , because the annuli used to compute the mean surface brightness profile can be much larger than the 2aperture . within each annulus , however , the noise in different pixels is correlated , since the same pixel from the original image may fall into the annulus several times from centering different sources . the correlations need to be accounted for in estimating the noise level at each radius for the stacked profile . in practice , given the low surface brightness of the extended emission and the high sky noise level , stacking the image before subtracting the continuum and sky may help to reduce errors caused by such subtraction and to increase the sensitivity . to understand the features in the stacked surface brightness profile and the origin of the two characteristic scales , we perform a test by assigning artificial surface brightness profiles for individual sources . the image of each source ( in halos above @xmath5 ) is assumed to be a circular disk of radius @xmath31 with uniform surface brightness normalized such that the total luminosity is the intrinsic luminosity from the source . that is , we replace the intrinsic point source with a uniform `` seeing '' disk or ( equivalently ) modify it by a tophat point spread function ( psf ) . we then follow the same procedure as above to form the stacked image for sources in @xmath17 halos and derive the surface brightness profile . no radiative transfer is performed in this artificial model , and we use it to illustrate the effect of source extent combined with source clustering . figure [ fig : stack_tophat_m11 ] shows the series of surface brightness profiles with different `` seeing '' disk radii @xmath31 . for a small `` seeing '' disk radius ( @xmath32 ) , sources are close to point - like . the central spike ( inside @xmath31 ) is clearly seen in the profile from the stacked image . outside of this radius , we still see signals , which are obviously contributed from neighboring sources as a result of clustering . the surface brightness profile has contributions from the central sources and from the neighboring clustered sources , analogous to the case of weak gravitational lensing around galaxies . in the spirit of the halo model ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , the profile can be decomposed into two components , that from the individual source ( the one - halo term ) and that caused by the clustering ( the two - halo term ) . formally , we can write the `` two - halo '' term of the stacked surface brightness profile @xmath33 as @xmath34 \sigma_{\rm 1h}({\mathbf r}-{\mathbf r^\prime}|m^\prime),\end{aligned}\ ] ] where @xmath35 is the surface number density of halos , @xmath36 is the 2d two - point correlation function ( essentially the angular correlation function ) , and @xmath37 is the surface brightness profile of individual sources ( the one - halo term ) . for simplicity , equation ( [ eqn:2halo ] ) assumes that @xmath37 depends only on halo mass , while in reality it depends on the environment around halos as well ( see below ) . we emphasize that the one - halo term in the stacked profile is defined as the emission from the central source alone ( i.e. , contributed from all the sources to be center - aligned and stacked ) . even though neighboring sources can have their photons scattered into the one - halo scales of the central source , these photons still belong to the contribution from the two - halo term . in the limit of compact sources ( small @xmath31 in the uniform `` seeing '' disk case ) , @xmath37 can be replaced with @xmath38 , where @xmath39 is the intrinsic luminosity and @xmath40 is the dirac delta function . the two - halo term then becomes @xmath41.\ ] ] for the small `` seeing '' disk case in figure [ fig : stack_tophat_m11 ] ( @xmath32 ) , the flattening below @xmath42 is a sign of the halo exclusion effect , which leads to a zero signal in the 3d two - point correlation function and a constant in the angular two - point correlation function @xmath36 . can be understood as follows . roughly speaking , @xmath43 with @xmath44 in the regime that @xmath45 , where @xmath460.3@xmath12 is the halo exclusion scale and @xmath47 is the 3d two - point correlation function ( zero on scales below @xmath48 ) . as @xmath49 increases , the competition between a decreasing @xmath47 on average and an increasing path length of integration ( i.e. , a decreasing lower limit @xmath50 ) leads to a small bump in @xmath36 around @xmath42 . if the seeing disk is comparable or larger than the size of the halo ( e , g . , the @xmath51 case in figure [ fig : stack_tophat_m11 ] for @xmath17 halos ) , the bump is smeared out . ] according to equation ( [ eqn:2halo_pointsource ] ) , we expect to see a flattened profile on large scales , where @xmath52 . on these scales , which are out of the range of the plot , it is hard in practice to disentangle such a uniform background from the sky background . we therefore limit ourselves to the regime where @xmath36 is still large . by varying the radius @xmath31 of the uniform `` seeing '' disk , we show the effect of the spatial extent of the emission on the surface brightness profile from the stacked image ( figure [ fig : stack_tophat_m11 ] ) . as the `` seeing '' disk size increases , the two - halo term of the profile is smoothed on scales below the `` seeing '' disk size , which results in a decrease in the amplitude . in our setup , the uniform `` seeing '' disk has a surface brightness of @xmath53 for @xmath54 , which is by definition the one - halo term contribution . however , the small - scale surface brightness ( in the one - halo regime ) in figure [ fig : stack_tophat_m11 ] decreases much slower than @xmath55 , as a consequence of the contribution from the smoothed two - halo term . for @xmath56 , the signal is already dominated by the two - halo term on all scales . the approximate plateau seen in the stacked profile from the radiative transfer model ( thick solid curve ) is close to the case with a `` seeing '' disk size of @xmath56 . the test suggests that the plateau feature between the two characteristic scales mainly originates from spatially extended , scattered emission from clustered sources around the central source . the uniform `` seeing '' disk only gives us a rough idea on the features in the extended emission . to be more accurate , we need to examine the surface brightness profile of individual sources . as an example , we show in figure [ fig : psf ] the surface brightness profiles of one individual source in the simulation box from the radiative transfer model . the two curves correspond to the profiles viewed along different directions , and the difference reflects the environment dependent radiative transfer , as discussed in detail in paper i. the surface brightness has a steep drop below @xmath42 . then it levels off until a cutoff is reached around 23@xmath12 . the inner surface brightness profile of one source roughly goes as @xmath57 and the luminosity increases logarithmically with radius . most of the luminosity comes from near the cutoff radius at @xmath02@xmath12 , which is the reason why the plateau up to such a scale in the total profile is produced ( see below ) . as discussed in paper i , the shape of the surface brightness profile largely depends on the velocity field . below @xmath42 , the contribution to the surface brightness profile mainly comes from the photons that are last scattered in the infall region around the halo . on larger scales , where the profile levels off , the last scattering of photons occurs in the igm that undergoes hubble flow . the cutoff radius corresponds to those photons that are the bluest after escaping the infall region . they can travel far away from the source before being redshifted into the line center to be significantly scattered in the igm . the exact shape of the surface brightness profile also depends on the initial frequency of photons ( see [ sec : line_prof ] ) . as emphasized in paper i and paper ii , radiative transfer depends on circumgalactic and intergalactic environments ( density and velocity structures ) . even for sources in halos of the same mass , there is a broad distribution of environments . therefore , the one - halo term of the surface brightness profile is not identical for sources in halos of fixed mass or for the same source viewed along different directions ( figure [ fig : psf ] ) . the one - halo term in equation ( [ eqn:2halo ] ) should have dependence on environment in addition to halo mass . to obtain the mean one - halo term for the stacked image and decompose the surface brightness profile , we perform radiative transfer calculations separately for each source with halo mass in the @xmath17 mass bin . we then stack the individual one - halo terms to obtain the mean . the dotted curve in figure [ fig : stacksb_1h2h ] is the mean one - halo term for sources in @xmath17 halos . the shape is similar to that seen in figure [ fig : psf ] . the dashed curve is the two - halo term , inferred from subtracting the one - halo term from the total surface brightness profile . the two - halo term can be expressed as a convolution ( equation [ eqn:2halo ] ) , with the convolution kernel being the one - halo term of the clustered star - forming halos of _ all _ masses . the shape of the one - halo term at different halo mass is similar to that shown in figure [ fig : stacksb_1h2h ] . in the two - halo term , a plateau reaches out to a radius of 12@xmath12 , which is around the cutoff scale of the mass - averaged one - halo term . this is not a coincidence . the scale where the plateau in the two - halo term ends marks the spatial extent of the extended emission of the clustered sources , since on scales larger than this the smoothing effect from the source psf is small . as the two - halo term dominates on large scales , the above scale is just the outer characteristic scale @xmath22 seen in the total surface brightness profile . the inner characteristic scale @xmath20 marks the transition from the one - halo term dominated regime to the two - halo term dominated regime . the realistic decomposition of the surface brightness profile confirms the results of the test in figure [ fig : stack_tophat_m11 ] that the origin of the plateau between the two characteristic scales in the total surface brightness profile is the extended emission . the environment dependent radiative transfer couples the observed emission with the circumgalactic and intergalactic environments . many consequences of such a coupling are studied in paper i and paper ii . for the stacked surface brightness profile , the environment dependence also appears , as detailed below . at fixed halo mass , which corresponds to fixed intrinsic luminosity or uv luminosity in our model , the observed luminosity has a broad distribution , reflecting the distribution of environments . in our model , for each source we link together the pixels above a surface brightness threshold similar to that used in @xcite to define the observed emission , and if none of the pixels around the central source exceeds the threshold , we simply take the flux in the pixel at the source position as the observed emission ( see paper i for details ) . the observed luminosity is only a fraction of the extended emission , coming from the central , high surface brightness part of the source . the one - halo term of the surface brightness profile is expected to be closely related to the environment . we divide the sources into two samples of equal size , according to the observed luminosity . in figure [ fig : stacksb_halfhalf ] , we show the stacked surface brightness profile for the bright half and faint half , respectively . the dotted curves show the corresponding one - halo terms . we note that the fluxes inferred from the one - halo terms of the two cases are not necessarily the same . the reason is that the scattered emission is not isotropic , being sensitive to the density and velocity structures around the source . overall , the sample of sources with lower observed luminosity has a shallower stacked surface brightness profile at small radius and there is no clear scale of the transition from the inner cusp to the approximate plateau . the result implies that the stacked narrowband image for star - forming galaxies with weaker observed emission would appear to be less compact . so far we have focused on the surface brightness profile for sources in halos of @xmath17 . the profile is expected to be a function of the halo mass . we show the dependence on halo mass in figure [ fig : sb_mh ] , both for halo in narrow mass bins ( left panel ) and above different mass thresholds ( right panel ) . since in our model the uv luminosity is tightly correlated with halo mass , the plot also represents a sequence of profiles as a function of uv luminosity . the amplitude of the stacked profile increases with halo mass . on small scales this increase is largely a reflection of the fact that halos of higher masses host sources of higher luminosity , while on large scales there is an additional contribution from stronger source clustering around more massive halos . on small scales ( inside the inner characteristic radius ) , where the one - halo profile makes a substantial contribution , sources in higher mass halos show a steeper profile . the inner characteristic scale @xmath20 , where the profile starts to level off , increases with halo mass . it implies that sources in higher mass halos have more extended emission . the outer characteristic scale @xmath22 , where the profile starts to steepen again , also increases with halo mass . as discussed above , this scale is an indication of the spatial extent of the emission of the clustered sources . massive halos are typically located in dense environments , therefore the clustered sources around them tend to be in massive halos as well . on scales larger than @xmath22 , the extended tail of the profile for higher mass halos is steeper . this is also a consequence of the higher clustering amplitude , since the surface brightness profile is proportional to ( 1+@xmath36 ) as in equation ( [ eqn:2halo ] ) , and the unity factor tends to flatten the profile for weakly clustered sources . overall , the stacked surface brightness profile is steeper for sources in higher mass halos , and the two characteristic scales increase with halo mass . on small scales where the central cusp dominates , if we approximate the profile by a power law , we find that the power - law index decreases with halo mass , ranging from @xmath58 to @xmath59 ( from @xmath58 to @xmath60 ) for the mass bin ( threshold ) samples shown in figure [ fig : sb_mh ] . the halo mass ( uv luminosity ) sequence does not have a priori selection based on observed emission , so it applies to the case of lbgs . in practice , the brightness profile that can be easily measured in a narrowband survey is the stacked image of the identified laes , classified according to their apparent ( observed ) flux . figure [ fig : sb_lap ] shows the stacked surface brightness profile for laes as a sequence of the apparent ( observed ) luminosity , for both luminosity - bin samples ( left panel ) and luminosity - threshold samples ( right panel ) . the apparent luminosity couples to the circumgalactic and intergalactic environment around sources as a result of the radiative transfer , which imposes a strong selection effect for sources that are identified as laes . on small scales ( inside @xmath20 ) , the stacked profile for laes appears to be steeper than those seen in cases in figure [ fig : sb_mh ] for sources without selection . in figure [ fig : sb_lap ] , the central cuspy profile has a slope of @xmath61 @xmath62 ( @xmath63 @xmath62 ) for luminosity bin ( threshold ) samples , with steeper slopes for higher observed luminosities . the amplitude increases with the apparent luminosity . on scales above the inner characteristic scale , the amplitude also increases with luminosity , but the dependence is much weaker . this weak dependence is a consequence of the weak dependence of the projected clustering on apparent luminosity , as discussed in paper ii . there is also a trend that the two characteristic scales increase with the apparent luminosity , but it is much weaker than that seen in figure [ fig : sb_mh ] for lbgs . in all our discussions above , the intrinsic line profile is assumed to be gaussian with the width determined by velocity dispersion in the halo ( paper i ) . more specifically , the standard deviation of the gaussian profile is set to equal the 1d thermal velocity dispersion for gas in the halo , @xmath64^{1/3 } \kms$ ] . as discussed in paper i and paper ii , the intrinsic line profile is one of the main uncertainties in our radiative transfer modeling . the spatial extent of the emission depends on the initial line profile . for example , in the extreme case where all the photons are shifted far to the red from the line center within the emitting galaxy ( e.g. , as a possible consequence of a strong galactic wind ; @xcite ) , they would not be further scattered in the surrounding medium and the emission would not extend beyond the size of the galaxy . to test the effect of the initial line profile on the spatial distribution of photons , we perform radiative transfer calculations with different initial values of wavelength for an individual source in a halo of @xmath17 ( the same source as in figure [ fig : psf ] ) . figure [ fig : initfreq ] shows the surface brightness and cumulative luminosity profiles for cases with different initial line shifts ( in units of the thermal velocity dispersion @xmath27 mentioned above ) . on average , bluer photons travel farther in the igm before they redshift into the line center to encounter significant scatterings . therefore , the emission would dilute to a larger spatial extent . this is clearly seen in figure [ fig : initfreq ] the spatial distribution becomes more and more extended as the initial wavelength shift varies from @xmath65 to @xmath66 . if we start with red photons , the final surface brightness profile becomes more compact . in figure [ fig : initfreq ] , the intrinsic luminosity for all curves is the same . the surface brightness curves for initially red photons appear to be of lower amplitude , because they have a central spike ( within 0.02@xmath12 ) that is not shown in the left panel . the contribution of the central spike to the total luminosity can be figured out from the cumulative luminosity curves shown in the right panel . overall , the emission appears to be close to point - like if photons start from the far red side of the line center . these red photons can still have a probability of experiencing resonant scatterings in the infall region around the halo and being shifted to the blue side . the surface brightness profile on scales @xmath67 is expected to be contributed by these photons . the probability of such scatterings depends on the magnitude of the shift and the infall velocity around the halo . for sources in @xmath17 halos , even with a @xmath68 initial line shift ( thin dashed curves in figure [ fig : initfreq ] ) , the flattening part in the surface brightness profile can still be clearly seen ( left panel ) , although the majority of the observed photons are from the central region ( right panel ) . we verify that for a @xmath69 initial line shift ( not shown in the plot ) the contribution from the central region to the total luminosity is @xmath7040% , less than that from the extended emission . if we take this as a threshold for the profile to change from point - like to extended , it would correspond to @xmath71 . if the wavelength shift of the initial photons is in the range of @xmath72 , the resultant surface brightness distribution is not sensitive to the initial wavelength , as shown by the solid curves in figure [ fig : initfreq ] . this is easy to understand : photons starting at a wavelength shift @xmath72 undergo many scatterings in the core , leading to a distribution that loses memory of the original wavelength and to a nearly constant surface brightness profile . but photons with a larger initial wavelength deviation are not scattered enough times to be redistributed in this way before they escape . therefore , the predicted profile should be reliable except when many photons are emitted from the central galaxy with a wavelength shift of several times the velocity dispersion . to eliminate most of the extended emission around the galaxies , the initial line shifts ( after photons escape the ism ) need to be larger than @xmath73 for halos of @xmath17 . galactic winds can have the effect of shifting emission toward red ( e.g. , * ? ? ? * ; * ? ? ? * ) . from observations of @xmath74 galaxies , @xcite find that the interstellar absorption line shifts with respect to the h@xmath75-defined galaxy systemic redshift by @xmath76 ( mean @xmath77 standard deviation ) for galaxies of baryon mass ( stars plus cold gas ) ranging from @xmath78 to @xmath79 . the uncertainty of velocity measured for individual galaxies is estimated to be @xmath80 . if the wind is confined in a region not far from the ism , the above observational results indicate that the initial shift in emission could be above @xmath81 . however , even in such a case , the shift may not be sufficient for photons to completely decouple from the circumgalactic and intergalactic environments , depending on the spatial distribution of the wind . observationally , the outflowing neutral gas of the wind is not isotropic , usually collimated in a bipolar fashion ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? it is likely that emission escaping in directions other than the wind would gain only a smaller redward shift , and the source would still become extended in the end . the wind effect on the spatial distribution of emission is also expected to be less significant for sources before reionization than those after reionization , since the higher neutral density boosts the scattering optical depth . it is also possible that the wind could reach scales much larger than the ism ( e.g. , * ? ? ? in such a case , the initial line shifts ( after photons escape the ism ) is not much affected by the wind , but the wind alters the distribution and dynamics of neutral gas in the circumgalactic region and therefore the details in the radiative transfer process . while the surface brightness profile of emission may change accordingly , the extended emission should remain as a generic feature . such a possibility deserves further investigations using high resolution galaxy formation simulations with a wind prescription . a change in the initial line profile can lead to a change in the observed luminosity and therefore the observed luminosity function . however , as discussed in paper i , a combination of changes in the star formation prescription used , the stellar imf adopted , and the dust extinction is able to keep the predicted luminosity function to match the observed one . on the other hand , the change in the initial line profile can alter the strength of the coupling between the observed emission and the circumgalactic and intergalactic environments . this would show up in the clustering of laes ( paper ii ) , which provides an interesting way to constrain the initial line profile and the effect of galactic wind . for the main results in this paper , our radiative transfer calculation is performed in a @xmath82 simulation box , where gas properties ( density , temperature , and peculiar velocity ) are sampled with a 768@xmath13 grid . all photons are originally emitted from point sources and the scatterings of photons inside individual cells are also followed . as shown in paper i , the dynamics of the infall region around halos plays a significant role in determining the distribution of photons , so we expect that the grid resolution in our calculation is able to predict the generic features in the extended emission . we perform a resolution test to show that this is the case . it would be ideal to do such a test with a higher resolution grid . however , the available reionization simulation outputs prevent us from doing this . therefore , we degrade the default 768@xmath13 grid used in this paper to a lower resolution 384@xmath13 grid by binning the gas density , temperature , and peculiar velocity . we then perform the radiative transfer calculation with such a grid for @xmath17 halos and compare the resultant average surface brightness profile ( the one - halo term ) with that with the @xmath4 grid . the comparison is shown in figure [ fig : resolution ] . for each resolution case , the thin and thick arrows mark the half and full size of a grid cell , respectively . for a source located at the center of a cell , the initial photons see a uniform gas distribution within a radius of about half cell size . however , the final surface brightness distribution inside this resolution radius is determined by the gas properties not only in this central cell but also in the surrounding cells . the reason is that photons propagating into surrounding cells have chances to be scattered back into the central cell . that is , the surface brightness profile is not a local quantity as a result of radiative transfer . therefore , the small scale surface brightness profile ( in particular the slope ) is not completely a result of the grid resolution , as indicated by the fact that the powerlaw - like inner profile naturally extends to scales a few times larger than the resolution scale . the transition to a plateau is not an artifact of grid resolution , either . with the low resolution grid , the general features in the surface brightness profile remain the same , a steep drop on small scales and an extended plateau on large scales before a sharp cutoff . however , the amplitude of the profile on small scales is about a factor of two higher with the low resolution grid , and it is compensated by a slightly lower amplitude on large scales . therefore , a lower resolution has the effect of making the emission more concentrated . if the grid were a factor of two finer than the default grid , we expect the predicted small scale surface brightness profile to be lower and the difference should be smaller than a factor of two . the emission would become _ more _ extended . we therefore expect that the generic prediction of the extended emission would be strengthened with an improved grid resolution . even though for the one - halo term of the surface brightness profile the overall cusp - to - plateau behavior does not depend on the grid resolution , the transition scale decreases for a higher resolution . figure [ fig : resolution ] seems to show that the transition happens at a scale about twice the cell size , indicating a resolution effect . although cells around the source cell contribute to the determination of the profile of the central cusp , the cuspy profile itself is inevitably expected to also depend on the exact density , temperature , and velocity distributions of neutral gas within the source cell . the effects discussed in [ sec : line_prof ] add further uncertainties in the cuspy profile . despite the above uncertainties , our prediction on the existence of the extended emission around high redshift star - forming galaxies remains valid , as long as photons encounter scatterings in the circum - galactic and inter - galactic media . the predicted two characteristic scales in the surface brightness profile from stacked images should also remain valid . the outer scale , which is mainly determined by the one - halo plateau region , is less sensitive to the exact central cuspy profile . the inner transition scale , on the other hand , depends on the amplitude and slope of the central cusp , which may not be properly resolved in our model . if the grid resolution is the only source of uncertainty , the inner transition scale of 0.10.2 mpc ( comoving ) seen in the stacked profiles in [ sec : extended ] should be regarded as upper limits . the emission from star - forming galaxies is extended because of resonant scattering in the circumgalactic and intergalactic environments . in the stacked narrowband image , the change from a declining to nearly flat surface brightness profile at a characteristic scale of a few tenths of mpc ( possibly an upper limit , see [ sec : resolution ] ) . and that from the flat profile to a declining profile at a few mpc are evidence for the resonant scattering nature of the extended emission . according to our radiative transfer modeling , the visibility of emission from star - forming galaxies depends on the circumgalactic and intergalactic environments around sources , and only the central , high surface brightness part can be detected as laes . the extended emission around the star - forming galaxies is predicted to exist ( as long as photons escape from the ism ) , no matter whether they can be detected as laes or not . this implies that the extended emission can be detected from both lbgs and laes . in practice , to detect the extended emission around lbgs , we need to know their redshifts and take their narrowband images . extended emission from laes is more readily detected , since laes are typically discovered in narrowband surveys . our radiative transfer modeling is performed for @xmath83 star - forming galaxies . the generic picture should remain valid at other redshifts as long as photons encounter significant scatterings in the circum - galactic / intergalactic media , the emission would become extended . however , the physical conditions in and around star - forming galaxies evolve with redshift . for example , compared to @xmath83 galaxies , we expect a lower neutral hydrogen density around @xmath74 galaxies , as a result of the expansion of the universe and the evolution of the uv background . the hubble expansion rate decreases , a change that may partially compensate the decrease in neutral density for scattering . the extended emission ( e.g. , the surface brightness profile ) at @xmath74 may not follow the details of our @xmath83 prediction , although it is expected to show up . we reserve the investigation of the redshift dependence of the extended emission for future work . we do not include dust in our model . the effect of ism dust on the line depends on whether dust distribution is clumpy or not @xcite . optical , uv , and observations of local star - forming galaxies show evidence that ism kinematics and geometry play a more significant role than dust in affecting the emission ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? we also expect the dust content to decrease towards high redshift ( e.g. , @xmath84 ) , and there is observational evidence for low dust content in @xmath85 galaxies ( e.g. , @xcite ) . it is likely , however , that the existence of dust can lead to variations in the amount of photons escaping from the ism , and thus add additional scatter in the extended emission among individual galaxies . keeping all the above caveats in mind , we discuss the current status of observing the extended emission . observations have long been suggesting the existence of extended emission . based on _ hubble space telescope ( hst ) _ broad - band and _ new technology telescope ( nnt ) _ narrow - band images , @xcite found that three lbg - like galaxies near a @xmath86 quasar are more extended in emission ( half - light radius of 0.50.7 ) than in the continuum ( half - light radius of 0.13 ) . observations with nnt and very large telescope ( vlt ) of a larger sample of laes ( @xmath87 ) in quasar fields at @xmath74 further strengthen the case for extended emission ( e.g. , @xcite ; also see @xcite for comparing one of their objects with the results from a model with radiative transfer ) . in particular , @xcite found the size of emission to be 0.65(_fwhm _ ) in the stacked narrow - band images of seven laes . although the spatial extent of the emission from the above observations is small compared to what we predict , owing to the surface brightness detection limit , it is clear that emission is more extended than the underlying uv emission . there are efforts to use stacked narrowband images in deep surveys to study the compactness of the emission . @xcite find that 24 spectroscopically confirmed @xmath74 lbgs in @xcite fall within the redshift range of their deep narrowband image observed with subaru telescope ( with only two of them being identified as laes ) . the uv luminosity of these lbgs ranges from -20 mag to -22.5 mag ( based on the observed @xmath88 magnitude ; @xcite ) . after removing two gigantic blobs , the composite narrowband ( continuum subtracted ) image of the 22 lbgs shows emission that extends at least to @xmath89 ( about @xmath90 comoving ; see their fig.8 ) . the surface brightness drops from @xmath91 at @xmath92 to a few times @xmath93 at @xmath94 , corresponding to a power - law index of @xmath95 . they further note that 13 lbgs with no individual detection of extended emission also show extended emission on their composite image . the findings of @xcite are in line with our prediction that extended emission exists for lbgs , and the slope of the surface brightness profile is also in broad agreement with our prediction for uv bright sources . @xcite present the result of stacking analysis of deep narrow - band observation of 92 uv continuum - selected star - forming galaxies at @xmath96 . some sources are the same ones as in @xcite but with much deeper narrow - band observation . the stacked narrow - band image shows diffuse emission around star - forming galaxies , with surface brightness drops from a few @xmath97 at the center to @xmath98 at a radius of @xmath99 ( @xmath080 proper kpc ) . the extended emission exists in the stacked image of sub - samples lbgs , laes , and absorbers all show qualitatively similar diffuse emission . the overall surface brightness amplitude of the extended emission increases with the central emission . although the redshift difference prevents us from a direct comparison between the observation and our prediction , the detected diffuse emission supports our generic picture that resonant scatterings in the circum - galactic and intergalactic media causes emission originated from star formation in the central galaxy to become extended . if we make a naive comparison , the emission is detected up to the inner characteristic scale in the surface brightness profile ( e.g. , figures [ fig : sb_mh ] and [ fig : sb_lap ] ) , not reaching the plateau yet . in @xcite , the clustered sources are intentionally masked before image stacking , which may change the shape of plateau if it could be detected . we plan to extend our model to @xmath74 star - forming galaxies and make direct comparison with the observation in @xcite . @xcite use stacked multiband images of laes at @xmath65.7 and 6.6 in the sxds to investigate the stellar population of laes . for the narrowband stacked images , they find that the fluxes level off for apertures larger than @xmath89 in diameter . such a behavior in the surface brightness profile is predicted in our model as the transition to the two - halo term dominated regime . it is likely that the feature seen by @xcite is caused by the clustering of emitting sources . @xcite detect emission from three spectroscopically confirmed @xmath100 laes in the hst / acs f658n narrowband imaging data . they find that emission appears to be more extended than the uv emission . however , the half - light radii of emission are quite small , in the range of [email protected]@xmath101 . one thing to notice is that the observation is not deep enough to reach the sensitivity of detecting the much larger extended halo predicted by our model . this is clear from the fact that the surface brightness level in the _ hst _ narrowband observation turns out to be above a few times @xmath102 . it is also supported by the fact that the three objects have significantly higher fluxes detected from ground - based narrowband image ( with larger telescope apertures and reduced sensitivity to read noise ; @xcite ) . with _ narrowband images of eight @xmath103 laes , @xcite find half - light radii of emission to be in a similar range , [email protected]@xmath101 . the surface brightness level ( a few times @xmath104 ) reached by the observation is still too high to probe the extended emission predicted by our model . extended emission can also be detected through deep spectroscopic observations . @xcite conduct a long - slit search for @xmath105 low surface brightness emission from a 92-hour long exposure with the vlt , reaching a @xmath106 surface brightness detection limit of @xmath107 in a 1 arcsec@xmath24 aperture . they find 27 objects with possible emission lines . these objects are typically fainter than the laes and lbgs in @xcite . about one third of them show clear signatures of extended emission . if the surface brightness profile at large radii is approximated by a power - law , the slope is generally in the range of @xmath108 to @xmath109 ( their figure 17 ) . in the stacked surface brightness profile ( their figure 20 ) , the emission clearly extends to @xmath110 ( about comoving 90@xmath14 ) , dropping from @xmath111 at a radius of 1to @xmath112 at large radii . the profile roughly follows a power law with an index in the range of @xmath113 to @xmath114 . this is in the ballpark of the predicted extended emission , although a direct comparison is not possible because of the redshift difference . the spatially resolved spectra of the extended emission remain as an interesting aspect to be explored in our model . the extended halo from scatterings predicted by our model shares some similarities with the one around a quasar before reionization described in @xcite . unlike @xcite , who assume a uniform , zero temperature igm undergoing hubble expansion , we use a realistic distribution of gas density , temperature , and velocity around a star - forming halo from cosmological reionization simulation . we expect that extended emission around star - forming galaxies also exists before reionization , which merits further study . however , the surface brightness is expected to be proportional to the source luminosity and to not depend on the neutral fraction in the igm as long as the majority of photons are scattered , while it should decline with redshift as @xmath115 , making the detection increasingly difficult at higher redshift . our radiative transfer calculation is done for sources residing in halos above @xmath5 . star formation in lower mass halos , down to @xmath0@xmath116 at @xmath7 ( with virial temperature @xmath117@xmath118 for gas to cool ; see @xcite ) , can also contribute to the two - halo term of the stacked profile . according to @xcite ( see their fig.1 ) , the total star formation rate in @xmath119@xmath5 halos can contribute about two thirds of the global rate . the amplitude of the clustering with these low mass halos is lower , and we expect that including the emission from the @xmath119@xmath5 halos would boost the two - halo term of the surface brightness profile presented in this paper at the factor of two level and that the two characteristic scales would shift inward . our model does not resolve subhalos , but we do not expect a large subhalo ( or satellite galaxy ) fraction for galaxies at redshift @xmath84 . if we consider a mass threshold sample of halos with @xmath120 and assume that the mean number of subhalos increase linearly with halo mass as @xmath121 @xcite , we find that the satellite fraction is about 15% for @xmath122 . as subhalos are distributed inside virial radius of the parent halo , including star formation in subhalos would make the central cuspy surface brightness profile in the stacked image more extended , which would smooth the transition to the plateau part of the profile and increase the inner characteristic scale . investigating the exact magnitude of the effect is beyond the scope of this paper . besides star formation , there are other sources of emission that increase in the dense regions around galaxies , including fluorescence from the uv background and cooling radiation . at @xmath7 , the fluorescent emission caused by the uv background is expected to be at the level of a few times @xmath123 ( e.g. , * ? ? ? * ; * ? ? ? * ) , much fainter than the signal of interest in this paper . cooling radiation may have a noticeable contribution to the stacked surface brightness profile , but there are uncertainties in the theoretical prediction ( e.g. , * ? ? ? * ) . further work is needed to see whether emission from cooling radiation can change the profile and to see how one can separate emission caused by star formation from other sources of emission . finally , for observational efforts , sky subtraction is an important step in revealing the extended emission from star - forming galaxies . to compare the observation with the theoretical prediction from simulations , a detailed description of the method of sky subtraction used in the observation is necessary , which should be reproduced in the same way in the model . after escaping from the ism , photons converted from ionizing photons in star - forming galaxies experience scatterings in the circumgalactic and intergalactic media . such a radiative transfer process makes emission from these galaxies spatially extended . in this paper , built on the radiative transfer modeling of laes in @xcite and @xcite , we investigate the predicted spatial distribution of emission that can be measured from the stacked narrowband image of these galaxies . in general , the predicted surface brightness profile measured from the stacked image has two characteristic scales , an inner one at tenths of mpc and an outer one at about 1 mpc . the profile shows a central cusp inside the inner scale , an approximate plateau between the two scales , and an extended tail beyond the outer scale . the inner scale may possibly be an upper limit , as an effect of the grid resolution in the simulation ( see [ sec : resolution ] ) . the stacked surface brightness profile can be understood as a superposition of the brightness distribution from the stacked sources themselves ( one - halo term ) and that from neighboring clustered sources ( two - halo term ) . the two - halo term is the profile of the ( angular ) two - point correlation function ( plus one ) smoothed by the extended emission profile ( psf ) of individual sources . the smoothing makes the profile flattened on scales smaller than the spatial extent of emission of clustered sources . the outer characteristic scale in the stacked surface brightness profile marks this spatial extent , and the plateau in the profile is a consequence of the smoothing effect . the transition from one - halo term domination to two - halo term domination leads to the inner characteristic scale seen in the stacked surface brightness profile . for continuum selected galaxies ( lbgs ) , the amplitude of the stacked surface brightness profile increases with the source uv luminosity ( or halo mass ) on both small and large scales . the two characteristic scales also increase with the uv luminosity . the central cusp becomes steeper for higher uv luminosity . for line selected galaxies ( laes ) , the amplitude of the central cuspy profile increases with observed luminosity and the slope is steeper than that of lbgs . beyond the inner characteristic scale , the amplitude of the surface brightness profile only has a weak dependence on observed luminosity . the inner and outer characteristic scales do not show strong dependence on luminosity , either . because of the contribution from source clustering to the stacked surface brightness profile , the cumulative luminosity from the stacked image does not converge to the intrinsic luminosity of the stacked sources . it is therefore not straightforward to estimate the total emission from the stacked image . in detail , the surface brightness profile depends on the initial line profile of emission ( after photons escape the ism ) . the source would appear to be more compact in the narrowband image as the initial line shift toward red increases . if this shift is caused by galactic wind , the wind velocity needs to be comparable to a few times the virial velocity of the host halo and the wind needs to be largely isotropic to make emission point - like . conversely , from the measured surface brightness profile in the narrowband image , it may be possible to constrain the effect of galactic winds . it is worth further study along this line . our particular prediction here is for sources after reionization is complete ( @xmath7 ) . it is interesting to investigate how the prediction changes for star - forming galaxies at lower redshifts , because of the large amount of efforts in observing star - forming galaxies around redshift 24 ( e.g. , @xcite ) and in the local universe ( e.g. , @xcite ) . it is also necessary to study the extended emission for sources at higher redshifts , before reionization is complete , given the increasing observational efforts ( e.g. , @xcite ) . at the time of our initial submission , we conclude that deep narrowband photometry from large ground - based telescopes is on the verge of detecting the extended emission around star - forming galaxies , including lbgs and laes ( e.g. , * ? ? ? * ; * ? ? ? now the latest observation by @xcite indeed revealed the extended emission to large radii ( @xmath99 ) around lbgs and laes , starting to verify our prediction and provide stringent test to the theoretical model . the detection of the predicted emission supports the generic picture that radiative transfer in the circumgalactic and intergalactic environments produces extended emission . the extended emission opens a new window to study the circumgalactic and intergalactic environment of high - redshift star - forming galaxies . the surface brightness profile encodes information about cold baryons around galaxies , including their density , temperature , and velocity . all of these properties could be modified by galactic wind , so the extended emission in principle can put constraints on the galactic wind . the stacked image also includes contributions from faint galaxies that can not be detected in a single exposure , which provides an opportunity to study low luminosity star - forming galaxies . the extended emission gives us a better idea on the total amount of emission from galaxies . when compared with the uv emission or other optically - thin line emission ( e.g. , h@xmath75 emission ) , we may infer the dust distribution and its effect on photons . with integral - field - units observations of high - redshift star - forming galaxies , we would have stacked spectra for the extended emission and expect to learn more about galaxy environments . details on how to extract all the information encoded in the extended emission need to be investigated , and radiative transfer calculation , in combination with sophisticated models of star - forming galaxies and their environments , will play an irreplaceable role . zz thanks masami ouchi , tomoki saito , and chuck steidel for helpful discussions and the hospitality of the institute for the physics and mathematics of the universe . we thank andy gould for useful comments and the referee for constructive suggestions . zz gratefully acknowledges support from yale center for astronomy and astrophysics through a ycaa fellowship . zz and dhw thank the institute for advanced study for their hospitality during part of the work . rc acknowledges support from nasa grants nng06gi09 g and nnx08ah31 g . jm is supported by the spanish grant aya2009 - 09745 . atek , h. , kunth , d. , hayes , m. , stlin , g. , & mas - hesse , j. m. 2008 , , 488 , 491 atek , h. , schaerer , d. , & kunth , d. 2009 , , 502 , 791 berlind , a. a. & weinberg , d. h. 2002 , , 575 , 587 bland , j. , & tully , b. 1988 , , 334 , 43 bond , n. a. , feldmeier , j. j. , matkovi , a. , gronwall , c. , ciardullo , r. , & gawiser , e. 2010 , , 716 , l200 bouwens , r. j. , et al . 2010 , , 708 , l69 , cooray , a. , & sheth , r. 2002 , , 372 , 1 cowie , l. l. , & hu , e. m. 1998 , , 115 , 1319 dijkstra , m. , lidz , a. , & wyithe , j. s. b. 2007 , , 377 , 1175 dijkstra , m. , & wyithe , j. s. b. 2010 , , 408 , 352 faucher - gigure , c .- a . , kere , d. , dijkstra , m. , hernquist , l. , & zaldarriaga , m. 2010 , , 725 , 633 finkelstein , s. l. , et al . 2011 , , 735 , 5 fynbo , j. u. , mller , p. , & thomsen , b. 2001 , , 374 , 443 fynbo , j. p. u. , ledoux , c. , mller , p. , thomsen , b. , & burud , i. 2003 , , 407 , 147 gawiser , e. , et al . 2007 , , 671 , 278 giavalisco , m. , koratkar , a. , & calzetti , d. 1996 , , 466 , 831 gould , a. , & weinberg , d. h. 1996 , , 468 , 462 guaita , l. , et al . 2010 , , 714 , 255 hansen , m. , & oh , s. p. 2006 , , 367 , 979 hayashino , t. , et al . 2004 , , 128 , 2073 hayes , m. , stlin , g. , atek , h. , kunth , d. , mas - hesse , j. m. , leitherer , c. , jimnez - bailn , e. , & adamo , a. 2007 , , 382 , 1465 hu , e. , & mcmahon , r. g. 1996 , , 382 , 281 kashikawa , n. , et al . 2011 , , 734 , 119 keel , w. c. 2005 , , 129 , 1863 kollmeier , j. a. , zheng , z. , dav , r. , gould , a. , katz , n. , miralda - escud , j. , & weinberg , d. h. 2010 , , 708 , 1048 kravtsov , a. v. , berlind , a. a. , wechsler , r. h. , klypin , a. a. , gottlber , s. , allgood , b. , & primack , j. r. 2004 , , 609 , 35 kunth , d. , leitherer , c. , mas - hesse , j. m. , stlin , g. , & petrosian , a. 2003 , , 597 , 263 laursen , p. , & sommer - larsen , j. 2007 , , 657 , l69 loeb , a. , & rybicki , g. b. 1999 , , 524 , 527 madau , p. , pozzetti , l. , & dickinson , m. 1998 , , 498 , 106 malhotra , s. , & rhoads , j. 2004 , , 617 , l5 moller , p. , & warren , s. j. 1998 , , 299 , 661 neufeld , d. a. 1991 , , 370 , l85 ono , y. , ouchi , m. , shimasaku , k. , dunlop , j. , farrah , d. , mclure , r. , & okamura , s. 2010 , , 724 , 1524 stlin , g. , hayes , m. , kunth , d. , mas - hesse , j. m. , leitherer , c. , petrosian , a. , & atek , h. 2009 , , 138 , 923 ouchi , m. , et al . 2008 , , 176 , 301 ouchi , m. , et al . 2010 , , 723 , 869 partridge , r.b . , & peebles , p.j.e . 1967 , , 147 , 868 pieri , m. m. , frank , s. , weinberg , d. h. , mathur , s. , & york , d. g. 2010 , , 724 , l69 rauch , m. , et al . 2008 , , 681 , 856 rhoads , j. e. , et al . 2003 , , 125 , 1006 seljak , u. 2000 , , 318 , 203 shopbell , p. l. , & bland - hawthorn , j. 1998 , , 493 , 129 steidel , c. c. , adelberger , k. l. , shapley , a. e. , pettini , m. , dickinson , m. , & giavalisco , m. 2003 , , 592 , 728 steidel , c. c. , erb , d. k. , shapley , a. e. , pettini , m. , reddy , n. , bogosavljevi , m. , rudie , g. c. , & rakic , o. 2010 , , 717 , 289 steidel , c. c. , bogosavljevi , m. , shapley , a. e. , kollmeier , j. a. , reddy , n. a. , erb , d. k. , & pettini , m. 2011 , in press , arxiv:1101.2204 taniguchi , y. , et al . 2009 , , 701 , 915 trac , h. , & cen , r. 2007 , , 671 , 1 trac , h. , cen , r. , & loeb , a. 2008 , , 689 , l81 veilleux , s. , & rupke , d. s. 2002 , , 565 , l63 verhamme , a. , schaerer , d. , & maselli , a. 2006 , , 460 , 397 wang , j .- x . , malhotra , s. , rhoads , j. e. , zhang , h .- t . , & finkelstein , s. l. 2009 , , 706 , 762 zheng , z. , cen , r. , trac , h. , & miralda - escude , j. 2010 , , 716 , 574 zheng , z. , cen , r. , trac , h. , & miralda - escude , j. 2011 , , 726 , 38

in a resonance model the reactions and are studied . for the reactions and , the resonances , and are included as intermediate states . for the reactions , the resonances , and are considered , while for the reactions the intermediate resonances are and . besides these resonances in the-channel , the-channel-exchanges are also taken into account as a smooth background . the relevant coupling constants for the meson - baryon vertices are obtained ( except for ) from the experimental decay branching ratios of the relevant resonances . all isospin channels of the and cross sections are calculated . by comparing the calculated results with the available experimental data , we find that the total cross sections of the reactions can be explained by the resonance model . the cross sections , for which no experimental data are available , are predicted theoretically . parametrizations of the calculated total cross sections for all different isospin channels are given for the use of kaon productions in heavy ion collisions . the differential cross sections are also studied . = 2.5em

nucl-th9412011

the purpose of this paper is to study the @xmath15 ( @xmath16 stands for @xmath16(1232 ) , @xmath17 stands for either the @xmath18 or the @xmath19 ) reactions for kaon productions in heavy ion collisions . one of the main goals of intermediate energy heavy ion physics is to determine the equation of state ( eos ) of nuclear matter . theoretical studies show that the kaons produced in heavy ion collisions are sensitive to the eos @xcite@xcite . to calculate the kaon production in heavy ion collisions one needs two ingredients @xcite@xcite : \1 . a transport theory to describe the evolution of the heavy ion colliding systems . the elementary cross sections of kaon production in the @xmath20 and @xmath21 collisions ( here @xmath22 stands for either a nucleon @xmath23 or a @xmath24 ) . therefore , in order to obtain a definite conclusion about the eos one needs the reliable elementary cross sections for the kaon production as well as the dynamical transport models . one usually uses transport models such as buu / vuu @xcite and qmd @xcite to simulate the dynamical evolution of projectile and target systems . in these models collision terms and the nuclear mean field are included . by including the nuclear mean field one hopes to determine the eos . in the collision terms , the following reactions are included as main processes : @xmath25 @xmath26 @xmath27 @xmath28 @xmath29 @xmath30 then , kaons can be produced mainly through the following reactions : @xmath31 and @xmath32 the lorentz - invariant differential multiplicity for kaons for a given impact parameter @xmath33 in heavy ion collisions is connected with the elementary cross section by @xmath34 \frac{d\omega } { 4\pi } \nonumber \\ & & + \sum _ { \pi b_{coll } } e '' \frac{d^3\sigma _ { \pi b\rightarrow y k } ( \sqrt { s_{\pi b}})}{d^3 p''}/ \sigma _ { \pi b}^{tot}(\sqrt { s_{\pi b } } ) , \label{basic1}\end{aligned}\ ] ] where the primed and the double - primed quantities are in the center - of - momentum ( c.m . ) frames of the two colliding baryons ( @xmath20 ) and pion - baryon ( @xmath21 ) , respectively . the unprimed quantities are in the c.m . frame of the two nuclei . @xmath35 and @xmath36 are the total cross sections for two baryons @xmath37 and the pion - nucleon system ( @xmath38 ) as functions of c.m . energy @xmath39 and @xmath40 , respectively . the factor @xmath41 $ ] stands for the pauli blocking correction for the final nucleon @xmath23 in the @xmath42 reaction due to the surrounding nuclear medium . the lorentz - invariant differential kaon - production cross section in heavy ion collisions is then given by : @xmath43 from eq.([basic1 ] ) we see that the elementary cross sections are directly related to the kaon yields in heavy ion collisions . thus , it is important to have good elementary kaon - production cross sections . however , most elementary cross sections needed in studying heavy ion collisions are not well known experimentally . therefore one has to use the available experimental data as far as possible . in the case when no experimental data are available one has to rely on symmetry considerations and theoretical models to extrapolate the available experimental data . the elementary kaon - production cross section in @xmath20 collisions used in eq.([basic1 ] ) are usually taken from randrup and ko@xcite . in their works , the @xmath44and @xmath45 total cross sections are obtained from @xmath46 cross section by symmetry consideration in isospin space , and the @xmath16 is treated as spin-@xmath47 ( not @xmath48 ) and isospin-@xmath48 particle . moreover , the coupling constants relevant to the @xmath16 are assumed to be the same as for the nucleon @xmath23 . the elementary @xmath49-production cross section in @xmath21 collisions needed in eq.([basic1 ] ) are usually taken from cugnon and lombard s parametrization@xcite . in this parametrization the isospin averaged cross sections are obtained from only three available experimental data assuming proton and neutron @xmath50 symmetry . this symmetry is , in principle , valid for the light nuclei , and not appropriate for collisions between heavy nuclei with a large neutron excess , i.e. @xmath51 . furthermore the parametrization for the reaction @xmath1 is not given , hence the contribution of this channel to kaon production in the heavy ion collision is unclear . by using the elementary cross section of randrup and ko , one finds that @xmath52 gives the main contribution to the total kaon yield in heavy ion collisions@xcite . however , the elementary cross section for this channel is experimentally unknown . thus a more sophisticated study of this channel is still necessary . in order to determine the cross section for the reaction @xmath52 , one needs at first to study the reaction @xmath53 according to randrup and ko s model@xcite@xcite . since in heavy ion collisions many pions are produced , and the threshold of @xmath54 is lower than in @xmath42 , the reaction @xmath55 seems to be important as a secondary process besides the @xmath42 channel for kaon production . according to a study of the giessen group , the @xmath56 process is dominant for kaon production in proton - nucleus reactions@xcite . recently , in the simulation codes of buu and qmd the isospin dependence has been considered . one distinguishes the members of the nucleon isospin - doublet(@xmath57 and @xmath58 ) , of the pion isospin - triplet(@xmath59 , @xmath60 and @xmath61 ) and of the delta isospin - quartet ( @xmath62 , @xmath63 , @xmath64 , and @xmath65)@xcite@xcite . this enables us to calculate the kaon production in an isospin dependent way , hence one needs information on all the possible isospin channels for kaon production . this means that the isospin - averaged elementary cross sections of @xmath66 and @xmath67 can not be used for this purpose . for an isospin - dependent description of kaon production , one needs all the isospin channels . however , the cross sections for some isospin channels are not available in experiment . therefore one needs a theoretical model to evaluate all the unknown isospin channels . on the other hand , in order to study medium effects for kaon production , a dynamical model for the elementary cross sections is needed , since one can not simply obtain the in - medium cross sections from a parametrization of experimental data of the corresponding cross sections in free space . as a first step , we concentrate on the @xmath68 reactions . once we know this cross sections we can calculate the second term in eq.([basic1 ] ) . this channel is also the basis to study @xmath69 , since the parameters of the @xmath68 reactions are ingredients for calculating the reaction @xmath69 in the present model . in this paper , we present results for all @xmath15 reactions using a resonance model . part of this work has been briefly reported in @xcite . by a resonance model we mean that the @xmath70 reactions happen via intermediate resonance states , which decay both into @xmath71 and @xmath72 . the coupling constants in the meson - baryon - resonances can be determined from the relevant branching ratios of the resonances . the cross sections of @xmath70 are obtained coherently from the square of the sum of all resonance amplitudes . besides these @xmath11-channel resonances which give peaks in the total cross sections , the @xmath12-channel @xmath13-exchange is also included . it provides a smooth background . we study the reactions @xmath73 for intermediate energies , i.e. for the range of the @xmath74 invariant mass from the threshold for @xmath75 production to about 3 gev . according to the `` review of particle properties''@xcite@xcite , the relevant resonances to be included as intermediate states are @xmath4 , @xmath8 , @xmath6 and @xmath9 . + in this section we list the properties of the resonances and of @xmath13 relevant to our calculations . + ( 1 ) @xmath76 : + mass @xmath77 mev full width @xmath78=145 to 190 mev @xmath79 150 mev partial decay modes and branching fractions : @xmath80 \(2 ) @xmath81 : + mass @xmath82 mev full width @xmath78=50 to 250 mev @xmath79 100 mev partial decay modes and branching fractions : @xmath83 \(3 ) @xmath84 : + mass @xmath85 mev full width @xmath78=100 to 200 mev @xmath79 150 mev partial decay modes and branching fractions : @xmath86 \(4 ) @xmath87 : + mass @xmath88 mev full width @xmath78=150 to 300 mev @xmath79 200 mev partial decay modes and branching fractions : @xmath89 \(5 ) @xmath90 : + mass @xmath91 mev full width @xmath92=49.8 mev partial decay modes and branching fractions : @xmath93 in the present study , the mesons , baryons and baryon - resonances are treated as fundamental fields . due to the fact that the relevant hadrons have finite sizes , form factors are included at every vertex . in section 3.1 to 3.4 we give the effective interaction lagrangians , and the formulas for the transition amplitudes and cross sections . throughout this work su(2 ) symmetry in isospin space is assumed . in detail we treat @xmath56 , @xmath94 , @xmath95 , @xmath96 in the following subsections @xmath973.1 , @xmath973.2 , @xmath973.3 and @xmath973.4 separately . from the experimental data listed in @xmath97 2 , three resonances @xmath99 , @xmath8 and @xmath6 contribute to the @xmath2 reactions . in addition @xmath13-exchange can also be responsible for the same reaction . the relevant feynman diagrams are shown in fig . 1 : the effective interaction lagrangians needed in describing the vertices are given by : @xmath100 @xmath101 @xmath102 @xmath103 @xmath104 @xmath105 @xmath106 @xmath107 + in the above lagrangians , @xmath23 , @xmath108 , @xmath109 , @xmath19 stand for spin-@xmath110 dirac spinors to describe the corresponding particles . spin-@xmath48 rarita - schwinger fields @xmath111 with mass @xmath112 satisfy the equations@xcite : @xmath113 and @xmath114 stands for the pion field . in isospin space nucleon fields are expressed by @xmath115 , where the superscript `` @xmath116 '' means the transposition operation , similar expressions can also be written for the nucleon resonances . the meson field operators appearing in eqs . ( [ pinn1650])-([kskpi ] ) are related to the physical representations as follows , @xmath117 @xmath118 @xmath119 @xmath120 @xmath121 for the propagators @xmath122 of the spin-@xmath110 and the propagators @xmath123 of the spin-@xmath48 resonances@xcite we use @xmath124 @xmath125 with @xmath126 , \label{pmunu}\ ] ] respectively , where @xmath112 and @xmath127 stand for the mass and the full decay width of the corresponding baryon resonance . according to the lagrangians given by eqs . ( [ pinn1650])-([kskpi ] ) , the amplitudes for the @xmath2 reactions are given by @xmath128 @xmath129 where the resonance amplitudes @xmath130 ( @xmath108 intermediate state ) , @xmath131 ( @xmath109 intermediate state ) , @xmath132 ( @xmath133 intermediate state ) and the @xmath13-exchange amplitude @xmath134 are given by @xmath135 @xmath136 @xmath137 @xmath138 \hspace{2cm}\ ] ] @xmath139 here @xmath140 and @xmath141 in eqs . ( [ a1])-([d1 ] ) are the spinors of the nucleon @xmath23 and the lambda @xmath19 , with four - momentum @xmath142 , spin @xmath143 and four - momentum @xmath144 , @xmath145 , respectively . @xmath57 is the four - momentum of the resonance . @xmath146 , @xmath147 are the four - momenta of the pion and the kaon , respectively . we obtain the absolute values of the effective coupling constants by equating the experimental branching ratios given in @xmath97 2 with the theoretical calculations given in appendix @xmath97 5.1 . form factors ( denoted by @xmath148 and @xmath149 below ) are introduced . they reflect the finite size of the hadrons . these form factors must be multiplied to each vertex of the interactions . thus , the coupling constants are obtained from the branching ratios in the rest frame of the resonances . for the form factors in baryon - baryon - meson vertices we use @xmath150 with @xmath151 being the magnitude of three - momentum @xmath152 of the mesons , and @xmath153 being the cut off parameter . for the @xmath13-@xmath154-@xmath155 vertex we adopt the form factor of ref . @xmath156 in evaluating the cross sections in the c.m . frame of the @xmath157 system , each coupling constant @xmath158 , @xmath159 and @xmath160 appearing in eqs . ( [ a1])-([d1 ] ) must be replaced by @xmath161 , @xmath162 and @xmath163 . here @xmath164 stands for the nucleon resonances . @xmath22 indicates a baryon ( @xmath165 or @xmath19 ) and @xmath166 stands for pseudoscalar - meson ( @xmath155 or @xmath154 ) , respectively . the ( spin unpolarized ) differential cross section for @xmath167 in the c.m . frame of @xmath60 @xmath57 is given by @xmath168 where @xmath169 stands in the c.m . frame for the outgoing three - momenta , and @xmath170 for the incoming three - momenta of the two particles . we have , in the c.m . frame of @xmath157 system , @xmath171 and @xmath172 here @xmath173 is the mandelstam variable , and @xmath174 is defined by@xcite @xmath175 the total cross section for the reaction @xmath167 reads @xmath176 similar expressions for other isospin channels for @xmath49 and @xmath177-production can be obtained in the same way as in eqs . ( [ nlamamp1 ] ) and ( [ nlamamp2 ] ) . + besides @xmath179-exchange , there are three resonances contributing to the @xmath7 reactions ( see @xmath972 ) . they are @xmath8 , @xmath6 and @xmath9 . the feynman diagrams are shown in fig . in addition to eqs.([pinn1710])-([pinn1720 ] ) and eq.([kskpi ] ) , the interaction lagrangians needed in describing the vertices relevant for the reactions @xmath7 are given by : @xmath180 @xmath181 @xmath182 @xmath183 @xmath184 in the eqs . above @xmath18 stands for the spin-@xmath110 dirac spinor of the @xmath18 particle . @xmath185 stands for the spin-@xmath48 rarita - schwinger field of @xmath14 , which satisfies the equations ( [ r - s1])-([r - s3 ] ) . @xmath186 is the ratio of the tensor coupling constant to the vector coupling constant . @xmath187 is the isospin transition operator defined by @xmath188 with @xmath189 and @xmath190 are the pauli matrices . @xmath191@xmath192 . the @xmath18 field operators appearing in eqs . ( [ ksin1710])-([kssin ] ) are related to the physical representations by @xmath193 the required amplitudes for the cross section are given by @xmath194 @xmath195 @xmath196 @xmath197 @xmath198 where the single resonance amplitudes @xmath199 ( with the @xmath109 resonance ) , @xmath200 ( @xmath133 ) , @xmath201 ( @xmath202 ) , and the @xmath13-exchange amplitude @xmath203 in the above equations are given by @xmath204 @xmath205 @xmath206 @xmath207 \hspace{3cm}\ ] ] @xmath208 form factors are inserted at each vertex in eqs . ( [ a2])-([d2 ] ) in the same way as in @xmath97 3.1 . the differential cross section for @xmath209 in the c.m . frame of @xmath210 is given by @xmath211 where @xmath169 stands for the outgoing three - momenta of the @xmath212 and the @xmath49 in the c.m . system , and @xmath170 stands for the incoming three - momenta of the @xmath59 and the @xmath57 . we have , in the c.m . frame of @xmath157 system , @xmath213 and @xmath172 the total cross section for @xmath209 reads @xmath214 the same expressions as eqs . ( [ cross2 ] ) and ( [ total2 ] ) for other isospin dependent reactions can be obtained by eqs . ( [ nsiamp2])-([nsiamp5 ] ) . from eqs . ( [ nsiamp3 ] ) and eq.([nsiamp4 ] ) , we derive @xmath215 the @xmath177 and the @xmath49 production have the same amplitudes given by eqs . ( [ nsiamp1])-([nsiamp5 ] ) , for example , from eq . ( [ nsiamp1 ] ) , one finds : @xmath216 + from the experimental data listed in @xmath97 2 , three resonances @xmath99 , @xmath8 , @xmath6 contribute to the @xmath218 reactions . due to isospin conservation @xmath13-exchange can not contribute to the @xmath219 reactions . the relevant feynman diagrams are shown in fig . 3 : besides eqs.([klan1650])-([klan1720 ] ) , the following three interaction lagrangian are needed in describing the vertices of the @xmath3 reactions : @xmath220 @xmath221 @xmath222 the amplitudes for the reactions are given by @xmath223 @xmath224 @xmath225 + here the amplitudes @xmath226 ( with @xmath108 as the intermediate state ) , @xmath227(@xmath109 ) and @xmath228(@xmath133 ) corresponding to each of the diagrams ( a ) , ( b ) and ( c ) given in fig . 3 are given by(the coefficients in eqs . ( [ dlamamp1])-([dlamamp3 ] ) , are from isospin ) : @xmath229 @xmath230 @xmath231 the rarita - schwinger spinor - vector @xmath232 of the delta s with momentum @xmath233 and spin @xmath234 , @xmath232 satisfies in momentum space the eqs . ( [ r - s1])-([r - s3 ] ) . @xmath235 form factors are inserted at each vertex in eqs . ( [ a3])-([c3 ] ) in the same way as in @xmath97 3.1 and @xmath97 3.2 . the differential cross section for @xmath236 in the c.m . frame of @xmath237 is given by @xmath238 where @xmath239 and @xmath240 with @xmath241 the total cross section for @xmath236 reads @xmath242 the relationships between other isospin channels and the above channel can be easily obtained from eqs.([dlamamp1])-([dlamamp3 ] ) . + according to the data listed in @xmath97 2 , two resonances contribute to the @xmath96 reactions : @xmath8 and @xmath6 . further the t - channel @xmath13-exchange is also relevant for these reactions . the feynman diagrams are shown in fig . 4 . in addition to lagrangians given by eqs . ( [ ksin1710])-([ksin1720 ] ) , ( [ pidn1710])-([pidn1720 ] ) and eq . ( [ kskpi ] ) , one more interaction lagrangian needed in describing the @xmath10 reactions is given by @xmath244 the amplitudes for the @xmath10 reactions are given by : @xmath245 @xmath246 @xmath247 @xmath248 @xmath249 @xmath250 @xmath251 @xmath252 the amplitudes @xmath253 ( with @xmath109 as an intermediate resonance ) , @xmath254(@xmath133 ) and @xmath255 ( @xmath13-exchange ) corresponding to diagrams ( a ) , ( b ) and ( c ) in fig . 4 are given as follows(isospin factors are included by the coefficients in eqs . ( [ dsiamp1])-([dsiamp8 ] ) ) : @xmath256 @xmath257 @xmath258 note that the @xmath108 resonance gives contributions to the @xmath3 reactions but does not give contributions to the @xmath10 reactions . form factors are inserted at each vertex in eqs . ( [ a4])-([c4 ] ) in the same way as in @xmath97 3.1 , @xmath97 3.2 and @xmath97 3.3 . the differential cross section for the reaction @xmath259 in the c.m . frame of @xmath260 reads @xmath261 and @xmath262 other isospin channels and @xmath177 productions can be obtained by eqs.([dsiamp1])-([dsiamp8 ] ) . in this section we give the results and discuss the cross sections of the reactions given in section 3 . we treat in detail the reactions @xmath56 , @xmath94 , @xmath95 , @xmath263 in the subsections @xmath974.1 , @xmath974.2 , @xmath974.3 and @xmath974.4 , respectively . in each subsection we first discuss the total and then the differential cross sections . the total cross section for @xmath264 is given by eq . ( [ total1 ] ) . since all the resonances contributing to the @xmath98 reactions have isospin @xmath110 , different isospin channels of the @xmath98 reactions differ only by an overall factor . the cross section for @xmath265 is two times the one of @xmath266 according to eqs . ( [ nlamamp1 ] ) and ( [ nlamamp2 ] ) . the amplitudes appearing in eqs . ( [ nlamamp1 ] ) and ( [ nlamamp2 ] ) are given by eqs . ( [ a1])-([d1 ] ) . in evaluation of these amplitudes we need eight coupling constants ( i.e. @xmath267 , @xmath268 , @xmath269 , @xmath270 , @xmath271 , @xmath272 , @xmath273 , @xmath274 ) and cutoff parameters . from the eight coupling constants , the absolute values of seven can be determined by equating eqs . ( [ ratio511])-([ratio517 ] ) given in appendix 5.1 with the experimental branching ratios listed in @xmath972 . one coupling constant , i.e. @xmath159 , is treated as a free parameter in the present work . as mentioned in @xmath97 3 , we include form factors in the decay branching ratios as well as in the scattering amplitudes throughout this work ( see also eqs . ( [ form1 ] ) and ( [ form2 ] ) ) . by choosing the cutoff values as : @xmath275 gev for @xmath108 , @xmath109 and @xmath133 , we obtain the coupling constants in table 1 . for the @xmath276 vertex we use @xmath277 , @xmath278 gev . the cut off parameters @xmath279 and @xmath280 in @xmath149 are @xmath281 fm and @xmath282 @xmath283 used in ref . @xcite , and the resulting coupling constant @xmath284 is also shown in table 1 . according to eqs.([nlamamp1 ] ) , ( [ nlamamp2 ] ) and ( [ cross1 ] ) we sum the amplitudes of the different resonances , and obtain the cross sections by squaring the sum of the amplitudes . thus we include all interference terms.(the problem with the signs of the coupling constants is discussed below . ) first , we show the total cross section for the reaction @xmath265 without interference terms in fig . the experimental data are from ref@xcite . since the experimental branching ratios can determine the square of the coupling constants only , the relative signs of the interference terms are not determined ( in the present case @xmath285 possibilities ) . we have tested all these possibilities . 5(b ) shows the result with the interference terms with one choice of the relative signs . in selecting this sign combination we choose the one which gives the best results compared with the experimental data for both the total and the differential cross section , we then choose this sign combination in all further investigations . we also found that other relative sign combinations give a better total cross section , however do not give a good differential cross section . a single resonance or @xmath13-exchange can not explain the experimental data . the contribution from @xmath108 is larger than from the other resonances . but one must include all the resonances from the particle booklet which decay into @xmath157 and @xmath286 for the description of kaon production . the @xmath11-channel resonances gives a breit - wigner form as expected , whereas the @xmath12-channel @xmath179-exchange is necessary to explain the long tail in the total cross section . 6(a ) , 6(b ) and 6(c ) show the results for differential cross sections of the reaction @xmath265 at the pion beam momentum 0.980 gev / c , 1.13 gev / c and 1.455 gev / c respectively . the experimental data of knasel _ et al . _ @xcite and saxon _ et al . _ @xcite are shown in the figures with error bars . one sees that the theoretical results are in roughly good agreement with the experimental data . we mention here that in the work of brown _ et al._@xcite , the isospin - averaged total cross section of @xmath287 is parametrized in the breit - wigner form by including the @xmath108 , the @xmath109 , the @xmath133 . the interference terms between the resonances are not taken into account . since the amplitudes are not given , they can not study the differential cross sections . our parametrization for the @xmath288 cross section is given by @xmath289 for @xmath290 smaller than the @xmath291 production threshold of 1.613 gev , the above cross section is zero . other isospin channels can be related to the reaction of @xmath288 by eqs . ( [ nlamamp1 ] ) and ( [ nlamamp2 ] ) . the total and differential cross sections of the @xmath293 reaction are given by eqs . ( [ total2 ] ) and ( [ cross2 ] ) , respectively . the expressions for other isospin dependent reactions can be written in the same way . the amplitudes for each channel are shown in eqs . ( [ a2])-([d2 ] ) . in evaluation of the amplitudes we need eight coupling constants ( i.e. @xmath268 , @xmath269 , @xmath294 , @xmath295 , @xmath296 , @xmath297 , @xmath298 and @xmath274 ) and cutoffs . among these eight coupling constants , three ( i.e. @xmath268 , @xmath269 and @xmath284 ) are already given in table 1 . the coupling constant and the cut off parameter of the @xmath299 vertex is assumed to be the same as @xmath300 vertex in order to introduce a minimum of free parameters , i.e. @xmath298=@xmath301=0.45 . the other four coupling constants ( i.e. @xmath294 , @xmath295 , @xmath296 and @xmath297 ) are determined by comparing eqs . ( [ ratio521])-([ratio524 ] ) given in appendix 5.2 with the experimental branching ratios listed in @xmath97 2 . ( the values of @xmath302 and @xmath297 obtained in this way will be called `` set 2 '' , the reason will get clear below . ) by using the cutoff value @xmath303 gev for the @xmath109 , the @xmath133 as in @xmath97 4.1 , and @xmath304 gev for the @xmath202 , these four coupling constants are shown in table 2 . unlike the @xmath98 reactions described in @xmath97 3.1 and @xmath97 4.1 , the resonances @xmath109 , @xmath133 , @xmath14 contribute to these reactions , due to the mixture of the isospin-@xmath110 of the nucleon resonances ( n(1710 ) and n(1720 ) ) , and isospin-@xmath48 of the @xmath202 resonance , the different isospin channels of the @xmath292 reactions differ not only by a overall factor but also in shape . this point can be seen from eqs . ( [ nsiamp1])-([nsiamp5 ] ) . in these equations the amplitudes @xmath199 , @xmath200 , @xmath201 and @xmath203 contribute to different isospin channels of the @xmath292 reactions with different weights . namely the @xmath14 and @xmath13-exchange distinguish different isospin channels of @xmath178 reactions . the role of the @xmath14 in the @xmath178 reactions was studied in our previous work @xcite . for the @xmath178 reactions three experimental data sets exist in the literature : ( 1 ) . @xmath305 , ( 2 ) . @xmath306 , ( 3 ) . @xmath307 . all these three channels have indeed different shapes , i.e. a different energy dependence of the total cross section . first , we discuss the reaction @xmath308 . in this channel the initial and the final states have charge two . therefore only the doubly charged resonance @xmath309 and the @xmath13-exchange can contribute ( see eq . ( [ nsiamp1 ] ) ) . the result is shown in the dashed line of fig . the cross section is considerably underestimated . this means that the contributions from other @xmath16 resonances with masses near 1.9 gev can not be neglected contribute to this reaction , however the value of branching ratios are still not yet determined . ] . thus we consider the @xmath202 as an effective resonance that simulates also other @xmath16 resonances in this energy region that contribute to this reaction . in this work we scale the two coupling constants @xmath297 and @xmath294 of the @xmath202 to account for other @xmath16 resonance contributions by fitting the @xmath310 experimental data . their values are also shown in table 2 . these two scaled coupling constants @xmath311 and @xmath294 are also used in all other channels of our calculations . all other coupling constants remain unchanged . the calculations with these two scaled coupling constants are called set 1 " . the results with these two coupling constants unscaled as determined from the branching ratios are also shown in the fig . 7 for reference ( `` set 2 '' ) . we believe that the set 1 " parameters produce reliable results and should be used for kaon production in heavy ion collisions . 7(b ) shows the result of the @xmath306 reaction . we find that set 1 " results are in good agreement with the data , especially for the second peak , while `` set 2 '' can not reproduce the second peak . this means that the two scaled coupling constants give a reliable description also of these data , although fitted to an other reaction . according to eq . ( [ nsiamp3 ] ) and eq . ( [ nsiamp4 ] ) the cross sections for @xmath312 and @xmath313 reactions are the same as for the @xmath307 reaction . we thus compare the calculations of the reactions @xmath312 and @xmath313 with the experimental data for @xmath307 in fig . 7(c ) . from fig . 7(c ) we see that these reactions are also in good agreement with the experimental data . from figs . 7(a)-7(c ) we see that the experimental data for different isospin channels have indeed a different shape . the theoretical calculations reproduce these different shapes . for the reaction @xmath314 no data are available . since this cross section is important for heavy ion collisions , we give the results for `` set 1 '' parameters in eq([prmtpips4 ] ) . since in the present approach we scaled the two coupling constants to take into account other @xmath16 resonances around 1.9 gev , we can not demand that the differential cross section can also be reproduced . the differential cross section strongly depends on the quantum numbers of the resonance . the differential cross section can be studied in the future if the branching ratios of the relevant resonances are well determined . in order to use these cross sections in heavy ion collisions , we parameterize all the channels based on `` set 1 '' parameters as follows : they are : @xmath315 @xmath316 @xmath317 @xmath318 + where , all parametrizations given above should be understood to be zero below the threshold @xmath319 gev . these parametrizations are especially useful for kaon production simulation codes for those channels where no experimental data are available . the total and differential cross sections of the reaction @xmath321 are given by eqs . ( [ total3 ] ) and ( [ cross3 ] ) , respectively . the same expressions for other isospin dependent reactions can be written in the same way . the amplitudes for each channels are given by eqs . ( [ a3])-([c3 ] ) . we need six coupling constants ( i.e. @xmath322 , @xmath323 , @xmath324 , @xmath270 , @xmath271 and @xmath272 ) , three of them ( i.e. @xmath270 , @xmath271 and @xmath272 ) are given in @xmath97 4.1 and listed in table 1 . three other coupling constants are obtained from the experimental data listed in @xmath97 2 and eqs . ( [ ratio531])- ( [ ratio533 ] ) given in appendix 5.3 . by choosing the cutoff values as : @xmath275 gev for @xmath108 , @xmath109 and @xmath133 , i.e. in the same way as in @xmath97 4.1 and @xmath97 4.2 , we obtain these three coupling constants as listed in table 3 . since all the resonances that contribute to the reactions @xmath325 have isospin-@xmath110 , the different isospin dependent reaction channels have all the same shape ( see eqs . ( [ dlamamp1])-([dlamamp3 ] ) . this is the same as for the @xmath326 reactions . we find that the contribution from the @xmath109 resonance is larger ( about @xmath327 of the total cross section at peak position ) than from other resonances . for this reaction one has four possible relative sign combinations . eq.([prmtpidla ] ) shows the total cross section of the reaction @xmath328 without interference terms . a careful study did show that the interference terms do not affect the final results appreciably for this reaction . + @xmath329 this parametrization should be understood to be zero below the threshold ( @xmath330 gev ) . this parametrization is useful for codes which simulate kaon production since no experimental data are available . we find that the angular distribution @xmath236 is almost constant for all bombarding energies . we need six coupling constants ( i.e. @xmath323 , @xmath324 , @xmath295 , @xmath296 , @xmath332 , @xmath284 ) , three of them ( @xmath295 , @xmath296 , @xmath284 ) are given in @xmath97 4.2 and table 2 . two coupling constant ( @xmath323 , @xmath324 ) are given in @xmath97 4.3 and table 3 . one remaining coupling constant @xmath332 is assumed to be @xmath333 where @xmath334 comes from the different normalization of the operator @xmath190 for isospin-@xmath110 and @xmath335 for isospin-@xmath48 . the cut off parameter @xmath336 for vertex @xmath337 is the same as @xmath338 used in @xmath97 4.1 and @xmath97 4.2 . according to the different reaction amplitudes given by eqs . ( [ dsiamp1])-([dsiamp8 ] ) , one sees that the @xmath339-exchange distinguishes channels with different isospin projections . hence , among eight possible reaction channels there are five different independent reactions which differ not only by a overall factors , they are ( 1 ) . @xmath340 , ( 2 ) . @xmath341 , ( 3 ) . @xmath342 , ( 4 ) . @xmath343 and @xmath344 . we find that the cross section of the reaction @xmath345 , to which only the @xmath339-exchange contribute ( see eq.([dsiamp3 ] ) ) , is negligible ( smaller than 0.01 mb at its maximum value ) . we plot the reactions ( 1)-(3 ) in fig . the reaction @xmath346 has a maximum cross section of 0.15 mb with the peak position located at @xmath347 gev . due to the small contributions from the @xmath339-exchange , the shape from the different reactions of ( 1 ) -(4 ) do not differ much especially at the peak positions . we also find that the differential cross sections of the reactions @xmath331 are almost constant as a function of @xmath348 for all beam energies . now , we are in a position to give parametrizations of total cross sections @xmath349 , @xmath350 , @xmath351 , and @xmath352 which are sufficient to describe all channels given in eqs . ( [ dsiamp1])-([dsiamp8 ] ) . channel ( 5 ) is omitted , since only @xmath13-exchange contributes , which is negligible as mentioned above . since the signs of interference terms can not be fixed by experimental data , we parameterize the results obtained without interference terms . + @xmath353 @xmath354 @xmath355 @xmath356 + the parametrizations for the cross sections @xmath357 in eqs . ( [ prmtpids1 ] ) to ( [ prmtpids4 ] ) should be understood to be zero below the threshold ( @xmath319 gev ) . these parametrizations are useful for codes which simulate kaon production since no experimental data are available . in this paper , we have calculated the @xmath0 and @xmath1 cross sections using a resonance model on the hadronic level . all resonances with known experimental branching ratios to @xmath157 and @xmath72 are included . we constructed the interaction lagrangians needed in the description of the vertices . for the reactions @xmath358 our theoretical calculations are in reasonable good agreement with the experimental data both for differential and total cross sections . for the @xmath359 reactions , we scaled the coupling constant related to the @xmath202 resonance , to take into account also other @xmath16 resonances in this energy region . to get a complete understanding of these reactions a better determination of the branching ratios also of other @xmath16 resonances is necessary . for the @xmath1 reactions , we predict the total cross sections and find that the differential cross sections are almost isotopic . for all different reaction channels with different charges of the particapants the total cross sections are parameterized for the application in heavy ion collisions + * acknowledgement : * the authors express their thanks to prof . r. vinh mau , prof . h. mther and prof . e. oset for useful discussions and are indebted to prof . k. w. schmid for providing the code used for fitting the theoretically calculated cross sections to the simple parametrisations given in this work . in this appendix , we list all the theoretical branching ratios of the resonances used in this work . the branching ratios averaged over initial and summed over final spin and isospin states of the baryon resonances and @xmath13 are given : ( form factors have to be inserted at each vertex see @xmath97 3.1 ) @xmath360 with @xmath361 99 j. aichelin and c. m. ko , phys . 55 2661 ( 1985 ) . a. lang , w. cassing , u. mosel and k. weber , nucl . phys . a 541 507 ( 1992 ) ; s. w. huang , a. faessler , g. q. li , r. k. puri , e. lehmann , d. t. khoa and m. a. matin , phys . b 298 41 ( 1993 ) ; g. hartnack , j. jnicke and j. aichelin , preprint , universite de nantes , rapport interne lpn-93 - 11 ( 1993 ) . w. cassing , v. metag , u. mosel and k. niita , phys . 188 363 ( 1990 ) . g. q. li , a. faessler and s. w. huang , progresses in particles and nuclear physics 31 159 ( 1993 ) . g. f. bertsch , h. kruse , s. das gupta , phys . rev . * c29 * 673 ( 1984 ) ; h. kruse , b. v. jacak , h. stcker , phys . lett . * 54 * 289 ( 1985 ) ; g. f. bertsch , s. das gupta , phys . rep . * 160 * 189 ( 1988 ) . j. aichelin , phys . 202 235 ( 1991 ) . j. randrup and c. m. ko , nucl . phys . a 343 519 ( 1980 ) ; nucl . phys . a 411 537 ( 1983 ) . j. cugnon and r. m. lombard , nucl . a 422 635 ( 1984 ) . e. ferrari , phys . 120 988 ( 1960 ) ; t. yao , phys . 125 1048 ( 1962 ) . w. cassing , g. batko , u. mosel , k . niita , o.schult and gy . wolf , phys . b 238 25 ( 1990 ) . wolf , g. batko , w. cassing , u. mosel , k. niita and m. schfer , nucl . a517 615 ( 1990 ) . s. huber and j. aichelin , nucl . a573 587 ( 1994 ) . k. tsushima , s. w. huang , a. faessler , phys . b 337 245 ( 1994 ) ; k. tsushima , s. w. huang , a. faessler , j. phys . g. in press . particle data group , phys . d 45 ( 1992 ) . particle data group , phys . b 239 1 ( 1990 ) . see , for example , y. takahashi , _ an introduction to field quantization _ , pergamon press , 1969 ; d. luri , _ particles and fields _ , john wiley & sons , 1968 . m. benmerrouche , r. m. davidson and nimai c. mukhopadhyay , phys . c 39 2339 ( 1989 ) . c. gobbi , f. iachello and d. kusnezov , yale university preprint ( 1994 ) pacs numbers : 13.25.+m , 11.30.na , 12.40.aa . c. itzykson and j. b. zuber , _ quantum field theory , _ mcgraw - hill , new york ( 1980 ) . a. baldini , v. flamino , w. g. moorhead and d. r. o. morrison , _ landolt - brnstein , numerical data and functional relationships in science and technology , _ vol . 12 , ed . by h. schopper , ( springer - verlag , 1988 ) . t. m. knasel , et al . d 11 1 ( 1975 ) d. h. saxon et al . b162 522 ( 1980 ) . g. e. brown , c. m. ko , z. g. wu and l. h. xia , phys . rev . c 43 1881 ( 1991 ) ; h. w. barz and h. iwe , nucl phys . a 453 728 ( 1986 ) ; v. pantuev , phys . b 281 20 ( 1992 ) . cccccccc @xmath164(resonance ) & @xmath389 & @xmath390 & @xmath391 & @xmath392 & @xmath393 + + @xmath108 & 150 & 70.0 & 1.41 & 7.0 & 6.40 @xmath394 + @xmath109 & 100 & 15.0 & 2.57 & 15.0 & 4.74 @xmath395 + @xmath133 & 150 & 15.0 & 5.27 @xmath396 & 6.5 & 3.91 + & @xmath397 & & @xmath398 & + & 6.89 @xmath394 & & 2.03 @xmath394 & & + & ( @xmath399=50 mev,&@xmath400 ) & & & + + cccccc @xmath164(resonance ) & @xmath389 & @xmath401 & @xmath402 & @xmath403 & @xmath404 + + @xmath109 & 100 & 15.0 & ( see table 1)&6.0 & @xmath405 + @xmath133 & 150 & 15.0 & ( see table 1)&3.5 & 3.15 + @xmath202 ( set 1 ) & &&(1.44)&&(3.83 ) + @xmath202 ( set 2 ) & 200 & 12.5 & @xmath406&2.0 & 1.11 + + & @xmath407 & & @xmath408 & & + + & ( see table 1 ) & & @xmath409(=@xmath410 ) & & + + * table 2 * the calculated or fitted coupling constants and the data used for the calculations . the values in brackets stand for the coupling constants obtained by fitting to the total cross section for the @xmath411 reaction ( set 1 ) . the value of @xmath412 is the fitted value to the @xmath209 channel when zero tensor coupling for @xmath413 interaction ( @xmath414 ) is applied and the rescaled coupling constants for @xmath202 are used . cccccccc @xmath164(resonance ) & @xmath389 & @xmath415 & @xmath416 & @xmath392 & @xmath417 + + @xmath108 & 150 & 5.0 & 6.56 @xmath394 & 7.0 & ( see table 1 ) + @xmath109 & 100 & 17.5 & 1.85 @xmath396 & 15.0 & ( see table 1 ) + @xmath133 & 150 & 10.0 & 1.12 @xmath395 & 6.5 & ( see table 1 ) + + * fig . 1 * the processes contributing to the @xmath2 reactions . the diagrams are corresponding to the different intermediate resonance states : @xmath418 @xmath419 , @xmath420 and @xmath421t - channel @xmath13-exchange , respectively . 2 * the processes contributing to the @xmath7 reactions . the diagrams are corresponding to the different intermediate resonance states : @xmath422 , @xmath423 , @xmath424 and @xmath421t - channel @xmath13 exchange , respectively . * 4 * the processes contributing to the @xmath10 reactions . the diagrams are corresponding to the different intermediate resonance states : @xmath422 , @xmath423 and @xmath421t - channel @xmath13 exchange , respectively . * fig . 5 * the total cross section for the reaction @xmath288 the experimental data with error bars are taken from @xcite . the theoretical calculation are shown with the solid lines for the results without interference terms ( fig . ( a ) ) and with interference terms ( fig . ( b ) ) . * fig . 6 * the differential cross section for the reaction @xmath288 as a function of @xmath425 in the c.m . frame at different bombarding energies . ( a ) , ( b ) and ( c ) show the reactions at pion beam momenta 0.980 gev / c ( corresponding to @xmath426 gev ) , 1.13 gev / c ( @xmath427 gev ) and 1.455 gev / c ( @xmath428 gev ) , respectively . the experimental data with error bars for ( a ) and ( b ) are taken from @xcite . the experimental data with error bars for ( c ) are from @xcite . the theoretical calculations ( solid lines ) include interference terms . * fig . 7 * ( a ) . the calculated and experimental @xcite total cross sections for the @xmath209 ( @xmath429 ) reaction . the solid line and the dashed line stand for the set 1 and set 2 parameters , respectively . set 1 are the relevant coupling constants for @xmath202 and @xmath430 adjusted to the cross section @xmath209 with the @xmath202 as an s - channel intermediate state resonance and other coupling constants are determined from the branching ratios . set 2 uses the same value of @xmath430 as that of set 1 and all other coupling constants determined from the branching ratios . ( b ) . the calculated total cross sections for the @xmath431 ( @xmath432 ) reaction . ( c ) . the calculated total cross sections for the @xmath433 and @xmath434 ( @xmath435 and @xmath436 ) reactions . * fig . 8 * ( a ) . the calculated total cross sections for the @xmath349 ( @xmath437 ) reactions . the solid line and the dashed lines stand for the results without and with the inclusion the interference terms , respectively . note that the largest and the smallest results are displayed for the four possibilities arising from the possible signs of the coupling constants and thus the interference terms . ( b ) . the calculated total cross sections for the @xmath350 ( @xmath438 ) reactions . ( c ) . the calculated total cross sections for the @xmath439 ( @xmath440 ) reactions . ( 11,20 ) ( 0,11)(1,1)@xmath23 ( 0,17)(1,1)@xmath19 ( 2,10)(1,1)(a ) ( 4,11)(1,1)@xmath155 ( 4,17)(1,1)@xmath154 ( 3,14.5)(1,1)n(1650 ) ( 3,14)(1,1 ) @xmath441 ( 2.5,13.5)(-1,-1)1.5 ( 2.5,13.5)(0.55,-0.55)3(1,-1)0.4 ( 2.48,13.5)(0,1)2 ( 2.5,13.5)(0,1)2 ( 2.52,13.5)(0,1)2 ( 2.5,15.5)(-1,1)1.5 ( 2.5,15.5)(0.55,0.55)3(1,1)0.4 ( 5.5,11)(1,1)@xmath23 ( 5.5,17)(1,1)@xmath19 ( 7.5,10)(1,1)(b ) ( 9.5,11)(1,1)@xmath155 ( 9.5,17)(1,1)@xmath154 ( 8.5,14.5)(1,1)n(1710 ) ( 8.5,14)(1,1 ) @xmath442 ( 8,13.5)(-1,-1)1.5 ( 8,13.5)(0.55,-0.55)3(1,-1)0.4 ( 7.98,13.5)(0,1)2 ( 8,13.5)(0,1)2 ( 8.02,13.5)(0,1)2 ( 8,15.5)(-1,1)1.5 ( 8,15.5)(0.55,0.55)3(1,1)0.4 ( 11,11)(1,1)@xmath23 ( 11,17)(1,1)@xmath19 ( 13,10)(1,1)(c ) ( 15,11)(1,1)@xmath155 ( 15,17)(1,1)@xmath154 ( 14,14.5)(1,1)n(1720 ) ( 14,14)(1,1 ) @xmath443 ( 13.5,13.5)(-1,-1)1.5 ( 13.5,13.5)(0.55,-0.55)3(1,-1)0.4 ( 13.48,13.5)(0,1)2 ( 13.5,13.5)(0,1)2 ( 13.52,13.5)(0,1)2 ( 13.5,15.5)(-1,1)1.5 ( 13.5,15.5)(0.55,0.55)3(1,1)0.4 ( 0,2)(1,1)@xmath23 ( 0,8)(1,1)@xmath19 ( 2,1)(1,1)(d ) ( 4,2)(1,1)@xmath155 ( 4,8)(1,1)@xmath154 ( 2,6)(1,1)@xmath179(892 ) ( 2,5.5)(1,1 ) @xmath444 ( 0.8,3)(0,1)5 ( 4.2,3)(0,1)5 ( 0.8,5.5)(1,0)3.4 ( 11,20 ) ( 0,11)(1,1)@xmath23 ( 0,17)(1,1)@xmath18 ( 2,10)(1,1)(a ) ( 4,11)(1,1)@xmath155 ( 4,17)(1,1)@xmath154 ( 3,14.5)(1,1)n(1710 ) ( 3,14)(1,1 ) @xmath442 ( 2.5,13.5)(-1,-1)1.5 ( 2.5,13.5)(0.55,-0.55)3(1,-1)0.4 ( 2.48,13.5)(0,1)2 ( 2.5,13.5)(0,1)2 ( 2.52,13.5)(0,1)2 ( 2.5,15.5)(-1,1)1.5 ( 2.5,15.5)(0.55,0.55)3(1,1)0.4 ( 5.5,11)(1,1)@xmath23 ( 5.5,17)(1,1)@xmath18 ( 7.5,10)(1,1)(b ) ( 9.5,11)(1,1)@xmath155 ( 9.5,17)(1,1)@xmath154 ( 8.5,14.5)(1,1)n(1720 ) ( 8.5,14)(1,1 ) @xmath443 ( 8,13.5)(-1,-1)1.5 ( 8,13.5)(0.55,-0.55)3(1,-1)0.4 ( 7.98,13.5)(0,1)2 ( 8,13.5)(0,1)2 ( 8.02,13.5)(0,1)2 ( 8,15.5)(-1,1)1.5 ( 8,15.5)(0.55,0.55)3(1,1)0.4 ( 11,11)(1,1)@xmath23 ( 11,17)(1,1)@xmath18 ( 13,10)(1,1)(c ) ( 15,11)(1,1)@xmath155 ( 15,17)(1,1)@xmath154 ( 14,14.5)(1,1)@xmath16(1920 ) ( 14,14)(1,1 ) @xmath445 ( 13.5,13.5)(-1,-1)1.5 ( 13.5,13.5)(0.55,-0.55)3(1,-1)0.4 ( 13.48,13.5)(0,1)2 ( 13.5,13.5)(0,1)2 ( 13.52,13.5)(0,1)2 ( 13.5,15.5)(-1,1)1.5 ( 13.5,15.5)(0.55,0.55)3(1,1)0.4 ( 0,2)(1,1)@xmath23 ( 0,8)(1,1)@xmath18 ( 2,1)(1,1)(d ) ( 4,2)(1,1)@xmath155 ( 4,8)(1,1)@xmath154 ( 2,6)(1,1)@xmath179(892 ) ( 2,5.5)(1,1 ) @xmath444 ( 0.8,3)(0,1)5 ( 4.2,3)(0,1)5 ( 0.8,5.5)(1,0)3.4 ( 11,20 ) ( 0,11)(1,1)@xmath16 ( 0,17)(1,1)@xmath19 ( 2,10)(1,1)(a ) ( 4,11)(1,1)@xmath155 ( 4,17)(1,1)@xmath154 ( 3,14.5)(1,1)n(1650 ) ( 3,14)(1,1 ) @xmath441 ( 2.5,13.5)(-1,-1)1.5 ( 2.5,13.5)(0.55,-0.55)3(1,-1)0.4 ( 2.48,13.5)(0,1)2 ( 2.5,13.5)(0,1)2 ( 2.52,13.5)(0,1)2 ( 2.5,15.5)(-1,1)1.5 ( 2.5,15.5)(0.55,0.55)3(1,1)0.4 ( 5.5,11)(1,1)@xmath16 ( 5.5,17)(1,1)@xmath19 ( 7.5,10)(1,1)(b ) ( 9.5,11)(1,1)@xmath155 ( 9.5,17)(1,1)@xmath154 ( 8.5,14.5)(1,1)n(1710 ) ( 8.5,14)(1,1 ) @xmath442 ( 8,13.5)(-1,-1)1.5 ( 8,13.5)(0.55,-0.55)3(1,-1)0.4 ( 7.98,13.5)(0,1)2 ( 8,13.5)(0,1)2 ( 8.02,13.5)(0,1)2 ( 8,15.5)(-1,1)1.5 ( 8,15.5)(0.55,0.55)3(1,1)0.4 ( 11,11)(1,1)@xmath16 ( 11,17)(1,1)@xmath19 ( 13,10)(1,1)(c ) ( 15,11)(1,1)@xmath155 ( 15,17)(1,1)@xmath154 ( 14,14.5)(1,1)n(1720 ) ( 14,14)(1,1 ) @xmath443 ( 13.5,13.5)(-1,-1)1.5 ( 13.5,13.5)(0.55,-0.55)3(1,-1)0.4 ( 13.48,13.5)(0,1)2 ( 13.5,13.5)(0,1)2 ( 13.52,13.5)(0,1)2 ( 13.5,15.5)(-1,1)1.5 ( 13.5,15.5)(0.55,0.55)3(1,1)0.4 ( 11,20 ) ( 0,11)(1,1)@xmath16 ( 0,17)(1,1)@xmath18 ( 2,10)(1,1)(a ) ( 4,11)(1,1)@xmath155 ( 4,17)(1,1)@xmath154 ( 3,14.5)(1,1)n(1710 ) ( 3,14)(1,1 ) @xmath442 ( 2.5,13.5)(-1,-1)1.5 ( 2.5,13.5)(0.55,-0.55)3(1,-1)0.4 ( 2.48,13.5)(0,1)2 ( 2.5,13.5)(0,1)2 ( 2.52,13.5)(0,1)2 ( 2.5,15.5)(-1,1)1.5 ( 2.5,15.5)(0.55,0.55)3(1,1)0.4 ( 5.5,11)(1,1)@xmath16 ( 5.5,17)(1,1)@xmath18 ( 7.5,10)(1,1)(b ) ( 9.5,11)(1,1)@xmath155 ( 9.5,17)(1,1)@xmath154 ( 8.5,14.5)(1,1)n(1720 ) ( 8.5,14)(1,1 ) @xmath443 ( 8,13.5)(-1,-1)1.5 ( 8,13.5)(0.55,-0.55)3(1,-1)0.4 ( 7.98,13.5)(0,1)2 ( 8,13.5)(0,1)2 ( 8.02,13.5)(0,1)2 ( 8,15.5)(-1,1)1.5 ( 8,15.5)(0.55,0.55)3(1,1)0.4 ( 11,11)(1,1)@xmath16 ( 11,17)(1,1)@xmath18 ( 13,10)(1,1)(c ) ( 15,11)(1,1)@xmath155 ( 15,17)(1,1)@xmath154 ( 13,14.8)(1,1)@xmath179(892 ) ( 13,14.3)(1,1 ) @xmath444 ( 11.8,12)(0,1)5 ( 15.2,12)(0,1)5 ( 11.8,14.4)(1,0)3.4

the non - equilibrium self - consistent generalized langevin equation theory of irreversible processes in liquids is extended to describe the positional and orientational thermal fluctuations of the instantaneous local concentration profile of a suddenly - quenched colloidal liquid of particles interacting through non spherically - symmetric pairwise interactions , whose mean value is constrained to remain uniform and isotropic , . such self - consistent theory is cast in terms of the time - evolution equation of the covariance of the fluctuations of the spherical harmonics projections of the fourier transform of . the resulting theory describes the non - equilibrium evolution after a sudden temperature quench of both , the static structure factor projections and the two - time correlation function , where is the correlation _ delay _ time and is the _ evolution _ or _ waiting _ time after the quench . as a concrete and illustrative application we use the resulting self - consistent equations to describe the irreversible processes of equilibration or aging of the orientational degrees of freedom of a system of strongly interacting classical dipoles with quenched positional disorder .

1605.04485

the fundamental description of dynamically arrested states of matter is a crucial step towards understanding the properties of very common amorphous solid materials such as glasses and gels @xcite , and of more technologically specialized materials , such as spin glasses @xcite . the main fundamental challenge posed by these materials derives from their inability to reach thermodynamic equilibrium within experimental times , and from the fact that their properties depend on the protocol of preparation , in obvious contrast with materials that have genuinely attained thermodynamic equilibrium . understanding the origin of this behaviour falls outside the realm of classical and statistical thermodynamics , and must unavoidably be addressed from the perspective of a non - equilibrium theory @xcite . in fact , a major challenge for statistical physics is to develop a microscopic theory able to predict the properties of glasses and gels in terms not only of the intermolecular forces and applied external fields , but also in terms of the protocol of preparation of the material . _ first - principles _ theoretical frameworks exist , leading to _ quantitative _ predictions of the dynamic properties of structural glass forming liquids near their dynamical arrest transitions , one of the best - known being mode coupling theory ( mct ) @xcite . however , this theory , as well as the equilibrium version of the self - consistent generalized langevin equation ( scgle ) theory of dynamical arrest @xcite , are meant to describe the dynamics of _ fully equilibrated _ liquids . hence , the phenomenology of the transient time - dependent processes , such as aging , occurring during the amorphous solidification of structural glass formers , falls completely out of the scope of these _ equilibrium _ theories . thus , it is important to attempt their extension to describe these non - stationary non - equilibrium structural relaxation processes , which in the end constitute the most fundamental kinetic fingerprint of glassy behavior . in an attempt to face this challenge , in 2000 latz @xcite proposed a formal non - equilibrium extension of mct which , however , has not yet found a specific quantitative application . in the meanwhile , the scgle theory has recently been extended to describe non - stationary non - equilibrium processes in glass - forming liquids @xcite . the resulting non - equilibrium theory , referred to as the _ non - equilibrium _ self - consistent generalized langevin equation ( ne - scgle ) theory , was derived within the fundamental framework provided by a non - stationary extension @xcite of onsager s theory of linear irreversible thermodynamics @xcite and of time - dependent thermal fluctuations @xcite , with an adequate extension @xcite to allow for the description of memory effects . the ne - scgle theory thus derived , aimed at describing non - equilibrium relaxation phenomena in general @xcite , leads in particular @xcite to a simple and intuitive but generic description of the essential behavior of the non - stationary and non - equilibrium structural relaxation of glass - forming liquids near and beyond its dynamical arrest transition . this was explained in detail in ref . @xcite in the context of a model liquid of soft - sphere particles . the recent comparison @xcite of the predicted scenario with systematic simulation experiments of the equilibration and aging of dense hard - sphere liquids , indicates that the accuracy of these predictions go far beyond the purely qualitative level , thus demonstrating that the ne - scgle theory is a successful pioneering first - principles statistical mechanical approach to the description of these fully non - equilibrium phenomena . as an additional confirmation , let us mention that for model liquids with hard - sphere plus attractive interactions , the ne - scgle theory predicts a still richer and more complex scenario , involving the formation of gels and porous glasses by arrested spinodal decomposition @xcite . as we know , quenching a liquid from supercritical temperatures to a state point inside its gas - liquid spinodal region , normally leads to the full phase separation through a process that starts with the amplification of spatial density fluctuations of certain specific wave - lengths @xcite . under some conditions , however , this process may be interrupted when the denser phase solidifies as an amorphous sponge - like non - equilibrium bicontinuous structure @xcite , typical of physical gels @xcite . this process is referred to as _ arrested _ spinodal decomposition , and has been observed in many colloidal systems , including colloid - polymer mixtures @xcite , mixtures of equally - sized oppositely - charged colloids @xcite , lysozyme protein solutions @xcite , mono- and bi - component suspensions of colloids with dna - mediated attractions @xcite , and thermosensitive nanoemulsions @xcite . from the theoretical side , it was not clear how to extend the classical theory of spinodal decomposition @xcite to include the possibility of dynamic arrest , or how to incorporate the characteristic non - stationarity of spinodal decomposition , in existing theories of glassy behavior @xcite . in refs . @xcite and @xcite it has been shown that the ne - scgle theory provides precisely this missing unifying theoretical framework . recently the ne - scgle theory was extended to multi - component systems @xcite , thus opening the route to the description of more complex non - equilibrium amorphous states of matter . until now , however , the ne - scgle theory faces the limitation of referring only to liquids of particles with radially symmetric pairwise interparticle forces , thus excluding its direct comparison with the results of important real and simulated experiments involving intrinsically non - spherical particles @xcite and , in general , particles with non - radially symmetric interactions . the present work constitutes a first step in the direction of extending the ne - scgle theory to describe the irreversible evolution of the static and dynamic properties of a brownian liquid constituted by particles with non - radially symmetric interactions , in which the orientational degrees of freedom are essential . more concretely , the main purpose of the present paper is to describe the theoretical derivation of the ne - scgle time - evolution equations for the spherical harmonics projections @xmath10 , @xmath11 , and @xmath12 , of the non - equilibrium and non - stationary static structure factor @xmath13 and of the collective and self intermediate scattering functions @xmath14 and @xmath15 . for this , we start from the same general and fundamental framework provided by the non - stationary extension of onsager s theory , developed in ref . @xcite to discuss the spherical case . the result of the present application are eqs . ( [ relsigmadifnonspher13])-([flmsqframenonstats ] ) below , in which @xmath8 is the delay time , and @xmath9 is the evolution ( or `` waiting '' ) time after the occurrence of the instantaneous temperature quench . the solution of these equations describe the non - equilibrium ( translational and rotational ) diffusive processes occurring in a colloidal dispersion after an instantaneous temperature quench , with the most interesting prediction being the aging processes that occur when full equilibration is prevented by conditions of dynamic arrest . although this paper only focuses on the theoretical derivation of the ne - scgle equations , as an illustration of the possible concrete applications of the extended non - equilibrium theory , here we also solve the resulting equations for one particular system and condition . we refer to a liquid of dipolar hard - spheres ( dhs ) with fixed positions and subjected to a sudden temperature quench . this is a simple model of the irreversible evolution of the collective orientational degrees of freedom of a system of strongly interacting magnetic dipoles with fixed but random positions . although this particular application by itself has its own intrinsic relevance in the context of disordered magnetic materials , the main reason to choose it as the illustrative example is that eqs . ( [ relsigmadifnonspher13])-([flmsqframenonstats ] ) describe coupled translational and rotational dynamics , whose particular case @xmath16 coincide with the radially - symmetric case , already discussed in detail in refs . thus , the most novel features are to be expected in the non - equilibrium rotational dynamics illustrated in this exercise . just like in the case of liquids formed by spherical particles , the development of the ne - scgle theory for liquids of non - spherical particles requires the previous development of the _ equilibrium _ version of the corresponding scgle theory . such an equilibrium scgle theory for non - spherical particles , however , was previously developed by elizondo - aguilera et al . @xcite , following to a large extent the work of schilling and collaborators @xcite on the extension of mode coupling theory for this class of systems . thus , we start our discussions in section [ section2 ] with a brief review of the main elements of the non spherical _ equlibrium _ scgle theory and its application to dynamical arrest in systems formed by colloidal interacting particles with non spherical potentials . in section [ nonequilib ] we outline the conceptual basis and the main steps involved in the derivation of the non - equilibrium extension of the scgle theory for glass - forming liquids of non - spherical particles . in the same section we summarize the resulting set of self consistent equations which constitutes this extended theory . in section [ agingspinglass ] , we introduce a simplified model for interacting dipoles randomly distributed in space and apply our equations to investigate the slow orientational dynamics as well as the aging and equilibration processes of the system near its `` spin glass''-like transitions . finally in section [ conclusions ] we summarize our main conclusions . in this section we briefly describe the equilibrium scgle theory of the dynamics of liquids formed by non - spherical particles developed by elizondo - aguilera et al . we first describe the main properties involved in this description and then summarize their time - evolution equations , which constitute the essence of the scgle theory . let us start by considering a liquid formed by @xmath17 identical non - spherical colloidal particles in a volume @xmath18 @xcite , each having mass @xmath19 and inertia tensor @xmath20 . the translational degrees of freedom are described by the vectors @xmath21 and @xmath22 , where @xmath23 denotes the center - of - mass position vector of the @xmath24th - particle and @xmath25 is the associated linear momentum . similarly , the orientational degrees of freedom are described by the abstract vectors @xmath26 and @xmath27 , where @xmath28 denotes the euler angles which specify the orientation of the @xmath24th molecule , and @xmath29 is the corresponding angular momentum , so that @xmath30 denotes the angular velocity . let us now assume that the potential energy @xmath31 of the interparticle interactions is pairwise additivity , i.e. , that @xmath32 where @xmath33 is the interaction potential between particles @xmath24 and @xmath34 . in the particular case of axially - symmetric particles , that we shall have in mind here , the third euler angle is actually redundant , and hence , @xmath35 . the most basic observable in terms of which we want to describe the dynamical properties of a non - spherical colloidal system is the time dependent microscopic one - particle density @xmath36 given that @xmath37 , _ any _ function @xmath38 can be expanded with respect to plane waves and spherical harmonics as @xmath39 where @xmath40 thus , using eq . ( [ localcon ] ) in ( [ fromega ] ) and ( [ flmk ] ) , we may define the _ so - called _ tensorial density modes @xmath41 and hence , we can define the following two - time correlation functions , @xmath42}y_{lm}^*(\bm{\omega}_n(t+\tau ) ) y_{l'm'}(\bm{\omega}_{n'}(t))\big\rangle,\end{aligned}\ ] ] where @xmath43 . we also define for completeness the self components @xmath44 and the corresponding two - time correlation functions @xmath45}y_{lm}(\bm{\omega}_t(t+\tau ) ) y_{l'm'}(\bm{\omega}_t(t))\big\rangle,\label{selfcorrelm}\end{aligned}\ ] ] where @xmath46 denotes the position of the center of mass of _ any _ of the particles at time @xmath9 and @xmath47 describes its orientation . as indicated before , we will refer to @xmath8 as the delay ( or _ correlation _ ) time , whereas for @xmath9 we refer to the _ evolution _ time . the equal - time value of these correlation functions are @xmath48 and @xmath49 where @xmath50 are the tensorial components of the static structure factor @xmath51 . of course , the dependence of these quantities on the evolution time @xmath9 is only relevant if the state of the system is not stationary . under thermodynamic equilibrium , @xmath52 , @xmath53 , and @xmath50 can not depend on @xmath9 , and we should denote them as @xmath54 , @xmath55 , and @xmath56 . in ref . @xcite the generalized langevin equation ( gle ) formalism and the concept of contraction of the description were employed to derive exact memory function equations for @xmath54 and @xmath55 . these dynamic equations only involve the corresponding projections @xmath56 of the _ equilibrium _ static structure factor . for notational convenience , however , we shall not write the label @xmath57 in what follows , although for the rest of this section we shall only refer to these equilibrium properties . as explained in ref . @xcite , the referred exact memory function equations for @xmath54 and @xmath55 require the independent determination of the corresponding self and collective memory functions . in a manner similar to the spherical case , simple vineyard - like approximate closure relations for these memory functions convert the originally exact equations into a closed self - consistent system of approximate equations for the dynamic properties referred to above @xcite . these equations thus constitute the extension of the _ equilibrium _ scgle theory of the dynamic properties of liquids whose particles interact through non - spherical pair potentials . let us now summarize the set of self - consistent equations that constitute the _ equilibrium _ scgle theory for a brownian liquid of axially - symmetric non spherical particles . in the simplest version ( we refer the reader to ref . @xcite for details ) these equations involve only the diagonal elements @xmath58 and @xmath59 , and are written , in terms of the corresponding laplace transforms @xmath60 and @xmath61 , as @xmath62 and @xmath63 in these equations , @xmath64 is the rotational free - diffusion coefficient , and @xmath65 is the center - of - mass translational free - diffusion coefficient , whereas the functions @xmath66 and @xmath67 are defined as @xmath68 $ ] and @xmath69 , where @xmath70 , with @xmath71 being the position of the main peak of @xmath72 and @xmath73 . this ensures that for radially - symmetric interactions , we recover the original theory describing liquids of soft and hard spheres @xcite . on the other hand , within well defined approximations discussed in appendix a of ref @xcite , the functions @xmath74 ( @xmath75 ) may be written as @xmath76\left[1-s^{-1}_{l0}(k)\right]^2 f^s_{l0}(k;\tau)f_{l0}(k;\tau)\label{zetart}\end{aligned}\ ] ] and @xmath77h^2_{l0 } ( k)\left[a_{l;0m}\right]^2\left[s^{-1}_{lm}(k)\right]^2 f^s_{lm}(k;\tau)f_{lm}(k;\tau)\label{zetarqq}\end{aligned}\ ] ] where @xmath78 denotes the diagonal _ k_-frame projections of the total correlation function @xmath79 , i.e. , @xmath78 is related to @xmath80 by @xmath81 , and @xmath82 is the number density . finally , @xmath83 $ ] and @xmath84 . the closed set of coupled equations in eqs . ( [ flmqframe1])-([zetarqq ] ) constitute the equilibrium non spherical version of the scgle theory , whose solution provides the full time - evolution of the dynamic correlation functions @xmath85 and @xmath86 and of the memory functions @xmath87 . these equations may be numerically solved using standard methods once the projections @xmath80 of the static structure factor are provided . under some circumstances , however , one may only be interested in identifying and locating the regions in state space that correspond to the various possible ergodic or non ergodic phases involving the translational and orientational degrees of freedom of a given system . for this purpose it is possible to derive from the full scgle equations the so - called bifurcation equations , i.e. , the equations for the long - time stationary solutions of equations ( [ flmqframe1])-([zetarqq ] ) . these are written in terms of the so - called non - ergodicity parameters , defined as @xmath88 @xmath89 and @xmath90 with @xmath91 . the simplest manner to determine these asymptotic solutions is to take the long - time limit of eqs . ( [ flmqframe1])-([zetarqq ] ) , leading to a system of coupled equations for @xmath92 , @xmath93 , and @xmath94 . it is not difficult to show that the resulting equations can be written as @xmath95 \lambda_t^{(lm)}(k)\lambda_r^{(lm)}(k)}{s_{lm}(k)\lambda_t^{(lm)}(k ) \lambda_r^{(lm)}(k)+k^2\gamma_t\lambda_r^{(lm)}(k)+l(l+1)\gamma_r\lambda_t^{(lm)}(k)}\label{fc1}\ ] ] and @xmath96 where the dynamic order parameters @xmath97 and @xmath98 , defined as @xmath99 are determined from the solution of @xmath100\left[1-s^{-1}_{l0}(k)\right]^2 s_{l0}(k)f^s_{l0}(k)f_{l0}(k),\label{gat}\ ] ] and @xmath101[s_{l0}(k)-1]^2s^{-1}_{lm}(k)f^s_{lm}(k)f_{lm}(k)a_{l;0m}^2 .\label{gar}\ ] ] as discussed in ref . @xcite , fully ergodic states are described by the condition that the non - ergodicity parameters ( i.e. , @xmath92 , @xmath93 , and @xmath94 ) are all zero , and hence , the dynamic order parameters @xmath97 and @xmath98 are both infinite . any other possible solution of these bifurcation equations indicate total or partial loss of ergodicity . thus , @xmath97 and @xmath98 finite indicate full dynamic arrest whereas @xmath97 finite and @xmath102 corresponds to the mixed state in which the translational degrees of freedom are dynamically arrested but not the orientational degrees of freedom . the main reason for this brief summary of the scgle theory for liquids with non - spherical inter - particle interactions , is that this equilibrium theory contains the fundamental ingredients to develop a theoretical description of the genuine non - equilibrium non - stationary irreversible processes characteristic of glassy behavior , such as aging @xcite . thus , let us now outline the conceptual basis and the main steps in the derivation of the non - equilibrium version of the scgle theory for glass - forming liquids of non - spherical particles , which we shall refer to as the _ non - equilibrium _ generalized langevin equation ( ne - scgle ) theory . our starting point is the non - stationary version @xcite of onsager s theory of thermal fluctuations and irreversible processes @xcite , which states that : @xmath103 the mean value @xmath104 of the vector @xmath105^{\dagger } $ ] formed by the @xmath106 macroscopic variables that describe the state of the system is the solution of some generally nonlinear equation , represented by @xmath107 , \label{releq0}\ ] ] whose linear version in the vicinity of a stationary state @xmath108 ( i.e. , @xmath109=0 $ ] ) reads @xmath110 \cdot \mathcal{e } \left[{\overline { \textbf a } } ^{ss } \right ] \cdot \delta\overline { \textbf a } ( t ) , \label{releq1}\ ] ] with @xmath111 , and that : @xmath112 the relaxation equation for the @xmath113 covariance matrix @xmath114 of the non - stationary fluctuations @xmath115 can be written as @xcite @xmath116 \cdot \mathcal{e}\left[{\overline { \textbf a } } ( t)\right]\cdot \sigma ( t)\ ] ] @xmath117 \cdot\mathcal{l}^{\dagger } [ { \overline{\textbf a}}(t ) ] + \left ( \mathcal{l } [ { \overline{\textbf a}}(t ) ] + \mathcal{l}^{\dagger } [ { \overline{\textbf a}}(t ) ] \right).\ ] ] in these equations @xmath118 $ ] is a @xmath119 `` _ _ kinetic _ _ '' matrix , defined in terms of @xmath120 $ ] as @xmath118\equiv -\left ( \partial \mathcal{r}\left [ { \textbf a } \right]/ \partial { \textbf a } \right ) \cdot \mathcal{e}^{-1}\left[{\textbf a}\right]$ ] , whereas @xmath121 $ ] is the @xmath122 thermodynamic ( `` stability '' ) matrix , defined as @xmath123 \equiv -\frac{1}{k_b}\left ( \frac{\partial^2 s[{\textbf a } ] } { \partial a_i\partial a_j } \right)= -\left ( \frac{\partial f_i[{\textbf a } ] } { \partial a_j } \right ) \ \ \ \ ( i , j=1,2, ... ,\nu ) , \label{matrixe}\ ] ] with @xmath124 $ ] being the entropy and @xmath125 \equiv k_b^{-1 } \left ( \partial s [ { \textbf a } ] / \partial a_j \right)$ ] the conjugate intensive variable associated with @xmath126 . the function @xmath127 $ ] , which assigns a value of the entropy @xmath128 to any possible state point * a * in the state space of the system , is thus the so - called fundamental thermodynamic relation @xcite , and constitutes the most important and fundamental external input of the non - equilibrium theory . the previous equations , however , do not explicitly require the function @xmath127 $ ] , but only its second derivatives defining the stability matrix @xmath129 $ ] . the most important property of the matrix @xmath129 $ ] is that its inverse is the covariance of the _ equilibrium _ fluctuations , i.e. , @xmath130 \cdot \sigma^{eq}=i \label{oz},\ ] ] with @xmath131 , where the average is taken with the probability distribution @xmath132 $ ] of the equilibrium ensemble . in addition , the non - equilibrium version of onsager s formalism introduces the globally non - stationary ( but locally stationary ) extension @xcite of the generalized langevin equation for the stochastic variables @xmath133 @xcite , @xmath134\cdot \sigma^{-1 } ( t)\cdot \delta \textbf{a}(t+\tau ) \\ & -\int_0^\tau d\tau ' \gamma[\tau-\tau';\overline{\textbf{a}}(t ) ] \cdot \sigma^{-1}(t)\cdot \delta \textbf{a}(t+\tau ' ) + \textbf{f}(t+\tau ) , \end{split}\end{aligned}\ ] ] where the random term @xmath135 has zero mean and two - time correlation function given by the fluctuation - dissipation relation @xmath136 $ ] . from this equation one derives the time - evolution equation for the non - stationary time - correlation matrix @xmath137 , reading @xmath138\cdot \sigma^{-1 } ( t)\cdot c(\tau;t ) \\ & -\int_0^\tau d\tau ' \gamma[\tau-\tau';\overline{\textbf{a}}(t ) ] \cdot \sigma^{-1}(t)\cdot c(\tau';t ) , \end{split}\end{aligned}\ ] ] whose initial condition is @xmath139 . in these equations , $ ] represents conservative ( mechanical , geometrical , or streaming ) relaxation processes , and is just the antisymmetric part of @xmath141 $ ] , i.e. , @xmath140=(\mathcal{l } [ { \textbf a } ] -\mathcal{l}^{\dagger } [ { \textbf a } ] ) /2 $ ] . the memory function @xmath142 $ ] , on the other hand , summarizes the effects of all the complex dissipative irreversible processes taking place in the system . taking the laplace transform ( lt ) of eq . ( [ fluctuations3 ] ) to integrate out the variable @xmath8 in favor of the variable @xmath143 , rewrites this equation as @xmath144 \cdot \sigma^{-1}(t)\right\}^{-1}\cdot c(\tau=0;t ) \label{cdzt}\ ] ] with @xmath145 $ ] being the lt of @xmath146 \equiv 2\delta ( \tau ) \omega[\overline{\textbf{a}}(t ) ] + \gamma[\tau;\overline{\textbf{a}}(t)].\ ] ] to avoid confusion , let us mention that @xmath145 $ ] thus defined is not , of course , an angular momentum . in terms of @xmath145 $ ] , the phenomenological `` _ _ kinetic _ _ '' matrix @xmath147 $ ] appearing in eq . ( [ sigmadtirrev2 ] ) , is given by the following relation @xmath148=\mathbf{l}[z=0;\overline{\textbf{a}}(t ) ] \equiv \omega[\overline{\textbf{a}}(t ) ] + \int_0^{\infty } d\tau \gamma[\tau;\overline{\textbf{a}}(t ) ] , \label{okcssppp}\ ] ] which extends to non - equilibrium conditions the well - known kubo formula . the exact determination of @xmath149 $ ] is perhaps impossible except in specific cases or limits ; otherwise one must resort to approximations . these may have the form of a closure relation expressing @xmath150 $ ] in terms of the two - time correlation matrix @xmath151 itself , giving rise to a self - consistent system of equations , as we illustrate in the application below . these general and abstract concepts have specific and concrete manifestations , which we now discuss in the particular context of the description of non - equilibrium diffusive processes in colloidal dispersions . for this , let us identify the abstract state variables @xmath152 with the number concentration @xmath153 of particles with orientation @xmath154 in the @xmath155th cell of an imaginary partitioning of the volume occupied by the liquid in @xmath156 cells of volume @xmath157 . in the continuum limit , the components of the state vector @xmath158 then become the microscopic local concentration profile @xmath159 defined in eq . ( [ localcon ] ) and the _ fundamental thermodynamic relation _ @xmath160 $ ] ( which assigns a value of the entropy @xmath128 to any point * a * of the thermodynamic state space @xcite ) becomes the functional dependence @xmath161 $ ] of the entropy ( or equivalently , of the free energy ) on the local concentration profile @xmath159 . using this identification in eqs . ( [ releq0 ] ) and ( [ sigmadtirrev2 ] ) leads to the time evolution equations for the mean value @xmath162 and for the covariance @xmath163 of the fluctuations @xmath164 of the local concentration profile @xmath159 . these two equations are the non - spherical extensions of eqs . ( 3.6 ) and ( 3.8 ) of ref . @xcite , which are coupled between them through two ( translational and rotational ) local mobility functions , @xmath165 and @xmath166 , which in their turn , can be written approximately in terms of the two - time correlation function @xmath167 . a set of well - defined approximations on the memory function of @xmath168 , which extends to non - spherical particles those described in ref . @xcite in the context of spherical particles , results in the referred ne - scgle theory . rather than discussing these general ne - scgle equations , let us now write them explicitly as they apply to a more specific ( but still generic ) phenomenon , namely , to a glass - forming liquid of non - spherical particles subjected to a programmed cooling while _ constrained _ to remain spatially _ homogeneous _ and _ isotropic _ with fixed number density @xmath169 . thus , rather than solving the time - evolution equation for @xmath170 , we have that @xmath171 now becomes a control parameter . as a result , we only have to solve the time - evolution equation for the covariance @xmath172 . furthermore , let us only consider the simplest cooling protocol , namely , the instantaneous temperature quench at @xmath173 from an arbitrary initial temperature @xmath174 to a final value @xmath175 . at this point let us notice that it is actually more practical to identify the abstract vector @xmath176^{\dagger } $ ] of state variables not with the local concentration @xmath170 itself , but with _ only one _ of its tensorial modes , so that @xmath177 $ ] , with @xmath178 , defined in eq . ( [ nlm ] ) . under these conditions , the corresponding non - stationary covariance @xmath179 is just a scalar , denoted by @xmath6 , and defined as @xmath180 with @xmath181 . in other words , @xmath6 is a diagonal element of the matrix @xmath182 . the time - evolution equation of @xmath6 then follows from identifying all the elements of eq . ( [ sigmadtirrev2 ] ) . the first of such elements is the thermodynamic matrix @xmath121 $ ] , which in this case is also a scalar , that we shall denote by @xmath183 $ ] . it is defined in terms of the second derivative of the entropy @xmath184 $ ] ( in a contracted description in which the only explicit macroscopic variable is @xmath185 ) as @xmath186\equiv -\frac{1}{k_b}\left(\frac{d^2s[n_{lm}(\mathbf{k})]}{dn_{lm}^2(\mathbf{k})}\right ) . \label{matrixe}\ ] ] according to eq . ( [ oz ] ) , this thermodynamic property is just the inverse of the equilibrium value of @xmath187 of @xmath6 , @xmath186=1/ s_{lm}^{eq}(k ) . \label{ozlm}\ ] ] let us notice , however , that @xmath183 $ ] is not just the diagonal element of the matrix @xmath188 $ ] , defined in terms of the second partial derivative of the entropy @xmath189 $ ] ( in a _ non_-contracted description in which the explicit macroscopic variables are _ all _ the tensorial density modes @xmath185 of the microscopic one - particle density @xmath190 ) as @xmath191\equiv -\frac{1}{k_b}\left(\frac{\partial ^2s[n]}{\partial n_{lm}(\mathbf{k})\partial n_{l'm'}(\mathbf{k})}\right).\ ] ] however , according again to eq . ( [ oz ] ) , the inverse of this matrix yields the full equilibrium covariance @xmath192 , whose diagonal element @xmath193 does determine @xmath183 $ ] , according eq . ( [ ozlm ] ) . let us mention , however , that in reality @xmath183 $ ] is also a functional of the spatially non - uniform local temperature field @xmath194 . to indicate this dependence more explicitly we shall denote the thermodynamic matrix as @xmath195 $ ] . here , however , we shall impose the constraint that at any instant the system is thermally uniform , @xmath196 , and instantaneously adjusted to the reservoir temperature @xmath197 , which will then be a ( possibly time - dependent ) control parameter @xmath198 . the second element of eq . ( [ sigmadtirrev2 ] ) that we must identify is the kinetic matrix @xmath199 $ ] . for this , let us first compare the equilibrium version of eq . ( [ cdzt ] ) , namely , @xmath200 \cdot \sigma^{-1}\right\}^{-1}\cdot \sigma , \label{cdzteq}\ ] ] with its particular case in eq . ( [ flmqframe1 ] ) , in which the scalars @xmath201 and @xmath80 correspond , respectively , to @xmath202 and @xmath203 . this comparison allows us to identify @xmath204 $ ] with the scalar @xmath205 . \label{ldzeq}\ ] ] extending this identification to non - stationary conditions , we have that @xmath206=\left[{\frac{k^2d_t^0 } { 1+\delta\zeta_t^*(z;t)\lambda^{(lm)}_t(k;t)}+\frac{l(l+1)d_r^0 } { 1+\delta\zeta_r^*(z;t)\lambda^{(lm)}_r(k;t)}}\right ] , \label{ldznoneq}\ ] ] where the functions @xmath207 are defined as unity and the functions @xmath208 as @xmath209 $ ] , where @xmath210 , with @xmath211 being the position of the main peak of @xmath212 . the functions @xmath213 and @xmath214 , to be defined below , are the non - stationary versions of the functions @xmath215 , and @xmath216 . since @xmath217=\mathbf{l}[z=0;\overline{\textbf{a}}(t)]$ ] ( see eq . ( [ okcssppp ] ) ) , the general and abstract time - evolution equation in eq . ( [ sigmadtirrev2 ] ) for the non - stationary covariance becomes @xmath218 \\ & \times \left[\mathcal{e}_{lm}(k , t ) s_{lm}(k;t ) - 1 \right ] , \label{relsigmadifnonspher12 } \nonumber % \end{multiline}\end{aligned}\ ] ] where @xmath219 $ ] . in the present application to the instantaneous isochoric quench at time @xmath173 to a final temperature @xmath175 and fixed bulk density @xmath24 , this property is a constant , i.e. , for @xmath220 we have that @xmath221 = \mathcal{e}^{(f)}_{lm}(k)$ ] . in addition , in consistency with the coarse - grained limit @xmath222 in @xmath223 and @xmath224 , we have also approximated @xmath208 and @xmath207 by its @xmath225 limit @xmath226 and @xmath227 , which are actually unity . thus , the previous equation reads @xmath228 \mathcal{e}_{lm}^{(f)}(k ) \left[s_{lm}(k;t ) - 1/\mathcal{e}_{lm}^{(f)}(k ) \right ] , \label{relsigmadifnonspher13}\ ] ] where the translational and rotational time - dependent mobilities @xmath229 and @xmath230 are defined as @xmath231^{-1 } \label{btdt}\ ] ] and @xmath232^{-1}. \label{brdt}\ ] ] in terms of the non - stationary @xmath8-dependent friction functions @xmath233 and @xmath234 . in order to determine @xmath229 and @xmath230 , we adapt to non - equilibrium non - stationary conditions , the same approximations leading to eqs . ( [ zetart ] ) and ( [ zetarqq ] ) for the equilibrium friction functions @xmath235 and @xmath236 , which in the present case lead to similar approximate expressions for @xmath233 and @xmath234 , namely , @xmath237\left[1-s^{-1}_{l0}(k;t)\right]^2 f^s_{l0}(k,\tau;t)f_{l0}(k,\tau;t)\label{zetart2}\ ] ] and @xmath238 h^2_{l0}(k;t)\left[a_{l;0m}\right]^2\left[s^{-1}_{lm}(k;t)\right]^2 f^s_{lm}(k,\tau;t)f_{lm}(k,\tau;t),\label{zetarq}\ ] ] where @xmath239 are the non - stationary , @xmath8-dependent correlation functions @xmath240 , with @xmath241 being the corresponding _ self _ components . in a similar manner , the time - evolution equations for @xmath242 and @xmath243 are written , in terms of the laplace transforms @xmath244 , @xmath245 , @xmath213 , and @xmath214 , as @xmath246 @xmath247 for given specific thermodynamic functions @xmath248 $ ] , eqs . ( [ relsigmadifnonspher13])-([flmsqframenonstats ] ) constitute a closed set of equations for the non - equilibrium properties @xmath249 , @xmath250 , @xmath251 , whose solution provides the ne - scgle description of the non - stationary and non - equilibrium structural relaxation of glass - forming liquids formed by non - spherical particles . in a concrete application , these equations only require as an input the specific form of @xmath248 $ ] and of the ( arbitrary ) initial static structure factor projections @xmath252 . in the following section we illustrate the concrete application of the theory with a simple but interesting application . eqs . ( [ relsigmadifnonspher13])-([flmsqframenonstats ] ) describe the coupled translational and rotational dynamics of a brownian liquid of non - spherical particles in search of thermodynamic equilibrium after a sudden quench . a thorough application to a concrete system should then exhibit the full interplay of the translational and rotational degrees of freedom during this process . as mentioned in the introduction , however , carrying out such an exercise falls out of the scope of the present paper . instead , as an illustrative application here we discuss the solution of our resulting equations describing the irreversible evolution of the orientational dynamics of a system of strongly interacting dipoles with fixed but random positions subjected to a sudden temperature quench . for this , let us recall that two important inputs of eqs . ( [ relsigmadifnonspher13])-([flmsqframenonstats ] ) , are the short - time self - diffusion coefficients @xmath65 and @xmath253 , which describe , respectively , the short - time brownian motion of the center of mass and of the orientations of the particles . hence , _ arbitrarly _ setting @xmath254 implies that the particles are prevented from diffusing translationally in any time scale , thus remaining fixed in space . within this simplification eq . ( [ relsigmadifnonspher13 ] ) reduces to @xmath255 , \label{relsigmadifnonspher}\ ] ] whereas eqs . ( [ flmqframenonstat ] ) and ( [ flmsqframenonstats ] ) now read @xmath256 and @xmath257 also , the time - dependent translational mobility satisfies @xmath258 . hence , we only need to complement eqs . ( [ relsigmadifnonspher ] ) , ( [ flmqframenonstat2 ] ) and ( [ flmsqframenonstat2 ] ) with @xmath232^{-1}. \label{brdt2}\ ] ] and @xmath259 h^2_{l0}(k;t)\left[a_{l;0m}\right]^2\left[s^{-1}_{lm}(k;t)\right]^2 f^s_{lm}(k,\tau;t)f_{lm}(k,\tau;t),\label{zetarq2}\ ] ] where @xmath260 $ ] and @xmath261 . in the following subsections we report the simplest application of these equations . let us consider a system formed by _ n _ identical dipolar hard spheres of diameter @xmath203 bearing a point dipole of magnitude @xmath262 in their center , such that the dipolar moment of the @xmath24-th particle @xmath263 can be written as @xmath264 where the unitary vector @xmath265 describes its orientation . thus , the orientational degrees of freedom of the system , @xmath266 , are described by the set of unitary vectors @xmath267 , so that the pair potential @xmath33 between particles @xmath24 and @xmath34 is thus the sum of the radially - symmetric hard - sphere potential @xmath268 plus the dipole - dipole interaction , given by @xmath269.\label{potiential}\end{aligned}\ ] ] the state space of this system is spanned by the number density @xmath24 and the temperature @xmath197 , expressed in dimensionless form as @xmath270 $ ] and @xmath271 $ ] ( with @xmath272 being boltzmann s constant ) . from now on we shall denote @xmath270 $ ] and @xmath271 $ ] simply as @xmath24 and @xmath197 , i.e. , we shall use @xmath203 as the unit of length , and @xmath273 as the unit of temperature ; most frequently , however , we shall also refer to the hard - sphere volume fraction @xmath274 . the application of the ne - scgle equations starts with the external determination of the thermodynamic function @xmath275 . at a given state point @xmath276 the function @xmath277 can be determined using the fact that its inverse is identical to the projection @xmath278 of the _ equilibrium _ static structure factor @xmath279 at that state point . in the context of the present application , this equilibrium property will be approximated by the solution of the mean spherical approximation ( msa ) for the dipolar hard sphere ( dhs ) fluid developed by wertheim @xcite . the details involved in the determination of the resulting equilibrium static structure factor , whose only non zero projections are @xmath280 and @xmath281 , can be consulted in ref . @xcite . the equilibrium projections @xmath278 can also be used in the so - called bifurcation equations of the equilibrium theory . these are eqs . ( [ fc1])-([gar ] ) for the non - ergodicity parameters @xmath282 and @xmath283 . according to eq.([parameters ] ) , however , @xmath254 implies @xmath284 , so that in the present case we must only solve eq . ( [ gar ] ) for @xmath283 . if the solution is infinite we say that the asymptotic stationary state is ergodic , and hence , that at the point @xmath276 the system will be able to reach its thermodynamic equilibrium state . if , on the other hand , @xmath283 turns out to be finite , the system is predicted to become dynamically arrested and thus , the long time limit of @xmath6 will differ from the thermodynamic equilibrium value @xmath278 . the application of this criterion leads to the prediction that the system under consideration will equilibrate for temperatures @xmath197 above a critical value @xmath285 , whereas the system will be dynamically arrested for temperatures below @xmath286 . in this manner one can trace the dynamic arrest line @xmath287 , which for our illustrative example is presented in fig . [ diagram ] . for example , along the isochore @xmath288 , this procedure determines that @xmath289 . ) state space of the system of interacting dipoles with fixed positions . this line is the boundary between the region of ergodic states , at which the system is predicted to reach thermodynamic equilibrium , and the predicted region of dynamically arrested states . each of the two superimposed vertical dashed arrows represent the quench of the system from an initial temperature @xmath174 ( green dot ) to a final temperature @xmath175 ( blue dots ) , in one case above ( @xmath20 ) and in the other case below ( @xmath290 ) the dynamic arrest line.,title="fig:",width=336,height=336 ] we can now use the same thermodynamic function @xmath275 to go beyond the determination of the dynamic arrest line @xmath287 by solving the set of ne - scgle equations ( [ relsigmadifnonspher])-([zetarq2 ] ) to describe the rotational diffusive relaxation of our system . for this , let us notice that these equations happen to have the same mathematical structure as the ne - scgle equations that describe the translational diffusion of _ spherical _ particles ( see , e.g. , eqs . ( 2.1)-(2.6 ) of ref . although the physical meaning of these two sets of equations is totally different , their mathematical similarity allows us to implement the same method of solution described in ref . thus , we do not provide further details of the numerical protocol to solve eqs . ( [ relsigmadifnonspher])-([zetarq2 ] ) , but go directly to illustrate the resulting scenario . at this point let us notice that there are two possible classes of stationary solutions of eq . ( [ relsigmadifnonspher ] ) . the first class corresponds to the long - time asymptotic condition @xmath291 , in which the system is able to reach the thermodynamic equilibrium condition @xmath292 . equilibration is thus a sufficient condition for the stationarity of @xmath6 . it is , however , not _ a necessary _ condition . instead , according to eq . ( [ relsigmadifnonspher ] ) , another sufficient condition for stationarity is that @xmath293 . this is precisely the hallmark of dynamically - arrested states . in what follows we discuss the phenomenology predicted by the solution of eqs . ( [ relsigmadifnonspher])-([zetarq2 ] ) for each of these two mutually exclusive possibilities . let us now discuss the solution of eqs . ( [ relsigmadifnonspher])-([zetarq2 ] ) describing the non - equilibrium response of the system to an instantaneous temperature quench . for this , we assume that the system was prepared in an equilibrium state characterized by the initial value @xmath294 , of @xmath6 , and that at time @xmath173 the temperature is instantaneously quenched to a final value @xmath175 . normally one expects that , as a result , the system will eventually reach full thermodynamic equilibrium , so that the long time asymptotic limit of @xmath6 will be the equilibrium projections @xmath295 . such equilibration processes are illustrated in fig . [ ergo](a ) with an example in which the system was quenched from an initial equilibrium state at temperature @xmath296 , @xmath297 , to a final temperature @xmath298 , keeping the volume fraction constant at @xmath288 ( the first of the two quenches schematically indicated by the dashed vertical arrows of fig . [ diagram ] ) . under these conditions , and from the physical scenario predicted in fig . [ diagram ] , we should expect that the system will indeed equilibrate , so that @xmath299 . this , however , will only be true for @xmath300 and @xmath301 , since , according to eq . ( [ relsigmadifnonspher ] ) , @xmath302 must remain constant for @xmath220 , indicating that the artificially - quenched spatial structure will not evolve as a result of the temperature quench . for the same reason , eqs . ( [ flmqframenonstat2 ] ) and ( [ flmsqframenonstat2 ] ) imply that the normalized intermediate scattering functions @xmath303 and @xmath304 will be unity for all positive values of the correlation time @xmath8 and waiting time @xmath9 . for reference , the structure of the frozen positions represented by @xmath305 , is displayed in fig . [ ergo](a ) by the ( magenta ) dotted line , which clearly indicates that the fixed positions of the dipoles are strongly correlated , in contrast with a system of dipoles with purely random fixed positions , in which @xmath302 would be unity . in the same figure , the initial and final equilibrium static structure factor projections , @xmath306 and @xmath307 , are represented , respectively , by the ( red ) dashed and ( blue ) dot - dashed curves . the sequence of ( brown ) solid curves in between represents the evolution of @xmath300 with waiting time @xmath9 , as a series of snapshots corresponding to the indicated values of @xmath9 . static structure factor projection , @xmath300 , corresponding to the isochoric quench @xmath308 for @xmath288 , with @xmath296 and @xmath309 . the ( red ) dashed line is the initial structure factor @xmath310 . the ( blue ) dot - dashed line is the asymptotic limit @xmath311 . the sequence of thinner ( brown ) solid lines in between represents @xmath300 for @xmath312 . for reference we also include the non - evolving component @xmath313 , indicated by the dotted line . ( b ) snapshots of the orientational autocorrelation function @xmath314 as a function of correlation time @xmath8 ( thin brown solid lines ) , corresponding to the same isochoric quench and same sequence of waiting times @xmath9 as in ( a ) . the ( red ) dashed line represents the initial function @xmath315 and the ( blue ) dot - dashed line is the asymptotic limit @xmath316 . the inset plots the @xmath317-relaxation time , defined as @xmath318 , as a function of waiting time @xmath9.,title="fig:",width=288 ] static structure factor projection , @xmath300 , corresponding to the isochoric quench @xmath308 for @xmath288 , with @xmath296 and @xmath309 . the ( red ) dashed line is the initial structure factor @xmath310 . the ( blue ) dot - dashed line is the asymptotic limit @xmath311 . the sequence of thinner ( brown ) solid lines in between represents @xmath300 for @xmath312 . for reference we also include the non - evolving component @xmath313 , indicated by the dotted line . ( b ) snapshots of the orientational autocorrelation function @xmath314 as a function of correlation time @xmath8 ( thin brown solid lines ) , corresponding to the same isochoric quench and same sequence of waiting times @xmath9 as in ( a ) . the ( red ) dashed line represents the initial function @xmath315 and the ( blue ) dot - dashed line is the asymptotic limit @xmath316 . the inset plots the @xmath317-relaxation time , defined as @xmath318 , as a function of waiting time @xmath9.,title="fig:",width=288 ] for each snapshot of the static structure factor projections @xmath6 , the solution of eqs . ( [ relsigmadifnonspher])-([zetarq2 ] ) also determines a snapshot of each of the dynamic correlation functions @xmath250 and @xmath319 . these functions are related with other more intuitive and experimentally accessible properties , such as the time - dependent autocorrelation function @xmath320 of the normalized dipole vectors @xmath321 . in fact , since our dynamic correlators @xmath250 and @xmath322 were assumed to be described from the intermolecular @xmath323-frame @xcite , one can relate them with the time - dependent autocorrelation function @xmath324 directly through the following expression @xcite , @xmath325 let us notice that , according to eq . ( [ flmsqframenonstat2 ] ) , the three terms in the sum on the right hand side of eq . ( [ ctau ] ) , @xmath326 and @xmath327 , satisfy the same equation of motion ( which only depends explicitly on @xmath328 ) and thus , contribute exactly in the same manner to the @xmath8 and @xmath9 dependence of @xmath314 . thus , @xmath314 summarizes the irreversible time evolution of the orientational dynamics , as illustrated in fig . [ ergo](b ) with the snapshots corresponding to the same set of evolution times @xmath9 as the snapshots of @xmath300 in fig . [ ergo](a ) . we observe that @xmath314 starts from its initial equilibrium value , @xmath315 and quickly evolves with waiting time @xmath9 towards @xmath329 . this indicates that the expected equilibrium state at ( @xmath330 ) is reached without impediment and that the orientational dynamics remains ergodic at that state point . as mentioned before , the structure of eqs . ( [ relsigmadifnonspher])-([zetarq2 ] ) is the same as that of the equations in @xcite describing the _ spherical _ case . thus , one should not be surprised that the general dynamic and kinetic scenario predicted in both cases will exhibit quite similar patterns . for example , the non - equilibrium evolution described by the sequence of snapshots of @xmath314 can be summarized by the evolution of its @xmath317-relaxation time @xmath331 , defined through the condition @xmath332 . in the inset of fig . [ ergo](b ) we illustrate the saturation kinetics of the equilibration process in terms of the @xmath9-dependence of @xmath331 , as determined from the sequence of snapshots of @xmath333 displayed in the figure . clearly , after a transient stage , in which @xmath331 evolves from its initial value @xmath334 , it eventually saturates to its final equilibrium value @xmath335 . let us now present the ne - scgle description of the second class of irreversible isochoric processes , in which the system starts in an ergodic state but ends in a dynamically arrested state . for this , let us consider now the case in which the system is subjected to a sudden isochoric cooling , at fixed volume fraction @xmath288 , and from the same initial state as before , but this time to the final state point @xmath336 lying inside the region of dynamically arrested states ( the second of the two quenches schematically indicated by the dashed vertical arrows of fig . [ diagram ] ) . static structure factor projection , @xmath300 , corresponding to the isochoric quench @xmath308 for @xmath288 , with @xmath296 and @xmath337 . the ( red ) dashed line is the initial structure factor @xmath310 . the ( blue ) dot - dashed line is the ( now inaccessible ) equilibrium structure factor @xmath338 , whereas the ( black ) dotted line is the predicted asymptotic limit @xmath339 . the sequence of thinner ( brown ) solid lines in between represents @xmath300 for @xmath340 . ( b ) snapshots of the orientational autocorrelation function @xmath314 as a function of correlation time @xmath8 ( thin brown solid lines ) , corresponding to the same isochoric quench and same sequence of waiting times @xmath9 as in ( a ) . the ( red ) dashed line represents the initial function @xmath315 , the ( blue ) dot - dashed line is the expected ( but now inaccessible ) equilibrium correlation @xmath341 , and the ( black ) dotted line is the predicted long-@xmath9 asymptotic limit , @xmath342 . the inset plots the corresponding @xmath317-relaxation time as a function of waiting time @xmath9 , with the ( black ) dashed line representing the asymptotic power law @xmath343.,title="fig:",width=288 ] static structure factor projection , @xmath300 , corresponding to the isochoric quench @xmath308 for @xmath288 , with @xmath296 and @xmath337 . the ( red ) dashed line is the initial structure factor @xmath310 . the ( blue ) dot - dashed line is the ( now inaccessible ) equilibrium structure factor @xmath338 , whereas the ( black ) dotted line is the predicted asymptotic limit @xmath339 . the sequence of thinner ( brown ) solid lines in between represents @xmath300 for @xmath340 . ( b ) snapshots of the orientational autocorrelation function @xmath314 as a function of correlation time @xmath8 ( thin brown solid lines ) , corresponding to the same isochoric quench and same sequence of waiting times @xmath9 as in ( a ) . the ( red ) dashed line represents the initial function @xmath315 , the ( blue ) dot - dashed line is the expected ( but now inaccessible ) equilibrium correlation @xmath341 , and the ( black ) dotted line is the predicted long-@xmath9 asymptotic limit , @xmath342 . the inset plots the corresponding @xmath317-relaxation time as a function of waiting time @xmath9 , with the ( black ) dashed line representing the asymptotic power law @xmath343.,title="fig:",width=288 ] under such conditions , the long - time asymptotic limit of @xmath249 will no longer be the expected equilibrium static structure factor @xmath344 , but another , well - defined non - stationary structure factor @xmath345 . in fig . [ sk_arrest](a ) we illustrate this behavior with a sequence of snapshots of the non - equilibrium evolution of @xmath346 after this isochoric quench at @xmath288 from @xmath347 to @xmath348 . there we highlight the initial structure factor @xmath349 , represented by the ( red ) dashed line and the dynamically arrested long - time asymptotic limit , @xmath350 , of the non - equilibrium evolution of @xmath346 , described by the ( black ) dotted line . for reference , we also plot the expected , but inaccessible , equilibrium static structure factor @xmath351 ( blue dot - dashed line ) . finally , let us illustrate how this scenario of dynamic arrest manifests itself in the non - equilibrium evolution of the dynamics . we recall that for each snapshot of the non - stationary structure factor @xmath249 , the solution of eqs . ( [ relsigmadifnonspher])-([zetarq2 ] ) also determines a snapshot of all the dynamic properties at that waiting time @xmath9 . for example , in fig . [ sk_arrest](b ) we present the sequence of snapshots of @xmath314 , plotted as a function of correlation time @xmath8 , that corresponds to the sequence of snapshots of @xmath346 in fig [ sk_arrest](a ) . in this figure we highlight in particular the initial value @xmath315 ( red dashed line ) , the predicted non - equilibrium asymptotic limit , @xmath352 ( black dotted line ) and the inaccessible equilibrium value of @xmath341 ( blue dot - dashed line ) . notice that , in contrast with the equilibration process , in which the long - time asymptotic solution @xmath341 decays to zero within a finite relaxation time @xmath335 , in the present case @xmath341 does not decay to zero , but to a finite plateau . this arrested equilibrium correlation function , however , is completely inaccessible , since now the long-@xmath9 asymptotic limit of @xmath353 is @xmath354 , which is also a dynamically arrested function , but with a different plateau than @xmath341 . just like in the equilibration process , which starts at the same initial state , here we also observe that at @xmath173 , @xmath314 shows no trace of dynamic arrest , and that as the waiting time @xmath9 increases , the relaxation time increases as well . we can summarize this irreversible evolution of @xmath314 by exhibiting the kinetics of the @xmath317-relaxation time @xmath331 extracted from the sequence of snapshots of @xmath333 in the same figure . this is done in the inset of fig . [ sk_arrest](b ) . clearly , after the initial transient stage , in which @xmath331 increases from its initial value @xmath334 in a similar fashion as in the equilibration case , @xmath331 no longer saturates to any finite stationary value . instead , it increases with @xmath9 without bound , and actually diverges as a power law , @xmath355 , with @xmath356 . except for quantitative details , such as the specific value of this exponent , we find a remarkable general similarity between this predicted aging scenario of the dynamic arrest of our system of interacting dipoles , and the corresponding aging scenario of the structural relaxation of a soft - sphere glass - forming liquid described in ref . @xcite ( compare , for example , our fig . [ sk_arrest](b ) above , with fig . 12 of that reference ) . as said above , however , our intention in this paper is not to discuss the physics behind these similarities and these scenarios , but only to present the theoretical machinery that reveals it . thus , in summary , we have proposed the extension of the self - consistent generalized langevin equation theory for systems of non - spherical interacting particles ( ns - scgle ) , to consider general non - equilibrium conditions . the main contribution of this work consist thus in the general theoretical framework , developed in sec . [ nonequilib ] , able to describe the irreversible processes occurring in a given system after a sudden temperature quench , in which its spontaneous evolution in search of a thermodynamic equilibrium state could be interrupted by the appearance of conditions of dynamical arrest for translational or orientational ( or both ) degrees of freedom . our description consists essentially of the coarse - grained time - evolution equations for the spherical - harmonics - projections of the static structure factor of the fluid , which involves one translational and one orientational time - dependent mobility functions . these non - equilibrium mobilities , in turn , are determined from the solution of the non - equilibrium version of the scgle equations for the non - stationary dynamic properties ( the spherical - harmonics - projections of the self and collective intermediate scattering functions ) . the resulting theory is summarized by eqs . ( [ relsigmadifnonspher13])-([flmsqframenonstats ] ) which describe the irreversible processes in model liquids of non - spherical particles , within the constraint that the system remains , on the average , spatially uniform . this theoretical framework is now ready to be applied for the description of such nonequilibrium phenomena in many specific model systems . although in this paper we do not include a thorough discussion of any particular application , in section [ agingspinglass ] we illustrated the predictive capability of our resulting equations by applying them to the description of the isochoric and uniform evolution of non - equilibrium process of a simple model , namely , a dipolar hard sphere liquid with fixed random positions , after being subjected to instantaneous temperature quench . here we used this example mostly to illustrate some methodological aspects of the application of the theory , since this specific application allows us to easily implement the numerical methods described in detail in ref . the same illustrative example , however , also allows us to investigate the relevant features of the orientational dynamics during the equilibration and aging processes , but leaves open many relevant issues , such as the relationship between these predictions and the phenomenology of aging in spin - glass systems . similarly , the non - equilibrium manifestations of the coupling between translational and rotational dynamics , involved in the complete solution of eqs . ( [ relsigmadifnonspher13])-([flmsqframenonstats ] ) , will be the subject of future communications . thus , we expect that the general results derived in this paper will be the basis of a rich program of research dealing with these problems . this work was supported by the consejo nacional de ciencia y tecnologa ( conacyt , mxico ) through grants no . 242364 , no . 182132 , no . 237425 and no . l.f.e.a . also acknowledge financial support from secretara de educacin pblica ( sep , mxico ) through prodep and funding from the german academic exchange service ( daad ) through the dlr - daad programme under grant no .

josephson plasma resonance has been introduced recently as a powerful tool to probe interlayer josephson coupling in different regions of the vortex phase diagram in layered superconductors . in the liquid phase , the high temperature expansion with respect to the josephson coupling connects the josephson plasma frequency with the phase correlation function . this function , in turn , is directly related to the pair distribution function of the liquid . we develop a recipe to extract the phase and density correlation functions from the dependencies of the plasma resonance frequency and the axis conductivity on the _ ab_-component of the magnetic field at fixed _ c _ -component . using langevin dynamic simulations of two - dimensional vortex arrays we calculate density and phase correlation functions at different temperatures . calculated phase correlations describe very well the experimental angular dependence of the plasma resonance field . we also demonstrate that in the case of weak damping in the liquid phase , broadening of the jpr line is caused mainly by random josephson coupling arising from the density fluctuations of pancake vortices . in this case the jpr line has a universal shape , which is determined only by parameters of the superconductors and temperature .

cond-mat0010113

recent studies of josephson plasma resonance ( jpr ) absorption in highly anisotropic layered high - temperature and organic superconductors have proved that this tool provides unique information on josephson coupling of layers and on the structure of vortex phases in the presence of an applied magnetic field . jpr in layered superconductors was observed by applying a microwave electric field oriented along the _ c _ axis which excites charge oscillations between layers . in the superconducting state interlayer charge transfer is provided by tunneling of the cooper pairs . in the absence of a magnetic field the jpr frequency , @xmath3 , at zero temperature is determined by josephson interlayer coupling @xcite : @xmath4 where @xmath5 is the josephson current , @xmath6 is the high frequency dielectric constant for electric fields along the _ c _ axis , @xmath7 and @xmath8 are the london penetration lengths for supercurrent in the @xmath9 plane and in the @xmath1 direction respectively , @xmath10 is the anisotropy ratio and @xmath11 is the interlayer spacing . for the most investigated hts compound bi@xmath12sr@xmath12cacu@xmath12o@xmath13 ( bi-2212 ) the jpr frequency @xmath14 at low temperatures is in the range of several hundred gigahertz . at nonzero temperature , in the presence of thermally excited quasiparticles , eq . ( [ omega0 ] ) is valid as well , if the jpr frequency is well below the scattering rate of quasiparticles , because under this condition only the cooper pairs contribute to plasma oscillations @xcite . then , the jpr frequency drops with temperature proportional to @xmath15 . this allows one to observe resonance absorption as a function of temperature at a fixed frequency of uniform microwave electric field oriented along the _ c _ axis @xcite . recently the jpr frequency @xmath16 was also measured by sweeping frequency @xcite and the value @xmath17 ghz was obtained for slightly underdoped bi-2212 . the same jpr frequency was recently obtained for a bi-2212 whisker @xcite . transport measurements @xcite on the mesa fabricated from this wisker gave the critical current density @xmath18 a/@xmath19 . this allows one to extract the dielectric constant @xmath20 from eq . ( [ omega0 ] ) , @xmath21 . a dc magnetic field applied along the _ c _ axis , penetrates inside the superconductor in the form of pancake vortices @xcite . in an ideal material at low temperatures pancakes would form straight lines and have no influence on interlayer coupling . due to very weak coupling , pinning and thermal fluctuations easily misalign pancakes in different layers , which leads to suppression of effective josephson coupling @xcite and , as a consequence , to reduction of the josephson plasma resonance frequency @xcite . in real materials pancakes in neighboring layers are always misaligned due to pinning in the vortex glass phase ( below the irreversibility line ) and due to thermal disorder in the vortex liquid phase . misaligned pancakes induce a gauge - invariant phase difference @xmath22 between neighboring layers @xmath23 and @xmath24 which causes suppression of the jpr frequency according to the relation @xcite : @xmath25 where @xmath26 means thermal average and average over disorder caused by pinning centers . the jpr frequency is given by the simple expression ( [ omcos ] ) and is not sensitive to phase dynamics only if @xmath27 is much higher than the characteristic frequencies @xmath28 of interlayer phase slips induced by vortex motion @xcite . this frequency can be estimated from the _ ab _ component of static resistivity @xcite , @xmath29[1/s]@xmath30[k]@xmath31 . we will show that , more directly , @xmath28 can be extracted from both the _ c _ axis conductivity and plasma frequency . these estimates show that typically @xmath32 in the liquid region varies from @xmath33 at @xmath34 k to @xmath35 at @xmath36 k. thus , the plasma frequency is determined by a snapshot of vortex positions . for this reason jpr study provides information on the equilibrium properties of the vortex lattice though it is dynamic in nature . the plasma frequency is mainly sensitive to the average value of @xmath37 even if its local variations are very large , provided the random part of @xmath38 changes rapidly in space . in this case these rapid variations are averaged out by the smooth external oscillating electric field . space variations of @xmath39 result in mixing of phase collective modes with different momenta and frequency . such a mixing is responsible for resonance line broadening@xcite . mixing is determined by josephson coupling and is proportional to @xmath40 , where @xmath41 is the josephson length , and @xmath42 is the characteristic length of the phase correlation function @xmath43\rangle . \label{coscorr}\ ] ] averaging in space is effective for @xmath44 . this condition is fulfilled for fields @xmath45 in a vortex phase with strong disorder along the _ c _ axis , when @xmath46 . such a situation is realized in the vortex liquid phase where interlayer correlations of vortex positions are very weak and also in the bean critical state obtained by sweeping magnetic field after zero field cooling . @xmath42 is of the order of @xmath47 also in the vortex glass phase obtained by cooling in high magnetic fields above the `` second peak '' field ( @xmath48 g ) , where the three - dimensional vortex lattice is strongly disordered according to @xmath49sr @xcite and neutron scattering @xcite measurements . in this decoupled phase , interaction of pancake vortices in different layers is very weak , see discussion in ref . . therefore , the key parameter which determines field and temperature dependence of the jpr frequency is @xmath50 . this allows the use of jpr as a sensitive probe of the vortex state . the field dependence of @xmath51 allows one to observe jpr absorption as a function of magnetic field by sweeping the magnetic field @xmath52 at fixed frequency @xmath53 of applied microwave electric field . in this way the resonance magnetic field @xmath54 was obtained as a function of temperature and frequency @xmath53 , @xmath55 , see , e. g. , refs . . _ @xcite found the field and temperature dependence of @xmath51 in the vortex state of bi-2212 by measuring @xmath56 at several frequencies and using field sweeping ( fs ) after zero field cooling . they found that the resonance frequency depends nonmonotonically on temperature at fixed field and has a maximum at the irreversibility line . measuring @xmath56 in the field cooling ( fc ) and fs modes , matsuda _ et al . _ @xcite demonstrated that the plasma frequency is history dependent below the irreversibility line . recenly the josephson plasma resonance has also been measured in the vortex crystal state at small fields@xcite . field dependence of the jpr frequency in this phase is well described by the model based on the suppression of the josephson coupling due to elastic fluctuations of pancake vortices@xcite . in the vortex liquid phase , for the magnetic field @xmath57 oriented along the _ c _ axis , the field and temperature dependence of jpr frequency is approximately described as @xcite @xmath58 with @xmath59 being slightly smaller than unity ( @xmath60 ) . the influence of vortices on interlayer coupling in the vortex liquid was also probed by direct _ c_-axis transport measurements @xcite . in particular , pancake misalignment suppresses the josephson current @xmath61 as compared to its zero field value @xmath62 , as @xmath63 , so that @xmath61 is proportional to @xmath64 . the field dependence @xmath65 , similar to eq . ( [ lf ] ) was indeed found for the _ c_-axis critical current @xcite . behavior of jpr in the liquid phase described by eq . ( [ lf ] ) was explained in ref . in the framework of the high temperature expansion with respect to josephson coupling . this approach provides a general relation , connecting the average cosine factor @xmath50 with the phase correlation function @xmath66 defined by eq . ( [ coscorr ] ) , @xmath67 where thermal average @xmath26 in eq . ( [ coscorr ] ) is now done without the josephson coupling , @xmath68 is the energy of josephson interlayer coupling per unit area , @xmath69 , and @xmath70 is a dimensionless universal function of order unity . the relation ( [ cfun ] ) was derived assuming that : a ) @xmath71\rangle = 0 $ ] if the josephson coupling is absent , and b ) @xmath72 is determined by vortices only , and thus its correlation length @xmath46 . such a perturbative approach is valid if the josephson energy in the area @xmath73 is small compared with @xmath74 , i.e. @xmath75 . this condition is fulfilled in magnetic fields @xmath76 and , under such condition , one gets @xmath77 . for bi-2212 crystals with @xmath78 of the order of several hundred the high temperature expansion approach is valid practically at all fields above @xmath79 g. the same physical picture provides prediction for dependence of the jpr frequency on the in - plane field @xmath80 . in the decoupled phase at @xmath81 the in - plane field produces an additional linearly growing term in the phase difference , @xmath82 , which accounts for the phase difference due to the vector potential corresponding to @xmath80 . therefore , the dependence of @xmath83 on @xmath80 is simply determined by the fourier transform of the phase correlation function @xmath72 . it was also assumed in ref . that @xmath72 can be approximated by a gaussian , @xmath84 . in this case the dependence of jpr on the in - plane field can be presented in the analytical form : @xmath85 the exponential factor is the fourier component of @xmath86 . such dependence was found to describe fairly well the experimental angular dependencies of the resonance field @xcite . in this paper we develop a detailed theory of the josephson plasma resonance in the liquid phase based on perturbation theory with respect to the josephson coupling . this approach is justified because @xmath87 is well below unity in this state . we will extend the analysis of ref . by taking into account spin - wave type phase fluctuations in addition to the effect of vortices on jpr . in this approach @xmath88 , where @xmath89 is determined by spin - wave type phase fluctuations , while @xmath90 is the phase correlation function induced by vortices . we will calculate @xmath91 and express @xmath92 via the correlation function of vortex density in the system without josephson coupling . we show that , at temperatures above the melting point , the coupling between two dimensional pancake liquids in different layers is very weak and can be neglected . if , in addition , pinning inside the layers is neglected , then the density correlation function coincides with the density correlation function of the two - dimensional one - component coulomb plasma @xcite . accurate analysis shows that the gaussian dependence of @xmath93 used in refs . and actually exists only at @xmath94 ; at large distances @xmath95 decays exponentially , @xmath96 . this means that the dependence of jpr frequency on the in - plane field @xmath80 is also not exactly gaussian , and is given by the fourier transform of the phase correlation function , @xmath97 , @xmath98 with @xmath99 . as the function @xmath100 is directly connected to the phase and density correlation functions , this equation allows us to obtain information on the structure of the vortex state from experimental study of jpr . we calculated numerically the two - dimensional density correlation functions using langevin dynamics simulations and used these functions to calculate the phase correlation functions @xmath101 at different temperatures . influence of weak interlayer coupling and pinning is investigated perturbatively . we found that the function @xmath100 in eq . ( [ 5 ] ) calculated from numerical data gives a better description of the angular dependence of the resonance field than the naive gaussian ansatz ( [ alf ] ) . part of these results has been published in a short paper @xcite . we will investigate also the shape of the resonance line . we demonstrate that in the case of weak damping in the liquid phase the plasma resonance line acquires a universal shape due to fluctuation - induced inhomogeneities of the interlayer phase difference . this mechanism gives a natural explanation for experimentally observed asymmetric broadening of the line . the tail at high frequencies arises due to mixing of propagating plasma modes by random josephson coupling . the low frequency tail is caused by plasma modes localized in the rare large areas where the josephson coupling is suppressed due to fluctuations , similar to the well known lifshitz tail in the electron density of states in disordered semiconductors . these results were published in the short communication . in this paper we present their detailed derivations . at high frequences ( fields ) we calculate the shape of the line using perturbative expansion with respect to inhomogeneous josephson coupling . at frequencies below the resonance frequency we use the method of optimal fluctuation to calculate the exponential tail of the jpr line . the paper is organized as follows . in section ii we present general thermodynamic relations for a layered superconductor in the mixed state and discuss phase dynamics . section iii is devoted to josephson coupling in the liquid state . we obtain the high temperature expansion for the parameter @xmath102 , derive general relations connecting the jpr frequency with the pair distribution functions , and calculate the pair distribution functions numerically . we also establish a relation between the plasma frequency and the static c - axis conductivity . in section iv we calculate the shape of the jpr line in the liquid state due to the inhomogeneous broadening mechanism . we consider a layered superconductor in the vortex state . in the most part of the field - temperature phase diagram below @xmath103 the relevant degrees of freedom are the vortex coordinates @xmath104 and phase spin - wave type degrees of freedom @xmath105 . the energy functional in terms of these degrees of freedom can be written as @xcite : @xmath106 here @xmath107 is the functional which accounts for the two - dimensional energy of pancakes and their electromagnetic interaction in different layers @xcite : @xmath108}{k^{2}[1+\lambda _ { ab}^{-2}(k^{2}+q^{2})^{-1 } ] } , \label{f2}\ ] ] where @xmath109 . the contribution @xmath110 accounts for pinning : @xmath111 where @xmath112 is the pinning potential for vortices , see , e.g. , ref . . the last term is the energy of intralayer currents associated with spin - wave type of excitations and the josephson energy . it depends on the regular part of the phase difference and vortex coordinates : @xmath113 \right\}. \nonumber\end{aligned}\ ] ] here @xmath114 . the mutual inductance of layers , @xmath115 , is : @xmath116,\ \ \ l(q)= \frac{1 } { 2(1-\cos q)+s^{2}/\lambda _ { ab}^{2}}.\ ] ] the phase @xmath117 is the singular part of the phase difference induced by vortices at positions @xmath104 and @xmath118 when josephson coupling is absent ( @xmath119 ) : @xmath120 , \label{2d}\ ] ] where @xmath121 is the polar angle of the point @xmath122 . the average cosine of the total phase difference is given by the expression : @xmath123 .\end{aligned}\ ] ] here @xmath124 is an average over disorder ( pinning ) , @xmath125 is the total number of layers , and @xmath126 is the total area of the layer . time variations of the phase difference leading to excitations of quasiparticles and the deviations of the quasiparticle distribution function from equilibrium should be accounted for in the time - dependent equation for the phase difference . general equations describing phase dynamics are still under discussion and they depend on the mechanism of pairing , see refs . . at the phenomenological level , the dynamic equations for the phase difference @xmath127 can be derived as follows . introducing the in - plane superfluid momentum @xmath128 , we can express the in - plane magnetic field as @xmath129 where @xmath130 is the unit vector in z - direction , @xmath131 . using this representation the z - component of the maxwell equation @xmath132 can be written as @xmath133 where the current @xmath134 consists of the josephson and quasiparticle contributions , @xmath135 . from the in - plane component of the maxwell equation we obtain the independent equation for @xmath136 @xmath137 here @xmath138 is the high frequency dielectric constant for the in - plane electric field . we split the in - plane current into the quasiparticle and superconducting components , @xmath139 with @xmath140 . in the following we consider the case when the in - plane equilibration occurs at time scales much shorter than the inverse plasma frequency . this allows us to neglect the in - plane quasiparticle current and electric field in eq . ( [ divp ] ) . solving formally eq . ( [ divp ] ) with respect to @xmath141 and substituting it into eq . ( [ z - compmaxw ] ) we obtain @xmath142 to obtain a closed set of equations for @xmath143 , we have to connect @xmath144 with @xmath145 . in a common josephson junction these quantities are connected by the famous josephson relation @xmath146 . however , it was shown@xcite that in atomically layered superconductors this simple relation is violated due to the charging of layers . the charging effect leads to the variations of the chemical potential @xmath147 in the layer @xmath23 , and , in general , the time derivative of @xmath143 can be written as @xmath148 where @xmath147 is related with the variation of the charge @xmath149 as @xmath150 where @xmath151 is the debye screening length . the charge density is related to electric field by the poisson equation @xmath152 combining eqs.([modifjos ] ) , ( [ mu - rho ] ) , and ( [ poisson ] ) we obtain relation between @xmath144 and @xmath145 @xmath153 finally , the quasiparticle current is determined by the quasiparticle conductivity @xmath154 @xmath155 the interlayer quasiparticle conductivity in superconductors with d - wave pairing remains nonzero even at @xmath156 and it was found to be @xmath157 ( kohm cm)@xmath158 in bi-2212 crystals at 4 k and @xmath159 @xcite depending weakly on @xmath52 @xcite . eqs.([phasedyn1 ] ) , ( [ phi - e ] ) , and ( [ normj ] ) form a closed set of equations for @xmath160 . they describe phase dynamics when deviations of the quasiparticle distribution function from that at equilibrium can be neglected . deviations from equilibrium and charging effects may lead to the broadening of jpr resonance in addition to the effect of dissipation caused by quasiparticles . it is unclear at that moment how important are nonequilibrium effects because until now they were studied only in the framework of simple microscopic models @xcite , which may be unadequite for hts . as we are mainly interested in the effect of pancake vortices on jpr in the following , we will use simplified equations in which we do not account for quasiparticle conductivity @xcite and charging effects @xcite . this simplest form of equation for @xmath161 which describes phase variations at fixed vortex positions is : @xmath162 = \frac{1 } { 4\pi j_{0}}\frac{\partial } { \partial t}{\cal d } _ { z}(t),\ ] ] where @xmath163 is the homogeneous in space alternating external electric field , and we assume that @xmath117 is time independent , because characteristic frequencies of vortex motion are well below the plasma frequency as discussed above . for small amplitude variations we expand the phase difference as : @xmath164 here @xmath165 is the solution of the stationary equation . we obtain for the fourier component of the phase difference induced by the external electric field , @xmath166 , the equation @xmath167=-\frac { i\omega } { 4\pi j_{0}}{\cal d}_{z}(t ) , \label{linphasdyn}\ ] ] the homogeneous part of the averaged electric field between layers @xmath23 and @xmath168 is @xmath169 and jpr absorption is determined by the imaginary part of the inverse dielectric function , @xmath170 . expanding @xmath171 in terms of the eigenfunctions @xmath172 of the schrdinger - like equation @xmath173\psi_{\alpha n}=\frac{\omega_{\alpha } ^{2}}{\omega _ { 0}^{2}}\psi_{\alpha n } , \label{eigeneq}\ ] ] we obtain a useful expression for @xmath174 @xmath175 where @xmath176 is the average over configurations of @xmath177 . this expression will be used in sec . v to calculate the shape of the jpr line . in the vortex liquid phase we can use perturbation theory with respect to josephson coupling , which relates the cosine of the total phase difference with the phase correlation function @xmath178 as in eq . ( [ cfun ] ) . splitting the total phase difference @xmath179 into vortex , @xmath117 , and regular @xmath180 contributions , we obtain : @xmath181\rangle = \frac { e_{j}}{2t}\int d{\bf r}\langle \cos [ \varphi _ { n , n+1}({\bf r})-\varphi _ { n , n+1}(0)]\rangle= \frac{e_{j}}{2t}\int d{\bf r}s_{v}({\bf r})s_{r}({\bf r } ) , \label{cosgvgs}\ ] ] where the correlation function of phases @xmath182\rangle $ ] is determined by vortex positions , @xmath183\exp \{-(1/t)[{\cal f } _ { em}+{\cal f}_{pin}]\}\rangle _ { dis } , \nonumber \\ & & z_{v}=\int d{\bf r}_{n\nu } \exp \{-(1/t)[{\cal f}_{em}+{\cal f}_{pin}]\ } , \nonumber\end{aligned}\ ] ] and @xmath184\rangle $ ] describes the effect of spin - wave type phase fluctuations ( phase difference fluctuations in the absence of vortices ) , @xmath185\exp [ -(e_{0}/2t)\sum_{nm}l_{nm}\nabla \varphi _ { n , n+1}^{(r)}\cdot \nabla \varphi _ { m , m+1}^{(r ) } ] , \nonumber \\ & & z_{s}=\int d\varphi _ { n , n+1}^{(r)}\exp [ -(e_{0}/2t)\sum_{nm}l_{nm}\nabla \varphi _ { n , n+1}^{(r)}\cdot \nabla \varphi _ { m , m+1}^{(r ) } ] . \nonumber\end{aligned}\ ] ] in the lowest order in @xmath186 , spin - wave type of fluctuations and vortex coordinates degrees of freedom do not interact . in the gaussian approximation for spin - wave type of phase fluctuations , we obtain : @xmath187 we cut off the logarithmic divergence in the integration over @xmath188 is at @xmath189 , where @xmath190 is the superconducting correlation length . as we show below , a particular value of this cutoff has no influence on observable quantities next we express the phase correlation function @xmath92 via the density correlation function of pancake vortices induced by the @xmath57 component of the magnetic field . let us consider the phase difference between points @xmath122 and @xmath191 in the layer @xmath23 induced by pancakes with density @xmath192 . denoting the phase in the layer @xmath23 at the point @xmath122 by @xmath193 we obtain @xmath194 where @xmath195 is the angle at which the segment connecting points @xmath196 and @xmath197 is seen from the point @xmath198 . @xmath199 has a jump of @xmath200 when point @xmath201 intersects segment @xmath202 $ ] . choosing the @xmath203 axis along the segment @xmath204 $ ] , we can represent @xmath199 as @xmath205^{1/2}},\ ] ] for the phase difference @xmath206 between layers @xmath23 and @xmath24 in the presence of both components @xmath80 and @xmath57 we obtain @xmath207 , \label{vp } \\ & & \phi _ { v}({\bf r})=\int d{\bf r}[\rho _ { n}({\bf r})-\rho _ { n+1}({\bf r } ) ] \beta ( { \bf r},{\bf r } ) . \label{ph}\end{aligned}\ ] ] here @xmath208 is the vector potential for the @xmath80 component of magnetic field . we assume here that @xmath80 is large enough so that it penetrates almost freely into the sample and therefore has no influence on the phase fluctuations . in the following we will use the gaussian approximation for the random function @xmath209 , i.e. we neglect irreducible higher - order correlation functions . then @xmath210\rangle \approx \exp [ -f_{v}({\bf r})],\ \ \ f_{v}(r)=\langle \lbrack \phi _ { v}({\bf r})]\rangle ^{2}/2 . \label{appr}\ ] ] therefore the vortex - induced phase fluctuations are characterized by the the correlation function @xmath211\cos ( 2\pi sb_{x}y/\phi _ { 0 } ) . \label{gvfv}\ ] ] using eqs . ( [ ph ] ) and ( [ appr ] ) we connect @xmath212 with the following density correlation function @xmath213[\rho _ { n}(0)-\rho _ { n+1}(0)]\rangle \nonumber\\ & \equiv&k_{0}({\bf r})-k_{1}({\bf r } ) \label{dencorr } \end{aligned}\ ] ] by the relation @xmath214 where @xmath215 and @xmath216 is the deviation of vortex density from its average value @xmath217 in the layer @xmath23 at point @xmath122 . we will use the representation @xmath218 in the case of uncorrelated pancake arrangements in neighboring layers @xmath219 is the pair distribution function of the 2d liquid . using identity @xmath220 we rewrite eq . ( [ zero ] ) as @xmath221^{2}\ ] ] performing the integration over @xmath201 and averaging over orientation of @xmath222 we finally obtain @xmath223 where the universal function @xmath224 is given by @xmath225 , \ \ \ r_{1}>r , \label{jr } \\ & & j(r , r_{1})=4\pi \left [ 2rr_{1}-r_{1}^{2}-(r_{1}^{2}/2)\ln ( r / r_{1})\right ] , \ \ \ r_{1}<r . \nonumber\end{aligned}\ ] ] the function @xmath224 and its first three derivatives are continuous functions of @xmath226 and @xmath227 but the fourth derivative has jump @xmath228 at @xmath229 . ( [ fr_kr ] ) and ( [ jr ] ) represent a very important general relation connecting phase fluctuations with the vortex density fluctuations in the gaussian approximation for the pancake liquid . the function @xmath212 increases linearly at large @xmath226 , @xmath230 indicating exponential decay of phase fluctuations at large distances . here we introduced the dimensionless distance @xmath231 with @xmath232 . for small @xmath226 the function @xmath212 increases approximately quadratically with @xmath226 : @xmath233 note that @xmath234 depends on the temperature and pinning . the temperature dependence is determined by the dimensionless parameter @xmath235 if the effect of pinning is neglected . respectively , the function @xmath236 is also temperature and pinning dependent . the problem of calculating of @xmath50 , which determines the josephson plasma frequency is now reduced to calculation of the integral ( [ fr_kr ] ) with pair distribution functions of the liquid , @xmath219 . these functions are not available analytically . they can be calculated from monte carlo simulations @xcite or ( approximately ) using the density functional theory @xcite . as the first approximation we consider uncorrelated two dimensional liquids in layers . for weakly coupled superconductors the density correlations between the layers and weak pinning can be taken into account perturbatively . in the 2d approximation the vortex system is equivalent to the one - component two - dimensional coulomb plasma , which was studied extensively in the past @xcite . the pair distribution function @xmath237 within this approximation has an exact scaling property : @xmath238 i.e. , it depends on magnetic field only through the spatial scale . as follows from eq . ( [ fr_kr ] ) this scaling property is also transferred to the function @xmath212 , @xmath239 . if , in addition , fluctuations of the regular phase are neglected then at @xmath240 we obtain from eqs.@xmath241([cosgvgs],[gvfv ] ) the expression ( [ cfun ] ) for @xmath242 . correlation properties of the liquid determine the field independent universal function @xmath243 defined by eq . ( [ fexpr ] ) . therefore , the scaling property @xmath244 would be exact if we neglect ( i ) coupling between pancake liquids in different layers ( ii)pinning inside the layers , and ( iii ) regular phase fluctuations . all these effects are typically small in the pancake liquid phase . now , we further develop this approach and include fluctuations of the regular phase , magnetic interlayer coupling and pinning . first , we include the effect of regular phase fluctuations . as the temperature approaches @xmath103 the role of these fluctuations of regular phase progressively increases . the correlation function of regular phase @xmath245 is given by eq . ( [ scff ] ) . we substitute ( [ scff ] ) into eq . ( [ cosgvgs ] ) and take into account that the josephson coupling , @xmath246 , and the plasma frequency at zero field are renormalized by the phase fluctuations . their suppression is determined by the cosine factor at @xmath159 , @xmath247 , which was estimated in ref . as @xmath248 using the scaling property of @xmath95 we obtain for magnetic field along the _ c _ axis : @xmath181\rangle = f_{s}(t ) \frac{e_{0}(t)}{2t}\left ( \frac{b_{j}}{b_{z}}\right ) ^{1-\alpha } , \label{mr}\ ] ] where @xmath249 again depends only on reduced temperature @xmath235 within the 2d approximation . therefore , regular phase fluctuations reduce the power index in the field dependence of @xmath250 and make it temperature dependent . numerical calculations , described below in section [ sec - numeric ] , give @xmath251 at @xmath252 . using eqs . ( [ cosgvgs ] ) , ( [ scff ] ) and ( [ zero ] ) we obtain the jpr frequency for arbitrarily oriented magnetic field : @xmath253 where @xmath254 is the bessel function . we consider now the influence of weak interlayer magnetic coupling and weak pinning . in both cases it is necessary to calculate a correction to the density correlation function and substitute it into eq . ( [ fr_kr ] ) . the starting point of the calculation is the thermodynamic expression for the density correlation function @xmath255 here we split the electromagnetic part of the energy ( [ f2 ] ) into the intralayer two - dimensional part ( @xmath256 ) and the interlayer magnetic coupling ( @xmath257 ) part . it is convenient to represent all energy contributions in continuous form : @xmath258 with @xmath259 magnetic interaction between pancake vortices , @xmath260 , has been considered in refs . . at small distances @xmath261 a simple analytical expression can be obtained for @xmath262 @xmath263 .\ ] ] if both @xmath257 and @xmath264 are neglected then @xmath265 , with @xmath266 being the density correlation function of a 2d coulomb plasma , @xmath267 . first , we obtain the correction to the density correlation function induced by magnetic coupling , @xmath268 . expanding ( [ knmcont ] ) with respect to @xmath269 we obtain : @xmath270 interplane phase correlations are determined by the density correlation function in neighboring layers , @xmath271 ( see eq . ( [ dencorr ] ) ) . the integral ( [ deltak_magn ] ) decreases rapidly at @xmath272 , because the main contribution comes from the distances @xmath273 of the order @xmath47 and @xmath274 . for @xmath275 , the interaction potential for pancakes in neighboring layers @xmath276 increases linearly with @xmath226 , @xmath277 and the correction to @xmath278 is : @xmath279 using the scaling property of @xmath280 , we get a simple dimensional estimate of this integral @xmath281 where @xmath282 is a dimensionless function of order unity . an exact shape of this function can be obtained by numerical evaluation of the integral in eq . ( [ deltak_magn1 ] ) . using the scaling estimate ( [ deltak_scal ] ) we obtain the correction to @xmath250 given by eq . ( [ mr ] ) due to magnetic coupling , @xmath283 : @xmath284 with @xmath285 this correction is small because of the small parameter @xmath286 in atomically layered superconductors . let us turn now to the pinning - induced correction to the density correlation function , @xmath287 . expanding ( [ knmcont ] ) with respect to @xmath288 we obtain in the second order with respect to pinning potential : @xmath289 . \nonumber\end{aligned}\ ] ] this correction is determined by the quartic correlation function of density . such a perturbation approach is valid for weak pinning when the pinning energy of a single vortex , @xmath290 , is small in comparison with the temperature . an important observation following from eq . ( [ deltak_pin ] ) is that , in this temperature interval , the first pinning correction exactly coincides with the correction induced by the extra interaction potential @xmath291 . the range of this potential coincides with the correlation length of pinning , @xmath292 , that is much smaller than the intervortex spacing . it means that this potential has almost no influence on the density correlation function at @xmath293 , because the contribution to the correction comes only from the rare configurations when two particles approach at distance of the order of @xmath292 . we can conclude , therefore , that influence of pinning on the pair distribution function of the liquid is negligible at @xmath294 . this conclusion is in agreement with the more detailed study of influence of point pinning on correlation properties of the pancake liquid using the replica approach @xcite . note , that strong pinning sites with @xmath295 , such as columnar defects produced by the heavy ion irradiation , can not be treated perturbatively and may strongly influence density correlations of the pancake liquid . in the decoupled liquid phase correlations can be calculated using the pair distribution function @xmath219 of the two - dimensional vortex system . the phase correlation function @xmath296 is connected with @xmath219 by relation ( [ fr_kr ] ) . to study the behavior of phase correlations in the liquid state we performed numerical simulations of the motion of @xmath297 point vortices by numerical solution of the set of langevin equations for the vortex coordinates @xmath298 : @xmath299 here @xmath300 is the intervortex interaction force , @xmath301 is the langevin force , @xmath302 . the quantity @xmath303 , and the lattice constant @xmath304 of the ideal lattice are taken as the units of energy and length . to facilitate numerical calculations we model the real infinite - range @xmath305 interaction force by the finite - range interaction force @xmath306 , cut off at the finite length @xmath307 . to minimize the effect of cutting we have chosen the cutting length @xmath308 to reproduce the shear modulus @xmath309 for the uncut interactions ( i.e. , for @xmath310 ) . we solve the langevin equations ( [ langev ] ) at different values of @xmath74 ranging from @xmath311 to @xmath312 . simulations give a melting transition at @xmath313 , in agreement with the value of the melting point for the two - dimensional coulomb plasma @xcite . a set of pair distribution functions , @xmath219 , has been generated using averaging over a large number ( @xmath314 ) of vortex configurations ( see fig.[fig - hr ] ) . for each @xmath219 we calculated the phase correlation function , @xmath315 , integrating numerically eq . ( [ fr_kr ] ) . several phase correlation functions are presented in fig . [ fig - fr ] . note that the phase correlation function has surprisingly weak temperature dependence . to facilitate comparison with experiment , we also calculated the function @xmath316 that determines the angular dependence of the plasma frequency via eqs . ( [ result ] ) and ( [ int ] ) . this function is plotted in fig . [ fig - fb ] . we also calculated typical lengths , characterizing decay of @xmath315 : @xmath317 defined by eq . ( [ asimp1 ] ) , @xmath318 , and @xmath319 , where the last two lengths are defined by relations @xmath320 temperature dependence of the phase correlation length @xmath317 is shown in fig . [ fig - l0 ] . for comparison we also show two points extracted directly from the phase correlation functions , which were generated using monte carlo simulations of the uniformly frustrated xy model @xcite . we see that the results of the two models are in perfect agreement . we found that , in spite of considerable change in shape of @xmath219 with increasing temperature , the phase correlation function does not change much . in particular , we find that @xmath317 drops from @xmath321 to @xmath322 ( see fig . [ fig - l0 ] ) , @xmath323 , and @xmath324 , ( @xmath325 ) in the investigated temperature range . measurements of dependence of the resonance frequency @xmath3 on @xmath326 at fixed @xmath57 and @xmath74 allow one , in principle , to obtain the phase correlation function @xmath315 and the pair distribution function @xmath234 , i.e. , to extract quantitative information about correlations in the vortex liquid . using eq . ( [ result ] ) the function @xmath327 can be found from the dependence @xmath328 as : @xmath329 usually the angular dependence of the resonance field @xmath330 is measured at fixed microwave frequency @xcite rather than @xmath331 . this dependence also can be used to extract @xmath332 : @xmath333= \left[\frac{b_{rz}(\theta)}{b_{rz}(0)}\right ] ^{1-\alpha } \label{fb1}\ ] ] the function @xmath334 may be found by the reverse fourier transform from eq . ( [ int ] ) , @xmath335.\ ] ] in the final stage the density correlation function @xmath234 according to eq . ( [ fr_kr ] ) is given by the relation : @xmath336 \right\ } . \label{de}\ ] ] unfortunately it includes high order derivatives which makes the procedure very difficult for practical realization . it is more practical to compare the fourier transform of the phase correlation function @xmath332 , which can be directly extracted from the angular dependence of resonance field via eq . ( [ fb1 ] ) , with the theoretically predicted function obtained by numerical simulations ( fig . [ fig - fb ] ) . to make this comparison we used angular dependence of the resonance field for an optimally doped bi-2212 sample at 40 k from fig . 3 of ref . . we plot in fig . [ fig - fbcomp ] the function @xmath337 ( @xmath338 at this temperature can be neglected ) vs @xmath339 and the theoretical @xmath340 at @xmath341 , roughly corresponding to 40 k. one can see that the experimental @xmath332 is very close to the theoretical one . this indicates that phase fluctuations in real bi-2212 crystals are well described by the pancake liquid model . in this part we will establish a relation between @xmath51 and the _ c _ -axis conductivity @xmath342 in the vortex liquid which allows us to estimate the characteristic frequency of phase slips , @xmath32 , from experimental data for @xmath51 and @xmath342 . for the plasma frequency we obtain : @xmath343\rangle , \label{sig}\end{aligned}\ ] ] the _ c_-axis conductivity below @xmath344 consists of the quasiparticle and josephson contributions . the conductivity due to the interlayer josephson currents is finite in the liquid phase because of the phase slips induced by pancake thermal motion @xcite . when the temperature is decreased at fixed field , the system smoothly crosses over from the region of dominating quasiparticle transport to the region of dominating josephson transport . we consider the latter region , where the conductivity is related to the correlation function of the josephson currents via the kubo formula @xmath345 now the characteristic frequency @xmath32 involved in the _ c _ axis conductivity may be estimated using the relation @xmath346 if both conductivity and plasma frequency are known at a given temperature and magnetic field . in bi-2212 crystals with @xmath347 k a @xmath1-axis conductivity @xmath348 ohm@xmath158cm@xmath158 was found in ref . at @xmath350 k and @xmath351 t. jpr frequency in a crystal with similar @xmath344 is @xmath352 120 ghz at 4.2 k and @xmath353 ghz at 67 k @xcite . from these data , using eq . ( [ result ] ) we estimate @xmath354 ghz at @xmath350 k and @xmath351 t. with @xmath355 we obtain @xmath356 ghz at @xmath357 t. thus we estimate @xmath358 which justifies the static approach for calculation of plasma frequency . consider now influence of the in - plane field on the c - axis conductivity . in the lowest order in josephson coupling , we split @xmath359 $ ] into the contribution induced by pancakes and that caused by the unscreened parallel component @xmath80 . assuming that @xmath80 is along the @xmath203 axis , we obtain a simple expression for the contribution of the parallel component to the phase difference : @xmath360_{b_{x}=0}- 2\pi sb_{x}y/\phi_0 . \label{me}\end{aligned}\ ] ] inserting this expression into eqs . ( [ sig ] ) and ( [ kubo ] ) we obtain @xmath361 where @xmath362 the function @xmath363 describes the dynamics of the phase difference caused by mobile pancakes . if @xmath364 is known in the vortex liquid , the correlation function @xmath365 may be found using the inverse fourier - bessel transform : @xmath366 this procedure was realized in ref . and the function @xmath367 was found for pristine and irradiated bi-2212 crystals . to find the line shape we need an equation for the phase difference which describes time variations of the interlayer phase difference and intralayer charges in the presence of an alternating electric field applied along the _ c _ axis . we will use the simplest version of this equation ( [ linphasdyn ] ) , which is similar to the schrdinger equation . thus the problem of inhomogeneous line broadening is very similar to the problem of the density of states of a quantum particle in a random potential , and one can use the same techniques . the role of the random potential in our problem is played by @xmath368-{\cal c}$ ] . neglecting the josephson coupling , we obtain : @xmath369 where @xmath86 is the phase correlation function ( [ coscorr ] ) . at large frequencies the random potential can be treated as a small perturbation . the exact criterion will be formulated below . perturbative analysis of eq . ( [ linphasdyn ] ) can be conveniently performed using the green function formalism . the green function @xmath370 is defined as the solution of the equation : @xmath371\delta _ { nm}+\lambda _ { j}^{2}l_{nm}\nabla ^{2}\right\ } \tilde{g}_{mn^{\prime } } \left ( { \bf r } , { \bf r}^{\prime } ; e\right ) = \delta _ { nn^{\prime } } \delta \left ( { \bf r}-{\bf r}^{\prime } \right ) .\ ] ] with @xmath372 . the homogeneous part of the alternating averaged phase is connected with @xmath373 by the relation : @xmath374 where @xmath375 is the fourier component of @xmath376 . the homogeneous part of the averaged electric field between layers @xmath23 and @xmath24 is : @xmath377 therefore , the dielectric function which describes the response to a homogeneous external electric field is connected with the green function as : @xmath378 the jpr absorption and shape of the jpr line is determined by @xmath379 $ ] . the green function at @xmath381 is @xmath382 this function determines the spectrum of the phase collective mode @xmath383 . using a standard diagrammatic approach , we can write : @xmath384^{-1 } , \label{green}\ ] ] where @xmath385 is the self - energy . in the lowest order with respect to @xmath386 ( the born approximation for scattering ) we obtain : @xmath387 the imaginary part of the self - energy , @xmath388 , determines line broadening , while the real part , @xmath389 , determines the shift of the resonance center due to inhomogeneities . for @xmath390 , one can neglect the @xmath188 -dependence of @xmath391 and we obtain : @xmath392 = ( s_{0}/4\lambda _ { j}^{2})\theta ( e- { \cal c } ) , \label{imsig } \\ \sigma _ { 1}({\bf k},q;e ) & = & -\frac{s_{0}}{4\pi \lambda _ { j}^{2}}\ln \frac { \lambda _ { j}^{2}/a^{2}}{\left| e-{\cal c}\right| } , \end{aligned}\ ] ] where @xmath393 at @xmath394 , and vanishes otherwise . @xmath395 is the fourier component of the function @xmath86 at @xmath396 , i.e. @xmath397 \right\rangle .\ ] ] it is of the order @xmath398 in both the liquid phase and the vortex glass phase with strong disorder . in addition , in the liquid phase , the high temperature expansion gives the relation @xmath399 , which allows us to connect the width of the line with its center . parameter @xmath400 gives the reduced scattering rate due to the random potential . perturbative expansion with respect to the random potential works if @xmath401 . in the real units this criterion is given by @xmath402 . in spite of strong local variations of the random potential @xmath386 , which are much larger than @xmath250 , the parameter @xmath403 is much smaller than @xmath250 . the reason is that typically the _ ac _ phase @xmath404 varies smoothly in space and effectively averages out these short - range variations . the dielectric function can be now represented as @xmath405 -i\omega _ { b}^{2}\theta ( \omega ^{2}-\omega _ { p}^{2})}.\ ] ] here @xmath406 is the line width for the vortex liquid at the right - hand side of the resonance line , and @xmath407 $ ] . the shape of the resonance line in the born approximation is given by the expression : @xmath408 = \frac{1}{\epsilon _ { c } } \frac{\omega ^{2}\omega _ { b}^{2}}{\left ( \omega ^{2}-\omega _ { p}^{2}\left [ 1-s_{1}(\omega ) \right ] \right ) ^{2}+\omega _ { b}^{4}}\theta ( \omega ^{2}-\omega _ { p}^{2}),\ ] ] the born approximation gives only the right - hand side of the resonance line because it describes transformation of the homogeneous plasma mode ( with @xmath409 and @xmath410 ) into the inhomogeneous phase collective modes with nonzero momenta @xmath411 and @xmath412 of the order unity in the presence of the random coupling @xmath413 . estimate for values @xmath414 involved in the line broadening follows from the relation @xmath415 at values @xmath416 of the order unity . this means that the main part of the jpr line near its center at @xmath417 comes from averaging over distances @xmath418 in the vortex liquid state . fluctuations of potential at smaller distances result in a far right - hand wing of the jpr line . note that @xmath419 in the liquid phase is small compared with the frequency of the line center due to the small factor @xmath420 . the absorption as function magnetic field is given by @xmath421 = \frac{1}{\epsilon _ { c } } \frac{\mu b_{r}b}{\left [ b - b_{r}(1-s_{1})\right ] ^{2}+\left ( \mu b_{r}\right ) ^{2}}\theta \left ( b - b_{r}\right ) , \label{fi}\ ] ] where @xmath422 is the resonance magnetic field in the liquid phase . ( [ fi ] ) is valid at @xmath423 , and , thus the term @xmath424 in the denominator may be dropped . then the dependence of absorption on @xmath52 is proportional to @xmath425 , and this dependence is in agreement with experiment @xcite . perturbative analysis of the previous section can be improved by using the renormalized green function ( [ green ] ) in the born expression for the self energy ( [ selfenborn ] ) . this approximation , known as the self - consistent born approximation , leads to the following equation for the self - energy @xmath426 introducing dimensionless function @xmath427 and @xmath428 of the dimensionless variable @xmath429 with @xmath430 as : @xmath431 we obtain the following equations for these functions @xmath432 \label{scba1}\end{aligned}\ ] ] the resonance absorption @xmath433 can be written in terms of these functions as : @xmath434 , \label{diel_freq } \\ & & f(\zeta ) = \frac{s_{2}(\zeta ) } { [ \zeta + s_{1}(\zeta ) ] ^{2}+s_{2}^{2}(\zeta ) } . \end{aligned}\ ] ] here @xmath435 is the resonance frequency shifted by the random josephson coupling , @xmath436 ) . this negative shift is due to the second order perturbative correction to the ground state `` energy '' , similar to the well known result of quantum mechanical perturbation theory . the scaling representation ( [ diel_freq ] ) is exact , i.e. , the shape of the line is fully determined by two dimensionless functions @xmath437 and @xmath438 . ( [ scba2 ] ) and ( [ scba1 ] ) allow us to calculate these functions approximately . [ fig - scborn ] shows dependencies of the reduced jpr absorption @xmath439 and the functions @xmath440 on the reduced frequency @xmath441 . the self - consistent born approximation corresponds to summation of the perturbation series , which includes only certain class of diagrams . omited terms have the same order and this approximation does not describe quantitatively the line shape in the region of maximum and below . in particular , within this approximation @xmath442 and resonant absorption still vanish at @xmath443 . in the real situation , there is a long absorption tail due to the localized plasma modes arising from exponentially rare fluctuation suppression of coupling at large areas . this tail is very similar to the lifshitz tail in the density of states in inhomogeneous semiconductors @xcite . to calculate the exponential tail due to localized plasma modes ( left - hand side of the resonance line ) one has to use the method of optimal fluctuation@xcite . we will follow the derivations present in the book of lifshitz , pastur , and gredeskul @xcite . from now on we make the replacement @xmath444 . the resonant absorption is determined by the averaged spectral density @xmath445 , \ ] ] which can be represented as : @xmath446 where @xmath447 and @xmath448 are eigenfunction and eigenvalue of the state @xmath59 in the system with the potential @xmath449 . the probability of realization of this potential is @xmath450 $ ] . the scale of variations of typical eigenfunctions @xmath447 for not too small @xmath451 is much larger than the scale of variation of @xmath452 . this allows us to treat @xmath453 as a short ranged gaussian random variable and approximate @xmath454 as @xmath455 ^{2}}{s_{0}}.\ ] ] this is a good approximation to find the line shape not far away from the line center because the eigenvalue @xmath456 not far away from @xmath457 comes from the effective random potential @xmath452 averaged over large length scales @xmath318 . this approximation breaks down only for very large and highly improbable fluctuations of the random potential . with exponential accuracy in the tail region we have @xmath458 with @xmath459 where @xmath460 is the lowest eigenvalue at given @xmath452 . following the standard line of reasoning @xcite , which we outline in the appendix , we obtain that @xmath461 is determined by solution of the nonlinear dimensionless eigenvalue problem : @xmath462 with condition @xmath463 . @xmath461 can be found from solution of this equation as : @xmath464 with @xmath465 . numerical solution of eq . ( [ dimeq ] ) gives @xmath466 and @xmath467 . asymptotic solution at large distances corresponds to a dipole field : @xmath468 because the time derivative of the phase difference @xmath469 gives the electric field @xmath144 between layers @xmath23 and @xmath24 . this electric field of the dipole type describes plasma oscillations with low frequencies when opposite charges are distributed between layers in the regions where josephson coupling is below the average value due to fluctuations in the vortex density . from eq . ( [ phie1 ] ) we obtain @xmath470 , which corresponds to the low - frequency tail ( low - magnetic - field tail ) of the dielectric function : @xmath471 \propto \exp \left [ -7.36 \frac{e_0(\omega_p^2-\omega^2)}{t\omega_p^2}\right]\ ] ] at fixed magnetic field ( @xmath51 depends on @xmath52 ) . at fixed frequency , the imaginary part of the inverse dielectric function as a function of magnetic field is : @xmath472 \propto\exp\left(-3.68\frac { b_r - b}{\mu b_r}\right ) . \label{tailb}\ ] ] where @xmath473 and the frequency dependent resonance field @xmath474 is defined after eq . ( [ fi ] ) . in real units the coarse - grained optimum cosine fluctuation at @xmath475 is : @xmath476 thus both sides of the resonance line are characterized by the width @xmath477 in the vortex liquid phase at @xmath478 . comparing the shape of the left - hand side of the resonance line , eq . ( [ tailb ] ) , with that of the right - hand side , eq . ( [ fi ] ) , we see that the low - field side is sharper in agreement with experiment @xcite . by using eq . ( [ fi ] ) and data of matsuda _ et al . _ @xcite for the resonance line in the liquid phase at @xmath479 k , we may estimate the parameter @xmath480 , while our theoretical estimate @xmath481 gives @xmath482 at @xmath483 and @xmath484 . we note that , actually , eq . ( [ fi ] ) may be invalid near @xmath485 , and , thus , our estimate of @xmath486 from experimental data is correct by order of magnitude only . the line was found to be anisotropic with a sharper left - hand side in agreement with eqs . ( [ fi ] ) and ( [ fi1 ] ) . thus , our model gives at least qualitative explanation of the jpr line form . we have obtained a quite complete description of the jpr for the vortex liquid phase in magnetic fields @xmath81 . we related the jpr frequency in the presence of a tilted magnetic field , @xmath487 , to the correlation function of the pancake liquid . numerical results for the density correlation function of the two - dimensional coulomb plasma allow one to predict the field dependence of the plasma frequency or _ vice versa _ ; experimental data for @xmath487 allow one to obtain information on the density correlations in the pancake liquid . we have explained the line shape of the jpr in the liquid phase by inhomogeneous broadening caused by disorder in pancake positions along the _ c _ axis . it is the excitation of the phase collective modes with nonzero momenta by a homogeneous - in - space alternating electric field in the presence of an inhomogeneous vortex distribution that leads to line broadening of the jpr line at @xmath417 . rare fluctuation suppressions of coupling in large areas due pancake density fluctuations result in the left - hand side ( @xmath488 ) of the resonance line broadening . these mechanisms lead to a jpr line that agrees well with experimentally observed asymmetric line shape . due to this inhomogeneous broadening of jpr line mechanisms associated with dissipation due to quasiparticles remain hidden when measurements are made in a disordered vortex state . we thank m. gaifullin for providing data of ref . which were used in fig . [ fig - fbcomp ] . we would like to thank v. m. vinokur for stimulating discussions and i. aranson for help in numerical calculations . this work in argonne was supported by the national science foundation office of the science and technology center under contract no . dmr-91 - 20000 . and by the u. s. department of energy , bes - materials sciences , under contract no . w-31- 109-eng-38 . in los alamos this work was supported by u.s . aek also would like to acknowledge support from the japan science and technology corporation , sta fellowship 498051 , and to thank national research institute for metals for hospitality . in this appendix we derive the eigenvalue equation ( [ dimeq ] ) which determines the spectral density @xmath489 in the lifshitz tail region . using the lagrange technique we represent @xmath461 as @xmath490 , \label{phielag}\ ] ] @xmath491 is the lagrange factor , which has to be determined by the condition @xmath492 . representing @xmath493 $ ] as @xmath494\ ] ] with the hamiltonian for the `` free particle '' @xmath495 we can rewrite eq . ( [ phie ] ) as : @xmath496 \right ) . \label{phie}\ ] ] optimal @xmath452 is determined by the condition @xmath497 which gives @xmath498 and eq . ( [ phie ] ) can be represented as : @xmath499 optimal @xmath500 is determined by the equation @xmath501 applying the operator @xmath502 to this equation and neglecting small terms of the order @xmath503 we can rewrite this equation in a simpler form @xmath504 where @xmath505 . using relation @xmath506 we obtain for @xmath461 : @xmath507 introducing dimensionless variables @xmath508 we obtain the dimensionless equation ( [ dimeq ] ) . @xmath461 can be found from solution of this equation as : @xmath509 with @xmath510 . using the relation @xmath511 we obtain the final relation ( [ phie1 ] ) . o. k. c. tsui , n. p. ong , y. matsuda , y. f. yan , and j. b. peterson , phys . lett . , * 73 * , 724 ( 1994 ) ; o.k . c. tsui , n. p. ong , and j. b. peterson , phys . rev . , * 76 * , 819 ( 1996 ) ; o. k. c. tsui , proceedings for spie s photonics west 96 , symposium on oxide superconductor physics and nano - engineering ii , san jose , ca . s. l. lee , p. zimmermann , h. keller , m. warden , i. m. savi , r. schauwecker , d. zech , r. cubitt , e. m. forgan , p. h. kes , t. w. li , a. a. menovsky , and z. tarnawski , phys . * 71 * , 3862 ( 1993 ) ; c. m. aegerter , s. l. lee , h. keller , e. m. forgan , and s. h. lloyd , phys . b * 54 * , r15 661 ( 1996 ) . a. yurgens , d. winkler , n. v. zavaritsky , and t. claeson , phys . lett . * 79 * , 5122 ( 1997);a . yurgens , d. winkler , t. claeson , g. yang , i. f. g. parker , and c. e. gough phys . b * 59 * , 7196 ( 1999 ) . l. n. bulaevskii , a. e. koshelev , v. m. vinokur , and m. p. maley , phys . b * 61 * , r3819 ( 2000 ) ; a. e.koshelev and l. n. bulaevskii , _ proceeding of the 6th international conference on materials and mechanisms of superconductivity and high - temperature superconductors _ , houston , february 20 - 25 , 2000 , to be published in physica c. l. n. bulaevskii , m. zamora , d. baeriswyl , h. beck , and j. r. clem , phys . b , * 50 * , 12831 ( 1994 ) ; r. kleiner , p. mller , h. kohlstedt , n. f. pedersen , and s. sakai , phys . b , * 50 * , 3942 ( 1994 ) . n. morozov , l. krusin - elbaum , t. shibauchi , l.n . bulaevskii , m. p. maley , yu . i. latyshev , and t. yamashita , rev . lett . * 84 * , 1784 ( 2000 ) ; i. vekhter , l. n. bulaevskii , a. e. koshelev , and m. p. maley , rev . lett . * 84 * , 1296 ( 2000 ) ; n. morozov , l. n. bulaevskii , and m. p. maley , yu . i. latyshev , t. yamashita , cond - mat/0005393 , unpublished . s. sengupta , c. dasgupta , h. r. krishnamurthy , g. i. menon , and t. v. ramakrishnan , phys . lett . * 67 * , 3444 ( 1991 ) ; g. i. menon , c. dasgupta , h. r. krishnamurthy , t. v. ramakrishnan , and s. sengupta , phys . b * 54 * , 16 192 ( 1996 ) . b. i. halperin and m. lax , phys . rev . * 148 * , 722 ( 1966 ) ; j. zittartz and j. s. langer , phys . rev . * 148 * , 741 ( 1966 ) ; i. m. lifshitz , zh . eksp . teor . fiz . * 53 * , 743 ( 1967 ) [ sov . phys.jetp , * 26 * , 462 ( 1968 ) ] ; p. van mieghem rev . mod . phys . * 64 * , 755 ( 1992 ) ;

the split superconducting transition of up - spin and down - spin electrons on the background of ferromagnetism is studied within the framework of a recent model that describes the coexistence of ferromagnetism and superconductivity induced by magnetic fluctuations . it is shown that one generically expects the two transitions to be close to one another . this conclusion is discussed in relation to experimental results on urhge . it is also shown that the magnetic goldstone modes acquire an interesting structure in the superconducting phase , which can be used as an experimental tool to probe the origin of the superconductivity .

cond-mat0312378

recently , the coexistence of ferromagnetism and superconductivity has been observed in a number of materials , including , uge@xmath0,@xcite urhge,@xcite and zrzn@xmath0.@xcite the experiments so far have ascertained the presence of bulk superconductivity , and various thermodynamic and transport properties have been measured , but little is known yet about the detailed nature of the superconducting state . an obvious possibility is spin - triplet pairing induced by ferromagnetic fluctuations , but other mechanisms have also been proposed.@xcite one interesting aspects of the experiments , which may provide a clue about the origin and the nature of the pairing , is that the superconductivity is observed only on the ferromagnetic side of the magnetic phase boundary . this is in sharp contrast to early theories of superconductivity induced by ferromagnetic fluctuations , which predicted that this type of superconductivity would be equally strong on the paramagnetic side of the magnetic phase boundary.@xcite however , in a recent paper,@xcite to be referred to as i , it has been shown that the presence of magnons in the ferromagnetic phase can lead to a drastic enhancement of the superconducting @xmath1 compared to the paramagnetic phase . clearly , more properties of the superconducting state must be studied in order to differentiate between different possible pairing mechanisms . in this context , an interesting result is the recent measurement of the specific heat in urhge.@xcite taken at face value , this experiment shows a single transition into a superconducting state , and a low temperature specific heat that is linear in the temperature , which suggests that a fraction of the electrons remain in a fermi - liquid state at low temperatures . similar behavior has been observed in uge@xmath0.@xcite to date all of the theoretical work has assumed only the simplest type of superconducting order for a given pairing mechanism , with the main goal being to determine the phase boundary for superconductivity . for example , in i the present authors assumed an ordering of spin - triplet cooper pairs with spins oriented in the direction of the magnetization . let us denote the gap function for this ordering by @xmath2 . previous work on helium-3 in a magnetic field@xcite suggests that at some point , the cooper pairs with spins oriented opposite to the magnetization , characterized by a gap function @xmath3 , will form as well . more generally , the complete phase diagram for these systems will likely be complicated and involve ferromagnetism coexisting with several types of superconducting order.@xcite our goal in this paper is three - fold . first , we develop a formal theory that enables us to consistently describe ferromagnetism coexisting with two types of superconducting order . the superconductivity in our theory is caused by ferromagnetic fluctuations . our goal is to derive an equation of state , analogous to the strong - coupling or eliashberg equations of conventional superconductivity , that describes both up - spin and down - spin superconducting order , as well as a consistent magnetic equation of state in the presence of superconductivity . a major complication is that , because the superconductivity is itself caused by a fluctuation effect , a simple mean - field theory is not sufficient , and fluctuations need to be taken into account . the fluctuations that cause superconductivity within our theory are described by the spin susceptibility tensor . our second goal is therefore to develop a theory for the spin susceptibility , both in a pure ferromagnetic phase , and in coexisting ferromagnetic and superconducting phases . the pairing potential for superconductivity involving only @xmath2 and @xmath3 is given by the longitudinal susceptibility , @xmath4 . however , the transverse susceptibilities , @xmath5 , are also needed because they enter the normal self energies , and because they couple to @xmath4 via mode - mode - coupling effects . in fact , as shown in i , within the framework of our theory it is this mode - mode - coupling mechanism , which exists only in a magnetically ordered state , that causes the superconductivity to be observable only in the ferromagnetic phase . we will see that the magnons , which are described by @xmath5 , have an interesting dispersion relation in the superconducting state and , under certain conditions , can become effectively massive . a striking feature of the mass is that it is proportional to the inverse square of the magnetization , and thus singular for small magnetization . we give estimates showing that this result should be observable with neutron scattering . our third goal is to discuss the phase diagram for superconductivity and magnetism , based on the above results . in a theory that allows for pairing of both up - spin and down - spin electrons , one expects a phase diagram that is qualitatively shown in fig . [ fig:1 ] . with decreasing temperature , one schematic phase diagram in the temperature - pressure plane showing a paramagnetic phase ( pm ) , a normal conducting ferromagnetic one ( ncfm ) , a superconducting ferromagnetic phase with up - spin pairing only ( scfm i ) , and a superconducting ferromagnetic phase with both up - spin and down - spin pairing ( scfm ii ) . the two superconducting phases have been drawn to end in the same point for simplicity , but this will not necessarily be the case.,width=226 ] oberves a transition from a paramagnetic ( pm ) phase to a normal conducting ferromagnetic ( ncfm ) one . with decreasing temperature , the system enters a ferromagnetic superconducting state where only up - spin electrons are paired ( scfm i ) . finally , with further decreasing temperature , down - spin electrons are paired as well ( scfm ii ) . in general the coupled magnetic and superconducting equations of state are very complicated to solve . for the points we want to make , however , a complete solution is not needed . our aim here is to compute , for our proposed pairing mechanism , the relative magnitudes of the transition temperatures @xmath6 and @xmath7 for the phase boundaries between the ncfm phase and the scfm i phase , and between the scfm i phase and the scfm ii phase , respectively . generically we find that these transition temperatures are close to one another . experimentally this suggests that , for example , any specific heat measurement should find two closely spaced transition signatures . this result is in conflict with the naive interpretation of the specific heat measurement in urhge noted above . a possible alternative interpretation of the data consistent with our theory will be discussed below . the plan of this paper is as follows . in section [ sec : ii ] we develop a formalism that allows for a consistent description of all components of spin - triplet superconductivity , induced by ferromagnetic fluctuations , in the presence of long - range ferromagnetic order . in section [ sec : iii ] we calculate the magnetic susceptibility in the ferromagnetic state in the presence of a non - unitary superconducting order parameter . in section [ sec : iv ] we solve the strong - coupling equations for superconductivity and determine the phase diagram containing phases with pure ferromagnetic order , ferromagnetic plus spin - up superconducting order , and ferromagnetic plus both spin - up and spin - down superconducting order , respectively . we discuss our results and their experimental implications in section [ sec : v ] . in appendix [ app : a ] we augment our microscopic approach with a more general landau - ginzburg - wilson theory and discuss the soft - mode structure of the magnetic superconducting phase . in appendix [ app : b ] we relate the physical spin susceptibility to the magnetization fluctuations that occur most naturally in the theory . some of our results have been reported before in ref .. our starting point is the same model of interacting electrons as in i. that is , we consider free electrons with a static , point - like spin - triplet interaction with amplitude @xmath8 . for simplicity we ignore the spin - singlet interaction . we do _ not _ include an explicit cooper channel interaction ; the pairing interaction will be generated by magnetic fluctuations . we then decouple the spin - triplet interaction by means of a hubbard - stratonovich transformation . bilinear products of fermionic fields are constrained to a bosonic field @xmath9 by means of a lagrange multiplier field @xmath10 with a transposed @xmath11 , and we integrate out the fermions . the resulting action is given by eq . ( 2.12a ) of i , with @xmath12 , [ m , g , ] & = & ( g_0 ^ -1 + - ^t ) + & & -15pt + ( g ) - dx m(x)(x ) . [ eq:2.1 ] here @xmath13 is the bare inverse green operator , _ 0 ^ -1 = - _ + _ 0(^2/2m_e + ) , [ eq:2.2 ] where @xmath14 is the chemical potential and @xmath15 is the electron mass . we denote the real - space and imaginary - time coordinates by @xmath16 and @xmath17 , respectively , combine these into a four - vector @xmath18 , and use the notation @xmath19 , with @xmath20 and @xmath21 the system s volume and temperature , respectively . @xmath22 is the hubbard - stratonovich field whose expectation value is proportional to the magnetization @xmath23 , and @xmath24 denotes three components of a four - vector of @xmath25 matrices , ( _ 0 , ) = ( _ 3_0,_3_1 , _ 0_2,_3_3 ) , [ eq:2.3 ] with @xmath26 and @xmath27 the pauli matrices and the @xmath28 unit matrix , respectively . @xmath29 is a @xmath25 matrix field which , in contrast to @xmath30 , depends on two space - time variables . its expectation value determines the various green functions of the electron system . specifically , [ eqs:2.4 ] _ 11(x , y)&= & - _ 33(x , y ) = _ ( x)_(y ) _ + & & g_(x - y ) , [ eq:2.4a ] + _ 22(x , y)&= & - _ 44(x , y ) = _ ( x)_(y ) _ + & & g_(x - y ) , [ eq:2.4b ] are the normal green functions , while _ 13(x , y)&= & _ ( x)|_(y ) _ f^+_(x - y ) , [ eq:2.4c ] + _ 24(x , y)&= & _ ( x)|_(y ) _ f^+_(x - y ) , [ eq:2.4d ] + _ 31(x , y)&= & _ ( x)_(y ) _ f_(x - y ) , [ eq:2.4e ] + _ 42(x , y)&= & _ ( x)_(y ) _ f_(x - y ) , [ eq:2.4f ] are the anomalous ones . here @xmath31 and @xmath32 denote averages with respect to the effective action @xmath33 and the underlying fermionic action , respectively . since we do not allow for mixed pairing of up- and down - spins , which one expects to be strongly suppressed , these are the only nonzero green functions . we decompose @xmath22 and @xmath34 into their expectation values and fluctuations [ eqs:2.5 ] ( x ) = m + ( x ) , [ eq:2.5a ] + ( x , y ) = ( x , y)+ ( x , y ) . [ eq:2.5b ] here @xmath23 is the magnetization , which we assume to be in @xmath35-direction . for the lagrange multiplier field @xmath10 we write , in analogy to eq.([eq:2.5b ] ) , ( x , y ) = ( x - y ) + ( x , y ) . [ eq:2.5c ] the matrix elements of @xmath36 represent the normal self energies , [ eqs:2.6 ] _ 11(x - y)&=&-_33(x - y ) _ ( x - y ) , [ eq:2.6a ] + _ 22(x - y)&=&-_44(x - y ) _ ( x - y ) , [ eq:2.6b ] and the anomalous ones , _ 13(x - y)&&^+_(x - y ) , [ eq:2.6c ] + _ 24(x - y)&&^+_(x - y ) , [ eq:2.6d ] + _ 31(x - y)&&_(x - y ) , [ eq:2.6e ] + _ 42(x - y)&&_(x - y ) . [ eq:2.6f ] despite the formal similarities in our treatments of @xmath30 and @xmath37 on one hand , and @xmath10 on the other , it is important to note that @xmath36 is _ not _ the expectation value of @xmath10 . rather , it will be determined self - consistently by the method explained below . we now expand the action in powers of the fluctuations @xmath38 and @xmath39 , as well as the quantity @xmath40 . to this end it is useful to define an inverse green operator ^-1 = _ 0 ^ -1 + m_t_3 - ^t [ eq:2.7 ] the zeroth - order or mean - field action then reads ^(0 ) = -m^2 + ( ) + ^-1 . [ eq:2.8 ] this mean - field action or landau theory , and its generalization to a landau - ginzburg - wilson ( lgw ) theory , are interesting in their own right , and we discuss them in appendix [ app : a ] . to bilinear order in the fluctuations , it is obvious from eq . ( [ eq:2.1 ] ) that @xmath40 couples to both @xmath41 and @xmath42 . this coupling can be eliminated by first shifting @xmath42 , and then @xmath40 . the diagonalized gaussian action then takes the form ^(2)&=&- ( ^-1 ) - ( ) + & & + ( ^^-1 ^ ) . [ eq:2.9 ] here we employ an obvious scalar product notation that implies summation and integration over all discrete and continuous indices , respectively , of the various fields . @xmath43 is an inverse gaussian magnetic susceptibility defined by [ eqs:2.10 ] ^-1_ij(x - y ) = _ ij(x - y ) + _ ^0_ij(x - y ) , [ eq:2.10a ] where ^0_ij(x - y ) = ( ( y - x)_i(x - y)_j ) . [ eq:2.10b ] in appendix [ app : b ] we show that @xmath44 is directly proportional to the magnetic spin susceptibility @xmath45 in a gaussian approximation . the matrix @xmath46 , which determines both the @xmath40 and the @xmath39 propagators , is given by [ eqs:2.11 ] = ^(1 ) + ^(2 ) , [ eq:2.11a ] where ^(1)_,(xy , xy)&= & _ (x - y)_(x-y ) , + [ eq:2.11b ] + ^(2)_,(xy , xy)&= & dzdz _ ij_,,, _ ( x - z ) + & & -70pt ( _ i ) _ _ ( z - y)_ij(z - z ) _ (x-z ) + & & -70pt ( _ j)_ _ (z-y ) . [ eq:2.11c ] notice that the @xmath39 vertex in eq . ( [ eq:2.9 ] ) is minus the inverse of the @xmath40 vertex . this property will be important in what follows . contributions to the linear part of the action , @xmath47 . solid lines denote @xmath38 , dashed lines @xmath40 , and dashed - dotted lines @xmath39.,width=302 ] we now consider the non - gaussian terms in the action , which also are affected by the shifts that diagonalize @xmath48 . the terms linear in the fluctuations are represented graphically in fig . [ fig:2 ] . explicitly , we find [ eqs:2.12 ] ^(1 ) = ( a. ) + ( b. ) + ( c.^ ) , [ eq:2.12a ] with a_i = , [ eq:2.12b ] and b_(x , y)&=&^_(x - y ) - _ ( x - y ) + dxdy_ij _ _ ( x - x)(_i ) _ _ ( x-y)_ij(x-y)a_j , + [ eq:2.12c ] + c_(x , y)&=&_(x - y ) + dxdy _ b_(x,y ) ^-1_,(xy,xy ) . [ eq:2.12d ] for the cubic terms we find [ eqs:2.13 ] ^(3)&=&dxdydz_ijk_ijk(x , y , z ) m_i(x)m_j(y)m_k(z ) + dxdxdxdy_ij _ _ ^ij ( x , x;x,y ) m_i(x ) + & & m_j(x ) ^ _ ( x,y ) + ( ) , [ eq:2.13a ] the vertices read _ ijk(x , y , z ) = ( 2_)^(3/2 ) , [ eq:2.13b ] and ^ij ( x , x;x,y ) = dzdz_ _^ij(x , x;zz ) ^-1_,(zz,xy ) , [ eq:2.13c ] where ^ij ( x , x;x,y)&= & _ ^ij ( x , x;x,y ) + dzdz _ ij _ _ ijj(x , x,z ) ( _ i)__ij(z-z ) _ (z-y ) + & & _ (x-z ) , [ eq:2.13d ] with ^ij ( x , x;x,y ) = - _ _ . [ eq:2.13e ] we note that @xmath49 is the original @xmath50 vertex . via the shifts of @xmath51 and @xmath40 that diagonalize @xmath52 it generates the @xmath53 vertex @xmath54 . the ` other terms ' in eq . ( [ eq:2.13a ] ) we will not need explicitly . their structure is displayed in fig . [ fig:3 ] . contributions to the cubic part of the action , @xmath55.,width=302 ] we are now in a position to determine the magnetic and superconducting equations of state . in i this was done by means of integrating out the fluctuations and minimizing the free energy with respect to the order parameters . this procedure is not straightforward,@xcite and it becomes more confusing the more order - parameter fields one needs to consider . for our present purposes , where we have one magnetic and two superconducting order parameters , we therefore prefer a generalization of ma s method.@xcite following ma,@xcite we determine the magnetic equation of state by requiring ( ) = 0 . [ eq:2.14 ] this is a formal expression for the exact equation of state . by expanding the action in powers of the fluctuations , as we have done above , it can be evaluated order by order in a loop expansion . to one - loop order , we obtain the diagrams shown in fig . [ fig:4 ] . the magnetic equation of state to one - loop order.,width=302 ] consider the last two diagrams . the original action , eq . ( [ eq:2.1 ] ) , contained no @xmath39 legs at cubic order ; these were produced by the shift ^-1^ , [ eq:2.15 ] that decouples @xmath40 and @xmath39 . as a result , the vertices in these two diagrams , which in fig . [ fig:4 ] are denoted by @xmath56 and @xmath14 , respectively , are multiplicatively related by two factors of @xmath57 . symbolically , @xmath58 together with the relation between the @xmath40 and @xmath39 propagators that was mentioned after eq . ( [ eq:2.11c ] ) , this implies that the last two diagrams cancel each other : @xmath59 clearly , this mechanism is not restricted to these particular diagrams , but yields the following general diagram rule@xcite _ rule : _ @xmath40 loops and @xmath39 loops cancel each other . this is the reason why we did not need to explicitly determine the ` other terms ' in eq . ( [ eq:2.13a ] ) . an evaluation of the first two diagrams in fig . [ fig:4 ] yields m_i3&= & ( _ i ) + dxdy _ jk_jk(x - y ) + & & . [ eq:2.16 ] the first term on the right - hand side represents the mean - field magnetic equation of state , which was derived ( by a different method ) and discussed in i,@xcite while the second term represents one - loop fluctuation corrections . we now determine the superconducting equation of state by requiring ^(x , y)= 0 . [ eq:2.17 ] taking into account the diagram rule from the preceding subsection , this condition is graphically represented in fig . [ fig:5 ] . the superconducting equation of state to one - loop order.,width=302 ] a calculation yields [ eqs:2.18 ] _ ( x - y)&= & -dxdy_ + & & ^-1_,(xy,xy ) . [ eq:2.18a ] with @xmath44 from eq . ( [ eq:2.10a ] ) , @xmath49 from eq . ( [ eq:2.13e ] ) , @xmath46 from eqs . ( [ eqs:2.11 ] ) , and = ^- . [ eq:2.18b ] here we have used the magnetic equation of state , eq . ( [ eq:2.16 ] ) . the equations ( [ eqs:2.18 ] ) for @xmath36 must be supplemented by a relation between @xmath60 and @xmath61 . we stress again that @xmath62 , so one must _ not _ require @xmath63.@xcite rather , we calculate @xmath61 directly . going back to the underlying fermionic formulation of the action , it is easy to show that ^(x , y ) = _ , ( x , y ) , [ eq:2.19 ] where @xmath64 is @xmath60 , eq . ( [ eq:2.7 ] ) , but with @xmath23 and @xmath36 replaced by the full fluctuating fields @xmath30 and @xmath10 . in evaluating @xmath65 , we restrict ourselves to one - loop order , and also to linear order in the magnetic susceptibility @xmath44 , as contributions quadratic in @xmath44 are indistinguishable from two - loop contributions . we find ^= + o(^2,2- ) . [ eq:2.20 ] the first term in brackets in eq . ( [ eq:2.18a ] ) thus vanishes , and the second term we again evaluate to linear order in @xmath44 . we finally obtain the superconducting equation of state in the form _ ( x - y ) = _ _ ij_ij(y - x ) _ . [ eq:2.21 ] for later reference , we note the following . we are analyzing the equation of state in a loop expansion , and our treatment constitutes a systematic one - loop evaluation . however , since the superconductivity does not occur at all unless one goes to one - loop order , while the magnetism appears already at zero - loop order , the resulting equations of state are not on the same level physically . specifically , the magnetic equation of state , eq . ( [ eq:2.16 ] ) , contains fluctuation effects , while the superconducting one , eq . ( [ eq:2.21 ] ) , does not , except for the magnetic fluctuations that cause the superconductivity in the first place . in fact , the generalized eliashberg equations that follow from eq . ( [ eq:2.21 ] ) are analogous to conventional eliashberg theory , which neglects all superconducting fluctuations . writing eq . ( [ eq:2.21 ] ) explicitly yields the desired eliashberg equations for superconductivity induced by magnetic fluctuations . by using eqs.([eq:2.7 ] ) and ( [ eq:2.3 ] ) we obtain a set of coupled equations for the matrix elements of @xmath36 , eqs . ( [ eqs:2.6 ] ) , [ eqs:2.22 ] _ ( k)&=&__q _ ( q - k ) _ ( q)/d_(q ) , [ eq:2.22a ] + _ ( k)&=&__q _ ( q - k ) _ ( q)/d_(q ) , [ eq:2.22b ] + _ ( k)&=&__q _ ( q - k ) g_^-1(q)/d_(q ) + & & -35pt + 2__q g_^-1(q)/d_(q ) , + [ eq:2.22c ] + _ ( k)&=&__q _ ( q - k ) g_^-1(q)/d_(q ) + & & -35pt + 2__q g_^-1(q)/d_(q ) . + [ eq:2.22d ] the @xmath66 obey the same equation as the @xmath67 . here we have performed a fourier transform to fermionic matsubara frequencies @xmath68 and wave vectors @xmath69 , and we use the notation @xmath70 , @xmath71 . we have introduced d_(k ) = g_^-1(k)g_^-1(-k ) + _ ( k ) ^+_(k ) , [ eq:2.23 ] and the @xmath72 are the inverse ` normal ' green functions ( see eqs . ( [ eq:2.20 ] ) , ( [ eq:2.7 ] ) , and ( [ eq:2.4a ] , [ eq:2.4b ] ) , g_^-1(k ) = i_n - _ , - _ ( k ) , [ eq:2.24 ] with @xmath73 . here @xmath74 , and @xmath75 is the stoner gap or exchange splitting . finally , we have used the fact that the magnetic susceptibility tensor @xmath44 in the presence of a magnetization has the structure@xcite = _ , + & _ , - & 0 + - _ , - & _ , + & 0 + 0 & 0 & _ . [ eq:2.25 ] this structure holds in general , and in particular for the explicit approximate expression for @xmath44 given by eqs . ( [ eqs:2.10 ] ) . in a superconducting phase , with non - vanishing gap functions , @xmath44 will depend on the gap . this gives rise to a complicated feedback mechanism that is characteristic of any purely electronic mechanism for superconductivity.@xcite we will discuss this feedback in the following sections . the eliashberg equations ( [ eqs:2.22 ] ) require the magnetic susceptibility @xmath44 as input , just like the eliashberg equations for conventional superconductivity require the phonon propagator as input . there are two possible attitudes one can take at this point . in principle , one could use experimental results for @xmath44 , in analogy to experimental phonon spectra being used as input for solving the conventional eliashberg equations . sufficiently detailed information for @xmath44 , however , is not available . alternatively , one can calculate or model @xmath44 in an effort to construct a self - contained theory . this is what we will do now . for the purpose of determining the phase diagram for up - spin and down - spin superconductivity , we need @xmath44 in two phases . the up - spin superconducting @xmath1 is determined by @xmath44 in the normal conducting ferromagnetic phase , ncfm in fig . [ fig:1 ] . this was discussed in i , and we briefly recall the result in sec . [ subsec : iii.a ] . for the down - spin superconducting @xmath1 , we need @xmath44 in the phase that has both magnetic and up - spin superconducting order , scfm i in fig . [ fig:1 ] . this is discussed in sec.[subsec : iii.b ] . the gaussian theory , sec . [ subsec : ii.b ] , yields an explicit expression for @xmath44 , viz . , eqs . ( [ eqs:2.10 ] ) . for the normal conducting ferromagnetic phase , this was evaluated in i. for the transverse susceptibility tensor at small wave vectors and frequencies , the result is [ eqs:3.1 ] _ t,+(k ) & = & ( . - ) , + [ eq:3.1a ] + _ t,-(k)-1pt & = & -1pt -2pt ( -1pt . + & & . + ) . + [ eq:3.1b ] here @xmath76 with @xmath77 a bosonic matsubara frequency . @xmath78 is the mean - field distance from the magnetic critical point , with @xmath79 the density of states per spin at the fermi surface , and @xmath80 is the fermi energy . @xmath81 in our free - electron approximation , but more generally it is a number of order unity . this result displays the magnons , or magnetic goldstone modes , that are a consequence of the spontaneously broken spin rotation symmetry in a ferromagnetic phase . it provides a qualitatively correct expression for the transverse spin susceptibility in such a phase . for the longitudinal susceptibility , the gaussian approximation yields [ eqs:3.2 ] _ l(*k*,i0 ) = , [ eq:3.2a ] with @xmath82 and @xmath83 constants of @xmath84 . popular model calculations give @xmath85 and @xmath86.@xcite in contrast to the transverse channel , however , eq . ( [ eq:3.2a ] ) is _ not _ qualitatively correct . the reason is the mode - mode coupling effect that couples @xmath4 to @xmath5 and leads to a longitudinal susceptibility that diverges at @xmath87 everywhere in the ferromagnetic phase.@xcite as was discussed in i , the one - loop expression _ l^(1)(k ) & = & _ l(k)_q _ l(k ) , [ eq:3.2b ] takes this effect adequately into account . the one - loop approximation _ ( k ) = _ ^(0)(k ) + _ ^(1)(k ) , [ eq:3.2c ] thus correctly reflects the behavior of @xmath4 at small wave numbers and frequencies . we now consider the magnetic susceptibility @xmath44 in the ferromagnetic phase with @xmath88 , @xmath89 , scfm i in fig.[fig:1 ] . we first consider the gaussian approximation defined by eqs.([eqs:2.10 ] ) , and discuss the validity of this approximation later . in terms of green functions , the five nonzero matrix elements of @xmath90 read , [ eqs:3.3 ] _ 33 ^ -1(k ) & = & 1 + _ _ q\ { + & & + ( k - k ) } , [ eq:3.3a ] + _ 22 ^ -1(k ) & = & 1 + _ _ q g_11(q ) + & = & _ 11 ^ -1(k ) , [ eq:3.3b ] + _ 12 ^ -1(k ) & = & i__q g_11(q ) + & = & -_21 ^ -1(k ) . [ eq:3.3c ] we need the susceptibility only at zero external frequency , and we perform the integrals in an approximation that neglects the normal self energy as well as the frequency dependence of @xmath2 . that is , we approximate @xmath91 , with @xmath92 the @xmath35-component of the unit wave vector . it is convenient to do the summation over frequencies first , and to replace the wave number summation by an integral over @xmath93 . the calculations are very similar to those of susceptibilities in an s - wave superconductor.@xcite one readily finds that , as in the case of s - wave superconductors , @xmath94 is independent of the gap . for @xmath95 in gaussian approximation , eq.([eq:3.2a ] ) is therefore still valid . in the transverse channel the situation is more complicated . it is illustrative to first consider the case of zero external frequency and wave number . from eq . ( [ eq:3.3b ] ) , and using the magnetic equation of state , eq . ( [ eq:2.16 ] ) , in zero - loop approximation ( i.e. , neglecting the second term on the right - hand side ) , we have _ 22 ^ -1(k=0 ) & = & 1 + 2__q g_11(q ) g_22(q ) + & & -50pt = - _ q _ [ eq:3.4 ] we see that , within the framework of our approximations , the transverse susceptibility has a mass proportional to @xmath96 . for small values of the frequency , the wave number , and @xmath2 , we thus obtain , instead of eqs . ( [ eqs:3.1 ] ) , [ eqs:3.5 ] _ , + ( k , i ) & = & ( - ( - ) ) , [ eq:3.5a ] + _ , -(k , i)&= & ( + ( - ) ) . [ eq:3.5b ] here g(/2 t ) = _ q _ ( q)^2g_(q)^2 g_(q ) . [ eq:3.5c ] in the limit @xmath97 we obtain g(y ) = _ 0^ dx ( xy ) = y & , + ay^2 & , [ eq:3.6 ] where @xmath98 . the above results have been obtained in an approximation that neglects all fluctuations of the superconducting order parameter , see the remark at the end of sec . [ subsec : ii.d ] ) . this raises the question whether the mass is generic , or a result of our approximations . to answer this question we consider , in appendix [ app : a.2 ] , a lgw theory that treats magnetic and superconducting fluctuations on equal footing . both a general symmetry analysis and an explicit calculation within the framework of the lgw theory show that the result obtained above is correct for all wave numbers only in the limit @xmath99 , where @xmath100 and @xmath101 is the superconducting and magnetic ( zero - temperature ) coherence length , respectively . for a finite value of this ratio , @xmath5 as a function of the wave number displays a shoulder , but still diverges for asymptotically small values of @xmath102 . we conclude that , first , the presence of @xmath2 qualitatively changes the dispersion relation of the transverse magnons , and second , the eqs . ( [ eqs:3.5 ] ) constitute an upper bound on this change , which becomes exact in the limit @xmath103 . we will discuss this point further in sec . [ sec : v ] below . equations ( [ eqs:3.2 ] ) for the longitudinal susceptibility in one - loop approximation remain valid . this expression , with @xmath104 from eqs . ( [ eqs:3.5 ] ) , and used in the eliashberg equations , will provide an upper limit for the effect of @xmath2 on @xmath7 . we now are in a position where we can solve the gap equations ( [ eqs:2.22 ] ) for the down - spin superconducting critical temperature , @xmath7 . we will do so in an approximation that is analogous to the one employed in i for @xmath6 . we start by linearizing the gap equations in @xmath3,@xcite while keeping the full dependence on @xmath2 . equations ( [ eqs:2.22 ] ) become [ eqs:4.1 ] _ ( k)&=&__q _ ( q - k ) _ ( q)g_(q)^2 , [ eq:4.1a ] + _ ( k)&=&__q _ ( q - k ) g_(-q ) + & & -35pt + 2__q g_^-1(q)/d_(q ) , + [ eq:4.1b ] + _ ( k)&=&__q _ ( q - k ) _ ( q)/d_(q ) , [ eq:4.1c ] + _ ( k)&=&__q _ ( q - k ) g_^-1(q)/d_(q ) + & & -35pt + 2__q g_(-q ) , + [ eq:4.1d ] notice that the down - spin and up - spin equations are coupled in two ways ; directly via the explicit dependence of the transverse spin - fluctuation contribution to @xmath105 on @xmath2 , and indirectly via the dependence of the @xmath5 on @xmath2 . the former effect means that in the second contribution to @xmath105 , the wave number is not strictly pinned to the up - spin fermi surface . however , this effect is only on the order of @xmath106 , and we neglect it . the up - spin and down - spin equations then formally decouple , except for the dependence of @xmath44 on @xmath2 . we can then solve for @xmath7 , in terms of @xmath44 , in the same mcmillan - allen - dynes approximation that was employed in i for @xmath6 . we find t _ = t_0(t)e^-(1+d^0 _ + 2d^0_)/d^1_. [ eq:4.2 ] here @xmath107 is the same temperature scale that was used in i , viz.,@xcite t_0(t ) = t_0 , [ eq:4.3 ] with @xmath108 a microscopic temperature on the order of the fermi temperature . the coupling constants @xmath109 are given by [ eqs:4.4 ] d^1 _ & = & _ n _ _ 0 ^ 2 dx x(1-x^2/2)_(xk_,i0 ) , + [ eq:4.4a ] + d^0 _ & = & _ n _ _ 0 ^ 2 dx x_(xk_,i0 ) , [ eq:4.4b ] + d^0 _ & = & _ n _ _ y_-^y_+ dx x_,+(xk_,i0 ) , [ eq:4.4c ] with @xmath110 . here @xmath111 and @xmath112 , @xmath113 , are the fermi wave number and the density of states at the up - spin and down - spin fermi surface , respectively . for calculating @xmath1 numerically , we use the same mean - field relation between @xmath114 and @xmath23 or @xmath115 as in i , namely,@xcite [ eqs:4.5 ] t & = & 1 - ( 1 + 3 ^ 2)^1/3/(1 + ^2/3 ) , [ eq:4.5a ] + m/_n _ & = & 3(1 + ^2/3)/(1 + 3 ^ 2 ) , [ eq:4.5b ] with @xmath116 the electron number density and @xmath117 the bohr magneton . in zero - loop approximation , with @xmath44 given by eqs.([eq:3.2a ] , [ eqs:3.5 ] ) , @xmath7 for a given @xmath2 is now easy to calculate . rather than solving the up - spin eliashberg equations for @xmath2 , we have approximated @xmath118 . @xmath6 is the same as in i.@xcite our results are very similar to those obtained by fay and appel@xcite and are shown in fig.[fig:6 ] . the effect of @xmath2 on @xmath7 is so small that it is not visible on the scale of the figure . at the zero - loop level , the feedback effect is thus unobservably small . @xmath6 and @xmath7 in zero - loop ( dotted lines , right scale ) and one - loop ( solid lines , left scale ) approximation . parameter values used are @xmath119 , @xmath120.,width=359 ] at the one - loop level , with @xmath4 given by eqs . ( [ eqs:3.2 ] ) with eqs . ( [ eqs:3.5 ] ) as input , the calculation is more complicated . we have approximated the effect of the mass in @xmath5 by a lower cutoff in the frequency summation , @xmath121 . this allows to perform the wave - number integral analytically , as was described in i. as in the zero - loop calculation , we replace @xmath2 by @xmath122 . this procedure will overestimate the effect of @xmath2 on @xmath7 . the calculation then proceeds as for @xmath6 . a representative result is shown in fig.[fig:6 ] . we see that the relative difference between @xmath6 and @xmath7 in one - loop approximation is on the same order as in zero - loop approximation , viz . , at most about 10% . the theory thus predicts two superconducting transitions close to one another . in the notation of the schematic phase diagram in fig . [ fig:1 ] this means that the scfm i phase is very narrow . the above results for the critical temperatures have implications for the specific heat . the two superconducting transitions close to one another imply two features in the specific heat . at the mean - field level , the specific heat at a superconducting transition has a discontinuity , and for our choice of an order parameter the low - temperature specific heat is a quadratic function of the temperature.@xcite one therefore qualitatively expects the specific heat coefficient as a function of temperature to behave as shown in fig . [ fig:7](a ) . this should be compared to a situation where , as suggested in ref . , the down - spin electrons do not pair down to the lowest observable temperature , which would result in a specific heat coefficient as shown qualitatively in fig . [ fig:7](b ) . the data of ref . show a very broad feature , even for the best samples , which is consistent with either scenario . the prediction of the present theory is that , with further improving sample quality , the residual value of the specific heat coefficient will decrease , and the broad peak will be resolved into two discontinuities . the specific heat coefficient of urhge as observed in ref . ( circles ) . the solid line in ( a ) shows the qualitative theoretical prediction for a perfect sample based on the present analysis . the one in ( b ) would apply if the down - spin electrons remained unpaired to the lowest observable temperature . the size of the discontinuity is known theoretically only in the weak - coupling limit;@xcite for the purpose of this qualitative illustration we have adjusted it to fit the maximum in the data.,title="fig:",width=226 ] -5 mm let us further discuss several aspects of our theory and its implications . first , we consider in more detail our results regarding the magnetic susceptibility . in a normal conducting ferromagnetic phase , the transverse spin susceptibility is soft , displaying ferromagnetic magnons with a quadratic dispersion relation . the scale in the dispersion relation is set by the stoner gap or exchange splitting @xmath115 , which is proportional to the magnetization . in natural units , measuring wave numbers in units of the magnetic coherence length @xmath101 , which is on the order of the fermi wave number @xmath123 , we have @xmath124 at zero frequency , the dimensionless transverse magnetic susceptibility diverges for @xmath125 like @xmath126 for a given distance from the magnetic phase transition , @xmath127 is a number of order one whose exact value depends on the details of the model . the gaussian approximation we have employed correctly reflects these exact features , see eqs . ( [ eqs:3.1 ] ) . in the presence of an up - spin gap , the structure of the magnon changes qualitatively . the gaussian approximation for @xmath5 predicts a true gap , eqs . ( [ eqs:3.5 ] ) . this is a consequence of the fact that the gaussian approximation effectively treats the superconducting gap as a fixed external field . the lgw theory of appendix [ app : a ] shows that this does not correctly reflect the symmetry properties of the system , and that the static @xmath5 actually has the structure [ eqs:5.1 ] _ ( k ) = . [ eq:5.1a ] here @xmath128 is related to the coupling constant @xmath129 and the magnetic order parameter @xmath130 of the lgw theory by @xmath131 . this result shows that , in general , @xmath5 has a much more complicated structure than in the absence of spin - triplet superconductivity . at asymptotically small wave numbers , @xmath132 , @xmath128 drops out of the expression for @xmath5 , and we have _ ( k ) = . [ eq:5.1b ] restoring the frequency , the dispersion relation of the magnon then is @xmath133 \,{\bm k}^2\quad.\ ] ] we see that the magnon dispersion relation is substantially changed by the presence of @xmath2 as long as the condition expressed by eq.([eq : a.10 ] ) is fulfilled . namely , the magnon is much stiffer , i.e. , the frequency rises much faster with the wave number , than in a normal conducting ferromagnet . the magnitude of this effect will depend on the detailed parameters of the material in question , but it clearly can be substantial : assuming a value of @xmath2 on the order of the superconducting @xmath1 , or @xmath134 , a value of @xmath115 on the order of 10 times the magnetic @xmath1,@xcite or @xmath135 , and @xmath136 , we find that the prefactor in the dispersion relation is enhance by a factor of 100 over its value in the normal conducting phase . for larger wave numbers , @xmath137 , we have _ ( k ) . [ eq:5.1c ] using eqs . ( [ eqs : a.9 ] ) , we have the correspondence @xmath138 , so we recover the result of the gaussian theory , eqs.([eqs:3.5 ] ) . as long as eq . ( [ eq : a.10 ] ) is valid , @xmath5 thus displays a pronounced plateau as a function of the wave number , followed by a very steep increase at asymptotically small wave numbers . the gaussian theory approximates this behavior by a true gap , ignoring the asymptotic regime . in the limit @xmath99 the asymptotic regime shrinks to zero , and the gaussian theory becomes qualitatively correct . however , in this context one should note that the lgw theory of appendix [ app : a ] is valid only for wave numbers @xmath139 . within the lgw theory , the shoulder will lie in that regime provided the coupling constant @xmath129 is sufficiently small . while the above identification of @xmath129 with the parameters of the microscopic theory makes @xmath129 effectively of order unity , given the initial sharp rise of the magnon frequency with the wave number in the regime where the lgw is valid , a pronounced shoulder in the dispersion relation at intermediate wave numbers is inevitable . the wave number region where one expects this shoulder is given by g_^-2 < k^2 < g(_/)^2_m^-2 . [ eq:5.2 ] if we use the same numbers as above , and @xmath140 , @xmath141 , the upper limit of this wave - number range is given by @xmath142 . this is a factor of 3 below the smallest wave numbers currently observable with neutron scattering.@xcite however , in materials with smaller values of the exchange splitting @xmath115 the plateau should be in an observable regime . second , we add some comments about our prediction of two superconducting transitions that are separated only by a small temperature interval . within our model , we have found this prediction to be very robust , especially given that all of our approximations have a tendency to overestimate the suppression of @xmath7 compared to @xmath6 . even an ( artificial ) increase of the gap in the transverse susceptibility by a factor of 10 does not make a visible change in fig . [ fig:6 ] . this reflects the fact that the up - spin and down - spin pairing are mediated by the same effective potential , viz . , @xmath4 , see eq . ( [ eq:4.4a ] ) and the corresponding eq . ( 3.14a ) in i. while @xmath4 is modified by @xmath2 , the effect is not sufficiently large to lead to a substantial separation of @xmath6 and @xmath7 . this in turn means that the lower value of @xmath7 is overwhelmingly due to the lower value of the density of states at the down - spin fermi level . in this context we need to keep in mind that we have used a free - electron model with parabolic bands . a complicated band structure could lead to a drastically reduced value of @xmath143 , which in turn would lead to a much lower value of @xmath7 . if experiments on samples of improved quality should fail to show two transitions , this would be the most likely explanation . this would be of interest also with regard to distinguishing between the two proposed explanations for why the superconductivity is observed in the ferromagnetic phase only : sandeman et al.@xcite have proposed a mechanism based on an intricate structure of the density of states , while the explanation proposed in i is based on properties of the magnetic susceptibility . third , we come back to our discussion of the specific heat . the experiment of ref . shows that the observed superconductivity is indeed a bulk effect . what is not clear _ a priori _ is the origin of the large residual value of the specific heat coefficient . while the down - spin electrons remaining unpaired , as was suggested in ref . , is a possibility , normal - conducting regions within the sample would have the same effect and would also lead to the observed smearing of the discontinuity in the specific heat . with increasing sample quality , the discontinuity should become sharper , and a crucial question will be whether the residual value drops correspondingly . of course , the emergence of the predicted double feature from the narrowing peak will be the most direct test of our predictions regarding the specific heat . in this context it is interesting to note that such a split transition , with two closely spaced discontinuities in the specific heat , has been observed in upt@xmath144,@xcite but only after a long period of gradually increasing sample quality . ( upt@xmath144 is not ferromagnetic , though , and the split transition has a physical origin that is very different from what we have discussed . ) even in the best samples that show the split transition , however , there is a substantial residual specific heat coefficient , the origin of which is not quite clear . in a ferromagnetic superconductor , one also has to keep in mind that the ground state will not be homogeneous , due to the formation of a spontaneous vortex state.@xcite normal electrons in the vortex cores are a possible source of a residual specific heat coefficient . this effect has been neglected in the current theory and will be pursued in a future publication . fourth , we briefly comment on the fact that the mass , or pseudo - mass , induced in the transverse magnetic susceptibility by the superconductivity , is a singular function of the magnetization . this can be seen in eqs.([eqs:3.5 ] ) , and it also leads to the factor of @xmath145 in the relation between the phenomenological coupling constant @xmath129 in the lgw theory and the microscopic parameters , eq . ( [ eq : a.9c ] ) . this behavior can be traced back to the behavior of the integral in eq . ( [ eq:3.5c ] ) , and thus ultimately to the green functions and the soft particle - hole excitations that are characteristic for itinerant electron systems . the singularity is therefore a result of a coupling between the particle - hole excitations and the magnetic and superconducting goldstone modes . it is very similar in nature to , e.g. , the anomalous magnetization dependence of the magnon stiffness in a normal conducting ferromagnetic phase that was discussed in ref .. in conclusion , we summarize our results . we have presented a consistent and self - contained theory for the coexistence of superconductivity and ferromagnetism in itinerant electron systems . we have presented a field - theoretic formulation of this problem that allows for the determination of both the magnetic and the superconducting equation of state in a systematic loop expansion . this method , which utilizes ma s procedure for generating equations of state , remedies some shortcomings of the earlier theory presented in i , which relied on a minimization of the free energy . the self - contained character of the theory is achieved by means of explicit expressions for the magnetic susceptibility , which is needed as input for the generalized eliashberg equations . these expressions have been evaluated in ferromagnetic phases , both normal conducting and superconducting ones . this is a generalization of the theory for the magnetic susceptibility that was developed in i , and it explicitly takes into account the feedback effects that are characteristic for any purely electronic mechanism for superconductivity . we have explicitly evaluated this theory to one - loop order , and for a model that allows for two components of the superconducting order parameter , one each for cooper pairs consisting of up - spin electrons and down - spin electrons , respectively . the limitations of this one - loop approximation , which neglects superconducting fluctuations and uses a zero - loop expression for the magnetic susceptibility , have been discussed by means of a phenomenological lgw theory that complements our microscopic theory . we have found that , for generic parameter values , the two superconducting transitions that describe the pairing of up - spin and down - spin electrons , respectively , occur close to one another , with transition temperatures that typically differ by only on the order of 10% . this suggests that , if the superconductivity observed in urhge is indeed of a spin - triplet p - wave type mediated by ferromagnetic fluctuations , then the broad feature observed in the specific heat near the temperature of the resistive transition@xcite should contain two transitions that are close together . if samples of improved quality should show only one sharp discontinuity in the specific heat , then this would be a strong argument against the type of pairing we have assumed in this paper . a caveat is provided by our assumption of parabolic bands , however , as discussed above . the presence of spin - triplet superconductivity has further been shown to drastically change the structure of the dispersion relation of the ferromagnetic magnons . in materials with a small exchange splitting this effect is observable with neutron scattering , and can also be used as a probe for the nature of the superconductivity . we would like to thank meigan aronson , j. david cohen , and john toner for helpful discussions . we are indebted to thomas vojta for insisting that the question of goldstone modes warranted a more detailed discussion than we initially were prepared to give . this work was supported by the nsf grant under nos . dmr-01 - 32555 and dmr-01 - 32726 . in this appendix we discuss the mean - field action @xmath146 , eq.([eq:2.8 ] ) , or the corresponding landau free energy density @xmath147 , and its generalization to a phenomenological lgw theory . from eq . ( [ eq:2.8 ] ) we have f^(0 ) = m^2 - ( ) - ^-1 . [ eq : a.1 ] in what follows we neglect the normal self - energy contribution to the matrix @xmath36 , eqs . ( [ eqs:2.6 ] ) , which lead only to a trivial renormalization of the normal green function . an expansion in powers of the order parameters @xmath23 and @xmath36 then yields f^(0 ) & = & ( ^(^ - _ 0/2 ) ) - ( ^_0)^2 + & & -40pt + tm^2 - m(_0_3 _ 0^_0^ ) + [ eq : a.2 ] here @xmath78 is the mean - field distance from the magnetic critical point , and we have omitted cubic terms of order @xmath148 , @xmath149 , and @xmath150 , as well as all quartic terms . this mean - field free energy does not describe a superconducting transition ; this requires magnetic loops , as we have seen in the main text . the magnetic part , however , agrees with an expansion of the mean - field magnetic equation of state , eq . ( [ eq:2.14 ] ) . in particular , the cubic term of @xmath151 shown in eq . ( [ eq : a.2 ] ) has important consequences . for a nonzero superconducting order parameter @xmath2 , eq . ( [ eqs:2.6 ] ) , the quantity [ eqs : a.3 ] h _ = ( _ 0_3 _ 0^_0^ ) [ eq : a.3a ] acts like an effective external magnetic field . performing the trace , we have h _ = -__k _ ( k ) ^2_0(k)^2_0(k ) . [ eq : a.3b ] general arguments@xcite show that such a term in the free energy leads to a mass in the magnetic goldstone mode , i.e. , in the transverse spin susceptibility , _ ^-1(k=0 ) = h_/m . [ eq : a.4 ] if we expand eq . ( [ eq:3.4 ] ) in powers of @xmath2 and @xmath23 , we see that this is indeed the same result we obtained in sec . [ sec : iii ] by a direct calculation of the susceptibility . the current derivation makes it obvious that this result has been obtained while neglecting fluctuations . we consider the influence of fluctuations in the next subsection . the conclusion in the previous subsection , namely , that the transverse spin susceptibility is massive , can not be strictly correct . even in the presence of superconducting order , one expects the system to be invariant under rotations of all spins , and this symmetry must manifests itself via a goldstone mode to which the magnetic susceptibility must couple . to investigate this point , we consider the following phenomenological action,@xcite s[m , ] & = & dx + dx + & & -igdx m(x)((x)^*(x ) ) . [ eq : a.5 ] here @xmath152 and @xmath153 are the magnetic and the superconducting order parameter , respectively , and the latter has been represented as a complex 3-vector in spin space.@xcite @xmath154 and @xmath155 are the dimensionless distances from the magnetic and superconducting critical point , and @xmath101 and @xmath100 are the ( zero - temperature ) magnetic and superconducting coherence lengths , respectively . @xmath129 is a phenomenological coupling constant . at the mean - field level , @xmath156 , and @xmath157 , with @xmath158 and @xmath159 , and the theory is seen to have the structure of the landau theory derived from the miscroscopic theory in sec . [ app : a.1 ] above . this action is invariant under a symmetry group so(3)@xmath160u(1 ) , which represents co - rotations of the vectors @xmath152 and @xmath153 , and in addition a gauge transformation of the complex field @xmath153 . in the ordered state given above , this symmetry is spontaneously broken , and the ordered state is invariant only under rotations about the 3-axis . we thus expect dim(so(3)@xmath160u(1)/so(2 ) ) = 3 goldstone modes . an explicit calculation , involving an expansion to gaussian order about the ordered state , confirms this expectation . of the three soft modes , two are spin wave - like modes , namely , superpositions of the transverse components of @xmath152 and the 3-component of @xmath153 , while the third is an anderson - bogoliubov - like mode that reflects the broken gauge symmetry . specifically , the transverse @xmath161-susceptibility has an overlap with one of the spin - wave goldstone modes . the gaussian theory yields _ ( p ) & & m_1(p)m_1(-p ) + & = & [ eq : a.6 ] from this result we see that , in the momentum range m_^-2 < p^2 < g(^2/m)_m^-2 , [ eq : a.7 ] the transverse susceptibility is effectively massive , _ ( p ) /2g^2 . [ eq : a.8 ] @xmath130 and @xmath162 can be related to @xmath23 and @xmath2 , respectively , in the microscopic theory by comparing eq.([eq : a.5 ] ) with the landau theory in sec . [ app : a.1 ] above . one finds the correspondence [ eqs : a.9 ] m , [ eq : a.9a ] + _ . [ eq : a.9b ] @xmath129 does not corresponds to a simple constant in the microscopic theory , but rather to a function of @xmath23 ( or @xmath163 ) and the function @xmath164 in eq . ( [ eq:3.5c ] ) , g(/2 t ) . [ eq : a.9c ] this complicated behavior of the effective coupling between the magnetic and superconducting order parameters is not reflected by the lgw theory . it is a result of the itinerant nature of the electrons , and the corresponding soft particle - hole excitations , as is discussed in sec . [ sec : v ] . upon substituting eqs . ( [ eqs : a.9 ] ) in eq . ( [ eq : a.8 ] ) we see that the latter is the same result as eq . ( [ eq:3.5a ] ) for the magnetic susceptibility . the momentum range given by eq . ( [ eq : a.7 ] ) exists only if the condition @xmath165 is fulfilled . in terms of the stoner gap @xmath163 and @xmath2 this condition takes the form / _ _ /_m . [ eq : a.10 ] these results are further discussed in sec . [ sec : v ] . in this appendix we establish the relation between the physical spin susceptibility @xmath45 , as measured , e.g. , by neutron scattering , and the quantity @xmath44 that emerges as the @xmath38 propagator of our field theory , see eqs . ( [ eq:2.9 ] , [ eqs:2.10 ] ) . we start by adding a magnetic field @xmath166 to our action that couples linearly to the electron spin density and acts as a source field for spin - density correlation functions . the action , eq . ( [ eq:2.1 ] ) , becomes [ m , g , ; ] & = & ( ) - dx m(x)(x ) + & & -75pt + ( g_0 ^ -1 + - ^t + ) . [ eq : b.1 ] the partition function , z [ ] = d [ , , ] e^a[m , g , ; ] , [ eq : b.2 ] then serves as the generating functional for the magnetization , m = n_^3(x)= _ = 0 z[h ] , [ eq : b.3 ] and for the spin susceptibility , _ ^ij(x - y ) = n_^i(x ) n_^j(y ) = _ = 0 -5pt z[h ] , [ eq : b.4 ] where @xmath167 is the electron spin density , and @xmath168 . the simplest way to deal with the source field is to shift @xmath30 : @xmath169 . we then find m = m_3(x ) , [ eq : b.5 ] in agreement with eq . ( [ eq:2.5a ] ) . an evaluation of @xmath170 to one - loop order yields the magnetic equation of state , eq . ( [ eq:2.14 ] ) . similarly , _ ^ij(x - y ) = m_i(x)m_j(y ) - _ ij(x - y ) . [ eq : b.6 ] if we evaluate the @xmath171-correlation function in gaussian approximation , we find@xcite _ ^ij(x - y ) /_. [ eq : b.7 ] we see that the @xmath171-propagator of the field theory is simply related to the physical spin susceptibility . in particular , the transverse part of @xmath45 is massive if and only if that of @xmath44 is . for the fourier transform of the longitudinal part , we find from eq . ( [ eq : b.7 ] ) _ ^33(k ) = , [ eq : b.8 ] which has the expected rpa - type structure . notice that the `` contact term '' , which results from the term in @xmath172 that is quadratic in @xmath173 , provides the numerator in eq . ( [ eq : b.8 ] ) that is missing in @xmath4 .

we report the discovery of an extremely red planetary - mass companion to 2mass j22362452 + 4751425 , a.6 late - k dwarf likely belonging to the myr ab doradus moving group . 2m2236 + 4751b was identified in multi - epoch nirc2 adaptive optics imaging at keck observatory at a separation of 3.7 , or 230 20 au in projection at the kinematic distance of 63 5 pc to its host star . assuming membership in the ab dor group , as suggested from its kinematics , the inferred mass of 2m2236 + 4751b is 1114 . follow - up keck / osiris-band spectroscopy of the companion reveals strong co absorption similar to other faint red l dwarfs and lacks signs of methane absorption despite having an effective temperature of k. with a () color of 2.69 0.12 mag , the near - infrared slope of 2m2236 + 4751b is redder than all of the hr 8799 planets and instead resembles the myr isolated planetary - mass object pso j318.522 , implying that similarly thick photospheric clouds can persist in the atmospheres of giant planets at ages beyond 100 myr . in near - infrared color - magnitude diagrams , 2m2236 + 4751b is located at the tip of the red l dwarf sequence and appears to define the `` elbow '' of the ab dor substellar isochrone separating low - gravity l dwarfs from the cooler young t dwarf track . 2m2236 + 4751b is the reddest substellar companion to a star and will be a valuable benchmark to study the shared atmospheric properties of young low - mass brown dwarfs and extrasolar giant planets .

1611.00364

[ sec : intro ] l dwarfs with anomalously red near - infrared colors were first identified in the two micron all - sky survey ( 2mass ; @xcite ) over a decade ago ( @xcite ) and have steadily grown as a population ever since . with some notable exceptions ( e.g. , @xcite ; @xcite ) , these objects were generally found to possess low - surface gravity features in their optical and near - infrared spectra , pointing to ages much younger than the the vast majority of brown dwarfs in the field ( @xcite ; @xcite ; @xcite ) . in recent years the sample of red l dwarfs has ballooned ( @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) as a result of all - sky infrared surveys like the wide - field infrared survey explorer ( wise ) , panoramic survey telescope and rapid response system ( pan - starrs ) , ukirt infrared deep sky survey ( ukidss ) , and visible and infrared survey telescope for astronomy ( vista ) , inspiring new gravity - insensitive spectral classification systems , all - sky searches for young brown dwarfs , and infrared parallax programs to better determine their physical properties and relationship with young moving groups ( @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) . it is now clear that these low - gravity brown dwarfs lie on a parallel sequence redward of the standard l dwarf locus in near - infrared color - magnitude diagrams . liu et al . ( 2016 , in press ) show that the ensemble of late - m and l dwarfs with very low gravity spectral classifications represent a `` fanning out '' of old field objects with the same spectral type ; later - type young l dwarfs are progressively redder and fainter in @xmath7 than their higher - gravity counterparts . this systematic offset and broadening of the l dwarf locus is likely caused by a combination of youthful overluminosity for late - m and early - l dwarfs and unusually dusty photospheres at later types that redistribute emergent flux to longer wavelengths ( e.g. , @xcite ) . while this narrative may be qualitatively correct , no set of self - consistent atmospheric and evolutionary models can completely reproduce the colors and spectra of this fascinating population . in parallel with discoveries of young isolated brown dwarfs , adaptive optics imaging surveys have identified a similar pattern of redder colors and fainter absolute magnitudes among young extrasolar giant planets ( see , e.g. , @xcite ) . the archetypal companion in this class is 2m12073932 b ( @xcite ) ; with a @xmath5@xmath4 color of 3.1 @xmath30.2 mag ( @xcite ) , it is the reddest ultracool object known and lacks the deep methane absorption expected at the effective temperature predicted by cooling models ( @xcite ; @xcite ) . more generally , this empirically - discovered trait not predicted beforehand by atmospheric models is also seen in other red , low - temperature companions like hr 8799 bcde ( e.g. , @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) , 2m01222439 b ( @xcite ; @xcite ) , and vhs j12561257 b ( @xcite ) , as well as young free - floating brown dwarfs near the canonical l / t transition ( e.g. , @xcite ) and may be caused by vigorous vertical mixing and strong disequilibrium carbon chemistry ( @xcite ; @xcite ) . a natural question that has emerged from these studies is the degree to which young isolated brown dwarfs and extrasolar giant planets share common atmospheric properties and cooling pathways . the physical properties of the lowest - mass brown dwarfs appear to broadly overlap with the most massive giant planets . however , if the initial conditions or compositions of these populations are substantially different from one another as a result of disparate formation channels then this similarity may be superficial and free - floating young red l dwarfs may not be `` exoplanet analogs '' after all . unfortunately , the dearth of known red l dwarf companions currently prevents this kind of comparison . here we present the discovery of an extraordinarily red companion to the late - k dwarf 2mass j22362452 + 4751425 . the host star was identified by @xcite as a candidate member of the ab dor moving group on the basis of its proper motion and uv emission from @xmath8 . our new radial velocity measurements of the primary are consistent with ab dor membership , implying a mass of @xmath11114 for the companion . at a separation of 3@xmath97 ( 230 au ) , 2mass j22362452 + 4751425 b ( herinafter 2m2236 + 4751b ) belongs to two classes of rare and enigmatic objects : planetary - mass companions on extremely wide orbits ( @xmath10100 au ) , which have low occurrence rates of @xmath112% around young stars ( @xcite ) , and red l dwarf companions , of which only a handful are known . 2m2236 + 4751b is likely to be an important benchmark for both populations . [ sec : observations ] lccccccc 2014 nov 08 & 2014.852 & 5 @xmath12 1 @xmath12 60 & @xmath13+cor600 & 3692 @xmath3 3 & 135.8 @xmath3 0.2 & 7.3 @xmath3 0.5 & 69 @xmath3 9 + 2015 aug 27 & 2015.653 & 5 @xmath12 1 @xmath12 60 & @xmath13+cor600 & 3694 @xmath3 3 & 135.7 @xmath3 0.2 & 8.22 @xmath3 0.04 & 48 @xmath3 1 + 2016 jun 27 & 2016.489 & 10 @xmath12 1 @xmath12 60 & @xmath13+cor600 & 3696 @xmath3 3 & 135.3 @xmath3 0.2 & 8.2 @xmath3 0.3 & @xmath14 + 2016 jun 27 & 2016.489 & 10 @xmath12 1 @xmath12 60 & @xmath15+cor600 & 3705 @xmath3 3 & 134.8 @xmath3 0.2 & 9.8 @xmath3 0.4 & 41 @xmath3 1 + 2016 jul 18 & 2016.546 & 10 @xmath12 10 @xmath12 6 & @xmath15+cor600 & 3707 @xmath3 3 & 135.4 @xmath3 0.2 & 9.0 @xmath3 0.2 & @xmath14 + 2016 jul 19 & 2016.549 & 10 @xmath12 6 @xmath12 10 & @xmath5+cor600 & 3705 @xmath3 3 & 135.7 @xmath3 0.2 & 10.4 @xmath3 0.3 & 34 @xmath3 1 + 2016 aug 03 & 2016.590 & 2 @xmath12 1 @xmath12 60 & @xmath5+cor600 & 3690 @xmath3 3 & 135.4 @xmath3 0.2 & 9.93 @xmath3 0.12 & 36 @xmath3 1 -.6 in -1.2 in -.6 in -1.9 in -0.75 in -1.4 in [ sec : obs : igrins ] we first imaged 2m2236 + 4751 on 2014 november 8 ut in @xmath13-band with keck / nirc2 using natural guide star adaptive optics ( ngs ao ; @xcite ) as part of our ongoing efforts to find , characterize , and measure the statistical properties of giant planets around young low - mass stars . 2m2236 + 4751b was identified at a separation of 3@xmath97 ( figure [ fig : nirc2img ] ) and subsequently confirmed to be comoving with its host star through follow - up imaging in 2015 and 2016 ( table [ tab : obs ] ) . all observations were carried out in a similar manner with the narrow camera using the entire 1024@xmath121024 pix@xmath16 array , the partly - translucent 600 mas diameter coronagraph , and vertical angle ( pupil - tracking ) rotator setup . in addition to coronagraphic imaging , we also acquired 1020 unsaturated frames of the host star for photometric calibration immediately before or after the deeper imaging except on 2016 june 27 ut ( @xmath13 filter ) and 2016 july 18 ut ( @xmath15 filter ) ; for these data sets we use the coronagraph transmission measurement from @xcite to derive the flux ratios of 2m2236 + 4751ab . raw images were bias - subtracted , flat - fielded , and corrected for bad pixels and cosmic rays . field rotation was small ( between 26@xmath17 ) in our 510 min sequences and the companion is outside the speckle noise - limited region close to the star , so no angular differential imaging subtraction was necessary . each frame was corrected for optical distortions using the solution from @xcite for data obtained prior to april 2015 and from @xcite thereafter to account for the altered optical distortion following a nirc2 pupil realignment . the corresponding plate scales and north orientation angles are 9.952 @xmath3 0.002 mas pix@xmath18 and + 0.252 @xmath3 0.009@xmath17 from yelda et al . , and 9.971 @xmath3 0.004 mas pix@xmath18 and + 0.262 @xmath3 0.02@xmath17 from service et al . images were registered using the position of the host star visible behind the occulting mask and were then de - rotated , median - combined , and north - aligned to produce a final reduced frame ( figure [ fig : jhkimgs ] ) . astrometry of the companion is derived in a similar manner as in @xcite and takes into account positional uncertainties of the star behind the coronagraph , centroid errors of the companion , and systematic errors in the distortion solution and north alignment , the latter of which is the dominant term in the error budget . when unsaturated frames were obtained , relative photometry was measured by calculating the mean and standard deviation of counts from the host star in the unsaturated images and from the companion in the deep coronagraphic data using aperture photometry . photometric uncertainties are propagated analytically . our nirc2 astrometry and relative photometry is listed in table [ tab : obs ] . [ sec : obs : osiris ] we obtained a medium - resolution ( @xmath19 @xmath20 @xmath21/@xmath22 @xmath0 3800 ) 1.962.38 @xmath23 m spectrum of 2m2236 + 4751b with the oh - suppressing infra - red imaging spectrograph ( osiris ; @xcite ) mounted on keck i and coupled with ngs ao on 2016 june 23 ut . our observations benefited from a new grating installed in 2012 ( @xcite ) and a new spectrograph detector in early 2016 , a few months prior to our observing run . observations were taken with the @xmath24 filter and the 50 mas pix@xmath18 plate scale , resulting in a 16@xmath1264 spaxel@xmath16 ( spatial pixel ) lenslet geometry and 0@xmath98@xmath123@xmath92 rectangular field of view . the rotator was set orthogonal to the binary position angle ( p.a . ) to avoid contamination from the host star . we acquired six pairs of nodded data cubes in an ab pattern with individual exposures of 300 s , totaling 60 min of on - source integration time altogether . conditions were clear with 0@xmath94 seeing throughout our science observations . immediately prior to this we observed the a0v standard hd172728 for telluric correction . the raw 2d images were transformed to 3d data cubes with the osiris data reduction pipeline and the latest rectification matrices from keck . spectra were extracted using aperture photometry with local sky subtraction then scaled to a common level and median - combined . telluric correction was carried out with the ` xtellcor_general ` routine in the spextool reduction package for irtf / spex ( @xcite ; @xcite ) . finally , our spectrum was flux calibrated using a multiplicative scale factor corresponding to the apparent @xmath13 magnitude measured with nirc2 . lcccccccc 2m2236 + 4751a & 2016 jun 13 & cfht / espadons & @xmath14 & @xmath14 & @xmath14 & 15 & 68000 & @xmath14 + 2m2236 + 4751a & 2016 jun 14 & mcdonald 2.7-m / igrins & @xmath14 & 0.98 & @xmath14 & 32 & 45000 & hd 219290 + 2m2236 + 4751a & 2016 may 24 & irtf / spex ( sxd ) & @xmath14 & 0.3 & @xmath14 & 4 & 2000 & hd 209932 + 2m2236 + 4751b & 2016 jun 23 & keck / osiris & @xmath24 & @xmath14 & 50 & 60 & 3800 & hd 172728 [ sec : obs : espadons ] high - resolution optical spectroscopy was obtained for 2mj2236 + 4751a in queue service observing ( qso ) mode with espadons ( @xcite ) at cfht on 2016 june 13 ut . espadons was used in the star+sky spectroscopic mode and combined with the normal ccd readout mode to yield a resolving power of r@xmath168000 covering the 3700 to 10500 wavelength range . the total integration time was 900 seconds . the data were reduced by the qso team using the cfht pipeline upena1.0 , which uses the libre - esprit software package ( @xcite ) . a heliocentric radial velocity of 22.1 @xmath3 0.5 km s@xmath18 and a projected rotational velocity of 4 km s@xmath18 were measured using the same methodology as described in @xcite . [ sec : obs : igrins ] we observed 2m2236 + 4751a with the immersion grating infrared spectrometer ( igrins ) on the 2.7-meter harlan j. smith telescope located at mcdonald observatory . igrins is a high - resolution near - infrared spectrograph offering simultaneous coverage of the @xmath15 and @xmath4 bands ( 1.45 - 2.45 @xmath23 m ) at a resolution of r@xmath2545,000 ( @xcite ) . observations consisted of four 480 s exposures taken in an abba nod pattern . immediately after we observed the a0v star hd 219290 at a similar airmass for telluric correction . the igrins data were reduced using version 2.1 of the igrins reduction pipeline @xcite . the pipeline performs dark , bias , and flat corrections , followed by an optimal extraction of the one - dimensional spectrum for both the target and a0v star . a first order wavelength solution is generated using lines from a thar lamp ; a full wavelength solution is then generated using oh emission sky lines . finally , a more refined wavelength solution is calculated using telluric absorption lines in the raw a0 spectrum . the final spectrum has a s / n of @xmath1150 per resolution element . the broad spectral grasp of igrins provides more than 20,000 resolution elements at r@xmath2545,000 . cross - correlation of all of these elements provides robust statistical measurements of stellar radial velocities . as described in @xcite , the spectral stability of igrins is sub - pixel within an observing night , producing a radial velocity precision of @xmath11 200 m / s . radial velocities have been derived for nearly all igrins observations . briefly , the igrins spectrum of 2m2236 + 4751a was cross - correlated in pixel space against 144 other igrins spectra with spectral types between k3 and m2 . shifts in the spectrum are removed by cross - correlating the telluric spectra to determine pixel shifts between the target and the templates with similar spectral type . pixel offsets between stellar spectra are converted into velocities using the empirically derived spectral resolution . the relative radial velocity is determined from the mean and standard error of all radial velocity measurements from all 144 template spectra , after a sigma filter that removes @xmath104@xmath26 outliers . the relative radial velocity is converted into an absolute radial velocity using a zero - point shift based on @xmath10100 absolute radial velocities from the literature . our final barycentric radial velocity for 2m2236 + 4751a is 21.4 @xmath3 0.2 km s@xmath18 , consistent with ( and more precise than ) the cfht / espadons measurement . + [ sec : obs : igrins ] to better characterize the host star we obtained a medium - resolution 0.72.55 @xmath23 m spectrum of 2m2236 + 4751a on 2016 may 24 ut with the short cross - dispersed mode of the recently - upgraded spex ( @xcite ) spectrograph at the nasa infrared telescope facility ( irtf ) . the 0@xmath93 slit was used which resulted in an average resolving power of @xmath19@xmath02000 . four pairs of 30-s exposures were taken in an abba pattern along the 15@xmath2 slit . the a0v standard hd 209932 was observed immediately beforehand at a similar airmass . images were pairwise subtracted then the spectra were extracted , median - combined , and corrected for telluric features using spextool reduction package ( @xcite ; @xcite ) . [ sec : observations ] -.25 in -.25 in -.1 in -.4 in .1 in -.4 in [ sec : cpm ] our nirc2 astrometry of 2m2236 + 4751ab spans 1.7 years ( table [ tab : obs ] ) . throughout this period , the expected change in separation and p.a . for a stationary background star are 119 @xmath3 5 mas and 1.0 @xmath3 0.3@xmath17 , respectively , based on our first epoch of astrometry , the host star s proper motion from ucac4 ( @xmath27cos@xmath28 = 62.6 @xmath3 1.4 mas yr@xmath18 , @xmath29=30.5 @xmath3 1.9 mas yr@xmath18 ; @xcite ) , and its kinematic distance ( 63 @xmath3 5 pc ; section [ sec : age ] ) . the measured change between the initial and latest epochs is only 2 @xmath3 4 mas and 0.4 @xmath3 0.3@xmath17 , heavily favoring the common proper motion scenario ( figure [ fig : backtracks ] ) . the comoving versus stationary background hypotheses can be tested more quantitatively using the bayes factor . following @xcite , the @xmath30 values of the comoving ( @xmath31 ) and stationary ( @xmath32 ) scenarios are 30 and 2850 , respectively , for five degrees of freedom . assuming equivalent prior odds , the posterior odds are @xmath33(@xmath34/@xmath35 ) = 613 , indicating the pair are unambiguously comoving and very likely to be gravitationally bound . [ sec : age ] the age of 2m2236 + 4751 is critical to the interpretation of this system ; a young age would imply that the companion has a mass in the planetary regime and its red colors are a result of thick clouds associated with low surface gravity like 2m12073932 b , while an old age would indicate that it is a brown dwarf with an unusually dusty ( and potentially metal - rich ) photosphere like 2m2148 + 4003 ( @xcite ) . here we discuss two lines of evidence regarding the system age : a potential kinematic link to the @xmath0120 myr ab dor moving group , and other age indicators like activity and rotation period . @xcite identified 2m2236 + 4751 as a candidate member of the ab dor moving group from its proper motion and sky position . we also find consistency with ab dor using an alternative method to identify new members of young moving groups ( bowler et al . , in prep . ) . by adopting the @xmath36 kinematics and uncertainties of moving groups from @xcite , we can invert the standard calculation of galactic space velocities ( @xcite ) when only a proper motion is available to predict an expected radial velocity distribution as well as two distance distributions associated with each proper motion component ( @xmath27cos@xmath28 and @xmath29 ) . these two distance distributions should agree with one another if the object is a member of a particular group being tested ; subtracting one from the other and calculating the tail integral about zero offers a simple metric ( @xmath37 ) to assess this consistency . @xmath37 values near 0.5 are most consistent , while incompatible distributions have @xmath37 values near 0 . for 2m2236 + 4751 we find an @xmath37 value of 0.45 for ab dor and values of 0.0 for the twa , @xmath38 pic , tuc - hor , columba , and carina moving groups , indicating excellent proper motion consistency with the ab dor moving group ( figure [ fig : d3plot ] ) . to assess the false positive probability that non - moving group members would have @xmath37 values at least this high by chance , we ran the same analysis for a sample of over 2000 inactive m dwarfs from @xcite . the probability that a field star would share a similarly consistent proper motion with the ab dor moving group is 1.1% . we adopt this as an upper limit because the joint probability of a similar proper motion _ and _ radial velocity ( see below ) will be substantially smaller than this . note that one advantage of our method over bayesian techniques is that our approach does not rely on probabilistic assignments using kinematic ( @xmath36 ) and positional ( @xmath39 ) models for moving groups and field stars . for example , the banyan i algorithm developed by @xcite predicts a probability of 68% for ab dor membership , but banyan ii ( which uses updated membership lists and is tailored to low - mass moving group members ; @xcite ) gives an ab dor probability of only 0.11% . position of 2m2236 + 4751 compared to bona fide ab dor members , which we discuss in more detail later in this section . for reference , note that the banyan ii `` young field '' and `` old field '' probabilities are 52% and 47% . ] we find predicted ( kinematic - based ) radial velocity and distances of 22.8 @xmath3 1.2 km s@xmath18 and 63 @xmath3 5 pc . this is in good agreement with our measured radial velocities from espadons ( 22.1 @xmath3 0.5 km s@xmath18 ) and igrins ( 21.4 @xmath3 0.2 km s@xmath18 ) , bolstering a possible kinematic association with ab dor . similarly , the photometric distance to 2m2236 + 4751 is 74 @xmath3 10 pc based on the @xmath40 versus @xmath41@xmath13 relationship for the pleiades from @xcite , which is @xmath01@xmath26 from the kinematic distance . our results are also consistent with the radial velocity and distance predictions of 23.0 @xmath3 1.2 km s@xmath18 and 65 @xmath3 9 pc inferred by schlieder et al . figure [ fig : uvw ] shows the @xmath36 galactic space velocities and @xmath39 heliocentric positions for 2m2236 + 4751 based on its proper motion , weighted mean radial velocity ( 21.5 @xmath3 0.2 km s@xmath18 ) , and kinematic distance compared to confirmed members of nearby young moving groups from @xcite . the agreement with known ab dor members in @xmath42 , @xmath41 , @xmath43 , @xmath44 , and @xmath45 is excellent , but the @xmath46 position appears to be an outlier . it is unclear how problematic this is since the current census of moving group members is incomplete and heavily biased towards nearby ( @xmath4760 pc ) young stars owing to distance cuts in early searches ( see , e.g. , @xcite ) . soon @xmath48 will reveal the entire stellar population and @xmath36/@xmath39 distributions of these nearby moving groups . until then , we simply note this possible tension with the @xmath46 positional distribution of ab dor members and await a re - evaluation in the future . activity diagnostics offer additional independent ways to constrain the age of this system . 2m2236 + 4751 was not detected in the _ roentgen satellite _ ( @xmath49 ) all - sky survey ( rass ) . adopting an upper limit equal to the detection limit of rass ( 0.05 cnts s@xmath18 ; @xcite ) and assuming a hardness ratio 1 ( hr1 ) value of 0.0 , the implied x - ray flux from 2m2236 + 4751 is @xmath50 @xmath11 4.1@xmath1210@xmath51 erg @xmath52 s@xmath18 , the x - ray luminosity is log(@xmath53 ) @xmath11 29.3 erg s@xmath18 , and the fractional x - ray to bolometric luminosity is log(@xmath53/@xmath54 @xmath11 3.1 dex . @xmath11 3.4@xmath1210@xmath51 erg @xmath52 s@xmath18 , log(@xmath53 ) @xmath11 29.2 erg s@xmath18 , and log(@xmath53/@xmath54 @xmath11 3.2 dex . ] these values are near the saturation limit for x - ray emission in low - mass stars , so a non - detection in rass is not particularly useful to constrain the coronal emission and age of 2m2236 + 4751 . low - mass pleiades members span a wide range of x - ray luminosities , log(@xmath53/@xmath54 @xmath0 3 to 4 dex ( @xcite ) , so this x - ray upper limit is consistent with a pleiades ( and ab - dor)-like age of @xmath1120 myr . table [ tab : k7 ] lists the activity levels for known or suspected k7 members of ab dor ; like pleiades members , these objects have fractional x - ray luminosities between 3.0 and 4.0 dex , consistent with our upper limit for 2m2236 + 4751 . lcccccccccc 2m0034 + 2523 & 00 34 08.43 & + 25 23 49.8 & 48 & 3.23 @xmath3 0.10 & 0.82 & @xmath14 & 3.16 & 10.12 @xmath3 0.07 & 12.3 @xmath3 0.4 & 1 , 2 , 3 + hip 25283 & 05 24 30.17 & 38 58 10.7 & 18 & 3.64 @xmath3 0.04 & @xmath14 & 12 & 9.34 & @xmath14 & @xmath14 & 4 , 5 , 6 + hip 26369 & 05 36 55.10 & 47 57 48.1 & 26 & 3.36 @xmath3 0.21 & 0.9 & 70 & 4.54 & @xmath14 & @xmath14 & 4 , 5 , 7 + hip 31878 & 06 39 50.04 & 61 28 41.8 & 22.4 & 3.82 @xmath3 0.15 & @xmath14 & 50 & 9.06 & 10.41 @xmath3 0.03 & 13.1 @xmath3 0.13 & 4 , 6 , 7 , 8 + hip 86346 & 17 38 39.65 & + 61 14 16.1 & 33 & 3.04 @xmath3 0.06 & 1.3 & 40 & 1.842 & 9.36 @xmath3 0.04 & 11.14 @xmath3 0.03 & 4 , 8 , 9 , 10 + 2m2039 + 0620 & 20 39 54.60 & + 06 20 11.8 & 38.5 & 3.95 @xmath3 0.12 & 0.3 & @xmath14 & @xmath14 & @xmath14 & @xmath14 & 1 , 2 + hip 106231 & 21 31 01.71 & + 23 20 07.5 & 24.8 & 3.10 @xmath3 0.09 & @xmath14 & 215 & 0.423 & 9.04 @xmath3 0.03 & @xmath14 & 4 , 7 , 11 + hip 113597 & 23 00 27.92 & 26 18 43.17 & 30 & 3.68 & 0.1 & 10 & 8.0 & 10.47 @xmath3 0.02 & 12.97 @xmath3 0.13 & 4 , 11 , 12 , 13 + 2m2236 + 4751 & 22 36 24.52 & + 47 51 42.5 & 63 & @xmath113.1 & 0.18 & @xmath1150 & 11.2 & 10.9 @xmath3 0.2 & @xmath14 & 14 , 15 -.4 in .4 in -.4 in -.1 in -.4 in -.2 in uv photometry from the _ galaxy evolution explorer _ ( @xmath8 ; @xcite ; @xcite ) provides another age diagnostic for 2m2236 + 4751 . the host star is detected in the near - uv filter ( @xmath55 = 20.9 @xmath3 0.2 mag ) but not in the far - uv ( @xmath56 ) . @xcite show that older stars trace an envelope in @xmath55@xmath5 color as a function of near - infrared color ; the @xmath55@xmath5 color of 10.9 @xmath3 0.2 mag for 2m2236 + 4751 is generally consistent with hyades - like chromospheric emission but slightly redder than known k7 members in ab dor with @xmath8 detections ( table [ tab : k7 ] ) . similarly , the f@xmath57/f@xmath58 ratio ( 9 @xmath3 2 @xmath1210@xmath59 ) from @xcite and ( f@xmath57/f@xmath58)@xmath60 excess ratio above the photosphere ( 8 @xmath3 2 @xmath1210@xmath59 ) from @xcite indicate that 2m2236 + 4751 has lower chromospheric emission levels than most star at ab dor - like ages . however , the large intrinsic scatter makes it difficult to rule out the all but the youngest ages ( @xmath4710 myr ) for this object . the h@xmath61 equivalent width measured from our high - resolution espadons spectrum is + 0.18 @xmath3 0.02 in absorption ( figure [ fig : hostspec ] ) . a lack of h@xmath61 emission is consistent with 2m2236 + 4751 s apparently weak chromospheric and coronal emission . this value is also somewhat weaker than the envelope traced out by older main sequence stars ( @xmath00.6 for a @xmath62@xmath41 color of 1.2 mag ; see , e.g. , figure 5 of @xcite ) , indicating the h@xmath61 line may be slightly filled in for this star . h@xmath61 emission for k7 ab dor members rangers from + 0.3 to 1.3 ( table [ tab : k7 ] ) . a weak absorption line at about 6707.8 may be caused by either the doublet or possibly at 6707.4 . if it originates from lithium , its equivalent width ( @xmath63=40 @xmath3 10 m ) is comparable to other ab dor members with similar optical colors ( e.g. , @xcite ) . stars spin up as they contract to the zero - age main sequence and then slowly spin down over time through magnetic braking and winds , so in principle rotation periods offer another tool to constrain stellar ages . in practice , age determinations using rotation periods are imprecise for low - mass stars because they exhibit a large spread in initial angular momenta that continues to broaden over time ( e.g , @xcite ; @xcite ) . @xcite measure a period of 11.2 days for 2m2236 + 4751 ; compared to the pleaides rotation distribution , which has a comparable age to ab dor and a well - characterized stellar population , this period sits near the maximum envelope of rotation rates for stellar masses of @xmath00.6 ( @xcite ; @xcite ; @xcite ) . k7 members in ab dor exhibit a similarly broad range of rotation periods from 0.49.3 days ( table [ tab : k7 ] ) . -1.3 in -1 in -3.3 in -.3 in -1.4 in -.3 in altogether we find that the kinematics and rotation period of 2m2236 + 4751 are consistent with the ab dor moving group . there is some tension between the heliocentric galactic @xmath46 position compared to accepted members , and similarly the activity level of 2m2236 + 4751 appears to be lower than ab dor members of the same spectral type , but neither of these decisively rule out membership . we conclude that 2m2236 + 4751 is either an inactive member of ab dor , presumably in the quiet tail end of the activity distribution , or it is an older kinematic interloper in this group . the unusually red spectrum of the companion may suggest the system is indeed a member of this group , similar to the dusty l dwarf wise j0047 + 6803 ( @xcite ; @xcite ) . because of the excellent kinematic agreement with ab dor , we adopt the cluster age of @xmath0120 myr for this work ( @xcite ; @xcite ) but note that older ages are also possible if further investigation shows it not a member . [ sec : cpm ] wide stellar companions to exoplanet host stars provide a more complete view of the system architecture and offer an independent way to age - date the planet ( e.g. , @xcite ) . we searched for companions to 2m2236 + 4751ab by querying the ppmxl proper motion catalog ( @xcite ) at separations corresponding to 5000 au , 10@xmath64 au , and 10@xmath65 au . @xcite found that ppmxl proper motions have systematic errors of @xmath02 @xmath3 5 mas yr@xmath18 based on a large calibration using stationary quasars . we therefore add this systematic uncertainty in quadrature with all objects returned in our search , then compute the difference in proper motions between 2m2236 + 4751 and stars within each specified angular radius ( figure [ fig : pmpm ] ) . uncertainties are propagated analytically and all stars within 3-@xmath26 are considered to be candidate comoving companions . real stellar companions should sit on or above the main sequence at the distance to 2m2236 + 4751 , so we also impose a color - magnitude diagram cut to further vet the sample . @xmath41 and @xmath13-band magnitudes are queried from ucac4 ; objects falling below the main sequence as determined by @xcite - band magnitudes compiled by e. mamajek can be found at http://www.pas.rochester.edu/ @xmath1emamajek / eem_dwarf_ubvijhk_colors_teff.txt . ] in @xmath40 versus @xmath41@xmath13 are excluded from consideration . aside from the host star itself , no candidate companions emerged in our 5000 au and 10@xmath64 au searches . within 10@xmath65 au , 77 stars have proper motions within @xmath113@xmath26 of 2m2236 + 4751 . based on the same analysis for an offset field 2@xmath17 away , the expectation value for the number of stars having consistent proper motions by chance is 34 and the probability of finding at least one star consistent with 2m2236 + 4751 is effectively 1.0 from binomial statistics . however , none of these candidates also have @xmath41-band magnitudes and fall on or above the main sequence . the probability of at last one star passing all these criteria by chance ( absolute @xmath41-band magnitude above the main sequence and consistent proper motion ) based on our offset field is only 0.28 . altogether , no promising stellar companions emerged in our search . [ sec : cpm ] no spectral type for 2m2236 + 4751a is evident in the literature , but its optical through near - infrared colors suggest @xmath1k7 . for example , its @xmath62@xmath41 color of 1.2 @xmath3 0.2 mag , @xmath41@xmath13 color of 3.36 @xmath3 0.03 mag , @xmath66@xmath67 color of 0.59 @xmath3 0.09 mag are all consistent with k7m0 types ( @xcite ; @xcite ; @xcite ) . this agrees with the estimate of k7 from @xcite based on @xmath41@xmath5 color . furthermore , our spex spectrum of 2m2236 + 4751a shows slight steam absorption at @xmath11.4 @xmath23 m and @xmath11.9 @xmath23 m and is most similar to m0 template from the irtf spectral library ( figure [ fig : hostcomp ] ; @xcite ) . altogether we estimate a spectral type of [ k7 @xmath3 1 ] , where the brackets indicate it is predominantly photometrically - based . we estimate a bolometric luminosity of @xmath68/@xmath69 = 1.17 @xmath3 0.08 dex for the host star using the @xmath15-band bolometric correction from @xcite and its kinematic distance . similarly , we estimate an effective temperature of 4045 @xmath3 35 k from their ( @xmath41@xmath5 ) relation . the corresponding mass using the @xcite evolutionary models and based on the bolometric luminosity and age ( 120 myr ) is 0.60 @xmath3 0.05 . this inferred mass is independent of older ages so the same value is expected if this system is a kinematic interloper in ab dor . -.4 in -.7 in -.6 in -.4 in 0.3 in -.4 in [ sec : cpm ] with a @xmath5@xmath13 color of 2.62 @xmath3 0.12 mag and a @xmath5@xmath15 color of 1.4 @xmath3 0.2 mag , the near - infrared spectral energy distribution of 2m2236 + 4751b is among the reddest l dwarfs known . the @xmath5@xmath13 color in particular rivals the most extreme objects in the color - magnitude diagram like 2m12073932 b ( @xcite ) , pso j318.522 ( @xcite ) , hr 8799 b ( @xcite ) , and vhs j12561257 b ( @xcite ) whose unusual spectra are thought to be caused by extremely thick photospheric clouds . figure [ fig : vhscomp ] shows our nirc2 photometry and our @xmath4-band osiris spectrum of 2m2236 + 4751b . the 1.22.2 @xmath23 m spectral shape and @xmath4-band spectrum closely match the near - infrared spectrum of pso j318.522 , a 58 l7 member of @xmath38 pic ( @xcite ; @xcite ) . the 2.3 @xmath23 m @xmath70co bandhead and red 2.02.2 @xmath23 m spectral slope are evident in the osiris data . compared to field l dwarfs and young red l dwarfs in figure [ fig : speccomp ] , the overall @xmath4-band shape of 2m2236 + 4751b more closely resembles the latter population of objects with very low and intermediate gravity classifications . @xcite show that a @xmath4-band spectrum alone is not sufficient for precise spectral and gravity classifications of young l dwarfs ; lacking broader spectral coverage , we adopt late - l pec " for 2m2236 + 4751b . lccc @xmath27cos@xmath28 ( mas yr@xmath18 ) & 62.6 @xmath3 1.4 & @xmath14 & 1 + @xmath29 ( mas yr@xmath18 ) & 30.5 @xmath3 1.9 & @xmath14 & 1 + @xmath71 ( pc ) & 63 @xmath3 5 & @xmath14 & 2 + @xmath72 ( pc ) & 74 @xmath3 10 & @xmath14 & 2 + @xmath73 ( km s@xmath18 ) & 22.8 @xmath3 1.2 & @xmath14 & 2 + @xmath74 ( km s@xmath18 ) & 22.1 @xmath3 0.5 & @xmath14 & 2 + @xmath75 ( km s@xmath18 ) & 21.4 @xmath3 0.2 & @xmath14 & 2 + @xmath76 ( km s@xmath18 ) & 21.5 @xmath3 0.2 & @xmath14 & 2 + @xmath36 ( km s@xmath77 & & 2 + @xmath39 ( pc ) & & 2 + @xmath78 ( mag ) & & 2 + @xmath79 ( mag ) & & 2 + @xmath80 ( mag ) & & 2 + @xmath8 @xmath55 ( mag ) & 20.9 @xmath3 0.2 & @xmath14 & 3 + @xmath62 ( mag ) & 13.73 @xmath3 0.18 & @xmath14 & 4 + @xmath41 ( mag ) & 12.51 @xmath3 0.03 & @xmath14 & 4 + @xmath81 ( mag ) & 13.1 @xmath3 0.2 & @xmath14 & 4 + @xmath82 ( mag ) & 11.91 @xmath3 0.03 & @xmath14 & 4 + @xmath83 ( mag ) & 11.32 @xmath3 0.08 & @xmath14 & 4 + @xmath84 ( mag ) & 11.5 & @xmath14 & 5 + @xmath85 ( mag ) & 9.975 @xmath3 0.022 & 19.97 @xmath3 0.11 & 2 , 6 + @xmath86 ( mag ) & 9.388 @xmath3 0.021 & 18.54 @xmath3 0.18 & 2 , 6 + @xmath87 ( mag ) & [ 9.180 @xmath3 0.018 ] & [ 17.28 @xmath3 0.04 ] & 2 , 6 + @xmath88 ( mag ) & 9.148 @xmath3 0.018 & 17.35 @xmath3 0.04 & 2 , 6 + @xmath89 ( mag ) & 9.058 @xmath3 0.023 & @xmath14 & 7 + @xmath90 ( mag ) & 9.094 @xmath3 0.020 & @xmath14 & 7 + @xmath91 ( mag ) & 8.947 @xmath3 0.027 & @xmath14 & 7 + @xmath92 ( mag ) & 8.731 @xmath3 0.356 & @xmath14 & 7 + @xmath93 ( mag ) & 0.59 @xmath3 0.03 & 1.4 @xmath3 0.2 & 2 , 6 + @xmath94 ( mag ) & [ 0.132 @xmath3 0.001 ] & [ 1.26 @xmath3 0.18 ] & 2 , 6 + @xmath95 ( mag ) & [ 0.747 @xmath3 0.001 ] & [ 2.69 @xmath3 0.12 ] & 2 , 6 + @xmath96 ( mag ) & 0.24 @xmath3 0.03 & 1.19 @xmath3 0.18 & 2 , 6 + @xmath97 ( mag ) & 0.827 @xmath3 0.03 & 2.62 @xmath3 0.12 & 2 , 6 + @xmath98 ( mag ) & 5.97 @xmath3 0.17 & 15.97 @xmath3 0.2 & 2 + @xmath99 ( mag ) & 5.39 @xmath3 0.17 & 14.5 @xmath3 0.3 & 2 + @xmath100 ( mag ) & 5.15 @xmath3 0.17 & 13.35 @xmath3 0.18 & 2 + @xmath101 ( mag ) & [ 5.18 @xmath3 0.17 ] & [ 13.28 @xmath3 0.18 ] & 2 + separation ( @xmath2 ) & & 2 + separation ( au ) & & 2 + age ( myr ) & & 8 + log(@xmath102/@xmath103 ) & 1.17 @xmath3 0.08 & 4.57 @xmath3 0.06 & 2 + mass & 0.60 @xmath3 0.05 & 1114 & 2 + spectral type & [ k7 @xmath3 1 ] & late - lpec & 2 + a shallow indentation is visible in the pseudo - continuum of our osiris spectrum from @xmath02.22.3 @xmath23 m that is similar to the onset of methane absorption in the latest field l dwarfs . to more quantitatively assess whether this feature is indeed from methane , we compared our spectrum with molecular templates of pure h@xmath104o , co , and ch@xmath105 following @xcite and @xcite . the templates were generated by computing a customized thermal emission model using the scarlet atmospheric retrieval framework ( @xcite ) . the model self - consistently calculates the thermal structure and equilibrium chemistry at @xmath106 = 1200 k for a cloud - free atmosphere with solar elemental composition at a resolving power of @xmath19 @xmath10 250,000.o , ch@xmath105 , nh@xmath107 , hcn , co , co@xmath104 and tio from the high - temperature exomol database ( @xcite ) ; o@xmath104 , o@xmath107 , oh , c@xmath104h@xmath104 , c@xmath104h@xmath105 , c@xmath104h@xmath108 , h@xmath104o@xmath104 , and ho@xmath104 from the hitran database ( @xcite ) ; and h@xmath104-broadening following the prescription in @xcite . collision- induced broadening from h@xmath104/h@xmath104 and h@xmath104/he collisions is computed following @xcite . ] the molecular templates are then convolved with the instrument profile , flattened by fitting a high - order polynomial , and cross - correlated with the our flattened @xmath4-band spectrum . results of the cross correlation are shown in figure [ fig : molxcor ] along with the same analysis for m , l , and t dwarfs from the irtf spectral library ( @xcite ; @xcite ) . water shows a significant peak with no cross correlation lag ( velocity shift ) in all mlt templates . as expected , the transition from l to t dwarfs is marked by a reduced strength in co power and an increase in broad methane cross correlation peak . although the m dwarfs show substantial power from the ch@xmath105 template , this is a result of the strong 2.2 @xmath23 m methane absorption feature locking onto the 2.21 @xmath23 m doublet at high temperatures . for 2m2236 + 4751b , we find strong evidence of co , some evidence of h@xmath104o , and no sign of methane absorption . the slight trough from 2.22.3 @xmath23 m in our osiris spectrum is therefore probably not a result of methane absorption . -.4 in .3 in -1 in we estimate a bolometric luminosity for 2m2236 + 4751b using the 0.92.4 @xmath23 m spectrum of pso j318.522 from @xcite which bears a close resemblance to 2m2236 + 4751b flux - calibrated to the @xmath13-band photometry of 2m2236 + 4751b together with a [ @xmath106=1100 k ; @xmath109=4.5 dex ] bt - settl atmospheric model from @xcite as a bolometric correction at short ( @xmath21 @xmath110.9 @xmath23 m ) and long ( @xmath21 @xmath10 2.4 @xmath23 m ) wavelengths . @xcite show that the flux redistribution from short to long wavelengths for young l dwarfs pivots between 1.52.5 @xmath23 m , so the atmospheric model parameters for our luminosity calculation are chosen to correspond to the @xmath110- and @xmath100-band magnitudes of 2m2236 + 4751b predicted by cond evolutionary models ( @xcite).@xmath111300 k , @xmath109 @xmath11 4.2 dex , and @[email protected] . ] this effective temperature is consistent with the value from the @xcite young @xmath106(@xmath110 ) relation , which yields 970 @xmath3 150 k for 2m2236 + 4751b . uncertainties in @xmath102 are derived in a monte carlo fashion by randomly and repeatedly sampling new distances and absolute flux calibration scale factors from normal distributions based on our kinematic distance estimate ( 63 @xmath3 5 pc ) and apparent @xmath13 magnitude of 2m2236 + 4751b ( 13.35 @xmath3 0.18 mag ) . we estimate a luminosity of log(@xmath102/@xmath103 ) = 4.57 @xmath3 0.06 dex from the mean and standard deviation of 10@xmath64 trials . the inferred mass of 2m2236 + 4751b is 13.3 @xmath3 0.8 based on its age , its bolometric luminosity , and the `` hybrid '' evolutionary models of @xcite . ( the corresponding effective temperature from these models is 1170 @xmath3 40 k. ) using absolute magnitudes instead of bolometric luminosity , we find masses of @xmath08 , @xmath011 , and @xmath011 for @xmath7 , @xmath110 , and @xmath100-band magnitudes based on the cond evolutionary models of @xcite . dusty models from @xcite imply an upper limit of @xmath1112.5 . altogether we adopt a mass range of 1114 for 2m2236 + 4751b assuming the system is a member of ab dor . in the unlikely case it is a kinematic interloper , the inferred mass of the companion is 43 @xmath3 5 , 70 @xmath3 2 , and 74 @xmath3 1 at ages of 1 gyr , 5 gyr , and 10 gyr . the corresponding effective temperatures are 1310 @xmath3 50 k , 1390 @xmath3 40 k , and 1390 @xmath3 40 k for the same ages . additional photometry especially at mid - infrared wavelengths will help to better constrain the luminosity and mass of this remarkable companion . [ sec : discussion ] empirical sequences of substellar isochrones in color - magnitude diagrams are important tools to chart the physical and spectral properties of brown dwarfs and giant planets at a given age , map their entire cooling history at a given mass , and jointly test low - temperature atmospheric and evolutionary models . the positions of young ( @xmath110200 myr ) substellar objects in near - infrared color - magnitude diagrams have progressively come into focus with programs searching for low - mass members of young moving groups ( @xcite ; @xcite ) and gas giant planets with high - contrast imaging ( @xcite ) together with follow - up parallax programs to measure distances ( @xcite ; liu et al . 2016 ) . this is particularly true for the young late - m through mid - l dwarf populations , whose numbers and trigonometric distances have swelled in recent years , but the location of young late - l , l / t , and t dwarfs have largely remained elusive owing to the relative dearth of discoveries in this regime and their intrinsic scarcity . the ab dor moving group is among the oldest of the nearby ( @xmath47100 pc ) young ( @xmath47200 myr ) comoving associations in the solar neighborhood ( e.g. , @xcite ; @xcite ) . with an age of @xmath1120 @xmath3 10 myr , it straddles younger well - characterized moving groups like @xmath38 pic and tuc - hor at 2050 myr and older , more heavily populated clusters at 500800 myr like ursa majoris , the hydes , and coma ber ( @xcite ) , thereby serving as an important benchmark for stellar and substellar evolution . the population of confirmed or suspected free - floating brown dwarfs in ab dor now includes over a dozen objects spanning the entire l dwarf sequence down to @xmath1t1 ( e.g. , @xcite ; @xcite ) as well as the single mid - t dwarf candidate sdss j1110 + 0116 ( @xcite ) . many brown dwarf and planetary - mass companions have also been identified with masses ranging from roughly 9 to 40 ( figure [ fig : abdorcomp ] ) : 1rxs j2351 + 3127 b ( l0 ; @xcite ) , cd35 2722 b ( l3 ; @xcite ) , 2m01222439 b ( l4 ; @xcite ) , and gu psc b ( t3.5 ; @xcite ) . altogether , this is the largest set of substellar companions for any young moving group . with a mass of 1114 , 2m2236 + 4751b is the second lowest - mass companion discovered in ab dor after gu psc b ( 11 @xmath3 2 ; figure [ fig : abdorcomp ] ) . 2m2236 + 4751b is also the reddest known member of this group and resides in a prominent position at the `` elbow '' of its substellar isochrone ( figure [ fig : abdoriso ] ) . it has an even more extreme color than two red free - floating brown dwarfs in ab dor , wise j0047 + 6803 ( ( @xmath5@xmath4)@xmath6= 2.48 @xmath3 0.08 mag ) and 2m2244 + 2043 ( ( @xmath5@xmath4)@xmath6 = 2.43 @xmath3 0.04 mag ) , demonstrating that the tip of the l dwarf sequence extends to at least ( @xmath5@xmath4)@xmath6 = 2.7 mag even at the relatively old age of this cluster . -1.2 in -.2 in -.6 in .1 in -.1 in 0 in -0.1 in .6 in 0 in the shape of the ab dor substellar isochrone has important implications for other clusters with similar ages . the pleiades is an especially interesting example with a comparable age of about 125 myr ( @xcite ) and possible dynamical relationship with the ab dor group ( @xcite ) . although it is one of the best - characterized nearby open clusters with a population of over one thousand stars ( e.g. , @xcite ) , its very low - mass substellar members have been difficult to explore owing to their intrinsic faintness . nevertheless , several deep near - infrared imaging surveys have identified l dwarfs down to the planetary - mass regime ( e.g. , @xcite ; @xcite ) . @xcite show the pleiades l dwarf sequence in the @xmath5/@xmath5@xmath4 diagram generally resembles ab dor , progressively reddening to @xmath5@xmath4 values of @xmath12.5 mag , but then appears to turn over to bluer colors beyond @[email protected] mag ( @[email protected] mag ) and @[email protected] mag ( @[email protected] mag ; @xcite ) . the absolute magnitudes of 2m2236 + 4751b ( @xmath7=15.97 @xmath3 0.2 mag ; @xmath100=13.35 @xmath3 0.18 mag ) are over one magnitude _ fainter _ than the location of this apparent turnover . similarly , the two aforementioned red l dwarfs wise j0047 + 6803 and 2m2244 + 2043 in ab dor have absolute magnitudes nearly half a magnitude fainter than this possible l / t turnover , suggesting that there is either an age discrepancy between ab dor and the pleiades , a dramatic level of photospheric diversity exists among brown dwarfs with similar absolute magnitudes at this age , or that some of the faint blue planetary - mass candidates identified by @xcite are contaminants . compared to other companions , 2m2236 + 4751b appears to have even more extreme photospheric properties than all previously known objects except the @xmath05 object 2m12073932 b ( ( @xmath5@xmath4)@xmath6 = 3.0 @xmath3 0.2 mag ; figure [ fig : cmd ] ) . however , all isolated objects and companions in this region of the color - magnitude diagram show qualitatively similar atmospheric properties : red colors , faint absolute magnitudes , strong co absorption , and no signs of methane absorption despite having effective temperatures below the traditional l / t transition of @xmath012001400 k for the older field population ( e.g. , @xcite ) . @xcite and @xcite find that very thick clouds and rapid vertical mixing are required to reproduce the red colors and lack of methane absorption in 2m12073932 b , the most extreme object at the tip of the l dwarf sequence , and these traits are likely to apply to 2m2236 + 4751b as well . however , liu et al . ( 2016 ) show that another population of red l dwarfs like 2m2148 + 4003 ( @xcite ) and wise j2335 + 4511 ( @xcite ) which do not appear to be young can also reside in this part of the color - magnitude diagram , indicating that surface gravity is not the sole underlying explanation for the shared atmospheric traits of these objects . 2m2236 + 4751b reinforces the impression that there do not appear to be any obvious differences between the red , faint objects being discovered as companions and the analogous population of isolated free - floating objects . with a projected separation of 230 au , 2m2236 + 4751b is the latest example of a growing population of planetary - mass companions on wide orbits of several hundred au ( e.g. , @xcite ; @xcite ; @xcite ) . the origin of these objects is unclear ; planet - planet scattering ( @xcite ) , the tail end of disk instability ( @xcite ) , and turbulent fragmentation of molecular clouds ( @xcite ) have all been invoked to explain their origin . @xcite argue against scattering as the dominant origin of this population due to the lack of additional close - in companions in these systems and the dearth of multiple massive ( @xmath110 ) planets uncovered at small separations in radial velocity surveys . in fact , because of their low numbers and lack of robust statistics , it is unclear whether these objects make up their own population or are part of a broader distribution of giant planets or low - mass brown dwarfs spanning small separations ( tens of au ) out to extremely wide separations ( thousands of au ) . unlike most of the wide planetary - mass companions currently known , 2m2236 + 4751b is part of a well - defined survey ( to be discussed in more detail in a future publication ) and will help clarify the statistical properties of planetary - mass companions on wide orbits around low - mass stars . [ sec : conclusions ] we have presented the discovery of a faint red l dwarf companion to the late k dwarf 2mass j22362452 + 4751425 . the radial velocity and proper motion of the host star are consistent with the 120 myr ab dor young moving group . 2m2236 + 4751a is relatively inactive , showing partly - filled h@xmath61 absorption but no x - ray emission or uv excess , suggesting that it is either in the tail end of the ab dor activity distribution or it is an older kinematic interloper in that group . the unusually red colors of the companion are similar to ( but more extreme than ) other mid- to late - l dwarfs in ab dor , possibly bolstering membership likelihood , but the existence of high - gravity brown dwarfs in this part of the color - magnitude diagram means that this trait is not uniquely a sign of younth and low surface gravity . assuming the pair is indeed a member of ab dor , the mass of 2m2236 + 4751b is 1114 from hot start evolutionary models and resides at a projected separation of @xmath0230 au at the kinematic distance of @xmath065 pc to the host star . if the system is a older kinematic interloper then the implied mass of the companion can be as high as @xmath174 at 10 gyr . the near - infrared colors of 2m2236 + 4751b are among the reddest known of any brown dwarf or giant planet . in near - infrared color - magnitude diagrams , it appears to mark the elbow of the ab dor substellar isochrone between other faint red l dwarfs like wise j0047 + 6803 and 2m2244 + 2043 and the two t dwarfs gu psc b and sdss j1110 + 0116 . this implies that the transition from red l dwarfs to dust - free t dwarfs occurs between a narrow mass range of about 1113 at an age of @xmath0120 myr . the colors of 2m2236 + 4751b are redder than hr 8799 b but slightly bluer than 2m12073932 b , demonstrating that even intermediate - age giant planets can possess similarly extreme atmospheric properties beyond 100 myr . the 2m2236 + 4751ab system is well - suited for a broad range of follow - up studies . the age and mass of 2m2236 + 4751b will eventually be clarified with a parallax measurement of the host star from @xmath48 to establish whether the system is a member of ab dor . follow - up photometry of the companion at both shorter and longer wavelengths will better constrain its luminosity , spectral type , and atmospheric properties . both low- and high - resolution near - infrared spectroscopy of 2m2236 + 4751b can be used to derive spectral and gravity classifications , independently age - date the system through gravity - dependent absorption line depths , measure its projected rotational velocity , and determine its composition and abundance ratios . finally , 2m2236 + 4751b is also an excellent target for variability studies to measure its rotational period and search for signatures of heterogeneity and patchiness in its unusually thick clouds . we thank a. kraus for helpful discussions about the age of the host star and the entire keck observatory staff for their exceptional support . i.t . was supported by the department of astronomy at the university of texas at austin though the cox endowment and board of visitors funds , as well as a nasa wiyn pi data award administered by the nasa exoplanet science institute as part of nn - explore through the scientific partnership of the national aeronautics and space administration , the national science foundation , and the national optical astronomy observatory . this paper includes data taken at the mcdonald observatory of the university of texas at austin . based on observations obtained with espadons , located at the canada - france - hawaii telescope ( cfht ) . cfht is operated by the national research council of canada , the institut national des sciences de lunivers of the centre national de la recherche scientique of france , and the university of hawaii . espadons is a collaborative project funded by france ( cnrs , menesr , omp , latt ) , canada ( nserc ) , cfht and esa . we utilized 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 . nasa s astrophysics data system bibliographic services together with the vizier catalogue access tool and simbad database operated at cds , strasbourg , france , were invaluable resources for this work . this work used the immersion grating infrared spectrometer ( igrins ) that was developed under a collaboration between the university of texas at austin and the korea astronomy and space science institute ( kasi ) with the financial support of the us national science foundation under grant ast- 1229522 , to the university of texas at austin , and of the korean gmt project of kasi . 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 . finally , mahalo nui loa to the kamaina of hawaii for their support of keck and the mauna kea observatories . we are grateful to conduct observations from this mountain . cutri , r. m. , skrutskie , m. f. , van dyk , s. , et al . 2003 , the 2mass all - sky catalog of point sources , university of massachusetts and infrared processing and analysis center ; ipac / california institute of technology

among the many new irregular satellites that have been discovered in the last five years , at least six are in the so - called kozai resonance . due to solar perturbations , the argument of pericenter of a satellite usually precesses from 0 to . however , at inclinations higher than and lower than a new kind of behavior occurs for which the argument of pericenter oscillates around . in this work we will concentrate on the orbital history of the saturnian satellite s/2000 s5 kiviuq , one of the satellites currently known to be in such resonance kiviuq s orbit is very close to the separatrix of the kozai resonance . due to perturbations from the other jovian planets , it is expected that orbits near the kozai separatrix may show significant chaotic behavior . this is important because chaotic diffusion may transfer orbits from libration to circulation , and vice versa . to identify chaotic orbits we used two well - known methods : the frequency analysis method ( laskar 1990 ) and maximum lyapunov exponents ( benettin _ et al . _ 1980 ) . our results show that the kozai resonance is crossed by a web of secondary resonances , whose arguments involve combinations of the argument of pericenter , the argument of the great inequality ( ) , longitude of the node , and other terms related to the secular frequencies , and . many test orbits whose precession period is close to the period of the great inequality ( 883 yrs ) , or some of its harmonics , are trapped by these secondary resonances , and show significant chaotic behavior . because the great - inequality s period is connected to the semimajor axes of jupiter and saturn , and because the positions of the jovian planets have likely changed since their formation ( malhotra 1995 ) , the phase - space location of these secondary resonances should have been different in the past . by simulating the effect of planetary migration , we show that a mechanism of sweeping secondary resonances , similar to the one studied by ferraz - mello _ et al . _ ( 1998 ) for the asteroids in the 2:1 mean motion resonance with jupiter , could significantly deplete a primordial population of kozai resonators and push several circulators near the kozai separatrix . this mechanism is not limited to kiviuq s region , and could have worked to destabilize any initial population of satellites in the kozai resonance around saturn and jupiter .

astro-ph0404562

carruba _ et al . _ ( 2002 ) studied an analytical model of the kozai resonance ( kozai 1962 ) based on a secular development of the disturbing function , expanded in series of @xmath10 ( where the @xmath11 stands hereafter for quantities related to the perturber ) , truncated at second order in @xmath10 , and averaged over the mean anomalies of both perturber and perturbee ( see also innanen _ et al . two kinds of behavior are possible in this simplified model : for inclinations of less than 39.23@xmath12 ( or higher than 140.77@xmath12 for retrograde satellites ) , the argument of pericenter freely circulates from 0@xmath12 to 360@xmath12 , while at intermediate inclinations a new class of solutions , in which the argument of pericenter librates around @xmath1390@xmath12 , are possible . such behavior is called the kozai resonance . the value of the critical inclination of 39.23@xmath12 is , in fact , an artifact of the averaging ; the real boundary between the region in which libration is possible or not is actually a more complicated function of the satellite s semimajor axis and eccentricity ( uk _ et al . _ 2003 , uk and burns 2004 ) . currently , six irregular satellites of jovian planets are known numerically to be in the kozai secular resonance : two around saturn ( s/2000 s5 kiviuq , and s/2000 s6 ijiraq ) , two around jupiter ( s/2001 j10 euporie , and s/2003 j20 ) , and two around neptune ( s/2002 n2 and s/2002 n4 , holman _ 2004 ) . a third neptunian satellite , s/2003 n1 is on a chaotic orbit that may also be in the kozai resonance ( holman _ et al . _ 2004 ) . a question that carruba _ et al . _ left unanswered regarded the possible presence of chaos and chaotic diffusion at the separatrix between the regions of circulation and libration . since in that work the authors considered a three - body problem , but averaged over the mean anomalies of both perturber and perturbee , our analytical model was an integrable one - degree of freedom model . in the real system , however , perturbations from other jovian planets and short - period terms may produce chaos near the kozai separatrix . chaotic evolution might lead to the escape of objects whose orbits were originally inside the kozai resonance . here we test the hypothesis that the chaotic layer is due to the fact that the kozai resonance overlaps with secondary resonances near the separatrix . it is well known that resonance overlap is one of the major sources of chaos in dynamical systems ( chirikov 1979 ) . moreover , the positions of the planets may have changed after their formation , due to gravitational scattering of planetesimals ( malhotra 1995 ) . as a consequence of the different planetary positions , the shape of the chaotic layer and the locations of the secondary resonances inside and outside the kozai resonance should have been different in the past . this might have had consequences on the stability of any primordial population of kozai resonators . this paper will concentrate on identifying the chaotic layer at the transition between circulation and libration for kozai resonators , in particular for the saturnian satellite kiviuq , whose orbit is very close to the separatrix of the kozai resonance . our goal is to understand the origins of chaos and the effects of planetary migration . our investigation is divided in the following way : section 2 describes the major features of our simplified model of the kozai resonance and identifies the transition region between circulation and libration . section 3 recalls two methods to identify chaotic behavior in dynamical systems , the frequency analysis method and the maximum lyapunov exponents . section 4 investigates possible causes of the chaotic layer . section 5 shows how the kozai resonance is crossed by a series of secondary resonances of different strengths . in the final section we demonstrate how the effect of planetary migration , combined with the effect of the secondary resonances , might have operated to depopulate a possible primordial satellite group of kozai resonators . carruba _ et al . _ ( 2002 ) described a simple analytical model of the kozai resonance . this model is based on a development of the disturbing function in a series of @xmath14 , where @xmath15 and @xmath16 are the radial planetocentric distances of the satellite and the sun , truncated at the second order . the perturbing function @xmath17 is : @xmath18 where @xmath19 is the gravitational constant , @xmath20 is the solar mass , @xmath21 is the angle between directions to the satellite and the sun as seen from the planet , and @xmath22 is the second - order legendre polynomial . the perturbing function , when averaged over the mean anomalies of both sun and satellite , becomes : @xmath23 , \label{eq : 4_2}\ ] ] where @xmath24 and @xmath25 are the semimajor axis , eccentricity , and inclination of the satellite , respectively . @xmath26 is the sun s semiminor axis . the perturbing function , when expressed in terms of delaunay elements ( @xmath27 ; see murray and dermott 1999 , danby 1988 ) , is : @xmath28 . \label{eq : 4_3}\ ] ] since both @xmath29 and @xmath30 are cyclic , the conjugate quantities @xmath31 and @xmath32 are constants of the motion , as is the quantity : @xmath33 with @xmath31 and @xmath32 fixed , the perturbing function ( [ eq : 4_3 ] ) represents a one - degree of freedom model for the kozai resonance ( innanen _ et al . the following equations for the orbital elements hold : @xmath34,\end{aligned}\ ] ] where @xmath35 and t is time . the time has been rescaled in this way to eliminate the equations dependency on the satellite s semimajor axis . kinoshita and nakai ( 1999 ) found the solution of ( 5 - 7 ) in terms of elliptic functions . [ fig : c_fig1 ] shows the results of numerical integrations of equations ( 5 - 7 ) for @xmath36 and @xmath37 . for @xmath38 less than 0.6 , librating solutions are possible . this model has several limitations . first , short - period terms have been eliminated by averaging . one consequence of this averaging is that the limit between the zones in which libration is possible or not is @xmath39 ( or @xmath40 for retrograde orbits ) , regardless of the satellite s semimajor axis ( both inclinations correspond to @xmath38 = 0.6 for @xmath41 ) . when short - period terms are included , the boundary shifts slightly ( uk and burns 2004 ) . a case where this is observed is the orbit of the jovian satellite euporie that is a kozai resonator but has @xmath42 . a second limitation of the averaged model is that , since it is an integrable , one - degree - of - freedom model , no chaos can appear at the separatrix between circulation and libration . finally , since mutual perturbations among jovian planets are ignored , the orbital elements of the planet about which the satellites circle are constant . in the real solar system , the eccentricities of jupiter and saturn vary with several prominent periods associated with the secular frequencies @xmath7 and @xmath43 ( timescales of up to @xmath44 yr ) , great - inequality terms @xmath45 , where the suffices j and s hereafter stand for jupiter and saturn , ( with a timescale of 883 yr ) , and short - period frequencies , connected with the orbital periods of the planets ( 12.5 and 29.7 yrs for jupiter and saturn , respectively ) . the shape of the separatrix may change in time according to the variations of the planet s eccentricity and semimajor axis ( remember the presence of the factor @xmath46 in @xmath47 ) which might introduce chaos along the real system s separatrix . to understand the behavior of the real system , we concentrate on the case of kiviuq , a saturnian satellite in kozai resonance whose orbit is very close to the separatrix between libration and circulation in @xmath0 . we start by numerically searching for the transition between circulation and libration . to do so , we used the following procedure : we numerically integrated kiviuq s orbit over 100,000 yrs and found the maximum eccentricity , because then the satellite is closest to the separatrix . we computed @xmath38 according to the inclination of kiviuq at that time , and generated a grid of initial conditions for test particles ; this contains 19 values in @xmath0 , from @xmath48 to @xmath49 , separated by @xmath50 . all particles were given the average value of kiviuq s semimajor axis during the integration ( @xmath51 au ) , and a grid in inclination with 21 values , starting at @xmath52 with a step of @xmath53 . the values of the eccentricities were computed from @xmath54 . initial @xmath6 and @xmath55 ( mean anomaly ) are equal to those of kiviuq at its maximum eccentricity . this grid forms our `` low - resolution survey '' . these initial orbits were integrated for 2 myr under the influence of the four jovian planets ( outer solar system , oss hereafter ) with a modified version of swift - whm , the integrator that uses the wisdom and holman ( 1991 ) mapping in the swift package ( levison and duncan 2000 ) . we modified the integrator so that the planets orbits are integrated in the heliocentric system and the particles orbits are integrated in the planetocentric reference frame . this setup minimizes the integration error ( nesvorn _ et al . _ 2003 ) . to more easily identify secular oscillations in saturn s inclination , we refer all the positions to saturn s initial orbital plane . [ fig : particles_fate ] shows the results of our simulations for our low - resolution survey . the actual separation between circulating and librating behavior does not follow the solution of the secular model when the full perturbations from jovian planets are considered . moreover , some orbits alternate between circulation and libration ( and , in a few cases , spend some time in the other libration island around @xmath56 ) . the question that remains to be answered is whether there is a chaotic layer near the separatrix , and if chaotic diffusion might be effective in depleting a primordial population of kozai resonators . to address this question , and identify chaos , we will use the frequency analysis method ( fam , hereafter ; laskar , 1990 ) , and will compute the maximum lyapunov exponents ( mle hereafter ; benettin _ et al . _ , 1980 ) which we now describe . the frequency analysis method ( fam ) is essentially an analysis of the frequency power spectrum of appropriate combinations of numerically determined orbital elements . in our case , for example , we are interested in the frequency related to the precession of the argument of the pericenter @xmath0 . fig . [ fig : periods ] shows the periods in @xmath0 for the particles in our low - resolution survey . the periods peak sharply in the transition region from circulation to libration in @xmath0 , reaching a maximum of 1800 yrs . the minimum libration period is 480 yrs . the average period of libration in the kozai resonance in our survey is 640 yrs . this range in periods corresponds to a frequency range of @xmath57 700 - 3200 /yr . we took the fourier transform of @xmath58 ( where @xmath59 ) over two time intervals ( 0 - 1 myr and 1 - 2 myr ) and computed the quantity : @xmath60 where @xmath61 is the frequency over the first interval and @xmath62 the corresponding frequency on the second interval . if the motion is regular , @xmath63 will be nearly constant , with small variations produced by errors in the determination of the frequencies ( to be discussed later in this section ) . if the motion is chaotic , variations in @xmath64 will be related to changes in @xmath0 s precession period . while this is the main idea behind the fam , a few complications occur . first , the precision of a fast fourier transform ( fft hereafter ) is limited by its coarse resolution [ @xmath65 , where @xmath66 is the ( constant ) time separation of @xmath0-sampling and n is the number of data points used , which must be a power of 2 ( press _ et al . _ 1996 ) ] in recovering the frequencies having the largest amplitude . to overcome this problem , we used the frequency modified fourier transform method ( fmft ) , with quadratic corrections ( idlichovsky and nesvorn 1997 ) , which allows us to retrieve the frequencies with the largest amplitude to a better precision . to estimate the error associated with this method , we computed the frequencies with and without quadratic corrections , and computed the average value of the differences in @xmath64 . this procedure gives an upper limit on the error . in our case , the average error of fmft was @xmath67 , while a simple fft had a resolution of @xmath68 ( we used a time step @xmath66=30 yr , and 32,768 data points , so that each interval for the determination of the frequencies was @xmath57 1 myr ) . another problem to take into account is _ aliasing_. the fft can recover frequencies up to the nyquist frequency ( @xmath69 , that corresponds to two samplings per period ) . frequencies higher than the nyquist frequency are not lost , but are mapped into a smaller apparent value of @xmath70 , given by @xmath71 . this phenomenon is called aliasing and can be easily understood by considering a sinusoidal wave , sampled at isolated points : at least two points per period are needed to obtain a good estimate of the period . if the sampling is longer than one period , than the retrieved period is larger , and its `` ghost '' signature may appear in the frequency range of interest . to avoid this problem , we used an `` on - line '' low - pass filter , following the procedure of quinn _ et al . _ ( 1991 ) ( based on the work of carpino _ et al . our filter suppresses to a level of @xmath72 all fourier terms with periods smaller than 66.7 yr ( stop - band ) , and attenuates to a level of @xmath72 all fourier terms with periods between 66.7 and 200 yr . terms with periods larger than 200 yrs are in the passband ( see added information on website ) . by following this procedure , we have eliminated all frequencies connected with the orbital periods of saturn and the satellites . finally , the shape of the frequency peaks can be modeled as gaussian , but our measurements are made at isolated points . this introduces what in signal processing is called `` impulse noise '' , which can be attenuated by the use of a median filter ( press _ et al . _ 1996 , pitas 2000 ) . more details about the use of the median filter and of the @xmath73 test we devised to prove the efficacy of our filter can be found in the appendix . to find an alternate way to distinguish between regular and chaotic behavior , we also computed maximum lyapunov exponents ( mle ) for our set of initial conditions . a detailed explanation of the theory of lyapunov exponents goes beyond the scope of this paper ; instead we refer the reader to lyapunov ( 1907 ) and benettin _ et al . _ mle is a measure of exponential stretching of nearby orbits . the lyapunov exponents are equal to zero for regular orbits ( they tend to zero in finite - time calculations ) , while they assume positive values for chaotic orbits . the inverse of a lyapunov exponent is the lyapunov time @xmath74 . smaller values of @xmath74 indicate enhanced local stochasticity . to estimate mle for orbits we used a modified version of swift - lyap2 , a code that integrates the difference equation ( morbidelli 2002 ) in the swift package ( levison and duncan 1994 ) . the code was modified to reduce the integration error by integrating the planets in the heliocentric frame and the satellites in the planetocentric frame . for each of the test particles , we integrated the difference equation with an initial difference vector of modulus @xmath75 and determined the modulus @xmath76 of the displacement vector between the two vectors at time t. we constructed a series @xmath77}$ ] and performed a linear least - squares fit on this series . since @xmath78 , where @xmath31 is the lyapunov exponent , the slope of @xmath79}$ ] versus time is equal to the maximum lyapunov exponent . we computed the lyapunov exponents for all orbits in our low - resolution survey . we will show these results in the following section . we should point out that fam and mle , while both useful tools to investigate chaotic behavior , measure different things . fam measures macroscopic changes in frequencies ( i.e. , the speed at which chaotic motion changes the frequencies ) , while @xmath74 measures the rate of exponential stretching of nearby trajectories . these two techniques are complementary : fam is more practical in finding regions that are macroscopically unstable , and @xmath74 can reveal the presence of different chaos - generating mechanisms . we will discuss in more detail these aspects of the two methods in the following section . to investigate the presence of chaos , we applied the fam to our low - resolution survey , integrated under the influence of the four jovian planets . [ fig : fig_lr1 ] plots @xmath80 for our low - resolution survey when the time interval between the first and the second frequencies was about 1 myr . fig . [ fig : fig_lr1 ] shows that high values of @xmath64 are found near the actual separatrix . the chaotic layer is asymmetric between values of @xmath81 versus @xmath82 , and kiviuq seems to be on the border of the chaotic layer . having shown that a chaotic layer exists , the next logical step is to ascertain its origin(s ) . let us recall the analytical model presented in section 2 . in that model , the shape of the separatrix depends on saturn s semiminor axis @xmath83 , which is a function of the semimajor axis and the eccentricity . variations in saturn s semimajor axis and eccentricity may therefore generate complex orbital evolution near the separatrix . but on what time - scales ? and how can we prove this hypothesis ? to study whether a variable @xmath83 may generate chaos , we performed a numerical simulation with the sun ( saturn ) on an orbit of fixed eccentricity ( equal to the average value of the saturnian eccentricity ) and the set of test particles used for our low - resolution survey . the difference compared to the integrable model is the presence of short - period terms . we performed the fam on test orbits in our low - resolution survey and found that all orbits are regular in this case . in contrast , when only saturn and the sun were present , and saturn had a constant eccentricity equal to the mean value computed during the 46,000 yr period of oscillation (= @xmath84 ) , no chaos was observed . this seems to confirm our intuition that variations in saturn s eccentricity and semimajor axis may be responsible for the chaotic layer . to test this hypothesis , we performed a simulation in which both saturn and jupiter are present . in this case the layer of chaos appears ( fig . [ fig : saturn alone ] ) . the similarity between figs . [ fig : fig_lr1 ] and [ fig : saturn alone ] suggest that the effects of uranus and neptune are unimportant . to demonstrate this , we performed two simulations with our low - resolution survey , one with the complete outer solar system ( oss ) , and the other with just jupiter and saturn . to show that the results are nearly equal we applied the @xmath73 test ( see appendix ) to the filtered results of both simulations . we obtained a value of @xmath73 of 22.4 ( out of a maximum number of 399 ) indicating a very high probability that the two distributions are the same . while perturbations from uranus and neptune might play a role in modifying the chaotic layer , such perturbations are clearly less important than jovian perturbations . so , in the following , we only consider jovian perturbations . in doing so we of course modify the values of planetary frequencies and mean motions , and therefore the positions of any secondary resonances . but , since in this phase of our work we are interested in understanding on what timescales the chaotic behavior is driven , we believe this is an acceptable approximation . there are two prominent timescales for oscillations of saturn s @xmath85 and @xmath86 : the secular oscillations ( period of order @xmath87 yrs ) and the oscillations related to the great inequality ( 883 yrs ) . to study which of these perturbations is more relevant to the origin of chaos , we describe the oss orbital evolution , in terms of fourier series ( bretagnon and francou 1988 ) . in this model , the equinoctial orbital elements ( @xmath88 ) are developed in series of cosines and sines of combinations of twelve angles : four mean longitudes of the giant planets , four proper longitudes of pericenter , and four proper longitudes of the nodes . these proper angles are linear functions of time . this model assures a fractional precision better than @xmath89 for @xmath90 and p over 10 myr . to investigate what minimal model can create a chaotic layer similar to the one observed in the full problem , we used different solutions for the orbits of both jupiter and saturn . we modified our version of swift - whm so that both orbital evolutions of jupiter and saturn are computed `` on - line '' according to the bretagnon secular solution . the test particles are then subjected to perturbations of planets that evolve on such orbits . we considered four models for saturn s orbital evolution : ( 1 ) sec , in which we limit the expansion to terms containing the @xmath91 and @xmath9 frequencies for @xmath92 and @xmath93 , assume a constant semimajor axis , and take 12 terms for @xmath94 s development . ( 2 ) a21sec , in which we have included 12 terms associated with the 1:2 commensurability between jupiter s and saturn s mean motions , and terms associated with the great inequality for the development in _ a _ , plus the same terms as in the sec solution . ( 3 ) gisec , in which there were 8 terms for @xmath95 and @xmath32 , and 6 for @xmath93 and @xmath96 associated with the great inequality , and a constant semimajor axis ; and ( 4 ) a21gisec , which includes the 2:1 terms in _ a _ , the great - inequality terms and the secular variations . table [ table : bret_model ] gives the number of terms taken in @xmath97 for each solution . [ fig : sec ] shows the results of the integration with these models . all simulations used the same solution for jupiter ( 10 terms in the development of @xmath86 , 2 terms for @xmath95 and @xmath32 , 2 for @xmath96 and @xmath93 , and 12 for @xmath94 ) . these figures may be compared with fig . [ fig : saturn alone ] , where the traditional version of swift - whm was used . table [ table : bret_res ] presents the results of a @xmath73 test in which we compared simulations with `` bretagnon '' models with those of the full integration of jupiter and saturn . according to these results , terms associated with the great inequality and the 2:1 terms in the development of saturn s semimajor axis are the most important for the presence of the chaotic layer . the major sources for chaos are the variations in saturn s eccentricity connected with the great inequality . to further investigate this hypothesis , we computed the maximum lyapunov exponent for the particles of our low - resolution survey , in a model containing jupiter and saturn . [ fig : lyap_time ] shows the resulting values of @xmath74 ( left ) when the full problem is considered and ( right ) when the planets evolve according to the sec solution . a remarkable asset of fig . [ fig : lyap_time]a is the region of high values of @xmath74 ( i.e. , a smaller degree of local stochasticity ) near the vertex of the separatrix , which , according to the fam , is a region of high dispersion of the frequencies . in this region @xmath98 yrs , i.e. , twice the period of the great inequality ; thus our results suggest that the feature is connected to a secondary resonance with the great inequality . note , however , that all orbits seem to have quite small values of @xmath74 . orbits inside the libration island , which , according to fam , are macroscopically regular , have @xmath99 yrs . this appears to be a general characteristic of high - inclination irregular satellite orbits , as all seem to have @xmath100 yrs ( m. holman , private communication , 2003 ) . the existence of a region characterized by relatively long lyapunov times , on one hand , but high frequency dispersion , on the other , seems counter - intuitive at first . however , recall that the rate of frequency dispersion is a macroscopic characteristic of the orbit that is related to the global structure of the phase space region . in contrast , the lyapunov time is a local property whose value is dictated by the mechanism that is responsible for the chaotic motion . the existence of regions of large @xmath74 but rapid diffusion is not uncommon in the asteroid belt . for example , in the 7/3 resonance ( see tsiganis _ et al . _ , 2003 ) , the smaller values of @xmath74 are observed at the borders of the resonance where they are related to the pulsation of the separatrix due to secular precession . however , the macroscopically most unstable region is located at smaller libration amplitudes , is characterized by larger values of @xmath74 , and is related to the colocation of the @xmath101 secular resonance . similar results can be shown for other low - order mean - motion resonances ( e.g. , 2:1 and 3:1 ) , where the most unstable zones are generated by coexisting secular resonances . as @xmath74 s value is typically of the order of the forcing period ( e.g. , @xmath102 yrs , @xmath103 yrs , tsiganis and morbidelli 2004 ) , these regions generally have higher values of @xmath74 than the region associated with the pulsating separatrix of the mean - motion resonance . we are likely observing a similar phenomenon in the case studied here . we believe that this high-@xmath74 region is associated with a secondary resonance , involving the great - inequality period . this is also supported by the results shown in fig . [ fig : lyap_time]b ( sec solution ) , where all orbits inside the libration zone become regular ( note the different scale compared to fig . [ fig : lyap_time]a ) . to further support our claim , we selected a particle from this high-@xmath74 region , for which we plot the time evolution of the argument @xmath104 ( fig . [ fig : resonant_arguments ] ) . this corresponds to one of the possible critical arguments of the 1:1 resonance between the frequency of @xmath105 and the frequency of the great inequality , @xmath106 ( @xmath107 is added in order to fulfill the d alembert rules of permissible arguments , i.e. , the sum of coefficients of individual longitudes must be zero and the sum of coefficients of nodal longitudes be even ) . the behavior of the arguments switches erratically between intervals of libration and circulation , which indicates a transition through the separatrix of this secondary resonance . the following picture arises from our low - resolution surveys . two major sources of chaos exist in the system . one is connected to the transition region from circulation to libration and , in particular , is driven by variations in saturn s eccentricity , mainly associated with great - inequality terms , and variations in saturn s semimajor axis . the other is related to a secondary resonance involving the argument of pericenter @xmath0 and the great - inequality terms for circulating particles . in most cases this resonance is a stronger source of chaos than the transition region . better resolution is needed to understand the behavior and shape of this resonance ( see sec . 5 ) . one problem that remains to be explained is the asymmetry of the chaotic layer with respect to @xmath108 ( see figs . according to the secular model , the result should be symmetric . our simulations with the bretagnon model suggest that variations in saturn s semimajor axis may be partly responsible for the asymmetry . alternatively , the asymmetry in the chaotic layer may be related to our choice of initial conditions . to exclude this second possibility , we need to ascertain that our choice of initial conditions ( @xmath6 and @xmath55 ) for the test particles orbits is not introducing spurious effects . to our knowledge , three resonant configurations may alter the eccentricities of the test particles , and so put their orbits closer or farther away from the region of chaos that we are interested in . the three resonances are : `` nodal '' ( @xmath109constant ) , pericentric ( @xmath110 ) , and the evection inequality [ @xmath111constant ( @xmath112 , so as to have an intermediate value of the evection angle ) as evection inequality . ] . in our simulation , the sun s orbit starts with : @xmath113 , and @xmath114 . we computed @xmath6 for the test orbits according to the resonant configuration . for the pericentric resonance , we used two values of the resonant argument : @xmath115 and ( yokoyama _ et al . _ , 2003 ) , @xmath116 , @xmath117 . the second configuration maximizes the perturbing effect of the pericentric resonance . to have a quantitative measure of the asymmetry , we use an `` asymmetry coefficient '' that is the difference of @xmath64 s values for @xmath0 larger and smaller than @xmath118 . to compute this coefficient for orbits in the libration island ( @xmath119 ) , we take two columns symmetric about @xmath108 ( e.g. , the columns for @xmath120 and @xmath121 , with the obvious exclusion of the column for @xmath108 ) and compute the fractional difference for each pair . we repeat the process for all pairs of values and compute the average and standard deviation of the measurements . the average value is our `` asymmetry coefficient '' , and the standard deviation gives an estimate of the error . obviously , the lower is the asymmetry coefficient , the more symmetric is the distribution of @xmath64 around @xmath122 . table [ table : asy_res ] gives asymmetry coefficients for each simulation . since the secular resonance with resonant argument equal to @xmath123 seems to give the more symmetric configuration , we believe its effect might be the dominant one ( but large errors prevent our firm conclusion ) . the low - resolution survey has yielded information regarding the large - scale structure of chaos inside the kozai resonance . however , because most of the chaotic behavior is present near the boundary region between circulation and libration , we now concentrate on analyzing this region in more detail . for this purpose , we constructed a new set of initial conditions : we used 276 bins in inclination , starting at @xmath124 and separated by @xmath125 . we used one @xmath0 (= @xmath126 ) . the choice of the other orbital elements followed the same criteria used for the low - resolution survey . we called this new set of initial conditions a `` high - resolution '' survey . we concentrate on the region around @xmath122 to eliminate the problems associated with asymmetries in @xmath0 suggested by our low - resolution survey . results of integrations of test particles with @xmath0 equal to @xmath127 or @xmath128 are very similar to those with @xmath122 and are not shown . our choice of values for the particles inclinations allow us to sample periods in libration and circulation that go from a minimum of 450 yrs to a maximum of 820 yrs for librating particles , and from a minimum of 650 yrs to a maximum of 1820 yrs for circulating particles , thus covering the whole transition region . we integrated the test particles of our high - resolution survey under the influence of the four jovian planets for 2 myr and applied fam . [ fig : high_res_survey ] shows two regions of high chaoticity . the one at @xmath129 is clearly associated with the transition from circulation to libration : test particles switch from @xmath0-circulation to @xmath0-libration at this @xmath130 ( dots in the plot ) . the other region of strong chaotic behavior at @xmath131 is associated with the pericentric secular resonance ( @xmath132 ) . [ fig : res . arguments]a shows the time evolution of the resonant argument for an orbit in this resonance ( @xmath133 ) . more interesting for our purposes is the region at @xmath134 . a plot of the resonant argument @xmath135 shows that this feature is associated with the 1:1 resonance between @xmath0 and the great inequality ( other resonances involving @xmath136 and @xmath137 , instead of @xmath138 , are also present but are weaker . this property is generally shared by other resonances involving harmonics of the great - inequality period , like the 4:3 , 3:2 , etc . ) . the resonance we discovered is only present for circulating particles . a question that might arise is if a similar resonance could also be found for librating test particles . to answer this question we consider the periods of libration and circulation and estimate them by the frequency associated with @xmath0 s precession found via fam . [ fig : omega periods ] shows such periods as a function of @xmath130 . we also report the regions having high values of @xmath64 , possibly associated with secondary resonances ( vertical lines ) , and the commensurabilities between periods in @xmath0 and great - inequality periods ( for example , the 3:4 commensurability means that , for three periods in @xmath0 there are four periods in @xmath139 . we are conscious that the commensurabilities between the great inequality and @xmath0 are not real resonances , since they do not respect the d alembert rules . nevertheless , we think this is a useful plot , since gives a first - order estimate of the position of the real resonances . to compute the expected positions of the actual resonances we used the following procedure : for a resonance of argument @xmath140 ( we will use @xmath0 for @xmath141 hereafter ) , the frequency of @xmath0 s precession for which there is resonant behavior is given by : @xmath142 where @xmath143 is the frequency associated with the great inequality term ( @xmath106 , equal to 1467.72 `` /yr for the current position of the planets ) , @xmath144 is the frequency associated with @xmath138 (= -26.345''/yr , bretagnon and francou 1988 , the same source is used for the values of @xmath7 and @xmath43 ) , and @xmath145 is the precession frequency . analogous equations can be used for other resonances . from @xmath0 s frequency it is straightforward to determine the period of precession , and from that period we can determine the value of @xmath130 . [ fig : omega periods ] seems to be very instructive . for example , since the period of the 1:1 resonance is 840 yrs and since the first test particle in librating regime has a period of 820 yrs , at the end of the transition region , there is currently no equivalent of the great - inequality secondary resonance for librating particles . also , apart from the two regions with @xmath64 associated with the pericentric secular resonance and the great - inequality resonance , another interesting feature of high chaotic behavior , which seems to be associated with a strong secondary resonance , appears at @xmath146 . [ fig : omega periods ] shows how in this region the values of periods in @xmath0 are constant , which is an indication of librating behavior in a secondary resonance . other regions of high @xmath64 values can be found at @xmath147 and , in the librating region , at @xmath148 ( since this region is so close to the separatrix , the chaotic behavior here observed might be related to the pulsating behavior of the separatrix itself , when perturbations from other jovian planets are considered ) . we believe that the features of high @xmath64 at @xmath147 and @xmath149 are associated with two other secondary resonances , whose resonant arguments are given by @xmath150 and @xmath151 , respectively . [ fig : res . arguments]c and d show the behavior of the argument for orbits near these two resonances . other weaker resonances connected with other commensurabilities between @xmath0 and @xmath106 are also observed . the resonance of argument @xmath152 is expected to be stronger than the one associated with the 3:4 commensurability , but its location is so close to the separatrix that its effect is difficult to discern . table [ table : resonant arguments ] lists the resonances for which the librating behavior of the resonant argument is observed , and a few candidates that could explain features of weaker chaos . so far we have only discussed the case of circulating test particles . features of weaker chaos , however , exist also in the librating region . unfortunately , in this case plotting the resonant argument is not so easy , since , by the definition of libration , @xmath0 oscillates around @xmath4 instead of covering all values from @xmath153 to @xmath1 . a possible way to overcome this problem would be to make a change of variables , so as to put the origin of the system of coordinates for @xmath154 and @xmath155 at the libration center . the new angle @xmath156 would then rotate from @xmath153 to @xmath1 , and it would then be possible to check for the behavior of the resonant arguments by combining @xmath156 with gi and other terms . this procedure is rather cumbersome , especially since the libration point itself is not a fixed point , but , due to perturbations from the other jovian planets , oscillates with timescales associated with the great inequality , @xmath157 , etc . fortunately , there is an alternative method to determine whether a resonance has in its resonant argument terms associated with the great inequality . it is widely believed that planets have migrated since their formation ( malhotra 1995 , gomes 2003 , levison and morbidelli 2003 ) . in particular , the gravitational scattering of a planetesimal disk modified the positions of the jovian planets so that , while saturn , uranus and neptune migrated outward in semimajor axis , jupiter , being the most effective scatterer , migrated inward . as the work of malhotra suggested , the origin of the highly eccentric , inclined , and neptune - resonance - locked orbits of pluto and the plutinos might be explained in the context of sweeping resonant capture due to the changing position of neptune s orbit . our work has shown that a major source of chaos for test particles in kozai resonance around saturn is due to a secondary resonance between the argument of pericenter @xmath0 , the argument of the great - inequality ( i.e. , @xmath158 ) , and terms connected with the planetary frequencies @xmath91 and @xmath9 . it is well known that by altering the positions of jupiter and saturn , the period of the great inequality changes . [ fig : great - inequality period ] shows how , by varying the initial value of the osculating semimajor axis of jupiter by a positive amount ( and keeping the position of saturn fixed ) it is possible to modify the period of the great inequality from the current value of 883 years to values of @xmath159 yrs or less . ) . however , in this work , we are interested in the actual effect of planetary migration , which alters the semimajor axes of the planets , and so we prefer to use the first method . ] the average and minimum values of @xmath0 s libration periods for particles in kiviuq s region are 640 yrs and 480 yrs , respectively . presently , the great - inequality resonance has one location , just outside the separatrix , in the region of circulating particles . an interesting question would be to see what happened when the great - inequality resonance had a shorter period . in particular , when the value of the great inequality was shorter than @xmath160 yrs ( the maximum period of libration ) , a new secondary resonance should appear in the region of librating orbits . when the great - inequality s period was 640 yrs , several particles in the libration island would have been in the region of this new secondary resonance . this can be seen in fig . [ fig : great_inequality_change ] which shows the results of the integration of our low - resolution survey integrated under the influence of saturn and a jupiter whose orbit was modified so that the period of the great inequality was a ) 640 yrs , and b ) 480 yrs . we called these simulations `` static integrations '' , as opposed to other simulations to be discussed in the next section for which the period of the great inequality is not fixed but changes with time . note how the position of the high - chaoticity region , that our simulations show to be connected to the 1:1 resonance between @xmath0 and the great inequality , is displaced upward with respect to the traditional integration for the first case , and is very close to the center of the libration island for the second case . more important , a significant fraction ( 25% ) of orbits that stayed inside the libration island in the integration with the present configuration of the planets were on switching orbits when the period of the great inequality was 640 yrs . these simulations show that indeed a 1:1 resonance between @xmath0 and the gi ( and the other terms needed to satisfy the dalembert rules ) was present in the libration region when the gi s period was lower . coming back to our problem of identifying secondary resonances for the high - resolution survey , the method applied to our low - resolution survey to change the gi s period can give precious insights on the identity of those resonances . [ fig : frequency_x ] shows how the frequency with largest amplitude in the spectra obtained with fam ( just the frequency with largest amplitude , not necessarily the frequency associated with @xmath0 s precession ) for our high - resolution survey changes with @xmath130 . when a secondary resonance is encountered , values of the frequencies spread around rather than following a quasi - linear behavior ( the feature associated with the pericentric secular resonance at @xmath161 is most instructive ) . to test our interpretation of the secondary resonances , we modified jupiter s position so that the gi period became 810 , 883 ( current value ) , and 950 yrs , respectively , and integrated the high - resolution survey with the three configurations . figs . [ fig : frequency_x_3]a , b , and c illustrate what happens to the frequencies for these three values of the gi , for the region of the 1:1 resonance , the 3:4 , and for librating particles , respectively . fig . [ fig : frequency_x_3]a shows the behavior in the region of the 1:1 resonance . the resonance position ( computed using eq : [ eq : f_resonant ] and the different gi periods ) shifts toward larger @xmath130 when the gi period becomes longer . when the period equals 810 years , the position of the resonance corresponds to that of the pericentric secular resonance ( note how its position does not move when the gi period is changed ) , and it goes to higher values of @xmath130 when we increase the period . the vertical lines indicate the expected position of the resonance when the gi period is modified . the fact that there is an excellent agreement between the predicted position of the resonance and the results of our numerical simulation seems to further confirm our hypothesis for the source of chaos in this region . the middle row of fig . [ fig : frequency_x_3 ] shows the same plots , but for @xmath162 . this region contains two resonances of argument @xmath150 , and @xmath151 : these are given the short - hand notation of 4:3 and 3(1:1 ) . once again , the numerical simulations demonstrate that the chaos zones move with the resonances , i.e. , they agree with the predictions based on our resonant arguments ( vertical lines ) . finally , the bottom row of fig . [ fig : frequency_x_3 ] shows the values of the frequencies of largest amplitude for the region of librating satellites . the dashed vertical lines represents the transition region in which particles alternate from circulating to librating behavior . the left panel shows substantial frequency diffusion near the separatrix , when @xmath163 yrs , not observed for longer great - inequality periods . we believe that this is due to the appearance of the 1:1 resonance ( having argument @xmath164 ) in the region of librating particles ( see the vertical line on the plot ) . this resonance is simply out of the range of periods in @xmath0 for librating particles when @xmath165 yrs . regarding other resonances in the region , we are still limited by the problem of plotting the resonant argument . however , our simulations suggest at least two other resonances in this region : @xmath166 and @xmath167 . the bottom row of fig . [ fig : frequency_x_3 ] shows how the expected positions of these resonances shift when the great - inequality period is modified . in each case , the expected position is very close to regions of frequency variation . we believe this is strong circumstantial evidence in favor of the existence of these resonances . other resonances containing combinations of only @xmath137 or @xmath136 instead of @xmath138 ( plus any other term needed to satisfy the second dalembert rule ) in their argument are also possible , but , as observed in other cases , are associated with features of weaker chaos ( their locations are not shown in fig . [ fig : frequency_x_3 ] ) . regarding the bottom row of fig . [ fig : frequency_x_3 ] , we observe that the resonance positions shift toward the separatrix as the gi period increases . to conclude , we have identified several secondary resonances connected to the great inequality for the region near the separatrix of the kozai resonance . the likelihood that the jovian planets had different past positions means that the locations of these resonances might have also been different . this introduces interesting new prospectives for the stability of primordial satellites inside the kozai resonance , that we now investigate . we have just argued that a web of secondary resonances involving @xmath0 , the argument of the great inequality , and other terms exists in the region of phase space around the separatrix between libration and circulation . considering that the period of the great inequality might have been different in the past , we ask ourselves if a mechanism of sweeping secondary resonances inside the kozai resonance , not dissimilar from the one that acted in the kuiper belt , could also have acted inside the kozai resonance . if that is the case , what are the repercussions for the stability of satellites orbits in the kozai resonance ? to address these questions we need to a ) have a model of planetary migration and b ) have an integrator able to simulate the effect of planetary migration for jupiter and saturn . following malhotra ( 1995 ) , we assumed that the semimajor axis varied as : @xmath168 where @xmath169 is the semimajor axis at the current epoch , @xmath170 is the change in semimajor axis ( equal to -0.2 au for jupiter and to 0.8 au for saturn ) and @xmath47 (= @xmath171 ) is a characteristic timescale for migration . for the integrator to simulate planet migration , we followed this recipe . we modified swift - whm so that an additional drag - force was applied to each planet along the direction of orbital velocity ( by the symbol * @xmath172 * we identify an unit vector along the direction of orbital velocity ) . this produces an acceleration : @xmath173 where * v * is the acceleration and @xmath174 is the initial position of a planet ( malhotra 1995 ) . we neglected the effect of the planetesimals perturbations on the satellites . [ fig : planet_migr ] displays several computed evolutions of saturn s and jupiter s semimajor axes . the two curves agree well . these are to be compared to the results of beaug _ et al . _ ( 2002 ) who used swift - rmvs3 to simulate the evolution of a disk of 1000 massless planetesimals subjected to the gravitational perturbations of the four major outer planets . the evolution of the planet s semimajor axes generally follows the exponential law of eq . [ eq : a_vs_t ] , but in addition is affected by short - time variations due to the impulses of single encounters . having now the tools for our investigation , we must define initial conditions for our simulations . to investigate the effect of planetary migration on the stability of orbits in the kozai resonance , we first integrated kiviuq with the integrator that accounts for planetary migration backward in time for 10 myr . not surprisingly , kiviuq remained inside the kozai resonance for the full length of the integration . we then used the last 100,000 yrs of the integration to compute the averaged orbital elements of kiviuq and @xmath38 s value (= 0.4322 ) . using this information , we generated a low - resolution survey of test particles , exactly as in sec . we then integrated this new set of initial conditions forward in time for 10 myr . during this forward integration , the great - inequality s period increased from 90 yrs at the beginning of the simulation to the current value of 883 yrs at the end of the integration . [ fig : planet_migr_particles_fate ] shows the particles fates at left @xmath175 = 0 and right @xmath175 = 5 myr . 14% of the particles originally in the libration island were on circulating or switching orbits by @xmath175 = 5 myr . at the end of the simulation , 15% of the test particles originally in the libration island were no longer kozai resonators . the mechanism of sweeping secondary resonances seems to be effective in depopulating the resonance . this mechanism not only affects orbits near the separatrix , but also those particles well inside the kozai resonance . examples are the particles with @xmath176 = @xmath177 and @xmath178 = 0.80 and 0.83 , which are very close to kiviuq s orbital region . this shows that this region is also affected by the sweeping secondary resonances . we explain these results in the following way : when , owing to the planets migration , the secondary resonance s period approaches @xmath179 yrs , a few of the particle inside the kozai resonance ( but only a fraction , the process is not 100% efficient ) are captured into a secondary resonance , and move further out in the libration island until the secondary - resonance s period attains @xmath180 yrs . for such libration periods , the orbits are so close to the separatrix that escape becomes possible . this mechanism seems to be confirmed by the time evolution ( fig . [ fig : omega_period_50 ] ) in @xmath0 for one of these particles ( @xmath181 ) . the fact that most particles escape from the kozai resonance in the first half of the simulation is not surprising . the period of the great inequality reaches 600 yrs during the first 5 myr , i.e. , the average value of the precession period for orbits in the kozai resonance ( see fig . [ fig : periods ] ) . after 5 myr , the growing value of the great - inequality period sweeps fewer and fewer particles inside the kozai resonance , and the mechanism loses its efficacy . another interesting consideration is that not only librating particles are pushed toward the separatrix , also some circulators in general are captured by one of the secondary resonances and are carried toward the separatrix . the fact that circulators do not generally become librators is due to the fact that in order to cross the separatrix they have to reach very high values of the precession period in @xmath0 ( see fig . [ fig : omega periods ] ) . once close to the separatrix , they reach a highly chaotic region , and many particles are lost from the secondary resonances . this mechanism could explain why so many irregular satellites are currently found in the proximity of the kozai separatrix ( uk & burns 2004 for ) . a word of caution should be given about our results . we only used a smooth exponential evolution ( malhotra 1995 ) for the planet s migration . in the real solar system things were most likely different from this model ; e.g. , hahn and malhotra ( 2000 ) introduced short - time scale variations in neptune s outward expansion by adding some random jitter to the torque applied to neptune ( cf . the @xmath86 history by beaug _ ( 2002 ) shown in fig . [ fig : planet_migr ] ) . this jitter was parametrized by the standard deviation @xmath182 of the planet - migration torque in units of the time - averaged torque . for small values of @xmath182 ( @xmath183 ) , capture efficiencies in the 2:1 resonances were substantially unaltered , and only when @xmath182 reached values larger than 25 were capture probabilities reduced . in our case , we can assume that a similar mechanism should be at work , and that any significant amount of short - time variations in the semimajor axes of both jupiter and saturn might in principle reduce the capture efficiency into the great - inequality secondary resonance . since our work is , in many ways , exploratory , we believe the use of the exponential model is defensible at this stage of our study . but we acknowledge that a more realistic model for the motions of jupiter s and saturn s semimajor axes should be used , before realistic values of capture probabilities for our mechanism of sweeping secondary resonances could be computed . in any case , we believe that our mechanism of sweeping secondary resonances might be an effective scheme for destabilizing the primordial population of bodies originally in the kozai resonance . the mechanism should not be limited to kiviuq s region . the fact that the great - inequality s period swept through values starting from 90 yrs at the beginning of the simulation with planetary migration to the current 883 yrs , implies that an analogous mechanism should also have been at work in the regions of the other saturnian and jovian satellites presently in the kozai resonance , for @xmath0 libration periods of less than 1000 yrs . we have analyzed the orbital behavior of satellites inside and outside the kozai resonance for the phase space around the saturnian satellite kiviuq . our goal was to identify areas of chaotic behavior and to understand if chaotic diffusion might have played an important role , now or in the past , for orbital stability inside the kozai resonance . by applying the frequency analysis method ( fam ) and computing the maximum lyapunov exponents , we identified several secondary resonances . the major source of chaos for orbits in the circulating region are secondary resonances involving the argument of pericenter @xmath0 (= @xmath141 ) , the great - inequality s argument ( @xmath158 ) , the longitude of the node ( @xmath6 ) , and the planetary secular frequencies ( @xmath184 , and @xmath9 ) . the region of transition between circulation and libration of @xmath0 is strongly chaotic , when perturbations from the other jovian planets are considered . according to our secular model , variations in saturn s semiminor axis ( and , therefore in its eccentricity and semimajor axis ) are what drive the chaotic behavior , especially the transfer from circulation to libration . while secular variations modulated by the @xmath185 and @xmath9 frequencies are not negligible , once again our results indicate that the dominant effect is connected with perturbations having the great - inequality s period . also , contrary to what might be expected from the secular model , our results indicate that the chaotic layer at the boundary between circulation and libration is not symmetric , but shows an asymmetry between values of @xmath0 larger and smaller than @xmath118 . this is a consequence of other perturbations connected with secular pericentric resonance , evection inequality , and variations in saturn s semimajor axis , which are strongly dependent on the initial conditions for particles @xmath186 and the sun s position . our simulations suggest that the dominant effect is connected with the secular resonance , but variations in saturn s semimajor axis are not entirely negligible . from our simulations we can introduce the following scenario for chaotic orbits near kiviuq s kozai resonance . starting from librating orbits close to the libration center , we possibly identify two weak secondary resonances of argument @xmath187 and @xmath188 that introduce chaotic behavior . the first strong source of chaos is however connected with the transition region between circulation and libration : orbits start to switch back and forth from circulation to libration and the whole region is characterized by high values of @xmath64 . once in the circulating regime , a strong resonance of argument @xmath151 appears at @xmath130 = 0.73 . in addition to that , the dominant source of chaos , excluding the nearby pericentric secular resonance , is connected to a secondary resonance with the great inequality of argument @xmath164 ( these orbits have argument - of - pericenter periods of order 840 yrs ) . orbits in the region of this secondary resonance show some of the highest @xmath64 values among the test particles we studied . other secondary resonances connected with the great inequality are also observed . the fact that the dominant source of chaos for orbits in kiviuq s region is connected with the great - inequality secondary resonance opens interesting perspectives . it is believed that jovian planets formed in different locations than their current ones , and then migrated by gravitationally scattering the primordial population of planetesimals ( malhotra 1995 ) . the possibility that planets occupied different semimajor axes in the past has important repercussions , since the great - inequality s period is connected with the osculating semimajor axes of jupiter and saturn . in particular , when the two planets were closer together , the period of the great inequality was shorter . we simulated this effect first by changing the position of jupiter so that the great - inequality s period was 640 yrs and 480 yrs , respectively ( `` static integrations '' ) . these simulations showed that , indeed the position of the great - inequality s secondary resonance shifted higher . a twin resonance for librating particles appeared when the great - inequality s period equaled the average libration period ( 640 yrs ) , and the minimum ( 480 yrs ) ( this resonance is not visible today because the maximum period of libration inside the kozai resonance is of @xmath189 yrs ) . then we fully integrated jupiter and saturn with an integrator that simulates planet migration according to the exponential law of malhotra ( 1995 ) . our results show that a mechanism of sweeping resonances , analogous to the one that operated in the kuiper belt ( malhotra 1995 ) , or inside the 2:1 mean motion resonance with jupiter ( ferraz - mello _ et al . _ 1998 ) must have operated as well inside the kozai resonance . our simulations show that 15% of the satellites originally inside the kozai resonance became circulators or exhibited switching behavior when planets acquired their final orbits . indeed several particles were captured into the secondary resonance with the great inequality ( or other secondary resonances ) and , as a consequence , their amplitudes of oscillation in @xmath0 increased until the test particles reached the boundary between circulation and libration , where escape became possible . several particles on circulating orbits were also pushed toward the separatrix by an analogous mechanism of sweeping secondary resonances ( but none crossed the boundary to become a librator ) . we believe that a similar mechanism might have acted in the past also for the region of phase space associated with the other jovian satellites currently inside or near the kozai resonance . this process could explain why so many irregular satellites are currently found near the kozai separatrix ( uk and burns 2004 ) . in this work we used several numerical tools . in a few cases , like for example the median filter for our fam results , we introduced a technique , that , while employed in other fields ( for example in remote - sensing and in the analysis of data from martian probes ) , was not , to our knowledge , used previously in celestial mechanics for reducing fam results . still on the technical side , we created new integrators to account for the bretagnon solutions of the giant planets and to simulate planetary migration ( malhotra 1995 ) . we believe these tools might be useful for other studies , not only for the irregular satellites of jovian planets . several questions remain that might be the subjects of future research . in this work we could not compute reliable capture probabilities for the great - inequality resonances when planetary migration is considered , since short - period variations in the semimajor axes of jupiter and saturn should also be taken into account ( hahn and malhotra 2000 , beaug _ et al . . it would also be interesting to extend our studies to the neighborhoods of the other jovian satellites currently inside the kozai resonance , and see if a web of secondary resonances ( possibly involving the lesser inequality , the 2:1 quasi - resonance between neptune and uranus ) is also present for the case of those neptunian satellites in kozai resonance ( holman _ et al . we hope that this work might have opened interesting new prospects for the study of origins and stability of irregular satellites in kozai resonance . we are grateful to thomas loredo , sylvio ferraz - mello , richard rand , doug hamilton , and suniti karunatillake for useful comments and suggestions supported in part by nasa s planetary geology and geophysics program . more information about this work is available at : http://www.astro.cornell.edu/~valerio/fft . when measuring frequencies with a fmft , the shape of the frequency peaks can be modeled as gaussian , but our measurements are made at isolated points . this introduces what in signal processing is called `` impulse noise '' , which can be attenuated by the use of a median filter ( press _ et al . _ 1996 , pitas 2000 ) . in our case we took data on a running 3x3 matrix ; the value at the central point is the median of the nine elements . then we moved the matrix by one element and repeated the procedure . to compute values of @xmath64 for points along the boundary of our figure , we added two external rows and columns of points all with values of @xmath64 obtained from the average of all points in the figure . once all elements were computed , we calculated the average of the percentage difference between the values of @xmath64 before and after the median filter was applied . if the percentage difference was larger than 5@xmath190 , we repeated the procedure until convergence was assured . to test that our median filter is actually removing only noise and not actual data , we devised the following test . we applied our median filter to one of our low - resolution surveys , until the 5% convegence criterion was fulfilled . we then computed the difference @xmath191 between filtered ( @xmath192 ) and raw data ( @xmath193 ) , and computed the standard deviation of the difference @xmath194 . we then generated a fictitious matrix @xmath195 of noise following a gaussian distribution with standard deviation @xmath194 and zero mean , and added this fictitious noise to the filtered data , obtaining a matrix of @xmath64 values @xmath196 . the median filter was then reapplied to these new data until convergence was assured , and a new matrix @xmath197 obtained . to compare the two distributions @xmath192 of filtered data and @xmath197 ( data obtained after the filtration of the fictitious noise ) , we devised the following scheme . we computed a @xmath198 variable , whose expression is given by : where @xmath200 and @xmath201 are the errors on the values of the filtered data , intended as the difference between raw and filtered data . the variable so computed is supposed to follow a @xmath73 -like distribution . the problem is to determine the number of degrees of freedom of the distribution . this is not just given by the number ( 399 ) of data points in our low - resolution survey since , when taking the median , each value of @xmath64 is connected to the value of at least some of its eight neighbors . moreover , the process of filtering is applied several times , until convergence is assured , so that the number of degrees of freedom can be substantially smaller than the number of data points , in a way that is very difficult to estimate . however , as a rule of thumb , a value of @xmath198 considerably smaller than 399 is a relatively good measure of a good fit . in the case of our test , the value of @xmath198 was 55 , so we believe our test should be a convincing proof of the validity of our application of the median filter . holman j. m. , kavelars jj , grav t. , gladman b. j. , fraser w. , milisavljiec d. , nicholson p. d. , burns j. a. , carruba v. , petit j. m. , rousselot p. , mousis o. , marsden b. g. , jacobson r. a. 2004 , nature , in press .

the problem of soliton - soliton force is revisited . from the exact two solitons solution of a nonautonomous gross - pitaevskii equation , we derive a generalized formula for the mutual force between two solitons . the force is given for arbitrary solitons amplitude difference , relative speed , phase , and separation . the latter allows for the investigation of soliton molecules formation , dynamics , and stability . we reveal the role of the time - dependent relative phase between the solitons in binding them in a soliton molecule . we derive its equilibrium bond length , spring constant , frequency , effective mass , and binding energy of the molecule . we investigate the molecule s stability against perturbations such as reflection from surfaces , scattering by barriers , and collisions with other solitons .

1004.3000

the old - new interest in the problem of soliton - soliton intertaction and soliton molecules has been increasingly accumulating particularly over the past few years . this is mainly motivated by the application of optical solitons as data carriers in optical fibers @xcite and the realization of matter - wave solitons in bose - einstein condensates @xcite . one major problem limiting the high - bit rate data transfer in optical fibers is the soliton - soliton interaction . on the one hand , soliton - soliton interaction is considered as a problem since it may destroy information coded by solitons sequences . on the other hand , it is part of the problem s solution , since the interaction between solitons leads to the formation of stable soliton molecules which can be used as data carriers with larger `` alphabet '' @xcite . the interaction force between solitons was first studied by karpman and solovev using perturbation analysis @xcite , gordon who used the exact two solitons solution @xcite , and anderson and lisak who employed a variational approach @xcite . it was shown that the force of interaction decays exponentially with the separation between the solitons and depends on the phase difference between them such that in - phase solitons attract and out - of - phase solitons repel . this feature was demonstrated experimentally in matter - wave solitons of attractive bose - einstein condensates @xcite where a variational approach accounted for this repulsion and showed that , in spite of the attractive interatomic interaction , the phase difference between neighboring solitons indeed causes their repulsion @xcite . for shorter separations between the solitons , malomed @xcite used a perturbation approach to show that stationary solutions in the form of bound states of two solitons are possible . however , detailed numerical analysis showed that such bound states are unstable @xcite . stable bound states were then discovered by akhmediev @xcite and a mechanism of creating robust three - dimensional soliton molecules was suggested by crasovan _ recently , soliton molecules were realized experimentally by stratmann _ _ in dispersion - managed optical fibers @xcite and their phase structure was also measured @xcite . perurbative analysis was used to account theoretically for the binding mechanism and the molecule s main features @xcite . quantization of the binding energy was also predicted numerically by komarov _ _ @xcite . in refs.@xcite , a hamiltonian is constructed to describe the interaction dynamics of solitons . the mechanism by which the relative phase between the solitons leads to their force of interaction , and hence the binding mechanism , is understood only qualitatively as follows . for in - phase ( out - of - phase ) solitons , constructive ( destructive ) interference takes place in the overlap region resulting in enhancement ( reduction ) in the intensity . as a result , the attractive intensity - dependent nonlinear interaction causes the solitons to attract ( repel ) @xcite . a more quantitative description is given in refs . @xcite . in view of its above - mentioned importance from the applications and fundamental physics point of views , we address here the problems of soliton - soliton interaction and soliton molecule formation using the exact two solitons solution . this approach has been long pioneered by gordon @xcite where he used the exact two solitons solution of the homogeneous nonlinear schr@xmath0dinger equation to derive a formula for the force of interaction between two solitons , namely @xmath1 where @xmath2 is the solitons separation and @xmath3 is their phase difference . this formula was derived in the limit of large solitons separation and for small difference in the center - of - mass speeds and intensities , which limits its validity to slow collisions . with appropriately constructed hamiltonian , wu _ et al . _ have derived , essentially , a similar formula that gives the force between two identical solitons and reliefs the condition on slow collisions @xcite . here , we present a more comprehensive treatment where we derive the force between two solitons for arbitrary solitons intensities , center - of - mass speeds , and separation . we also generalize gordon s formula to inhomogeneous cases corresponding to matter - wave bright solitons in attractive bose - einstein condensates with time - dependent parabolic potentials @xcite and to optical solitons in graded - index waveguide amplifiers @xcite . many interesting situations can thus be investigated . this includes the various soliton - soliton collision regimes with arbitrary relative speeds , intensities , and phases . most importantly , soliton - soliton interaction at short solitons separations will now be accounted for more quantitatively than before . specifically , soliton molecule formation is clearly shown to arise from the time - dependence of the relative phase which plays the role of the restoring force . in this case , the force between the two solitons is shown to be composed of a part oscillating between attractive and repulsive , which arises from the relative phase , and an attractive part that arises from the nonlinear interaction . the time - dependence of the relative phase results in a natural oscillation of the molecule s bond length around an equilibrium value . the various features of the soliton molecule , including its equilibrium _ bond length _ , _ spring constant _ , _ frequency _ and _ amplitude _ of oscillation , and _ effective mass _ , will be derived in terms of the fundamental parameters of the solitons , namely their intensities and the nonlinear interaction strength . the two solitons solution is derived here using the inverse scattering method @xcite . although the two solitons solution of the homogeneous nonlinear schr@xmath0dinger equation is readily known @xcite , here we not only generalize this solution to inhomogeneous cases , but also present it in a new form that facilitates its analysis . the solution will be given in terms of the four fundamental parameters of each soliton , namely the initial amplitude , center - of - mass position and speed , and phase . the main features of the solution will be shown clearly such as the contribution of the nonlinear interaction to the actual separation and phase difference between solitons where it turns that the separation between the two solitons increases with logarithm of the difference between the amplitudes of the two solitons . furthermore , the general statement that _ a state of two equal solitons with zero relative speed and finite separation does not exist as a stationary state for the homogeneous nonlinear schr@xmath0dinger equation _ , will be transparently and rigorously proved . stability of soliton molecules is an important issue since , in real systems , perturbations caused by various sources such as losses , raman scattering , higher order dispersion , and scattering from local impurities , tend to destroy the molecules . to investigate the stability of the soliton molecules described by our formalism , we have considered three situations . first , we studied the reflection of the molecule from a hard wall and a softer one . while for the hard wall the molecule preserves its molecular structure after reflection , it generally breaks up for the softer ones due to energy losses at the interface . secondly , the scattering of the molecule by a potential barrier was also investigated . we show that the molecular structure is maintained only for some specific heights of the barrier . this suggests a quantization in the binding energy as predicted by komarov _ the oscillation period of the reflected molecule is noticed to be smaller than for the incident one . in addition , the outcome of scattering depends on the phase of the molecule s oscillation at the interface of the barrier . for instance , a dramatic change in the scattering outcome takes place if the _ coalescence _ point of the molecule lies exactly at the interface . in such a case , the otherwise totally transmitting molecule will now split into reflecting and tunneling solitons . thirdly , we have considered the collision between a single soliton with a stationary soliton molecule . the effects of different initial speeds , amplitudes , and phases of the scatterer soliton were studied . it turns out that for slower collisions , it is easier for the scatterer soliton to break up the soliton molecule , while for fast collisions the scatterer soliton expels and then replaces one of the solitons in the molecule . the phase of the scatterer soliton plays also a crucial rule in preserving or breaking the bond of the molecule , which can be used as _ key _ tool to code or uncode data in the molecule . the rest of the paper is organized as follows . in section [ exact_sec ] , we use the inverse - scattering method to derive the two solitons solution of the inhomogeneous nonlinear schr@xmath0dinger equation and present the solution in the above - mentioned appealing form . the main features of the solution will be discussed in subsection [ mainsec ] . the center - of - mass positions and relative phases will be derived in subsections [ com_sec ] and [ phases_sec ] , respectively . the force between solitons will be derived in section [ force_sec ] where gordon s formula will be extracted as a special case in subsection [ gordon_sec ] and our more general formula will be derived in subsection [ our_sec ] . in subsection [ num_sec ] , we compare our formula with the numerical calculation . in section [ molsec ] , we show the possibility of forming soliton molecules , derive their main features in subsection [ formation_sec ] , and investigate their stability in subsection [ stability_sec ] . we end in section [ conc_sec ] with a summary of results and conclusions . the details of the derivation of the two solitons solution and the center - of - mass positions are relegated to appendices a and b , respectively . matter - wave solitons of trapped bose - einstein condensates and optical solitons in optical fibers can be both described by the dimensionless gross - pitaevskii equation @xmath4 , time to @xmath5 , and the wavefunction @xmath6 to @xmath7 , where @xmath8 and @xmath9 are the characteristic frequencies of the quasi one - dimensional ( @xmath10 ) trapping potential in the axial and radial directions , respectively . in these units , the strength of the interatomic interaction will be given by the ratio @xmath11 , where @xmath12 is the @xmath13-wave scattering length . for the case of optical solitons , the function @xmath6 represents the beam envelope , @xmath14 is the propagation distance , @xmath15 is the radial direction , and the intensity - dependent term represents the kerr nonlinearity . in this case , scaling is in terms of the characteristic parameters of the fiber , as for instance in ref . the specific form of the prefactors of the inhomogeneous and nonlinear terms guarantees the integrability of this equation @xcite . for the special case of @xmath16 , the homogeneous case is retrieved . other interesting special cases have also been considered @xcite . as outlined in appendix [ appa ] , we use the darboux transformation method to derive the two solitons solution of this gross - pitaevskii equation , which can be put in the form @xmath17\nonumber\\ & + & \sqrt{\frac{n_2 \alpha_{22}(t)}{2}}e^{i ( \phi_{01}+\phi_{02}+\phi_2(x , t)+ \tan^{-1}({\alpha_2\over\alpha_1 } ) + \tan^{-1}({\alpha_4\over\alpha_3}))}\nonumber\\ & \times&{\rm{sech}}\left[\alpha_{22}(t)(x - x_{\rm cm2}(t ) ) + \frac{1}{2}\log\left(\frac{\alpha_1 ^ 2+\alpha_2 ^ 2}{\alpha_3 ^ 2 + \alpha_4 ^ 2}\right)\right]\label{exactsol},\end{aligned}\ ] ] where + + @xmath18 , @xmath19 , + + @xmath20 , + + @xmath21 , + + @xmath22 , + + @xmath23 , + + @xmath24 , + + @xmath25 , + + @xmath26 , + + @xmath27 , + + @xmath28 , + + @xmath29 , + + @xmath30 , + + @xmath31 , + + @xmath32 , + + @xmath33 , + + @xmath34 , + + @xmath35 , + + @xmath36 . + + the solution is put in this suggestive form to facilitate its analysis . the first sech part corresponds to the exact single soliton solution with center - of - mass position @xmath37 , width @xmath38 , phase @xmath39 , and normalization @xmath40 . hence , @xmath41 and @xmath42 correspond to the initial center - of - mass position and speed , respectively . the second sech term contains the same features in addition to a shift in both the center - of - mass position and phase . it should be noted , however , that @xmath43 , @xmath44 , @xmath45 , and @xmath46 correspond to the center - of - mass position and speed , normalization , and phase of the single noninteracting solitons . due to the interaction between solitons , these four characteristic quantities may not correspond exactly to the values of the same physical quantities as they did for the single soliton solution . for instance , @xmath37 will not correspond to the center - of - mass of one of the solitons . instead , the soliton may be shifted from that position due to the interaction with the other soliton . in the following , we present a detailed analysis of the locations and phases of the two solitons . inspection shows that there are two main regimes for the two solitons solution , namely the regime of resolved solitons and the regime of overlapping solitons . in the former case , the center - of - mass concept is well defined and analysis of the relative dynamics becomes feasible . the solitons are considered resolved as long as the two main peaks are not overlapping , which means that partial overlap may occur in this regime . the analysis in this section assumes the resolved solitons regime . in the resolved solitons regime , the argument of the second sech term of eq . ( [ exactsol ] ) , namely @xmath47 $ ] , simplifies to a function with three roots . the fact that the sech function is peaked at the roots of its argument , leads to that the second sech term corresponds to three `` solitons '' . this is shown in fig [ fig1 ] . we denote these solitons as the `` left '' , `` central '' , and `` right '' solitons with peak locations at @xmath48 , @xmath49 , and @xmath50 , respectively . we notice that the central soliton is located at the position of the soliton of the first sech term , namely near @xmath37 . further inspection shows that the two solitons at this location are out of phase and interfere destructively such that they do not appear in the total profile . therefore , the two solitons that our solution of eq . ( [ exactsol ] ) describes are in fact the left and right solitons arising from the second sech term . this is different from what one aught to conclude from the form of the exact solution , namely that the first sech term corresponds to one soliton and the second sech term corresponds to the other soliton . in general , the center - of - mass locations of the left and right solitons , @xmath48 and @xmath50 , do not match @xmath37 and @xmath51 . typically , @xmath48 will be shifted to the left of @xmath37 , @xmath50 will be shifted to the right of @xmath51 , while @xmath52 remains near @xmath37 . the amount of shift depends mainly on the normalization difference @xmath53 and the relative speed @xmath54 , as will be shown in the next subsection . an interesting general and exact result is that , for the homogeneous case @xmath16 , a state of two equal solitons , @xmath55 , with zero relative speed , @xmath56 , and finite separation , does not exist as an exact solution of the gross - pitaevskii equation . this can be proven by substituting @xmath55 and @xmath56 in @xmath57 to find that the right and left solitons migrate to @xmath58 and @xmath59 , respectively , while the center - of - mass of the central soliton matches exactly @xmath37 . furthermore , we show in section [ phases_sec ] that the phase difference between this central soliton and that of the first sech term equals , in this case , @xmath60 ; guaranteeing their destructive interference . thus , the three solitons disappear in such a special case and @xmath6 becomes the trivial solution . in view of the above , a need arises to derive formulae for the center - of - mass positions , @xmath48 and @xmath50 in terms of the solitons parameters , which will then be used to derive the force between the two solitons . in this section , we derive formulae for the three roots of @xmath57 which correspond to the locations of the left , central , and right solitons , @xmath48 , @xmath52 , and @xmath50 , respectively . to facilitate the derivation , we define @xmath61}$ ] , @xmath62}$ ] , @xmath63 , @xmath64 , @xmath65 , and @xmath66 . the equation @xmath67 is a third - order polynomial in both @xmath68 and @xmath69 . in principle , this equation can be solved algebraically for @xmath68 or @xmath69 . however , extracting @xmath15 from the resulting three roots will not be possible analytically for general @xmath70 . alternatively , and as can be seen from fig . [ fig1 ] , we can exploit the simple linear behavior of @xmath57 near its roots . assuming , without loss of generality , @xmath71 , and noting that in the resolved solitons regime the solitons separation @xmath72 is large , we argue in appendix [ appb1 ] that @xmath73 for all @xmath15 , @xmath74 for @xmath75 , @xmath76 for @xmath77 , and @xmath78 for @xmath79 . based on this , the center - of - mass position @xmath50 can be derived from a taylor expansion of @xmath57 in powers of large @xmath68 and @xmath69 . expanding @xmath57 in powers of large @xmath68 only and leaving @xmath69 arbitrary , accounts for @xmath52 and @xmath48 simultaneously . keeping terms up to first order of @xmath80 , we show in appendix [ appb1 ] that the position of the right soliton is given by @xmath81 where @xmath82 and the position of the the left soliton is given by @xmath83 where @xmath84 , @xmath85 , and @xmath86 and @xmath87 are given in appendix [ appb1 ] . the second and third terms on the right hand side of eqs . ( [ xreq2 ] ) and ( [ xleq2 ] ) account for the shift in the center - of - mass position with respect to the single soliton ones . the third terms are much smaller than the second ones since they decay exponentially with the solitons distance @xmath88 . in the limit @xmath89 and @xmath90 , both @xmath91 and @xmath92 take the form @xmath93}$ ] . thus , it is obvious that for @xmath94 , @xmath95 and @xmath96 . this agrees with our earlier result that two equal solitons with zero relative speed and finite separation , do not exist as an exact solution . in this section , we calculate the phases of the left , central , and right solitons in reference to the phase of the soliton of the first sech part in the two solitons solution . for simplicity , the special case of @xmath97 and @xmath98 will be assumed . the phase difference between the left , central , and right solitons on one hand and the soliton of the first sech term on the other hand is generally given by @xmath99 which by observing that for all @xmath15 @xmath100 reduces to @xmath101 this expression gives the phases of the right soliton @xmath102 , the central soliton @xmath103 , and the left soliton @xmath104 , for @xmath75 , @xmath52 , and @xmath48 , respectively . to calculate these phases we express the parameters @xmath105 in terms of @xmath68 and @xmath69 , as follows @xmath106 @xmath107 @xmath108 @xmath109 in general , @xmath73 for all @xmath15 , but @xmath74 only for @xmath110 , and @xmath111 for @xmath112 , as shown in appendix [ appb1 ] . + + * i. phase of the right soliton @xmath102 : * + + in this case @xmath113 which leads to @xmath114 and thus @xmath115 , @xmath116 , @xmath117 , @xmath118 . therefore + + + @xmath119 + + + and @xmath120 , which gives @xmath121 where @xmath122 is a function that depends on the ratio @xmath123 for nonzero @xmath124 and @xmath125 . + + * ii . phase of the central soliton @xmath103 : * + + in this case @xmath126 which gives @xmath127 , and hence @xmath128 , and @xmath129 . finally , we get @xmath130 the last result shows that , for @xmath131 , i.e. , two equal solitons with zero relative speed , the central soliton and the soliton of the first sech term of the two solitons solution are out of phase and therefore interfere destructively . + + + * iii . phase of the left soliton @xmath104 : * + + in this case @xmath132 which results in @xmath78 and @xmath133 , @xmath134 , @xmath135 , @xmath136 . therefore , @xmath128 and @xmath137 , which gives @xmath138 the phase difference between the left and right solitons is thus given by @xmath139 noting that @xmath140 for @xmath141 and small but finite @xmath125 , and @xmath142 for @xmath143 and small but finite @xmath124 , we finally conclude that the phase difference between the two solitons is given by @xmath144 the above results are verified in fig . [ fig2 ] , where we plot @xmath145 , @xmath102 , @xmath104 , and @xmath103 versus @xmath15 . the agreement between our estimated values and the exact curve is evident . in this section , we use the results of the previous two subsections to derive the force between the left and right solitons . the force is proportional to the acceleration of the solitons separation @xmath146\,e^{-\frac{1}{4 } g_0 { n_s } { x_d}(t ) e^{\gamma ( t)}}\label{deltaeq},\end{aligned}\ ] ] where @xmath147 and the coefficients @xmath148 are given in appendix [ appb2 ] . the first two terms on the right hand side of eq . ( [ deltaeq ] ) are the dominant ones since they correspond to the noninteracting solitons separation , @xmath149 , and their logarithmic shifts @xmath150 arising from the interaction between solitons . the time - dependence of @xmath151 originates from @xmath152 , @xmath153 , and @xmath149 . the acceleration can thus be derived @xmath154 \ , e^{-\frac{1}{4 } g_0 n_s { \delta}(t ) e^{\gamma ( t)}+3 \gamma ( t)}\label{force2},\ ] ] where the coefficients @xmath155 are given in appendix [ appb2 ] . in the last equation , we have used eq . ( [ deltaeq ] ) with only the first two terms of its right hand side to substitute for @xmath149 in terms of @xmath2 in the exponential factor . the first term on the right hand side of eq . ( [ force2 ] ) corresponds to the force due to the external potential which vanishes for the homogeneous case . the rest of the terms correspond to the force of interaction between the solitons . the interaction force depends , as expected , on the phase difference of the two solitons and decays exponentially with their separation . it should be noted that this equation is a generalization of gordon s formula @xcite in two aspects . first , it is derived for a time - dependent inhomogeneous medium . secondly we have , essentially , no restriction on the difference between the two solitons amplitudes and speeds ; apart from some extreme cases which were mentioned in appendix [ appb1 ] and will be discussed further below . for the homogeneous case , @xmath16 , and in the limits @xmath89 and @xmath90 , the acceleration formula , eq . ( [ force2 ] ) , simplifies considerably . an apparent inconsistency occurs when switching the order of these two limits , namely @xmath156 while @xmath157 which differ by an overall minus sign . the conflict is resolved by invoking eq . ( [ phasediff ] ) where it is shown that in the first case : @xmath158 , while in the second case : @xmath159 . therefore , the two approaches agree on the following result @xmath160 this is essentially gordon s result , eq . ( [ gordoneq ] ) , since in his derivation gordon took @xmath161 and soliton amplitude @xmath162 , @xmath163 , as can be seen in eqs.(1 ) and ( 6 ) of ref . @xcite . substituting @xmath161 and @xmath164 in the last equation , it becomes identical to eq . ( [ gordoneq ] ) . it should be mentioned here that eq . ( [ gordoneq ] ) was also derived in ref . @xcite using a perturbation analysis based on the inverse scattering method , and in ref . @xcite using a variational calculation . for nonzero @xmath152 and in the limits @xmath89 and @xmath90 , the acceleration formula takes the form @xmath165 this is a generalization to gordon s formula for the inhomogeneous case as modeled by eq . ( [ gp ] ) . depending on the specific form of @xmath152 , the two force terms , namely the external ( first term ) and the interaction ( second term ) , can be repulsive , attractive , or oscillatory . in addition , the phase difference @xmath3 also depends on @xmath152 . it is established in the homogeneous case , as will also be shown in section [ molsec ] , that the time - dependence of the phase difference is responsible for binding the two solitons in the soliton molecule . here , in addition to the possibility of forming soliton molecules , the dependence of @xmath3 on @xmath152 allows for controlling the parameters of the molecule such as its equilibrium bond length , period , and spring constant . this and possibly other interesting phenomena will be left for future investigation . to obtain an estimate of the accuracy of the general acceleration formula , eq . ( [ force2 ] ) , we calculate numerically the acceleration from the exact two solitons solution , eq . ( [ exactsol ] ) , and compare the two results . the distance between the two solitons of the function @xmath166 is determined using a numerical algorithm that employs our formulae for @xmath48 and @xmath50 given by eqs . ( [ xreq2 ] ) and ( [ xleq2 ] ) to calculate seed values . the distance is then differentiated numerically twice at @xmath167 . in fig . [ fig3 ] , we compare the two results . good agreement is obtained for @xmath168 . the analytical solution diverges at @xmath169 . the value at which divergence takes place is set by the specific choice of parameters in fig . as pointed out in section [ com_sec ] , the divergence occurs due to the merging of the central and left solitons . this artifact divergency can be remedied by associating the location of the local maximum of @xmath57 to @xmath48 once this maximum has reached the @xmath15-axis from above . restricting our study to the region where agreement is obtained , we interestingly notice that the acceleration is oscillating between positive and negative values . this means that the force between the solitons is oscillating between attractive and repulsive . the possibility of having attractive forces for finite @xmath124 is particularly interesting ; for two solitons with nonzero relative positive speed , i.e. the solitons are initially diverging from each other , the force between them is attractive . this suggests that , if the force remains attractive for sufficient time , the two solitons will slow down and eventually converge at some point . if true , this should occur at small distances since the force decays exponentially with distance and when the two solitons are allowed to diverge even for a short while , the force might be weakened such that the two solitons can not return back . to be able to judge on such a possibility , we need to know what happens to the acceleration at later times . to that end , we calculate numerically the acceleration in terms of @xmath124 and @xmath14 . the result is plotted in fig . [ fig4 ] , where it is clear that the acceleration indeed decays with time for all @xmath124 . this leads to that any nontrivial effect of the oscillating force is most likely to take place at short solitons separations . this is what we find in the next section where the possibility of forming stable soliton molecules is pointed out . we have shown in the previous section that , as a result of the solitons time - dependent relative phase , the force of interaction between solitons is oscillating between repulsive and attractive . since the force decays exponentially with the solitons separations , this oscillation will have a tangible effect only when the two solitons are close to each other . in this section , we investigate the force of interaction between solitons for short solitons separation . in such a special case , eq . ( [ deltaeq ] ) takes a simple form that accounts for the solitons separation in terms of their relative phase . using this formula , we show that the solitons will be bound to oscillate around some equilibrium distance where the phase plays the role of the restoring force . comparison with exact numerical calculations shows that this formula is accurate for almost the full range of the solitons separation , except at the coalescence point ( if any ) . in subsection [ formation_sec ] , we discuss the main features of the resulting soliton molecules , and in subsection [ stability_sec ] , we investigate numerically their stability in different scattering regimes . to focus on the role of relative phase , we simplify the analysis by restricting our treatment to the homogeneous case , @xmath16 , and zero relative speed , @xmath170 . we also set @xmath171 so that any separation between the solitons to be as small as possible which in this case arises only from the logarithmic shifts ( @xmath172 in eq . ( [ deltaeq ] ) ) . in this case , the solitons separation @xmath151 , given by eq . ( [ deltaeq ] ) , simplifies in the limit @xmath173 to @xmath174-\frac{4}{g_0 n_s}\log \left(\frac{g_0 n_d^2}{n_s}\right)\label{delmol},\ ] ] and the acceleration is given by @xmath175 it is noted that for @xmath176 or @xmath177 , gordon s formula is retrieved , but here with an explicit time - dependence of the phase , @xmath178 . this acceleration formula deviates considerably from gordon s formula for @xmath179 . specifically , for @xmath180 , @xmath151 diverges to @xmath59 at @xmath181 , which indicates that the two solitons coalesce . this is confirmed below by examining the exact solution at this condition . we note here that an approximate expression for the solitons separation was also derived in refs . in addition , our predicted molecule s oscillation frequency ( see eq . ( [ freqeq ] ) below ) agrees with these references . to verify this feature , we calculate numerically the distance between the two solitons directly from the exact solution , eq . ( [ exactsol ] ) , for different values of @xmath182 . for @xmath183 , the density plot in fig . [ fig5]a , shows a soliton molecule of two clearly resolved solitons with a separation oscillating around some nonzero equilibrium distance . approaching the solitons coalescence point with @xmath184 , the density plot in fig . [ fig6]a , shows the two solitons approaching each other more than the previous case . furthermore , this figure shows a slight bounce back by one of the solitons in the region of collision . approaching further the coalescence condition with @xmath185 , we indeed observe in fig . [ fig7]a that the two solitons merge almost completely . for a more quantitative comparison , we calculate numerically the center - of - mass trajectories of the two solitons . we show the trajectory curves in the density plots of figs . [ fig5]a-[fig7]a . in figs . [ fig5]b-[fig7]b , we plot the solitons separation obtained from formula ( [ delmol ] ) and the numerical trajectories obtained from the exact solution . it is clear from these figures that this formula agrees well with the exact soliton separation except near the collision region . in fig . [ fig5]b , the two solitons remain away from each other during the collision , and therefore good agreement is obtained with the exact result even in the collision region . in fig . [ fig6]b the two solitons approach each other further such that formula ( [ delmol ] ) does not account for the above - mentioned slight bounce of one of the solitons . in fig . [ fig7]b , agreement with the exact solution in the collision region is qualitative . we found that at the condition @xmath180 and for @xmath173 , the analytic curve overlaps with that of the exact solution ; apart from the horizontal segments where formula ( [ delmol ] ) diverges to @xmath59 . further insight is obtained by plotting the density profile of the soliton molecule at some specific times , as shown in figs . [ fig5]c-[fig7]c . in fig . [ fig5]c , we observe that the initial amplitude imbalance is never removed during the dynamics . instead , it becomes maximum when the two solitons are closest to each other . in addition , we notice that the oscillation amplitude of the larger soliton around its equilibrium position is larger . figure [ fig6]c shows clearly the soliton bounce which takes place in the time interval @xmath186 to @xmath187 . in these subfigures we plot two vertical dashed lines that indicate the position of the solitons at the closest approach . it is clear that after first closest approach at @xmath186 , the right soliton bounces back with a maximum displacement at @xmath188 . in fig . [ fig7]c , it is shown that , although the two solitons coalesce , two small symmetric @xmath189 appear . a detailed examination of these wings shows that they are the remnants of the two solitons after they coalesce and they both bounce back in the collision region similar to the case of fig . [ fig6 ] . it is also instructive to show the dynamics of the phase profile during the molecule s oscillation . this is shown with the contour plots of fig . [ fig8 ] which correspond to the molecules of figs . [ fig5]-[fig7 ] . in fig . [ fig8]a , which corresponds to fig . [ fig5 ] , the two solitons start initially in phase . by time the phase of the right soliton , which is the one with higher intensity amplitude and larger oscillation displacement , starts to exceed that of the left soliton . at the point of closest approach , the phase difference is exactly @xmath60 . after that point , the two solitons diverge again , the phase difference starts to decrease , and the cycle is repeated . similar behavior is seen in fig . however , in fig . [ fig8]c , where the two solitons coalesce for a considerable amount of time , the phase difference during the coalescence time is zero . it is thus not completely understood why , in this case , the two solitons still repel each other and eventually split . having established the existence of the soliton molecule from the exact two solitons solution and derived a formula that describes its bond length , here we use this formula to examine more closely the propertie of the soliton molecule and its mechanism of binding . it is clear from eq . ( [ delppmol ] ) that the sinusoidal time - dependence of the solitons relative phase leads to a force of interaction that oscillates between attractive and repulsive and hence allowing for soliton molecule formation . further details of the mechanism of binding will be uncovered by expressing the acceleration , @xmath190 in terms of @xmath2 by substituting for @xmath191 from eq . ( [ delmol ] ) into eq . ( [ delppmol ] ) , to get @xmath192 this shows that the interaction force between the two solitons is the resultant of an attractive part and a repulsive part . the equilibrium bond length , defined by @xmath193 , is given by @xmath194 in consistence with our previous result , the equilibrium bond length diverges as @xmath195 . solving the last equation for @xmath196 and then substituting in eq . ( [ delpp2mol ] ) , @xmath197 simplifies to @xmath198.\ ] ] for small amplitude oscillations , @xmath199 , the last equation gives @xmath200 the restoring force ( @xmath201 ) originates from the phase - dependent terms , @xmath202 . this appealing form of the force of interaction shows that the force between the solitons is of hooke s law type with a spring constant @xmath203 where @xmath204 is the _ bare mass _ of the molecule . expressed in terms of @xmath205 and @xmath125 , the spring constant takes the form @xmath206 which shows that @xmath207 for @xmath143 ; corresponding to a soliton molecule of infinite bond length . furthermore , @xmath208 diverges for @xmath180 , which signifies soliton coalescence , as we have pointed out in the previous subsection . since the frequency of the soliton molecule is given by @xmath209 and the spring constant is given in eq . ( [ keq ] ) , the effective mass @xmath210 will be given by @xmath211 which again diverges at the soliton coalescence condition , @xmath180 . having determined the main properties of the soliton molecule , we can now return back to eq . ( [ delmol ] ) to express @xmath151 as @xmath212 where @xmath213 is the initial value of @xmath151 , which is given by the solitons parameters through @xmath214 the amplitude of the oscillation @xmath215 is thus given by @xmath216 which gives an elastic potential energy @xmath217 it should be noted here that this is equal to the mechanical energy since the initial speed vanishes , @xmath218 . the fact that the potential energy diverges at the coalescence condition @xmath180 is a gain an artifact of the calculation , but it at least indicates that the bond is tighter than cases where @xmath176 or @xmath177 . here , we investigate the stability of the soliton molecule against break up in the following three collision regimes : @xmath219 reflection by a hard wall , @xmath220 crossing a finite potential barrier , and @xmath221 collision with a single soliton . to that end we solve the gross - pitaevskii equation , eq . ( [ gp ] ) , numerically . as an initial state , we use , for cases @xmath219 and @xmath220 the two solitons solution , eq . ( [ exactsol ] ) , which represents the soliton molecule . for case @xmath221 , we use the superposition of the exact single soliton , eq . ( [ psi1_exact ] ) , with the two solitons solution . before starting the discussion of results , we point out that in figs . [ fig9]-[fig14 ] , we present the results of this section using spacio - temporal density plots . since the solitons are too thin compared to the spacial range that we consider , a density plot with full spacial and time ranges will not show a clear solitons peak density or center - of - mass path , as fig . [ fig9]c shows . to solve this problem , we restricted the density plotting to a finite range of @xmath166 , namely between 0.025 and 0.15 corresponding to the upper part of the solitons peaks . this results in an easier tracking of both the solitons peak density and center - of - mass path , as shown in fig . [ fig9]a , b , d and the rest of subsequent figures . for reflection from a hard wall we solve eq . ( [ gp ] ) with a potential step of the form @xmath222 where @xmath223 and @xmath224 are the hight and location of the potential wall , respectively . the result of reflection from this hard wall , with @xmath225 , is shown in fig . the soliton molecule preserves its molecular structure but with different characteristics . the solitons in the reflected molecule do not coalesce as in the incident molecule . in other words , the equilibrium bond length becomes larger . the density plot shows that initially the two solitons are of comparable intensities . after reflection , the brighter color of the left soliton and darker color of the right soliton indicate that the left soliton acquires higher intensity on the expense of the right soliton . we also notice that the left soliton performs two reflections from the potential interface . after the first reflection , it collides with the right soliton and then collides with the potential interface for the second time . the picture becomes different when the hight of the wall is reduced to @xmath226 , as shown in fig . the soliton molecule breaks up after reflection . this is due to loss of energy at the interface of the potential . part of the soliton molecule transmits as a nonsolitonic pulse that broadens and decays in intensity by time . by plotting @xmath166 in fig . [ fig9]c with its full range , we can see the nonsolitonic part as the left- and right - going two red ejections corresponding to the transmitted and reflected nonsolitonic pulses , respectively . in fig . [ fig9]d , we combine fig . [ fig9]b and fig . [ fig9]c to show the locations of the nonsolitonic ejections with respect to the solitons centers . in the case of reflection from a hard wall , the nonsolitonic ejections are essentially not present which results in the stability of the molecular structure . for reflection from a potential barrier we solve eq . ( [ gp ] ) with the potential @xmath227 where @xmath223 , @xmath228 , and @xmath224 , are the height , width , and location of the right side of the barrier , respectively . in fig . [ fig10 ] , we show the many different possibilities that result when the height of the barrier is changed . the free evolution case with @xmath229 is shown as a reference plot . the full reflection case is shown for @xmath225 , which is similar to the previous case of reflection from a hard wall . reducing the height of the barrier to @xmath230 , we notice that the soliton molecule breaks up after reflection . as pointed above , this is due to the nonsolitonic ejections taking place at the interfaces of the potential . reducing the height of the barrier to @xmath231 , a sign of solitons recombining appears in the form of a soliton molecule of a short lifetime . at @xmath232 a stable molecule is remarkably formed with a considerably shorter period than for the incident molecule . we have confirmed numerically that this molecule remains stable for much longer time provided that the soliton molecule remains sufficiently far from the boundaries of the spacial grid . this unique structure remains for some small domain around @xmath232 , but is lost for @xmath233 , where the molecule breaks for long evolution times . decreasing the height of the barrier to @xmath234 , the molecule breaks at the interface and splits into a reflected and transmitted solitons . for @xmath226 , the molecule breaks at the interface , but both solitons transmit through the barrier . for @xmath235 , the transmitted solitons show a sign of recombining again but with shorter period than for the free evolution case and larger than in the case of @xmath232 . motivated by the fact that at the coalescence point the intensity of the molecule is considerably higher than at other instants , we expect to find different scattering dynamics when the soliton molecule meets the interface of the potential at different phases of its periodic oscillation . in fig . [ fig11 ] , this is investigated by fixing the height of the potential barrier at @xmath232 and changing the initial launching position of the molecule . starting at @xmath236 , the molecule breaks after reflection . a shortly - lived molecule is obtained at @xmath237 , and a stable molecule is found for @xmath238 , which corresponds to the @xmath232 case of fig . [ fig10 ] . at @xmath239 , the soliton splits at the interface into transmitted and reflected solitons . the coalescence point is , in this case , located at the interface . transmission takes place due to the high intensity of the soliton at the coalescence point . at @xmath240 , the two solitons still split as in the previous subfigure but with a weak transmitted soliton intensity less than 0.025 and hence will not be shown in our density plots which are restricted to the intensities between 0.025 and 0.15 , as pointed out previously . for @xmath241 , the coalescence point takes place before the molecule reaches the interface and both solitons reflect but the molecule breaks up . for @xmath242 the two reflected solitons start to recombine forming a stable molecule at @xmath243 . at @xmath244 the reflected molecule starts to break up since the second coalescence point becomes close to the interface . thus , the conclusion from this figure is that the soliton molecule is more vulnerable to break up when it meets the interface at the coalescence point . equivalently , soliton molecules with larger equilibrium bond length , such that coalescence does not occur , will be more stable against breakup post reflection from barriers . finally , we present in figs . [ fig12]-[fig14 ] the results of scattering of a soliton molecule by a single soliton described by eq . ( [ newsol1 ] ) with normalization @xmath245 , center - of - mass position and speed @xmath246 and @xmath247 , respectively . the effects of the phase , speed , and amplitude of the injected soliton are investigated separately . in fig . [ fig12]a , a soliton initially at @xmath248 is launched towards a stationary soliton molecule near @xmath249 . at the impact , the molecule brakes up , its right soliton is ejected in the direction of the positive @xmath15-axis , and the left soliton combines with the scatterer soliton to form a new stationary molecule shifted by about a bond length to the left . we point out here that for such an outcome to occur , it is essential that the amplitude of the scatterer soliton is nearly equal to that of the right soliton of the molecule . otherwise , a different outcome , as that of fig . [ fig14 ] , will be obtained . in fig . [ fig12]b , the same numerical experiment is repeated but with adding a @xmath60 to the phase of the scatterer soliton . clearly , this phase addition prevents the formation of a new molecule resulting in three solitons diverging from each other . from applications point of view , the phase of @xmath60 could be used as a `` key '' to `` unlock '' the molecule for the purpose of extracting stored data . in fig . [ fig13 ] , we show the effect of the initial speed of the injected soliton . in contrary to one s first judgment , the molecule preserves its structure for fast collisions , as in fig . [ fig13]a , and breaks up for slower collisions , as in fig . [ fig13]b . in fig . [ fig14 ] , the injected soliton has an amplitude that is approximately two times larger than any of the two solitons of the molecule . the injected soliton penetrates the molecule leaving it almost unchanged apart from a center - of - mass shift to the left . we have used the inverse scattering method to derive the two solitons solution of a nonlinear schr@xmath0diner equation with a parabolic potential and cubic nonlinearity with time - dependent coefficients , as given by eq . ( [ gp ] ) . the solution was then simplified and put in a suggestive form in terms of the fundamental parameters of the two solitons , namely their amplitudes , center - of - mass positions and speeds , and their phases . in this form , two different regimes of the solution , namely the resolved solitons and overlapping solitons , were distinguished and the main features such as the solitons separation and relative phase were extracted . from the expression for the solitons separation we find that for the homogeneous case and zero solitons relative speed , the solitons separation diverges logarithmically with the solitons amplitude difference such that , for equal solitons , the trivial solution is obtained . the force of interaction was then derived , essentially , for arbitrary solitons parameters . this resulted in generalizing gordon s formula @xcite to _ i _ ) the generalized inhomogeneous case considered here , _ ii _ ) arbitrary solitons relative speed and amplitudes , and _ iii _ ) short solitons separations ( compared to their width ) . with this formula , the possibility of forming soliton molecules emerged naturally , where the force at short distance was shown to be composed of an attractive part , resulting from the nonlinearity , and another part that oscillates between repulsive and attractive resulting from the time - dependent relative phase . the main features of the soliton molecule , including its equilibrium bond length and bond spring constant , frequency and amplitude of oscillation , effective mass , and its elastic potential energy , where then calculated in terms of the solitons parameters . it turns out that the amplitudes difference @xmath63 plays an important role in determining these quantities . furthermore , we show that at the condition @xmath250 , the two solitons coalesce while away from this condition , the solitons approach each other but remain resolved . at this condition , the molecule s effective mass and spring constant have maximum values . in our expressions eqs . ( [ keq ] ) and ( [ masseq ] ) diverge because these formulae were derived assuming the solitons remain resolved . to have a sense of its stability we investigated numerically _ i _ ) the collision of the soliton molecule with a hard wall and softer one , _ ii _ ) scattering by a potential barrier , and _ iii _ ) collision with a single molecule . the first case showed that while the molecular structure is preserved after reflection from a hard wall , it breaks when reflecting from a softer one . reflection from a finite barrier showed that the molecular structure is preserved only for specific heights of the barrier . for an incident molecule with the coalescence condition satisfied , the molecule will be most vulnerable to break up when the coalescence point takes place at the interface of the barrier . this is simply understood by the fact that the intensity of the soliton molecule is maximum when the two solitons coalesce such that tunneling becomes possible . stability of the molecule was also investigated by scattering the molecule with a single soliton . it turned that slower collisions tend to break up the molecule more easily than faster ones . in addition , the outcome of the collision depends on the phase of the incoming soliton such that a scatterer soliton which is in - phase with the molecule will typically preserve its molecular structure , but for an out - of - phase soliton , the molecule breaks up . the two solitons solution presented here and the analysis that shows how to extract the solitons separation and relative phase may constitute the basis for a more accurate and detailed investigation of the origin of the soliton - soliton force , especially for short separations and at coalescence . the results of this paper will be hopefully of relevance to possible future applications of soliton molecules as data carriers or memories . the author acknowledges helpful discussions with vladimir n. serkin . the lax pair associated with the gross - pitaevskii equation , eq . ( [ gp ] ) , is obtained using our lax pair search method @xcite and reads @xmath251 @xmath252 where + + @xmath253 , @xmath254 , @xmath255 , @xmath256 , + + + @xmath257 , + + + @xmath258 , @xmath259 , and @xmath260 and @xmath261 are arbitrary constants . here , @xmath6 is the solution of eq . ( [ gp ] ) and @xmath262 is the _ auxiliary _ field . the _ compatibility condition _ @xmath263 of the linear system , eqs . ( [ phix ] ) and ( [ phit ] ) , is equivalent to the gross - pitaevskii equation , eq . ( [ gp ] ) , and its complex conjugate . for a known _ seed _ solution , @xmath264 , of eq . ( [ gp ] ) the linear system will have the solution @xmath265 . the darboux transformation is defined as @xmath266={\bf\phi}\cdot{\lambda}-{\bf\sigma}\,{\bf\phi}$ ] , where @xmath266 $ ] is the transformed field and @xmath267 . requiring the linear system to be _ covariant _ under the darboux transformation imposes the transformation @xmath268={\bf p}+{\bf j}\cdot{\bf\sigma}-{\bf\sigma}\cdot{\bf j}$ ] , where @xmath268 $ ] is the transformed @xmath269 . this gives the new solution in terms of the seed solution as @xmath270 using the trivial solution @xmath271 as a seed , the darboux transformation generates the well - known sech - shaped single soliton solution @xcite @xmath272 where @xmath273 @xmath274 @xmath275 @xmath276 , @xmath277 , @xmath278 , and the constant @xmath279 corresponds to an arbitrary overall phase . this solution corresponds to a soliton density profile that is localized at @xmath37 , moving with center - of - mass speed @xmath280 , and normalized to @xmath40 . using this single soliton solution as a seed , the darboux transformation generates a two solitons solution . the solution of the linear system , eqs . ( [ phix ] ) and ( [ phit ] ) , can in this case be derived and simplified to the following form @xmath281 @xmath282 @xmath283 @xmath284 @xmath285 where , @xmath286 @xmath287 @xmath288 @xmath289 finally , the two solitons solution is obtained by substituting for @xmath290 and @xmath291 in eq . ( [ newsol1 ] ) , which upon substituting @xmath292 and further simplification can then be put in the form of eq . ( [ exactsol ] ) . since @xmath68 and @xmath69 are functions of @xmath15 , we start by examining their values near the roots of @xmath57 . this task can be simplified by rewriting @xmath68 as @xmath293 . for @xmath294 , we get @xmath295 , and for @xmath296 , we have @xmath297 . for @xmath113 the first term in the exponent of @xmath68 is larger than the second one , provided that @xmath40 is not much larger than @xmath298 , as remarked at the end of this section , hence @xmath73 . for @xmath132 the magnitude of the first term in the exponent of @xmath68 is smaller than the magnitude of the second one , which again leads to @xmath73 . for the region @xmath299 , the condition @xmath73 is always satisfied . in conclusion , @xmath73 for all @xmath15 , apart from situations with extreme values of @xmath300 . the situation is simpler for @xmath69 : @xmath74 for @xmath113 , @xmath301 for @xmath302 , and @xmath303 for @xmath132 . thus , to find @xmath50 , we expand @xmath57 in powers of large @xmath68 and @xmath69 and to find @xmath48 and @xmath52 , we expand @xmath57 in powers of large @xmath68 . expanding @xmath57 in powers of @xmath68 and @xmath69 around @xmath58 up to first order in @xmath80 and @xmath304 , we find @xmath305 where @xmath306 the root of this equation gives the center - of - mass position of the right soliton . the first two log terms equal @xmath307 . the third log term is constant and diverges for @xmath94 . the last term is small since it is proportional to @xmath80 , but it is needed because it contains the phase - dependent contributions . due to the combination of @xmath308 and @xmath68 , finding an algebraic expression for the root of this equation will not be possible . instead , we ignore at first the @xmath80 term to obtain the dominant contribution which will then be used to find the @xmath80 contribution . this gives @xmath309 substituting back in eq . ( [ q1 ] ) and solving for @xmath15 , we finally get @xmath310 where @xmath311 . for the left and central solitons , we expand @xmath57 in powers of @xmath68 ; leaving @xmath69 arbitrary . in this manner , we account for the two roots , @xmath48 and @xmath52 , simultaneously . expanding @xmath57 in powers of large @xmath68 , we get @xmath312 where @xmath313\nonumber\\&/&{\left(g_0 ^ 2 e^{2 \gamma_0 } ( { n_d}+{n_s } { y } ) ^2 + 4 ( { y } + 1)^2 { \eta_d}^2\right)}.\end{aligned}\ ] ] similar to the above procedure for @xmath50 , we solve first eq . ( [ q2 ] ) without the @xmath80 term , which gives @xmath314 - 8 { \eta_d}^2}{2 g_0 ^ 2 { n_s}^2 e^{2 \gamma_0}+8 { \eta_d}^2},\ ] ] where @xmath315 in the limit @xmath89 and @xmath90 , the solution @xmath87 approaches 0 , which corresponds to @xmath316 . the solution @xmath317 approaches 1 , which corresponds to @xmath318 . thus , the solutions @xmath87 and @xmath317 correspond the left and central solitons , respectively . since we will be interested only in the left and right solitons , we take for the rest of this section the @xmath87 . to find the contribution of the @xmath80 term , we substitute for @xmath87 in the @xmath80 term of eq . ( [ q2 ] ) , and then solve for @xmath69 to get a corrected expression for @xmath87 @xmath319 -8 { \eta_d}^2}{2 g_0 ^ 2 { n_s}^2 e^{2 \gamma_0}+8 { \eta_d}^2},\ ] ] where @xmath320 , @xmath321 , @xmath322 , @xmath323 , and @xmath324 . expanding for large @xmath325 and then solving for @xmath15 , we finally get @xmath326 final remarks about the validity of the above derivation are in order . the condition @xmath73 will be met in the region @xmath113 only for @xmath327 . note that the numerator of the right hand side of this inequality is less than the denominator by at least @xmath328 . for @xmath132 , the condition @xmath73 will be met only for @xmath329 . here , the numerator of the right hand side of this inequality is larger than the denominator by at least @xmath328 . a more quantitative estimate for the ratio @xmath300 can be obtained using the above results for @xmath330 and @xmath331 . furthermore , the above - derived formula for @xmath48 is limited to values of the parameters for which the quantities @xmath332 and @xmath333 are positive . at @xmath334 , the two roots @xmath48 and @xmath52 coincide . in fig . [ fig1 ] , this corresponds to the maximum of the @xmath57-curve lying on the @xmath15-axis . for @xmath335 , the maximum of the curve is below the @xmath15-axis and the two roots become nonreal . noting that : + @xmath341 + @xmath342 + @xmath343 + @xmath344 + @xmath345 + @xmath346 , + + the acceleration , @xmath190 , can be calculated to take the form of eq . ( [ force2 ] ) with coefficients @xmath347 @xmath348 @xmath349 @xmath350 a. hasegawa and y. kodama , _ solitons in optical communications _ ( oxford university press , new york , 1995 ) ; l.f . mollenauer and j.p . gordon , _ solitons in optical fibers _ ( acadamic press , boston , 2006 ) ; g.p . agrawal , _ nonlinear fiber optics _ ( academic press , san diego , 2001 ) , 3rd ed . ; n.n . akhmediev and a. ankiexicz , _ solitons : nonlinear pulses and beams _ ( chapman and hall , london , 1997 ) . l. khaykovich , f. schreck , g. ferrari , t. bourde , j. cubizolles , l.d . carr , y. castin , c. salomon , science * 296 * , 1290 ( 2002 ) . m. stratmann , t. pagel , and f. mitschke , phys . * 95 * , 143902 ( 2005 ) . ) , @xmath47 $ ] . thick ( green ) curve : the soliton intensity @xmath351 . light ( red ) curve : the soliton intensity with only the second sech term of eq . ( [ exactsol ] ) . the parameters used are : @xmath16 , @xmath352 , @xmath353 , @xmath354 , @xmath355 , @xmath356 , @xmath357 , @xmath358 , and @xmath167.,width=566 ] ) . dotted ( blue ) curve corresponds to the phase of the right soliton , @xmath102 , calculated from eq . ( [ phaser ] ) . thick dashed curve corresponds to the phase of the central soliton , @xmath103 , calculated from eq . ( [ phasec ] ) . dashed - dotted curve corresponds to the phase of the left soliton , @xmath104 , calculated from eq . ( [ phasel ] ) . the solid curves correspond to the density profiles as in fig . the parameters used are : @xmath16 , @xmath352 , @xmath353 , @xmath354 , @xmath355 , @xmath359 , @xmath360 , @xmath357 , @xmath358 , and @xmath167.,width=566 ] , versus @xmath124 . thick ( blue ) curve is calculated from eq . ( [ force2 ] ) . light ( red ) curve is calculated numerically from the exact two solitons solution , eq . ( [ exactsol ] ) . the parameters used are : @xmath16 , @xmath352 , @xmath353 , @xmath354 , @xmath355 , @xmath359 , @xmath357 , @xmath358 , and @xmath167.,width=566 ] ) . * ( b ) * thick ( green ) curve is solitons separation calculated from the exact solution ( [ exactsol ] ) . light ( red ) curve is the solitons separation calculated from formula ( [ delmol ] ) . * ( c ) * density profile at some specific times . the parameters used are : @xmath16 , @xmath361 , @xmath362 , @xmath363 , @xmath356 , @xmath364 , @xmath358.,width=566 ] -[fig7 ] . subfigures * ( a ) * , * ( b ) * , and * ( c ) * , correspond to fig . [ fig5 ] , fig . [ fig6 ] , and fig . [ fig7 ] , respectively . the blue ( lower ) and green ( upper ) curves correspond to the center - of - mass trajectories of the left and right solitons , respectively.,width=566 ] ) . the dashed vertical line represents the interface of the potential step . the parameters used are : @xmath364 , @xmath366 , @xmath367 , @xmath368 , @xmath369 , @xmath370 , @xmath371 . * ( a ) * @xmath225 . * ( b),(c),(d ) * @xmath226 . ,

following the idea of a field quantization of gravity as realized in group field theory , we construct a minisuperspace model where the wavefunction of canonical quantum cosmology ( either wheeler dewitt or loop quantum cosmology ) is promoted to a field , the coordinates are minisuperspace variables , the kinetic operator is the hamiltonian constraint operator , and the action features a nonlinear and possibly nonlocal interaction term . we discuss free - field classical solutions , the quantum propagator , and a mean - field approximation linearizing the equation of motion and augmenting the hamiltonian constraint by an effective term mixing gravitational and matter variables . depending on the choice of interaction , this can reproduce , for example , a cosmological constant , a scalar - field potential , or a curvature contribution .

1201.4151

despite much recent progress @xcite , background - independent approaches to quantum gravity face several open challenges . these concern : ( i ) the definition of the quantum dynamics of the fundamental degrees of freedom of spacetime that they identify , and the full control over it ; ( ii ) the recovery of an effective description in terms of a smooth spacetime and geometry , once the dynamics is somehow defined , and in particular when the fundamental degrees of freedom are not continuous geometric data ; ( iii ) the contact with the effective dynamics of general relativity and quantum field theory , and with phenomenology . an example is given by loop quantum gravity ( lqg ) , a background - independent framework aiming to quantize the gravitational degrees of freedom in a nonperturbative way @xcite . to this purpose , a canonical quantization scheme is employed where the constraints are written in terms of the densitized triad and of the ashtekar barbero connection . the end result at the kinematical level is a hilbert space of ( spin network ) states associated with graphs embedded in the spatial manifold and labeled by algebraic data ( lorentz group elements or corresponding representation labels ) . as in any canonical scheme , while geometry is fully dynamical , the topology of the universe is fixed by construction , at least at the beginning . in general , however , one may ask whether it is possible to build a quantum theory inclusive of topology change or , in other words , if one can envisage an interacting multiverse scenario obeying a set of quantum rules . since the degrees of freedom of a single universe are already fields , eventually to be quantized , such a scenario is sometimes said to be one of `` third quantization . '' this can be achieved , at least at a formal level , by defining a field theory over the space of geometries , for given spatial topology @xcite . other ( albeit inconclusive ) arguments from a canonical quantum gravity perspective in favor of going to a `` third - quantization '' setting were also offered in @xcite . beside the issue of topology change , the main difficulty faced by the lqg approach is the complete definition of the quantum dynamics and the proof that the resulting theory leads back to einstein s gravity in an appropriate limit . a tentative but complete definition of the quantum dynamics of spin network states is obtained , via spin foam models @xcite ( a covariant definition of lqg dynamics ) , by embedding lqg states into the larger framework of group field theories ( gft ) @xcite , in turn strictly related to tensor models @xcite . these are quantum field theories on group manifolds whose states are indeed spin networks and whose feynman amplitudes are spin - foam models . this embedding has several advantages , from the lqg point of view . first of all , as said , it provides a complete definition of the quantum dynamics . second , it defines such dynamics as the superposition of interaction processes ( creation and annihilation ) of spin - network vertices , forming complexes of arbitrary topology , such that topology is naturally made dynamical ; it provides , in other words , a sort of local field - quantization scheme @xcite . third , the field - theory framework offers powerful mathematical and conceptual tools for tackling the issue of the continuum limit and of the extraction of effective dynamics for better contact with phenomenology . in doing so , however , one has to abandon the familiar framework of canonical quantization of a classical ( and local ) field theory of gravitation , and is forced to face new types of conceptual and mathematical difficulties . this program is just as ambitious as the original lqg one , if not more , and is difficult to realize in a complete and rigorous way , despite many recent advances . toy models inspired by the full theory then become very important . in fact , they fulfill three main purposes : ( i ) they offer a simplified testing ground for ideas and techniques developed in the full theory ; ( ii ) as such , they also have an important pedagogical value ; ( iii ) they may represent , in principle , an effective , approximate framework to which the full quantum dynamics may reduce , in some limit , and thus they may be directly applicable to phenomenological studies . obviously , due to their simplicity , one should be cautious in interpreting the result obtained in the context of such toy models as truly physical , and their validity can be assessed only once the relation between toy model and full theory has been understood . an important type of simplified scenario has been developed in the context of lqg , in a symmetry - reduced setting of interest for cosmology . in fact , in order to understand certain features of loop quantum gravity , one often resorts to a minisuperspace model , loop quantum cosmology ( lqc ) , where degrees of freedom are drastically reduced @xcite . in a pure friedmann robertson walker ( frw ) universe filled with a massless free scalar field , the classical and quantum dynamics of the universe as a whole can be described by the same formalism used for a free particle . in particular , the path integral is well defined @xcite and two - point correlation functions admit the usual classification @xcite . by now , a wealth of interesting results have been obtained in this context @xcite . given this analogy with the free particle , it is all the more natural to ask oneself if one can construct a sensible `` interaction '' among frw universes and , once this is done , to change the interpretation of the two - point function from particle transition amplitude to field propagator as in the usual field quantization . one would then write down a field theory on minisuperspace , to obtain a field - quantized lqc framework . another way to see the same field theory would then be as a toy model for group field theory , in which many of the difficult features of the latter are absent due to the global nature of the formulation and to the simplification provided by symmetry reduction , but where some ideas and techniques can still be applied . as with any toy model , one would then use it as a pedagogical testing ground and keep it available as a possible effective description of the full theory . we propose such a field theory for ( loop ) quantum cosmology in this paper , with the above motivations . the presentation is organized as follows . we review some basic features of lqc in sec . [ loqc ] , but the wheeler dewitt case is also easily recovered . in sec . [ gufc ] the field theory is defined by promoting the quantum hamiltonian constraint to the kinetic operator of a ( real ) scalar field @xmath0 on minisuperspace . we analyze the relation between different kinetic operators and the gauge choice , with particular focus on exactly solvable free theories . we discuss the various possible choices for the interaction term . following this general definition , we move on to analyze some consequences of the formalism . we analyze the free propagator of the theory first , corresponding to the evolution of a single universe ( sec . [ fft ] ) . we show how the embedding into a field - theory setting has immediate interesting consequences also for the single - universe dynamics . then , we consider how the presence of interactions affects this single - universe evolution . approximating the interaction as a mean - field term , we find an effective equation linear in the field @xmath0 , correcting the hamiltonian quantum constraint equation by an extra term ( sec . [ mfa ] ) . the latter mixes , in general , gravitational and matter degrees of freedom , and its exact form depends on the chosen initial interaction as well as on the mean - field configuration considered . we conclude with a discussion of other possible applications of this formalism . our starting point is the description within loop quantum cosmology of the spatially flat , homogeneous and isotropic universe with a massless scalar field as matter , which we summarize briefly in this section . in the canonical analysis of dimensionally - reduced general relativity , one restricts integrations to a fixed fiducial three - dimensional cell of comoving volume @xmath1 , with a flat metric @xmath2 which may be taken to be @xmath3 in cartesian coordinates . the four - dimensional metric is then ds^2 = -n^2(t ) dt^2 + a^2(t)^0 q_abdx^adx^b , [ frwmetric ] where @xmath4 is the scale factor , spatial indices are labeled by latin indices @xmath5 , and there is a freedom in the choice of the lapse function @xmath6 . indices @xmath7 will denote directions in the tangent space . with a choice of frame @xmath8 and dual @xmath9 , orthonormal with respect to @xmath2 , the physical triad @xmath10 ( @xmath11 ) and @xmath12 are orthonormal with respect to ( [ frwmetric ] ) , and the levi - civita connection is _ 0 = ^0 e^i,_j=0 , [ levi ] where a dot denotes time derivative . from this one computes the variables used in loop quantum gravity , the ashtekar barbero @xmath13 connection @xmath14 and the densitized triad @xmath15 , via [ lqgvaria]a_a^i & = & ( ^i_0)_a , + e^a_i & = & ( e)e^a_i = a^2 ^ 0 e_i^a , where @xmath16 is the barbero immirzi parameter . a shortcut to the standard canonical analysis is to substitute the frw metric and the ricci scalar stemming from ( [ frwmetric ] ) , r=6(-+ ) , into the einstein hilbert and matter action , which then depend on @xmath17 , @xmath18 , and @xmath19 . the conjugate momenta are @xmath20 and @xmath21 , and the conservation in time of the primary constraint @xmath22 ( the symbol @xmath23 denotes weak equality ) leads to the friedmann equation : = -0 , [ friedm ] which should be imposed as a constraint on quantum states in quantum cosmology . for the hamiltonian constraint because it will eventually be regarded as a kinetic operator . although this differs from the more standard choice of symbol @xmath24 or @xmath25 , it has the further advantage of avoiding confusion with the hubble parameter . ] focusing on the gravitational sector for now , the crucial difference from traditional minisuperspace ( wheeler dewitt ) approaches to quantum cosmology in lqc is that one follows the kinematics of full loop quantum gravity , where not the connection but only its holonomies are defined as operators @xcite . it is convenient to introduce new conjugate variables @xmath26 and @xmath27 , where c=_0 ^ 1/3=-,p = a^2_0 ^ 2/3 , so that @xmath14 only depends on @xmath26 , and powers of @xmath28 have been introduced to make @xmath26 and @xmath27 invariant under the residual symmetry @xmath29 , @xmath30 in eq . ( [ frwmetric ] ) . instead of @xmath31 and @xmath32 one now defines @xmath32 and @xmath33 as operators , where @xmath34 can be a real parameter or a function of @xmath27 chosen by means of a suitable procedure . the kinematical hilbert space @xmath35 is taken to be the space of square - integrable functions on the bohr compactification of the real line . one can work in a basis where @xmath32 is diagonal , with orthonormality relation @xmath36 , so that one is dealing with a nonseparable hilbert space . in this representation , if @xmath34 is taken to be a nontrivial function of @xmath27 the action of the holonomy operator @xmath33 takes a rather complicated form , and it is convenient to choose a different representation . in the improved dynamics scheme @xcite , where @xmath37 , this is a basis @xmath38 of eigenstates of the volume operator @xmath39 measuring the kinematical volume of the fiducial cell , @xmath40 , |= 2 g || | , where @xmath41 has dimensions of length . the states @xmath38 can be normalized to by multiplication , and @xmath42 , where @xmath43 is conjugate to @xmath44 [ and is proportional to the hubble parameter @xmath45 and @xmath46 , which acts as a shift in @xmath44 . these satisfy the standard heisenberg algebra . for the matter sector , one chooses the usual schrdinger quantization with a natural representation of the hilbert space @xmath47 , the space of square - integrable functions on @xmath48 , on which @xmath49 acts by multiplication and @xmath50 by derivation , and with an orthonormal basis given by [ phino ] | = ( - ) . the hilbert space of the coupled system is then just the tensor product @xmath51 . as in traditional approaches to quantum cosmology , the variable @xmath19 is removed from the configuration space because the primary constraint @xmath22 would mean that wavefunctions are independent from @xmath19 . we note that the full constraint would be a multiple of @xmath19 , so that in situations where the resulting quantum constraint depends on the choice of lapse function ( as below ) the choice @xmath52 seems more natural when considering that @xmath19 is also originally in the configuration space . in fact , in cosmology the lapse can be regarded as a function of the scale factor , @xmath53 , which is an independent variable . a choice of the form @xmath54 in the full theory is somewhat less justified before solving the constraints . the quantum analogue of the friedmann equation ( [ friedm ] ) is obtained by starting with the hamiltonian constraint of full general relativity in terms of the variables ( [ lqgvaria ] ) and expressing the curvature of @xmath14 through the holonomy around a loop , taking account of the _ area gap _ the result in lqg that the area of such a loop can not assume arbitrarily small nonzero values . the hamiltonian constraint is ( , ) : = -b()(+_^2)(,)= 0 , [ wdw ] where @xmath55 is a wavefunction on configuration space and @xmath56 is a difference operator only acting on @xmath35 and of the form -b ( ) ( , ) & : = & a()(+_0,)+c ( ) ( , ) + & & + d()(-_0 , ) , [ thet ] where @xmath57 , and @xmath58 are functions which depend on the details of the quantization scheme ( _ inter alia _ , on the choice of lapse function ) and @xmath59 is an elementary length unit , usually defined by the square root of the area gap ( the planck length up to a numerical factor ) . the physical states are the solutions of eq . ( [ wdw ] ) . due to the structure of eq . ( [ thet ] ) , in lqc one has an interval s worth of superselection sectors in @xmath35 : @xmath56 preserves all subspaces spanned by @xmath60 for some @xmath61 . we may restrict ourselves to one of these subspaces , i.e. , assume that wavefunctions only have support on a discrete lattice which we take to be @xmath62 [ for a generic gauge choice , there may be issues with the definition of ( [ wdw ] ) at the most interesting point @xmath63 . this restriction picks out a separable subspace to which we will limit our analysis . we now define our field theory on ( mini)superspace . the hamiltonian constraint ( [ wdw ] ) of the first - quantized theory is the natural starting point for the free action of the field theory . we define this action to be s_f[]=_d ( , ) ( , ) , [ akshn ] where in the simplest setting we take @xmath0 to be a real scalar field . if @xmath64 is as in eq . ( [ thet ] ) , we must assume that the combination @xmath65 is symmetric in @xmath44 , i.e. , that [ da ] d()=a(-_0 ) in eq . ( [ thet ] ) , in order to reproduce the equation of motion ( [ wdw ] ) . put differently , for any constraint ( [ thet ] ) the action ( [ akshn ] ) projects out its self - adjoint part with respect to the measure given by the kinematical inner product of lqc . taking this measure as given , possible manipulations of the constraint @xmath66 are restricted by this requirement . notice , however , that eq . ( [ da ] ) does hold in lqc for the usual choices of gauge , so we do not need to impose it as an additional requirement . to give an example , for the preferred lapse choice @xmath52 and in improved dynamics , the functions @xmath57 take the form @xcite & & a()=|+|| |+|-|+| & & b()=|| | |+|^-|-|^ |^3 , + & & c()=-a()-a(-_0 ) , [ complicated ] where now @xmath67 , whereas for the solvable `` slqc '' model in @xcite [ which uses @xmath68 and the symmetry ( [ symmetry ] ) ] , & & a()=(+),b()= , + & & c()=-a()-a(-_0)=- . [ slqcfunc ] equation ( [ slqcfunc ] ) can be shown to agree with the previous expressions ( [ complicated ] ) in the `` semiclassical '' limit @xmath69 . in lqc , one normally assumes symmetry of the wavefunction @xmath55 under orientation reversal , ( , ) = ( - , ) , [ symmetry ] since the kinematical hilbert space can be split into symmetric and antisymmetric subspaces which are superselected ( in other words , the physics should not depend on the frame orientation ) . from the field theory perspective , such a requirement is less natural , in particular if interactions are taken into account ; we will allow for general field configurations without assuming eq . ( [ symmetry ] ) . by definition of second quantization , and by construction in our case , the classical solutions of the free field theory will correspond to the quantum solutions of the first - quantized model . we now complete the definition of the field theory on minisuperspace with the addition of an interaction term for our field . the first - quantized theory , that is ( loop ) quantum cosmology , does not offer indications on how this interaction should be defined , so one has to proceed in a rather exploratory way guided only by general intuition ( and by the results obtained following various choices ) . one could take an arbitrary functional , but we opt for an @xmath70th - order polynomial in @xmath0 not necessarily local in the minisuperspace variables . we will specialize to simpler , concrete choices in the following , in order to study some consequences of the formalism . we get the general form for the interacting theory s_i[]&=&_d(,)(,)+_j=2^n[intact ] + & & _ _ 1 _jd_1 d_j f_j(_i,_i ) _ k=1^j(_k,_k),where @xmath71 are unspecified functions depending on @xmath72 . this gives the equation of motion ( , ) + _ j=2^n _ _ 1 _j-1d_1 d_j-1 & & [ fieldeq ] + _ l=1^j-1(_l,_l)_k=1^j _ k(_i,_i ; , ) & = & 0,where _ k(_i,_i ; , ) : = f(_i,_i ) , with @xmath73 and @xmath74 . in writing down the field theory action we have included possible nonlocal ( in @xmath44 ) quadratic terms in the interaction part ( specified by the interaction kernel @xmath75 ) rather than in the kinetic term , in order to emphasize the fact that @xmath66 is usually chosen to be a local operator in the geometry in quantum cosmology . these could however be also included in the definition of @xmath66 . the above is rather general . explicit analyses require choosing specific interaction terms , that is , functions @xmath71 . each of them could correspond to a choice of a physical quantity to be conserved through the `` interaction '' of the universes , and of a conjugate physical quantity in terms of which the fields interact , instead , `` locally . '' that means that the conjugate variable is identified across `` incoming '' and `` outgoing '' universes . different choices for the function @xmath76 , and hence for the quantities conserved in the interaction , can thus be motivated by different physical considerations . in particular , the choice will be influenced by the interpretation of such an interaction as true topology change , that is , splitting / merging of universes , or rather as a merging / splitting of homogeneous and isotropic patches within a single inhomogeneous and anisotropic universe . the detailed definition of the second scenario in physical terms is not easy , and we will confine our treatment to a brief discussion of it at the end . however , it is important to keep in mind that its consideration would sensibly affect the very definition of the field theory to analyze . examples of interactions are : * if , e.g. , only @xmath77 is nonvanishing and we implement locality in ( @xmath78 ) by choosing @xmath79 , we have the field equation ( , ) + ^2(,)=0 . such a potential , which implements locality both in the 3-volume ( i.e. , the scale factor ) and in the scalar field @xmath17 , implies the existence of conservation laws for the conjugate quantities @xmath80 and @xmath81 . these conservation laws are nothing but the ( modified ) second friedmann and klein gordon equations , respectively . since interactions allow for topology change , in this case for nonzero @xmath77 there is a process where two `` universes '' merge into one , and the metric in both ingoing patches as well as in the outgoing patch is required to be the same ( analogous to the interaction term proposed in @xcite ) . because of the conservation law , there is a discontinuity in the hubble parameter @xmath82 . * a different conservation law , namely conservation of hubble volume @xmath83 or locality in the conjugate quantity @xmath84 , is suggested if one interprets the interaction as the topology - changing process just described with a discontinuity in the causal past of an observer passing the `` merging point '' ( this is typically considered `` bad '' topology change ; see the review @xcite ) . * another possibility would be to take the conserved quantities of the classical ( wheeler dewitt ) hamiltonian @xmath81 and @xmath85 as also conserved in interactions , which would then be local in @xmath17 and @xmath86 . for the slqc model detailed below , the second quantity would be modified to @xmath87 and we would require locality in the conjugate variable @xmath88\}$ ] . in the wheeler dewitt case , we can choose the type of interaction analogously , since the only difference is in the kinetic operator ( choice of first - quantization scheme ) and not in the choice of minisuperspace variables . before starting our analysis of the model , we introduce a reformulation of the same that is advantageous for practical manipulations . since @xmath66 is not diagonal in @xmath44 , it may be convenient to use a fourier transform as in @xcite , & & ( b,):=_e^b ( , ) , + & & ( , ) = _ 0 ^ 2/_0 dbe^-b(b , ) . as @xmath89 is real , @xmath90 , and ( [ symmetry ] ) holds , then @xmath91 . the action ( [ intact ] ) becomes s_i[]&=&ddb(b,)(b,)+_j=2^n + & & ( ) ^j_l=1^j d_ldb_l|f(b_i,_i)_l=1^j(b_l,_l),where @xmath92 is the ( complex conjugate of the ) fourier transform of the function @xmath76 appearing in ( [ intact ] ) , and @xmath93 is an appropriate differential operator . in the above conventions , & = & e^_0 ba(-_b)+c(-_b)+a(-_b)e^-_0 b[cofb ] + & & + b(-_b)_^2 . the inverse of this operator will give the propagator . in the slqc case , eq . ( [ slqcfunc ] ) , in order to avoid nonlocal expressions such as @xmath94 in the action we could multiply the expression for @xmath56 by @xmath44 . however , this would lead to a `` nonsymmetric '' form such that eq . ( [ da ] ) is not respected [ obviously , @xmath95 and @xmath64 is not self - adjoint . a symmetrized version of @xmath64 , which has not been previously considered in the literature , has a()=(+)^2 , b()=1 , c()=-^2 . [ symme ] when the nonsymmetric form resulting from multiplication by @xmath96 is used in lqc , the kinematical inner product is modified accordingly to keep the constraint self - adjoint ( see , e.g. , @xcite ) . the action ( [ akshn ] ) , which involves both the constraint and the kinematical inner product of lqc , is left unchanged by such a redefinition . as mentioned above , the lqc setting is chosen because of the initial motivation to obtain a toy model for a gft construction , in turn a field theory formalism for lqg states . this is , however , not essential for our general purposes . one could define an analogous set of models for wheeler dewitt quantum cosmology , which is actually simpler from a technical point of view . the usual ordering for the quantum friedmann equation is @xcite : = ( a_a)^2-p_^2 , [ wdwcon ] and the scale factor @xmath18 is now a continuous variable . it is then convenient to define @xmath97 so that @xmath18 is restricted to be nonnegative and the constraint is simply the klein gordon operator @xmath98 . one ends up with a scalar field theory in 1 + 1 dimensions with standard kinetic operator and an unusual potential term . as any ordinary field theory , the above field theory on minisuperspace can be quantized and a fock space of quantum states can be constructed . the construction of the fock space rests on a choice of a complete basis on the space of fields . as customary , we can choose a basis of solutions of the free field theory , that is single - particle states in @xmath99 whose elements correspond to classical ( expanding or contracting ) solutions of the modified friedmann dynamics . ) ] . both operations can be motivated by the first - quantization framework of lqc , but they are performed already at the classical level . ] these solutions are labeled by @xmath100 , i.e. , 3-geometries ( described by @xmath44 ) embeddable into a four - dimensional frw universe by means of @xmath100 and a choice of lapse @xmath6 . in solvable lqc , the deparametrized solutions are of the form [ lit ] b()= , and @xmath100 is the value of the dirac observable @xmath81 . equation ( [ lit ] ) , specifying the classical trajectories , will later reappear as the `` light cone '' @xmath101 of the propagator of the theory , see eq . ( [ prpx ] ) . in the gauge @xmath68 , consider the modified hamiltonian [ moh ] = -6g\{(t)}^2 . hamilton s equations give @xmath102 and @xmath103 , solved by b(t)&= & , + ( t)&=&,[cosh ] so that the bounce is apparent already in each classical history ( we have treated @xmath44 as a continuous parameter for the purpose of simplicity here ) . in fact , eq . ( [ cosh ] ) is consistent both with ( [ lit ] ) ( @xmath104 are solutions of the constraint @xmath105 ) and with the expectation value of the volume operator ( [ coshv ] ) . the construction of the fock space proceeds as usual . one defines creation operators @xmath106 and annihilation operators @xmath107 from the mode decomposition of generic fields into the above basis , and builds generic elements of the hilbert space of states from their combined action on the fock vacuum state @xmath108 . this state , as in the complete gft formalism @xcite , would correspond to a very degenerate `` no - space '' state , in which no geometric and no topological structure at all is present . topological and geometric structures are created out of it by the action of the creation operators . a crucial ingredient in the definition of the fock space is the choice of quantum statistics . in the following , we fix the statistics to be bosonic , @xmath109\propto \delta_{k , k'}$ ] . this seems physically natural if quantum states are associated with classical geometries and whole universes . in group field theory , where the fock space would be constructed out of microscopic `` building blocks '' of space @xcite , the choice of statistics is less obvious and it is the focus of current research ( see for example the discussion in @xcite ) . we may also expect the situation to be subtler if the objects created and annihilated in our field theory are to be interpreted as local homogeneous and isotropic patches of a single universe . however , as anticipated , we do not discuss in detail this possibility in this work , and thus stick to the simplest choice of statistics . we now begin the analysis of the field theory we defined . we limit most of our considerations to solvable slqc , where explicit calculations can be performed . however , we try to maintain a certain level of generality in the presentation , to leave room for the study of other cases . we start with the analysis of the free field theory . we have seen that , already at this level , the field - theory setting implies certain restrictions on and modifications of the lqc hamiltonian constraint operators , and thus of the single - universe dynamics . we focus on these first . for slqc , the free equation ( [ wdw ] ) can be solved analytically . we reserve the symbol @xmath55 for field solutions of the free theory and the symbol @xmath0 for the classical field solutions of the interacting model . inserting eq . ( [ slqcfunc ] ) into ( [ cofb ] ) , one obtains _ b=\{_^2 - 12g^2}. [ slqc ] setting @xmath110 and choosing an ( irrelevant ) integration constant to be zero yields = -\{_^2 - 12g^2}=0 , with general solution @xmath111+\chi_-[\phi+x(b)]$ ] , where x(b)= , [ transf ] so that @xmath112^{-1}$ ] . note that @xmath80 parametrizes the circle @xmath113 from which the point @xmath114 ( or @xmath115 ) must be removed to make the coordinate transformation ( [ transf ] ) well defined . as @xmath116 , we see that @xmath117 . we will encounter this singular behavior when giving the expression of the propagator of the theory in terms of @xmath80 , see eq . ( [ prpx ] ) below . we may expand the general solution in fourier modes , ( b,)=_-^+dk\{_l(k)e^k[-x(b)]+_r(k)e^k[+x(b)]},[chi ] and hence ( , ) & = & _ 0 ^ 2/_0 dbe^-b_-^+dkx(b ) + & & \{a_l(k)e^k[-x(b)]+a_r(k)e^k[+x(b ) ] } + & = & _ -^+ dxe^-b(x)_-^+dk [ solution ] + & & , where @xmath118 and @xmath119 and @xmath120 are chosen so that @xmath55 is real . in lqc ( first - quantized theory ) , the existence of a bounce can be proven analytically with the state @xmath121 . noting that the volume operator in @xmath122 representation acts as @xmath123 , one can compute its expectation value on the state ( [ chi ] ) , with the scalar product | |=-idx ( ^*_||-||_^ * ) . since @xmath121 obeys a klein gordon equation in @xmath17 and @xmath122 and its left and right sectors depend on the combinations @xmath124 , it is easy to see that the cosh in the volume operator factorizes as a cosh in the scalar field , so that [ coshv ] | constraint , @xmath125 is some measure determined by the normalization of single - particle states @xmath126 [ by imposing @xmath127 , and @xmath128 is the fock vacuum . these states are eigenstates of the annihilation operators @xmath107 with eigenvalue @xmath129 , so that the field operator @xmath130 has expectation value |(,)|= ^2 , where @xmath131 is the solution to eq . ( [ wdw ] ) labeled by @xmath100 . clearly , this expectation value can then be equivalently viewed as a first - quantized ( real ) wavefunction @xmath132 , eq . ( [ vev ] ) . in both viewpoints there is a normalization condition : for @xmath133 to be in the fock space , one must have @xmath134 , while @xmath132 defines a first - quantized state if it has finite norm in the appropriate physical inner product , e.g. , ^2=_b()|(,_0)|^2 , at some fixed @xmath135 @xcite . for the solvable slqc model , where the general solution is eq . ( [ solution ] ) , we have to choose appropriate functions @xmath136 and @xmath137 in the mean - field approximation . if we consider only a single right - moving mode @xmath138 [ @xmath139 , @xmath140 , _ k_0(,)&=&_-^+ dxe^-b(x)e^k_0(+x)+ + & ~ & + & & ( k_0 ) . here we have used a stationary - phase approximation dxe^g(x)~_g(x_0)=0e^g(x_0)in the limit @xmath141 , in which we find that @xmath55 does not decay and , in general , does not satisfy a normalization condition . alternatively , we might consider a gaussian , giving a wave packet centered around a classical trajectory @xmath142 , similar to the wave packets studied by kiefer for wheeler dewitt quantum cosmology @xcite . the stationary - phase method then gives ( , ) & = & _ -^+ dxdke^-b(x)-(k - k_0)^2+k(+x)+ + & = & _ -^+ dxe^k_0(+x)-(+x)^2-b(x)+ + & ~&()^-()_k_0 ( , ) , + where we pick up an extra factor from the gaussian . for large @xmath44 , the field now falls off faster than any power of @xmath44 . the scalar field @xmath17 not only acquires an effective periodic potential [ @xmath143 for @xmath144 , but it also becomes nonminimally coupled with gravity via a nontrivial function @xmath145 . we can redo the analysis for wheeler dewitt cosmology using a wave packet of the usual form , ( a , ) & = & _ -^+dke^-(k - k_0)^2e^k[-n(a ) ] + & = & e^k_0[-n(a)]e^-[-n(a)]^2 , so that for large @xmath18 ( and at fixed @xmath17 ) we find again a fall - off behavior ( a , ) ~a^-a faster than any power of @xmath18 . in general , any statement about the explicit form of the contribution to @xmath66 will strongly depend not only on the chosen field configuration @xmath146 , but also on the form of the interactions in our model , which are not strongly constrained . let us discuss possible results for this effective hamiltonian dynamics , and how to obtain them . a possibility which has been suggested by studies in lqc @xcite is that , for a `` monomial '' interaction , the gft coupling constant @xmath147 is related to the cosmological constant @xmath148 . this possibility had been also considered previously @xcite , but , in analogy with matrix models and tensor models and with the `` third - quantization '' model of @xcite , the coupling @xmath147 would be expected to be related to the _ exponential of the cosmological constant _ rather than @xmath148 itself . for example , the formal arguments of @xcite suggest that a second - quantized field - theory feynman amplitude should correspond to @xmath149 evaluated on the classical spacetime represented by the feynman diagram . our simplified scheme suggests another way to obtain a relation between a fundamental field - quantized coupling and an effective cosmological constant . under the assumption that our toy cosmological model comes out from some more fundamental gft dynamics , and thus that the coupling constant of the toy model can be related to the fundamental gft interactions , the mean field approximation can relate very directly the coupling constant of the cosmological model to an effective cosmological constant in the single - universe dynamics ( free theory ) , that is , in an effective hamiltonian constraint . let us see how this can happen . the effective contribution to @xmath66 should be of the form @xmath150 . this term grows with large @xmath44 and , for a trivial polynomial interaction ( that is , trivial interaction kernels @xmath151 ) , presumably it could not come from a normalizable @xmath0 . however , a general choice of interaction functional , such as that in eq . ( [ intact ] ) , can accommodate an effective cosmological constant . actually , it can even reproduce a nonconstant scalar - field potential term , _ v = v()b()^2 , for example of the type that would be needed for inflation in the early universe . take , again , @xmath152 and eq . ( [ heff ] ) . the contribution in @xmath75 is nontrivial only if the effective hamiltonian is nonlocal , otherwise it would just be an extra piece defining the initial hamiltonian constraint . since we are interested in a local constraint , we can ignore it , set @xmath153 , and absorb @xmath77 in the potential @xmath154 . then , a function @xmath155 that happens to match the behavior of the mean - field vacuum and contains the appropriate dependence on @xmath44 and @xmath17 on top of it would be f_3(_i,_i)&= & ( _ 1-_2)(_2-_3)__1,_2__2,_3 + & & v(_3)b(_3)_3 ^ 2[_0(_3,_3)]^-1 . [ f3 ] notice that , by construction , @xmath155 is symmetric with respect to all its arguments . clearly , this is rather _ ad hoc _ and one should have an independent justification ( and possibly a full derivation from a fundamental gft ) for a given choice of interactions @xmath155 , _ such that _ the wished - for effective hamiltonian constraint comes out , _ for a reasonable choice of nonperturbative vacuum _ which should also be independently justified . however , the above derivation proves an intriguing possibility , to be explored further : an underlying , more fundamental dynamics of creation / annihilation of universes , i.e. , topology change , or of merging / splitting of homogeneous and isotropic patches within a single universe could result , at an effective level , in a nontrivial potential term for the ( homogeneous and isotropic ) scalar field , and a cosmological constant term . notice also that , in the first - quantized lqc and wheeler dewitt frameworks , the presence of a potential spoils the separation of positive - frequency and negative - frequency sectors , as recalled , e.g. , in @xcite . this is not an issue in our second - quantized model ( being a field theory ) , and we are able to generate a scalar potential ( nontrivial in the inflationary early universe ) without formally changing the structure of the free theory . with the same procedure , we can obtain also other types of effective contributions , for instance a nonvanishing curvature @xmath156 ( closed and open universe , respectively ) . it is sufficient to replace v()v()-^1/3 in eq . ( [ f3 ] ) to obtain an effective spatial curvature term of the form one finds in the classical friedmann equation . it is also straightforward to obtain an effective hamiltonian constraint corresponding to , e.g. , the @xmath157 model in lqc @xcite . in fact also in that case the term corresponding to spatial curvature only acts by multiplication with a function of @xmath44 . in this paper we outlined the construction of a field theory of universes drawing inspiration from the perspective advanced in quantum gravity by group field theory , and by the general idea of `` third quantization of gravity , '' that had been advanced in the early days of the subject @xcite . we have also studied various aspects of the formalism , in particular the consequences it has for the single - universe dynamics , that is , for the standard ( loop ) quantum cosmology setting . these come already from the embedding of the canonical dynamics within a field theory , as encoded in the free field theory . more interesting consequences , of course , come from the existence of interactions , which , we showed , can be taken into account via mean field approximation . given the subject , and the current level of understanding of the fundamental theory ( in either the lqg or in the gft formulation ) , our goal was then in many respects necessarily of an exploratory nature . in particular , the least developed point of the discussion concerns the choice of interaction . as in all simplified models for cosmology , however , the main task will be to better justify the assumptions and the dynamics chosen from the fundamental theory , and to possibly show how such a simplified model can emerge naturally in some sectors of the full theory . understanding this issue will also clarify the role and physical significance of conserved currents within the model . before concluding , we should mention an alternative view of the physics of this `` group field cosmology , '' that we anticipated in passing in the course of our presentation . while loop quantum or wheeler dewitt cosmology describe a single evolving quantum universe , in the field cosmological model one has many - particle interacting states . instead of interpreting them as @xmath70 distinct universes merging and splitting in topology - changing `` scattering '' processes , one could think of them as @xmath70 frw patches which , collected together , approximate a single inhomogeneous universe . this is reminiscent of the separate universe approach @xcite , where inflationary large - wavelength perturbations are represented as spatial gradients among homogeneous patches of hubble size . in each patch centered at some spatial point @xmath158 , one has a `` local '' scale factor @xmath159 , hubble parameter @xmath160 , and so on . in particular , the local scale factor @xmath161 $ ] encodes both the minisuperspace variable @xmath18 and the nonlinear scalar perturbation @xmath162 . linear perturbations can then be identified with gradients . at the linear level , @xmath163 $ ] ; call @xmath164 . one has @xmath165/(x_1^i - x_2^i)\sim \partial_i a(t,{\bf x})$ ] , so that , up to a numerical factor , for a perturbation of wavelength @xmath147 we get @xmath166 . in our field - theory picture , the present - day universe would resemble some configuration of many particles ( regions of linear size @xmath167 ) that looks homogeneous to high precision at large scales . this could be a condensate phase of the theory where discrete translation invariance in the @xmath44 variable is spontaneously broken , as is known for systems in condensed matter physics . then , the challenge would be to define a model whose collective behavior agrees with the standard cosmological perturbation theory . however , the resulting action would be nonlocal in order to accommodate the infinite multiplicity of spatial points into a finite - dimensional , minisuperspace - like phase space . for instance , following the above - mentioned gradient expansion , a linear perturbation would be defined via the `` interaction '' of two patches in a nonlocal quadratic term : da dda d f(a , a,,)(a,)(a, ) , where @xmath168 is a function which should encode the correct dynamics to match with the perturbed hamiltonian constraint . just as we did for the effective hamiltonian constraint in the previous section , one could explicitly calculate @xmath168 from this perturbed hamiltonian constraint . the main difficulty to overcome , in developing properly the separate universe perspective of our second - quantized cosmology , is to develop first a proper quantum cosmology version of the separate universe approach to cosmological perturbations . admittedly , it is unclear at the present stage whether applying our framework to the problem of cosmological perturbations has practical advantages over conventional strategies , unless some conservation law be implemented . it is yet another possibility worth exploring , however . on the other hand , the implementation of the separate universe idea within a field - theoretic formalism like the one we propose could have a better chance of being derived from fundamental formulations of quantum gravity such as gft . this would give a more solid ground to this way of dealing with cosmological perturbations and also offer a way to test fundamental models via their cosmological predictions . this research was supported by perimeter institute for theoretical physics . research at perimeter institute is supported by the government of canada through industry canada and by the province of ontario through the ministry of research & innovation . support from the a. von humboldt stiftung , via a sofja kovalevskaja award , is gratefully acknowledged . 99 d. oriti ( ed . ) , _ approaches to quantum gravity _ ( cambridge university press , cambridge , u.k . , 2009 ) . c. rovelli , _ quantum gravity _ ( cambridge university press , cambridge , u.k . , 2007 ) . t. thiemann , _ modern canonical quantum general relativity _ ( cambridge university press , cambridge , u.k . , 2007 ) ; t. thiemann , http://arxiv.org/abs/gr-qc/0110034[gr-qc/0110034 ] . s.b . giddings and a. strominger , http://dx.doi.org/10.1016/0550-3213(89)90353-2[nucl . phys . b * 321 * , ] http://dx.doi.org/10.1016/0550-3213(89)90353-2[481(1989 ) ] . s. coleman , http://dx.doi.org/10.1016/0550-3213(88)90097-1[nucl . b * 310 * , 643 ( 1988 ) ] ; t. banks , http://dx.doi.org/10.1016/0550-3213(88)90455-5[nucl . b * 309 * , 493 ( 1988 ) ] ; v.a . rubakov and p.g . tinyakov , http://dx.doi.org/10.1016/0370-2693(88)91373-1[phys . b * 214 * , 334 ( 1988 ) ] ; m. mcguigan , http://dx.doi.org/10.1103/physrevd.38.3031[phys . d * 38 * , 3031 ( 1988 ) ] ; a. hosoya and m. morikawa , http://dx.doi.org/10.1103/physrevd.39.1123[phys . rev . d * 39 * , 1123 ( 1989 ) ] ; m. mcguigan , http://dx.doi.org/10.1103/physrevd.39.2229[phys . d * 39 * , 2229 ( 1989 ) ] ; w. fischler , i.r . klebanov , j. polchinski , and l. susskind , http://dx.doi.org/10.1016/0550-3213(89)90290-3[nucl . b * 327 * , 157 ( 1989 ) ] ; y .- m . xiang and l. liu , http://dx.doi.org/10.1088/0256-307x/8/1/014[chin . * 8 * , 52 ( 1991 ) ] ; h.j . pohle , http://dx.doi.org/10.1016/0370-2693(91)90324-j[phys . b * 261 * , 257 ( 1991 ) ] ; s. abe , http://dx.doi.org/10.1103/physrevd.47.718[phys . rev . ] http://dx.doi.org/10.1103/physrevd.47.718[d * 47 * , 718 ( 1993 ) ] ; t. horiguchi , http://dx.doi.org/10.1103/physrevd.48.5764[phys . d * 48 * , ] http://dx.doi.org/10.1103/physrevd.48.5764[5764 ( 1993 ) ] ; m.a . castagnino , a. gangui , f.d . mazzitelli , and i.i . tkachev , http://dx.doi.org/10.1088/0264-9381/10/12/008[class . quantum grav . * 10 * , 2495 ] http://dx.doi.org/10.1088/0264-9381/10/12/008[(1993 ) ] ; a. vilenkin , http://dx.doi.org/10.1103/physrevd.50.2581[phys . d * 50 * , 2581 ( 1994 ) ] [ http://arxiv.org/abs/gr-qc/9403010[gr-qc/9403010 ] ] ; l.o . pimentel and c. mora , http://dx.doi.org/10.1016/s0375-9601(01)00048-2[phys . ] http://dx.doi.org/10.1016/s0375-9601(01)00048-2[lett . a * 280 * , 191 ( 2001 ) ] [ http://arxiv.org/abs/gr-qc/0009026[gr-qc/0009026 ] ] ; s. gielen and d. oriti , in _ quantum field theory and gravity _ , edited by f. finster _ et al . _ ( springer , basel , switzerland , 2012 ) [ http://arxiv.org/abs/1102.2226[arxiv:1102.2226 ] ] . e. fuchs and m. kroyter , http://dx.doi.org/10.1016/j.physrep.2011.01.003[phys . rep . * 502 * , 89 ( 2011 ) ] [ http://arxiv.org/abs/0807.4722[arxiv:0807.4722 ] ] . g. calcagni and g. nardelli , http://dx.doi.org/10.1016/j.nuclphysb.2009.08.004[nucl . b * 823 * , 234 ] http://dx.doi.org/10.1016/j.nuclphysb.2009.08.004[(2009 ) ] [ http://arxiv.org/abs/0904.3744[arxiv:0904.3744 ] ] . g. calcagni and g. nardelli , http://dx.doi.org/10.1007/jhep02(2010)093[jhep * 1002 * , 093 ( 2010 ) ] [ http://arxiv.org/abs/0910.2160[arxiv:0910.2160 ] ] . k. kuchar , in _ proceedings of the fourth canadian conference on general relativity and relativistic astrophysics _ , edited by g. kunstatter _ et al . _ ( world scientific , singapore , 1992 ) ; http://dx.doi.org/10.1142/s0218271811019347[int . j. mod . d * 20 * , ] http://dx.doi.org/10.1142/s0218271811019347[3 ( 2011 ) ] . c. isham , in _ integrable systems , quantum groups , and quantum field theories _ , edited by l.a . ibort and m.a . rodrguez ( kluwer , dordrecht , the netherlands , 1993 ) [ http://arxiv.org/abs/gr-qc/9210011[gr-qc/9210011 ] ] . d. oriti , http://dx.doi.org/10.1088/0034-4885/64/12/203[rep . phys . * 64 * , 1489 ( 2001 ) ] [ http://arxiv.org/abs/gr-qc/0106091[gr-qc/ ] http://arxiv.org/abs/gr-qc/0106091[0106091 ] ] . a. perez , http://dx.doi.org/10.1088/0264-9381/20/6/202[class . quantum grav . * 20 * , r43 ( 2003 ) ] [ http://arxiv.org/abs/gr-qc/0301113[gr-qc/ ] http://arxiv.org/abs/gr-qc/0301113[0301113 ] ] . d. oriti , in @xcite [ http://arxiv.org/abs/gr-qc/0607032[gr-qc/0607032 ] ] . d. oriti , in _ quantum gravity _ , edited by b. fauser _ et al . _ ( birkhuser , basel , switzerland , 2008 ) [ http://arxiv.org/abs/gr-qc/0512103[gr-qc/0512103 ] ] . d. oriti , in _ foundations of space and time : reflections on quantum gravity _ , edited by j. murugan _ et al . _ ( cambridge university press , cambridge , u.k . , 2012 ) [ http://arxiv.org/abs/1110.5606[arxiv:1110.5606 ] ] . r. gurau and j.p . ryan , http://dx.doi.org/10.3842/sigma.2012.020[sigma * 8 * , 020 ( 2012 ) ] [ http://arxiv.org/abs/1109.4812[arxiv : ] http://arxiv.org/abs/1109.4812[1109.4812 ] ] ; v.rivasseau , [ http://arxiv.org/abs/1112.5104[arxiv:1112.5104 ] ] . a. ashtekar and p. singh , http://dx.doi.org/10.1088/0264-9381/28/21/213001[class . quantum grav . * 28 * , ] http://dx.doi.org/10.1088/0264-9381/28/21/213001[213001 ( 2011 ) ] [ http://arxiv.org/abs/1108.0893[arxiv:1108.0893 ] ] . k. banerjee , g. calcagni , and m. martn - benito , http://dx.doi.org/10.3842/sigma.2012.016[sigma ] http://dx.doi.org/10.3842/sigma.2012.016 [ , 016 ( 2012 ) ] [ http://arxiv.org/abs/1109.6801[arxiv:1109.6801 ] ] . a. ashtekar , m. campiglia , and a. henderson , http://dx.doi.org/10.1016/j.physletb.2009.10.042[phys . ] http://dx.doi.org/10.1016/j.physletb.2009.10.042[lett . b * 681 * , 347 ( 2009 ) ] [ http://arxiv.org/abs/0909.4221[arxiv:0909.4221 ] ] . a. ashtekar , m. campiglia , and a. henderson , http://dx.doi.org/10.1088/0264-9381/27/13/135020[class . ] http://dx.doi.org/10.1088/0264-9381/27/13/135020[quantum grav . * 27 * , 135020 ( 2010 ) ] [ http://arxiv.org/abs/1001.5147[arxiv:1001.5147 ] ] . g. calcagni , s. gielen , and d. oriti , http://dx.doi.org/10.1088/0264-9381/28/12/125014[class . quantum ] http://dx.doi.org/10.1088/0264-9381/28/12/125014[grav . * 28 * , 125014 ( 2011 ) ] [ http://arxiv.org/abs/1011.4290[arxiv:1011.4290 ] ] . a. ashtekar , t. pawlowski , and p. singh , http://dx.doi.org/10.1103/physrevd.74.084003[phys . d ] http://dx.doi.org/10.1103/physrevd.74.084003 [ , 084003 ( 2006 ) ] [ http://arxiv.org/abs/gr-qc/0607039[arxiv:gr-qc/0607039 ] ] . a. ashtekar , a. corichi , and p. singh , http://dx.doi.org/10.1103/physrevd.77.024046[phys . d * 77 * , ] http://dx.doi.org/10.1103/physrevd.77.024046[024046 ( 2008 ) ] [ http://arxiv.org/abs/0710.3565[arxiv:0710.3565 ] ] . f. dowker , http://arxiv.org/abs/gr-qc/0206020[gr-qc/0206020 ] . w. kamiski and j. lewandowski , http://dx.doi.org/10.1088/0264-9381/25/3/035001[class . quantum ] http://dx.doi.org/10.1088/0264-9381/25/3/035001[grav . * 25 * , 035001 ( 2008 ) ] [ http://arxiv.org/abs/0709.3120[arxiv:0709.3120 ] ] . s.w . hawking and d.n . page , http://dx.doi.org/10.1016/0550-3213(86)90478-5[nucl . b * 264 * , 185 ] http://dx.doi.org/10.1016/0550-3213(86)90478-5[(1986 ) ] . a. baratin , f. girelli , and d. oriti , http://dx.doi.org/10.1103/physrevd.83.104051[phys . d * 83 * , ] http://dx.doi.org/10.1103/physrevd.83.104051[104051 ( 2011 ) ] [ http://arxiv.org/abs/1101.0590[arxiv:1101.0590 ] ] . i.s . gradshteyn and i.m . ryzhik , _ table of integrals , series , and products _ ( academic press , london , u.k . , 2007 ) . birrell and p.c.w . davies , _ quantum fields in curved space _ ( cambridge university press , cambridge , u.k . , 1982 ) . d. oriti and l. sindoni , http://dx.doi.org/10.1088/1367-2630/13/2/025006[new j. phys . * 13 * , 025006 ( 2011 ) ] [ http://arxiv.org/abs/1010.5149[arxiv:1010.5149 ] ] ; e.r . livine , d. oriti , and j.p . ryan , http://dx.doi.org/10.1088/0264-9381/28/24/245010[class . quantum grav . * 28 * , 245010 ( 2011 ) ] [ http://http://arxiv.org / abs/1104.5509[arxiv:1104 . ] http://http://arxiv.org / abs/1104.5509[5509 ] ] ; w. fairbairn and e.r . livine , http://dx.doi.org/10.1088/0264-9381/24/20/021[class . quantum ] http://dx.doi.org/10.1088/0264-9381/24/20/021[grav . * 24 * , 5277 ( 2007 ) ] [ http://arxiv.org/abs/gr-qc/0702125[gr-qc/0702125 ] ] ; f. girelli , e.r . livine , and d. oriti , http://dx.doi.org/10.1103/physrevd.81.024015[phys . d * 81 * , 024015 ( 2010 ) ] [ http://arxiv.org/abs/0903.3475[arxiv:0903.3475 ] ] . a. altland and b. simons , _ condensed matter field theory _ ( cambridge university press , cambridge , u.k . , 2006 ) . c. kiefer , http://dx.doi.org/10.1103/physrevd.38.1761[phys . d * 38 * , 1761 ( 1988 ) ] . a. ashtekar , t. pawlowski , p. singh , and k. vandersloot , http://dx.doi.org/10.1103/physrevd.75.024035[phys . d * 75 * , 024035 ( 2007 ) ] [ http://arxiv.org/abs/gr-qc/0612104[gr-qc/0612104 ] ] . d. wands , k.a . malik , d.h . lyth , and a.r . liddle , http://dx.doi.org/10.1103/physrevd.62.043527[phys . d * 62 * , 043527 ( 2000 ) ] [ http://arxiv.org/abs/astro-ph/0003278[astro-ph/0003278 ] ] . g.i . rigopoulos and e.p.s . shellard , http://dx.doi.org/10.1103/physrevd.68.123518[phys . d * 68 * , ] http://dx.doi.org/10.1103/physrevd.68.123518[123518 ( 2003 ) ] [ http://arxiv.org/abs/astro-ph/0306620[astro-ph/0306620 ] ] . lyth and a.r . liddle , _ the primordial density perturbation _ ( cambridge university press , cambridge , u.k . e. wilson - ewing , http://arxiv.org/abs/1108.6265[arxiv:1108.6265 ] .

using analytic arguments and numerical simulations , we examine whether chondrule formation and the fu orionis phenomenon can be caused by the burst - like onset of gravitational instabilities ( gis ) in dead zones . at least two scenarios for bursting dead zones can work , in principle . if the disk is on the verge of fragmention , gi activation near to 5 au can produce chondrule - forming shocks , at least under extreme conditions . mass fluxes are also high enough during the onset of gis to suggest that the outburst is related to an fu orionis phenomenon . this situation is demonstrated by numerical simulations . in contrast , as supported by analytic arguments , if the burst takes place close to au , then even low pitch angle spiral waves can create chondrule - producing shocks and outbursts . we also study the stability of the massive disks in our simulations against fragmentation and find that although disk evolution is sensitive to changes in opacity , the disks we study do not fragment , even at high resolution and even for extreme assumptions .

0806.1740

gravitational instabilities can activate in a disk when the toomre ( 1964 ) parameter @xmath2 for a thick disk ( see durisen et al . 2007 ) , where @xmath3 is the sound speed , @xmath4 is the surface density , and @xmath5 is the epicyclic frequency . as indicated by @xmath6 , a disk is unstable against gis when it is cold and/or massive . the resulting spiral waves driven by self - gravity efficiently transfer angular momentum outward and mass inward ( e.g. , lynden - bell & kalnajs 1972 ; durisen et al . another mechanism that can efficiently transfer angular momentum outward is the magnetorotational instability ( mri ; see balbus & hawley 1991 ; desch 2004 ) . in contrast to gis , the mri only requires a weak magnetic field coupled to the gas . these mechanisms , either separately or in combination , are likely to be the principal way t tauri stars accrete gas from a disk ( hartmann et al . 2006 ) . in order for the mri to occur , ionized species must be present in the gas phase . thermal ionization of alkalis occurs wherever @xmath7 k , but depletion of ions by dust grains may move the temperature threshold closer to @xmath8 k ( desch 1999 ; sano et al . 2000 ) , where the dust sublimates completely . elsewhere , the ionization must be driven by a nonthermal source , e.g. , energetic particles ( eps ) . for this discussion , eps refers to any particles that could ionize the gas , e.g. , x - rays . gammie ( 1996 ) proposed that disks may have active and inactive mri layers due to attenuation of eps by the gas . in the inner regions of a disk ( at radii @xmath9 few au ) where the column densities are large , mri may only be active in a thin layer , resulting in _ layered accretion_. as one moves outward , column densities drop , and the entire disk can become mri active . the region where the mri is mostly absent is called the _ dead zone_. eps are attenuated by a surface density of only about 100 g @xmath10 ( stepinski 1992 ) , and so even a minimum mass solar nebula ( mmsn ) will likely exhibit layered accretion ( desch 2004 ) . even if mass accretion is only reduced and not altogether halted as a result of reynolds stresses ( fleming & stone 2003 ; oishi et al . 2007 ) , mass may still pile up in the dead zone . if enough mass accumulates , then even for an otherwise low - mass disk , gis can activate . the fu orionis phenomenon is characterized by a rapid ( 1 - 10s yr ) increase in optical brightness of a young t tauri object , typically by 5 magnitudes , and is driven by sudden mass accretion of the order @xmath11 from the inner disk onto the star ( hartmann & kenyon 1996 ) . because fu ori objects appear to have decay timescales of about 100 yr , the entire mass of a mmsn ( @xmath12 ) can be accreted onto the star . to date , the best explanation for the optical outburst is a thermal instability ( e.g. , bell & lin 1994 ; kley & lin 1999 ; see discussions in hartmann & kenyon 1996 , green et al . 2006 ; and zhu et al . armitage et al . ( 2001 ) suggested that gis in a bursting dead zone might be able to trigger an fu ori outburst by rapidly increasing the accretion into the inner disk ( @xmath13 au ) and initiating an mri through thermal ionization . hartmann ( 2007 , in private communication ) and zhu et al . ( 2007 ) also suggest that the heating due to gravitational torques might drive an mri for @xmath14 au , which would then feed mass inside 0.1 au until a thermal instability sets in . the fu ori phenomenon may be a result of a _ cascade _ of instabilities , starting with a burst of gi activity in a dead zone , followed by accretion due to an mri , followed finally by a thermal instability ( cf kley & lin 1999 ) . indeed , recent observations of fu ori indicate that very large mass fluxes are present out to at least @xmath15 au ( zhu et al . 2007 ) . although details are still being debated , most chondrules formed in the first 1 to 3 myr of the solar nebula s evolution ( bizzarro et al . 2004 ; russell et al . chondrule precursors were flash melted from solidus to liquidus , where high temperatures @xmath8 k were experienced by the precursors for a few minutes . the melts then cooled over hours , with the actual cooling time depending on chondrule type . chondrule collisional histories and isotopic fractionation data , chondrule - matrix complementarity , fine - grained rim accumulation , and petrological and parent body location arguments ( krot et al . 2005 ) suggest that chondrules formed in the solar nebula ( wood 1963 ) in strong , localized , repeatable heating events . the shock wave model for chondrule formation can accommodate these observational constraints and reproduce heating and cooling rates required to form chondrule textures ( iida et al . 2001 ; desch & connolly 2002 ; cielsa & hood 2002 ; miura & nakamoto 2006 ) . one plausible source of chondrule - producing shocks is a global spiral wave ( wood 1996 ) . harker & desch ( 2002 ) suggest that spiral waves could also explain thermal processing at distances as large as 10 au , which may be necessary to explain observations of comets ( wooden et al . 2005 ) and recent _ stardust _ results ( e.g. , mckeegan et al . it has been suggested that bursts of gis may be able to produce the required shock strengths ( boss & durisen 2005 ) and provide a source of turbulence and mixing ( boss 2004b ; boley et al . global spiral shocks are appealing because they fit many of the constraints above . they may be repeatable , depending on the formation mechanism for the spiral waves ; they are global , but produce fairly local heating ; they can form chondrules in the disk ; and they can work in the inner disk as well as the outer disk . knowing under what conditions protoplanetary disks can fragment is crucial to understanding disk evolution inasmuch as a fragmented disk may produce gravitationally bound clumps . this has become known as the _ disk instability _ hypothesis for the formation of gas giant planets ( kuiper 1951 ; cameron 1978 ; boss 1997 , 1998 ) . the strength of gis is regulated by the cooling rate in disks ( tomley et al . 1991 , 1994 ; pickett et al . 1998 , 2000 , 2003 ) , and if the cooling rate is high enough in a low-@xmath6 disk , a disk can fragment ( gammie 2001 ) . gammie quantified that a disk will fragment when @xmath16 for a disk with a @xmath17 , where @xmath18 is the two - dimensional adiabatic index , such that @xmath19 , where @xmath20 is the gas pressure and @xmath21 is the vertical direction in the disk . here , @xmath22 is the local cooling time and @xmath23 is the angular speed of the gas . this criterion was approximately confirmed in 3d disk simulations by rice et al . ( 2003 ) and meja et al . rice et al . ( 2005 ) showed through 3d disk simulations that this fragmentation criterion depends on the 3d adiabatic index @xmath24 and , for @xmath25 or 7/5 , the fragmentation limit occurs when @xmath26 or 12 , respectively . these results show that a change by a factor of about 1.2 in @xmath27 has a factor of two effect on the critical cooling time . in addition , these results indicate that the cooling time must be roughly equal to the dynamical time of the gas for the disk to be unstable against fragmentation when @xmath28 . do such prodigious cooling rates occur in disks when realistic opacities are used with self - consistent radiation physics ? this question is heavily debated in the literature ( e.g. , nelson et al . 2000 ; boss 2001 , 2004a , 2005 ; rafikov 2005 , 2007 ; boley et al . 2006 , 2007b ; mayer et al . 2007 ; stamatellos & whitworth 2008 ) . the simulations to date use a wide variety of numerical methods , including very different approximations for radiation physics , and only two groups have published results of radiative transfer tests appropriate for disks ( boley et al . 2007b ; stamatellos & whitworth 2008 ) . nelson et al . ( 2000 ) used 2d sph simulations with radiation physics to study protoplanetary disk evolution . because their simulations were evolved in 2d , they assumed that the disk at any given moment was in vertical hydrostatic equilibrium . using a polytropic vertical density structure and pollack et al . ( 1994 ) opacities , they cooled each particle according to an appropriate effective temperature . in their simulations , the cooling rates are too low for fragmentation . in contrast , boss ( 2001 , 2005 ) employed radiative diffusion in his 3d grid - based code ; fragmentation occurs in his simulated disks . besides the difference in dimensionality of the simulations , boss assumed a fixed temperature structure for rosseland mean optical depths less than 10 , as measured along the radial coordinate ( cf recent flux - limited simulations by boss 2008 ) . boss ( 2002 ) found that the fragmentation in his disks is insensitive to the metallicity of the gas and attributed this independence to fast cooling by convection ( boss 2004a ) . however , it must be noted that nelson et al . ( 2000 ) assumed a vertically polytropic density structure . because the specific entropy @xmath29 , where @xmath30 is the polytropic equation of state and @xmath31 is the gas density , the nelson et al . approximation effectively assumes efficient convection . except for extremely massive and extended disks ( stamatellos et al . 2007 ) , recent simulations with radiative physics by cai et al . ( 2006 ) , boley et al . ( 2006 , 2007b ) , and stamatellos & whitworth ( 2008 ) find long cooling rates and no efficient cooling by convection . cai et al . also show that the strength of gis is dependent on the metallicity . furthermore , boley & durisen ( 2006 ) suggest that shock bores , which can cause a rapid vertical expansion in the post - shock region of spiral shocks , could be misidentified as `` convective '' flows . in both contrast and support of these studies , mayer et al . ( 2007 ) use sph simulations with 3d flux - limited diffusion and find that fragmentation only occurs once the mean molecular weight of the gas @xmath32 . however , the simulations presented by meja ( 2004 ) , cai et al . ( 2006 ) , and boley et al . ( 2006 ) unintentionally were evolved with @xmath33 due to an error in the inclusion of he in the opacity tables , and their disks do not fragment . the issue of fragmentation in radiatively cooled disks thus remains unsettled . for this study , we adopt the hypothesis , which we refer to as the _ unified theory _ , that bursts of gi activity in dead zones drive the fu ori phenomenon and produce chondrule - forming shocks . this hypothesis is an amalgamation of ideas presented in wood ( 1996 , 2005 ) , gammie ( 1996 ) , armitage et al . ( 2001 ) , and boss & durisen ( 2005 ) . in order to investigate this scenario , we designed a numerical experiment to evolve a massive , highly unstable disk with an initial radial extent between 2 and 10 au . commensurately , we investigate the stability of these massive , gravitationally unstable disks against fragmentation to assess the feasibility of the disk instability hypothesis for gas giant formation . the shock speed @xmath34 for a fluid element entering a global , logarithmic spiral wave with pitch angle @xmath35 is approximately described by @xmath36 where @xmath37 is the corotation radius of the spiral pattern and where we have assumed that the gas azimuthal motion is keplerian and @xmath38 is negligible . when @xmath39 , the shock speed limits to the keplerian speed times @xmath40 ; whether spiral waves with large @xmath37 produce chondrules is mainly dependent on the pitch angle of the spiral wave . in contrast , when @xmath41 , the speed increases as @xmath42 , and even shallow pitch angles can produce chondrules . for example , consider the mmsn , where the midplane @xmath43 g @xmath44 ( hayashi et al . figure 1 shows the @xmath34-@xmath31 plane with heavy , solid curves indicating shock speeds for @xmath45 , 2.5 , 5 , and @xmath46 au and with @xmath47 and 30@xmath48 . the colored regions on the plot highlight where chondrule formation is expected in a mmsn ( yellow ) and a 10@xmath49mmsm ( orange ) . the chondrule - forming curves are based on the results of desch & connolly ( 2002 ) , which we summarize by @xmath50 . the boundaries set by these curves are meant to be illustrative , not definitive . spiral shocks along shallow pitch angle spiral waves ( @xmath51 ) are mostly if not entirely out of the chondrule - forming range between @xmath52 and 5 au for shocks inside corotation . however , spirals with corotation in the inner disk can produce chondrule - forming shocks for a wide range of radii , depending on the actual @xmath37 . if the spiral waves have very open pitch angles ( @xmath53 ) , then spiral waves with corotation near @xmath54 au can produce chondrule - forming shocks near the asteroid belt and at cometary distances . the mass of the disk does not change these general behaviors . a major caveat for this simple - minded approach is that the fluid elements motions may be poorly described by equation ( 1 ) if vertical and radial excursions induced by shock bores ( boley & durisen 2006 ) can not be neglected . the possibility of producing chondrules in the asteroid belt and at comet distances during the same burst is attractive . moreover , we would also like to use these simulations to study other phenomena such as radiation transport , convection , and disk fragmentation in the 2 - 10 au region . for these reasons , we tried to design a numerical experiment to investigate the connection between gis , fu orionis events , and chondrule formation that is biased toward producing strong shocks with a corotation near 5 au and with open pitch angles . if chondrule - producing shocks can not be created in simulations biased in their favor , then it would present a serious problem for gis as the source of chondrule processing . the reader should keep in mind that the model , at this stage , is not necessarily meant to be representative of a typical disk , although the massive disk we present may be plausible for early class i objects ( e.g. , l1551 ; osorio et al . 2003 ) . to create the initial conditions ( ics ) , consider a disk that is vertically polytropic , is axisymmetric , has a constant @xmath27 , and is keplerian ( @xmath55 ) . also assume that self - gravity is negligible . with these assumptions , the vertical disk structure can be calculated by using the following equations : @xmath56}{\gamma[\gamma/(\gamma-1)]}\right)^ { 2/(\gamma+1)},\end{aligned}\ ] ] where @xmath57 for @xmath30 is the polytropic coefficient at any given @xmath42 , @xmath58 is the gravitational constant , @xmath23 is the keplerian angular speed , and @xmath59 is the disk scale height . using equations ( 2 ) through ( 4 ) , one can show that @xmath60}{\gamma[(3\gamma-1)/(2\gamma-1)]}\right)^2 \right)^{(1-\gamma)/4}\\\nonumber & \times&\left(\gamma k(r)\right)^{1/2}\left(gq(r)\right ) ^{-(\gamma+1)/2}\omega(r)^\gamma.\end{aligned}\ ] ] to select a model , one needs to specify power laws for @xmath61 and @xmath6 , @xmath61 and @xmath57 , or @xmath57 and @xmath6 . for calculating @xmath6 , the midplane sound speed @xmath3 is used . note that when @xmath62 constant , there exists a lower limit for @xmath63 , @xmath64 , where the radial entropy gradient remains positive . because the specific entropy @xmath29 , equation ( 5 ) can be solved to show that @xmath65 for constant @xmath6 . the critical surface density power law is then @xmath66 for @xmath28 and @xmath67 for @xmath68 . when @xmath69 , radial thermal convection may set in for a constant @xmath6 disk . however , the radial entropy gradient is not the only stability criterion for radial thermal convection in disks ( klahr & bodenheimer 2003 ) . equations ( 2 ) through ( 5 ) are used to generate the ics for this study . the disk is massive , @xmath70 , with @xmath71 , initially between about 2 and 10 au in radius and with a @xmath72 everywhere . although the surface density power law is steep , it is consistent with the nice - mmsn ( desch 2007 ) . because the analytic model ignores self - gravity , assumes a constant @xmath27 ( see 3.2 below ) , and has sharp disk truncation edges , the model must be relaxed to a new equilibrium structure in the hydro code . once the ics are generated , they are loaded into our code and evolved without cooling at very low azimuthal resolution , essentially two - dimensionally . the radial and vertical momenta are damped to allow the disk to relax calmly to an equilibrium configuration , which takes about ten orbits at 10 au . to create a mass concentration in the disk , mass is then added to the disk linearly in time with a gaussian profile . the specific internal energy is held constant . the peak of the gaussian is centered on 5 au . the radial fwhm of the gaussian is chosen to be 3 au , which is roughly the size of the most unstable radial wavelength @xmath73 ( binney & tremaine 1987 ) . mass is added until nonaxisymmetry is visible in midplane density images , which takes approximately 190 yr , i.e. , 6 orp of evolution ( 1 orp is defined as 1 outer rotation period at 10 au for this model , which is about 33 yr ) . the total mass added is about 0.08 @xmath74 , and so if one were to imagine a corresponding accretion rate , it would be about @xmath75 yr@xmath76 . it is probably unphysical to add so much mass so quickly to the disk without increasing the temperature however , we remind the reader that the study is a numerical experiment . figures 2 and 3 illustrate the density distributions of the disk before and after the mass accretion . the mass buildup increases the total mass of the simulated disk to @xmath77 m@xmath78 and it decreases @xmath6 below unity in a narrow region around 5 au just before strong gi activation . because the ring is grown over 6 orp , the instability is not obviously overshot . the mass - weighted average @xmath79 over the fwhm of the gaussian centered at 5 au is approximately unity when the ring becomes unstable . this implies that @xmath6 must approach unity for a spatial range of at least the most unstable radial wavelength before gis will activate in a disk . the disk is evolved with the chymera code ( boley 2007 ) , which employs the boley et al . ( 2007b ) radiation physics algorithm . chymera is an eulerian code that solves the hydrodynamics equations on a fixed , cylindrical grid . the code includes self - gravity , a ray+flux - limited diffusion hybrid scheme , and a variable first adiabatic index @xmath24 . this code has been extensively tested . in response to criticisms of boss s ( 2001 , 2002 , 2004a , 2005 ) simulations by boley et al . ( 2007a ) , boss ( 2007 ) questions the accuracy of the indiana university code , which is the predecessor of chymera . boss raises concerns that we address here . _ potential solver : _ chymera uses a direct poisson solver to calculate the potential due to mass on the grid . to provide the boundary conditions for the direct solver , a spherical harmonics expansion for @xmath80 is used to calculate the potential on the grid boundary ( e.g. , pickett et al . 2003 ; boley et al . boss argues that the boundary solver used in chymera may be too low - order to capture clump formation . tests by pickett et al . ( 2003 ) and meja et al . ( 2005 ) and the results from the wengen test 4 code comparison ( wt4 ; mayer et al . 2008 , in prep . ) demonstrate that this is not the case . during the initial analyses of wt4 , the potential solver was rigorously tested in an attempt to explain some deviations noticed between chymera and other codes . these deviations were finally attributed to differences in the initial conditions , and new simulations with consistent ics show very good agreement between codes . to test the potential solver in chymera , a high - order bessel function expansion boundary solver ( cohl & tohline 1999 ) was employed with legendre half - integer polynomials expanded out to @xmath81 , which showed convergence to better than @xmath82 when compared with @xmath83 . the spiral structure and clump formation in the simulation with this high-@xmath84 boundary solver were indistinguishable from the normal lower - order spherical harmonics expansion . the wengen test is very demanding , with strong nonaxisymmetric structure and a highly flattened disk , and yet the lower - order spherical harmonics expansion for @xmath80 used in chymera ( e.g. , boley et al . 2007b ) captures clump formation . _ resolution & numerical heating : _ the resolution for the high - resolution simulations presented in this paper are at higher resolution than the base simulations presented for wengen test 4 . as will be demonstrated here , chymera shows fragmentation when it is expected based on experimentally determined fragmentation criteria ( gammie 2001 ; rice et al . the numerical heating reported by boley et al . ( 2006 ) affected a small region of their inner disk , and is understood to be a resolution effect . the simulations presented here are at high enough resolution , as determined by the boley et al . ( 2007b ) tests , that this effect should be negligible if present at all . _ radiative transfer boundary conditions : _ chymera has been shown to relax to analytic solutions for the flux through an atmosphere as well as the temperature profile for high , moderate , and low optical depths . the code allows convection when it should , and does not when it should not ( boley et al . 2007b ) . for the treatment of h@xmath85 , we use a frozen ortho / para ratio of 3:1 in this study for reasons presented in boley et al . ( 2007a ) . the grain size distribution is assumed to follow the ism power law , @xmath86 ( dalessio et al . 2001 ) , where the maximum grain size @xmath87 is chosen to be 1 mm to account for grain growth ( dalessio et al . 2006 ) . to transition smoothly between the low and high optical depth regimes , a weighted combination between rosseland and planck mean opacities @xmath88 is used @xmath89 , where @xmath90 is the vertical midplane rosseland optical depth . different methods for interpolating @xmath91 and @xmath92 _ along the ray _ are being explored , but they have not been implemented here . the particles are assumed to be well - mixed with the gas . the effect of dust settling is heuristically considered by evolving the disk in several ways : standard opacity and @xmath93 , @xmath94 , and @xmath95 of the standard opacity , referred to as the @xmath96 , @xmath97 , and @xmath98 simulations . this choice in opacities varies the vertical midplane rosseland mean optical depths from about @xmath99 to unity , and spans the limits where advection of photons becomes important ( @xmath100 ; krumholz et al . 2007 ) and where radiative cooling becomes most efficient ( @xmath101 ) . no external radiation is assumed to be shining onto the disk , except for a 3 k background temperature . although such rapid dust settling and a naked disk are likely to be unrealistic , the assumptions are made to bias the disk toward strong shocks inasmuch as irradiation and opacity affects gi strength ( cai et al . 2006 , 2008 ) . for a base comparison , an adiabatic simulation , i.e. , where shock heating is allowed but not cooling , is evolved for about 33 yr . the simulations are evolved at two resolutions : @xmath102 and @xmath103 cells , respectively , above the midplane . the lower resolution simulations ( 512 sim ) have a grid spacing of 0.05 au per cell in @xmath42 and @xmath21 and the higher 0.025 au ( 1024 sim ) . for both resolutions , @xmath104 at about @xmath105 au . these parameters are summarized in table 1 . for each disk , 1000 fluid tracer elements are randomly distributed in @xmath42 between about 3 and 7 au , in @xmath106 over @xmath107 , and in @xmath21 roughly within the scale height of the disk . due to inefficient cooling in the standard simulations , high enough temperatures in the disk are reached after 80 and 55 yr for the 512 and 1024 simulations , respectively , to create radiative timescales that are too short to resolve ; the simulations are stopped . as described below , the reduced opacity simulations are able to cool much more efficiently , and the disk temperatures remain manageable with the time - explicit algorithms used in chymera . however , the reduced opacity simulations are stopped after about 110 yr due to the development of a strong , one - arm spiral , which is treated incorrectly with our fixed - star assumption . the hydrodynamics scheme in chymera is eulerian , and does not give direct information on the histories of tracer fluid elements . in order to derive detailed and statistical shock information and to capture the complex gas motions in unstable disk simulations , we have combined a tri - akima spline interpolation algorithm with a runge - kutta integrator ( e.g. , press 1986 ) for tracing a large sample of fluid elements during the simulation . velocities and thermodynamic quantities , e.g. , temperature and density , are interpolated once each time step . because the hydrodynamics code explicitly solves the equation of motion for the gas , the algorithm only needs time and velocity information to integrate the positions of the fluid elements . the runge - kutta integrator was intended to be fourth - order accurate ; unfortunately , an implementation error when combining the integration scheme into chymera resulted in a first - order accurate scheme in time . because tracing fluid element histories is an important point of this paper , we reran a stretch of one of the simulations with a second - order fluid element tracing routine . after about one orbit at @xmath108 au , the mean fractional differences , @xmath109 , between the first and second - order schemes are @xmath110 , @xmath111 , and @xmath112 for @xmath42 , @xmath106 , and @xmath21 , respectively . the sample standard deviations are @xmath113 , @xmath93 , and @xmath114 for @xmath42 , @xmath106 , and @xmath21 , respectively . the mean differences are small , and show that there are no obvious systematic effects . the sample deviations indicate that the variations between the the two schemes are also marginal for the radial and azimuthal directions . there is considerably larger scatter in the vertical direction , where spiral shocks can create complicated vertical flows ( boley & durisen 2006 ) . because the fluid elements are used to gather shock information , the scatter in the vertical positions between the two schemes is not a major concern for this study . the effect is similar to redistributing the fluid elements vertically , keeping the same @xmath42 and @xmath106 , every few orbits . an akima spline is similar to a natural cubic spline , but typically yields better results for curves with sudden changes ( see akima 1970 ) , as are expected in shock profiles . although the spline fits a curve to a one - dimensional set of data points , the interpolation can be extended to data in three dimensions . first , consider a cubic volume with data at the vertices . a value anywhere in the volume can be approximated by seven linear interpolations : four to calculate values between each vertex in a particular direction , two to calculate the values along the projections of the desired point onto the interpolated lines , and a final interpolation through the point of interest . extending this from a simple tri - linear fit to a tri - akima spline is relatively straightforward . instead of using the eight nearest points that enclose a volume , one uses the 125 closest points , with five data points used for every akima spline fit . the central data point is the closest data value to the point of interest . we use the gnu scientific library akima spline algorithm for performing fits . as these disks become gravitationally unstable , low - order spiral waves develop . these waves drive a sudden increase in mass flux throughout the disk and drive strong spiral shocks . following meja et al . ( 2005 ) , we refer to this phase of disk evolution as a burst of gi activity . surface density plots for the 512 simulations are shown in figures 4 and 5 at @xmath115 and 77 yr , respectively . the images at @xmath116 yr indicate that except for the @xmath98 simulation , the burst has a well - defined , three - arm spiral and that the spiral arms become denser as the opacity is lowered . by 77 yr , the disk structures are noticeably different . each disk has a visually distinct two - arm spiral , again , except for the @xmath95 @xmath5 simulation . for comparison with figures 4 and 5 , the 1024 simulations are shown in figures 6 and 7 at @xmath116 and 78 yr , respectively . the higher resolution simulations have additional fine structure , but they do not develop clumps or dense knots . in fact , some of the knotty structure that develops in the 512 10@xmath117 @xmath5 simulation is absent at higher resolution . the lower opacity simulations appear to have stronger amplitudes , with denser spiral arms and stronger mid - order fourier components ( see below ) . as will be discussed , the @xmath95 @xmath5 simulation is on the verge of fragmentation . the visual differences between the simulations can be quantified by a fourier mode analysis . define @xmath118 , where @xmath119 and @xmath120 ( cf . boley et al . 2006 who use the volume density ) . in addition , define @xmath121 . as shown in table 1 , @xmath122 tends to increase with both resolution and decreased opacity . this is also seen in the fourier component spectrum , shown in figure 8 , where squares indicate the 1024 simulations and crosses lines the 512 simulations . the @xmath97 simulation is not shown for readability , but if falls between the @xmath93 and @xmath98 simulations and has the least amount of divergence between the 512 and 1024 simulations at large @xmath84 . for all runs , the fourier @xmath123 through 6 dominate , and the power in the @xmath124 spectra drops off steeply around @xmath125 . the 1024 simulations have a shallower profile at large @xmath84 , so resolution effects probably play a role for @xmath126 , even at resolutions of lmax=512 . the convergence of the @xmath124 profile for large @xmath84 is a topic for future investigation ( steinman - cameron et al . , in preparation ) . the energetics of the disk simulations are shown in figures 9 and 10 , where cumulative energy loss by radiation ( blue ) and cumulative energy dissipation by shocks ( red ) are plotted . these quantities are integrated during the evolution , so they represent the actual heating and cooling in these disks . for both resolutions of the standard case , the shock heating clearly dominates over the cooling ; the disk is heating up over the entire evolution . energy is transported inefficiently for the highly optically thick disk , and the disk evolves like an adiabatic simulation . when the opacity is reduced by a factor of 100 , the cooling becomes much more important than it is in the standard simulation , with radiative energy losses becoming comparable to heating by shocks . in addition , the shock heating rate becomes somewhat larger during the burst . the 1024 standard and @xmath93 @xmath5 show additional heating . this heating is quite strong in the 1024 standard simulation , and it must be stopped even earlier than the 512 standard due to high temperatures , as discussed above . the 1024 @xmath93 @xmath5 is very similar to its 512 counterpart . there is additional heating and cooling during the burst , but the heating and cooling curves for both resolutions roughly track each other . the adiabatic curve , which is only shown to 33 yr ( 1 orp ) , is difficult to see because it tracks the standard cases extremely closely . when the opacity is reduced by a factor of @xmath127 , the radiative cooling is even more efficient and surpasses shock heating . the @xmath128 opacity reduction shows the strongest shock heating and the fastest disk cooling . the 1024 @xmath95 @xmath5 simulation curves track the 512 curves , with the burst contributing the most to the offset of the curves . unlike the higher opacity simulations , for @xmath98 , there is less shock heating and less total cooling at higher resolution . the 1024 @xmath94 @xmath5 simulation roughly tracks the 512 counterpart , with the strongest deviation near 66 yr ( 2 orp ) . the adiabatic curve is shown again as a reference . as one would expect from analytic arguments , the opacity has a profound effect on disk cooling and , consequently , on spiral shocks . the effect of opacity on energy losses is also demonstrated in figures 11 and 12 . in these figures , brightness temperature maps for the 1024 simulations are shown for similar times as in figure 6 , where @xmath129 and @xmath130 is vertical outward intensity as calculated by the ray solution in chymera . in addition , we define a cooling temperature @xmath131dz/\sigma)^{1/4}$ ] , which is the temperature that corresponds to a given column s effective flux if all of the energy were to leave the column vertically . if the column is being heated , @xmath132 is set to zero and appears as black in the images . as the opacity is lowered , the spiral structure becomes more clearly outlined , and the disk becomes brighter because the photons can leave from hotter regions . the brightness maps also demonstrate that sustained fast cooling by convection is absent . if hot gas near an optically thick midplane is quickly transported to altitudes where @xmath101 , convection can , in principle , enhance disk cooling . however , the efficacy of convective cooling is controlled by the rate at which that energy can be radiated from the photosphere of the disk . if localized convective flows were responsible for fast cooling in the optically thick disks , one would expect to find strong , localized @xmath133 and @xmath132 enhancements . for example , the standard and @xmath96 simulations would need to have regions as bright as the @xmath98 simulation . there are thin regions along the spiral shocks in the @xmath96 simulation that show large @xmath132 , but they do not correspond to enhancements in @xmath133 as well . this suggests that the energy from the spiral shocks is not transported efficiently to altitudes where it can be quickly lost from the disk . this assessment is supported by the net radiative heating in regions surrounding the strong shocks , e.g. , the black outlines ( areas experiencing net radiative heating ) around the thin , hot spiral waves . the corresponding energy flux that is needed for convection to be sustained and to cool the disk does not occur ( see also boley et al . 2006 , 2007b , nelson 2006 , & rafikov 2007 ) . the importance of the opacity for disk cooling is also shown in figures 13 through 15 . on these plots , three quantites are shown : the toomre @xmath6 , the mass - weighted @xmath24 , and the @xmath134 profiles . for the @xmath22 curves , the cooling times are calculated by dividing the azimuthally and vertically integrated internal energy by the azimuthally and vertically integrated radiative cooling rate for each annulus in the disk . the @xmath135 is the critical value of @xmath136 , below which fragmentation is expected ( gammie 2001 ) , for the corresponding @xmath24 . rice et al . ( 2005 ) demonstrated that @xmath137 and @xmath138 . based on these values , we assume @xmath139 for the stability analysis presented here . one should keep in mind that @xmath135 is not a strict threshold and that the relation is approximate , especially for simulations that permit the cooling time to evolve with the disk ( johnson & gammie 2003 ; clarke et al . regardless , it serves as a general indicator for disk fragmentaiton . as figures 13 through 15 demonstrate , the disk is approaching a state of constant @xmath6 for a wide range of radii , as expected in an asymptotic phase ( meja et al . the mass - weighted average @xmath6s between 3 and 6 au are given in table 1 and tend to decrease , as expected , when the opacity decreases . they do not seem to be greatly affected by resolution . the variation in the average @xmath6 with opacity is roughly consistent with the @xmath140 measurements . one should keep in mind that the @xmath6 profiles represent snapshots of the disk , while the @xmath140 measurements are time averages . the standard simulation is not shown in the stability plots , but the analysis of the 512 standard disk shows that it is highly stable against fragmentation . because the cooling times fluctuate rapidly , we average the integrated cooling rates and the internal energies for the , roughly , 65- and 70/80-year snapshots . the cooling time profiles for the appropriate @xmath24 only drop substantially below unity over extended regions where @xmath6 is high ; these disks are stable against fragmentation . as the opacity is lowered , the cooling times decrease as well , with the @xmath95 @xmath5 simulation close to the fragmentation limit . it should also be noted that the only disk that behaves like a constant @xmath24 disk is the standard opacity simulation ( not shown ) , and for all other simulations , a constant @xmath24 is inappropriate ( see boley et al . 2007a ) . mass in these disks is redistributed efficiently during the activation of gis . time - averaged mass fluxes are calculated by differencing the mass inside a cylinder at two different times . this represents the average mass flux as calculated by the second - order hydrodynamics scheme , and it is independent from the fluid element tracer . the inward and outward accretion rates vary , and in all simulations , can be well above @xmath141 yr@xmath76 ( fig . 16 ) . due to our resolution , shocks are spread over scales of about @xmath142 cm or larger , whereas 1d calculations of chondrule - producing shocks give structures @xmath143 cm ( desch & connolly 2002 ) . to overcome this difficulty , we measure the pressure difference between the pre- and post - shock regions to calculate the mach number @xmath144 . using the pressure and temperature histories for each fluid element , a possible shock is identified whenever the @xmath145 changes from negative to positive then back to negative . the pre - shock flow is taken to be the first sign switch , and the post - shock flow is the second sign switch . let @xmath146 be the fractional pressure change , where @xmath147 and @xmath148 are the pre- and post - shock pressures , respectively , and where @xmath27 is the average of the pre- and post - shock first adiabatic index . if @xmath149 , the event is counted as a shock . once @xmath144 is determined , the pre - shock velocity @xmath150 is calculated , where @xmath151 is the pre - shock adiabatic sound speed of the gas . the shock strengths are also derived using the ratio of the post - shock to pre - shock temperatures : @xmath152 . both approaches should give the same answer whenever the shocks are cleanly defined and adiabatic . however , if radiative cooling becomes very efficient , then assuming adiabatic shock conditions may be incorrect . furthermore , secondary waves or shocks ( boley & durisen 2006 ) may make identifying pre- and post - shcok regions very difficult . neither method is clearly favored , and both will only provide crude estimates . we estimate a shock to be chondrule - forming when @xmath34 lies in a 1 km s@xmath76 band between 5 km s@xmath76 and 11 km s@xmath76 and when the pre - shock density is between @xmath153 = -8.5 and -10.5 . this estimate is based on the desch & connolly ( 2002 ) 1d calculations , similar to the relation used to locate chondrule - producing shocks in figure 1 . chondrule - forming shocks should be thought of as having the potential to form chondrules , but may not yield chondritic material due to incompatible dust to gas ratios , too high or low cooling rates , and fractionation . table 2 indicates shock information as extracted from the pressure histories of the fluid elements . estimating shock strengths with the temperature ratio yields comparable numbers except for the chondrule - forming shocks ( see below ) . column two shows approximately the total number of detected shocks ( ts ) with @xmath154 , and column three displays the average number of shocks per fluid element . columns four ( ts mass ) and five ( cs mass ) show the total dust mass ( see below ) that goes through a shock with @xmath155 and the total dust mass that encounters a chondrule - forming shock , respectively . to estimate how much dust is processed in shocks , we assign each fluid element a mass by calculating the total disk mass within some @xmath156 and distributing that material evenly among all fluid elements in that @xmath156 . a gas to solids ratio of 100 is assumed everywhere for the dust processing calculations . although this is inconsistent with the growth and settling of solids assumed in six of the simulations , we do not worry about this detail inasmuch as the midplane will process more solids per shock and the high - altitude shocks will process less . finally , column six shows the total number ( total fe ) of fluid elements that remain after the 70 yr time period , i.e. , the elements that were not accreted onto the star or fluxed into background density regions . we check whether the total number of detected shocks is reasonable by evaluating the expected number of shocks in a keplerian disk with @xmath84-arm spiral waves . if the corotation radius for the pattern is @xmath37 , the number of shocks during some time period @xmath157 for a fluid element orbiting at @xmath42 is @xmath158 evaluating equation ( 6 ) for 1000 fluid elements evenly distributed in annuli between 2 and 8 au , with @xmath159 au , with @xmath160 , and with @xmath161 yr yields approximately 8000 shocks . so the number of detected shocks in these simulations is reasonable . the total number of shocks per fluid element are roughly consistent for each simulation except for the @xmath98 runs . the more flocculent spiral structure in these simulations not only produces a larger number of shocks in the disk , but also creates candidate chondrule - producing shocks for four fluid elements ; the 1024 @xmath97 simulation has one fluid element that experiences a possible chondrule - forming shock . several thousand @xmath162 of dust are pushed through shocks in each simulation , implying that almost all dust experiences a shock with @xmath154 . in addition , a few @xmath162 of dust experience chondrule - forming shocks in the @xmath98 runs . we remind the reader that these are only crude , order of magnitude estimates . when the temperature ratio is used , shocks with chondrule - forming strengths are _ undetected _ , but the total dust mass pushed through shocks remains the same to about one percent . in this section , we review the implications of the results of this study for disk fragmentation , the effects of opacity on disk cooling , the fu ori phenomenon , and chondrule formation . we remind the reader that the simulations presented here are meant to be a numerical experiment that explores the possible connection between chondrules , fu ori outbursts , and bursts of gi - activity . moreover , this experiment provides a systematic study of the effects of opacity on disk cooling , which complements the cai et al . ( 2006 ) metallicity study and provides a test bed for disk fragmentation criteria . after the onset of the burst , none of the disks fragments , and only the 512 @xmath95 @xmath5 simulation shows dense knot formation during the peak of the burst ( 10 - 30 yr ) . these knots do not break from the spiral wave even in the 1024 simulation , and so clump formation does not seem to be missed due to poor resolution . one should also keep in mind that for these simulations , the disk is first relaxed to equilibrium with standard opacity and then the opacity is suddenly dropped by a factor of 10@xmath163 at @xmath164 . in effect , the dust settling is treated as instantaneous . knot formation may not occur in a disk with more realistic settling timescales . as discussed below , this is also a caveat for the chondrule formation results . on the other hand , it does suggest that disk fragmentation might occur inside 10 au under even more extreme , but perhaps unphysical , conditions than we have modeled . the stability of these disks against fragmentation is supported by figures 13 through 15 . the cooling rates for the standard and @xmath96 simulations are too low to cause disk fragmentation . the @xmath94 and @xmath98 simulations do have areas where @xmath165 , but the disks approach stability against gis in those regions ( @xmath166 ; durisen et al . 2007 ) . as noted in 4.2 , @xmath167 is not a strict instability criterion , but it does serve as an estimate for disk stability against fragmentation . for no simulation is @xmath168 well below unity where @xmath6 is also @xmath169 . lowering the opacity increases the cooling rates . the @xmath95 @xmath5 simulations exhibit the most rapid cooling because the midplane optical depths are near unity , which results in the most efficient radiative cooling possible . _ changes in dust opacity have a profound effect on disk cooling_. although not modeled , if the opacity were to continue to drop such that the midplane optical depth becomes well below unity , cooling would once again become inefficient . in addition , supercooling of the high optical depth disks ( standard and @xmath96 ) by convection does not occur . based on our results , we believe that hydraulic jumps as a result of shock bores ( boley & durisen 2006 ) rather than convection are a better explanation for the upwellings around spiral arms reported by boss ( 2004a ) and mayer et al . ( 2007 ) . these findings are consistent with analytic arguments by rafikov ( 2005 , 2007 ) , with numerical studies of disk fragmentation criteria by gammie ( 2001 ) , johnson & gammie ( 2003 ) , and rice et al . ( 2005 ) , and with global disk simulations where tests of the radiation algorithm and/or careful monitoring of radiative losses were performed ( nelson et al . 2000 ; boley et al . 2006 , 2007b ; stamatellos & whitworth 2008 ) . the radiative algorithm used in chymera has passed a series of radiative transfer tests that are relevant for disk studies , including permitting convection when expected . our conclusions regarding fragmentation are based on mulitple analyses : surface density rendering ( figures 4 - 7 ) , an energy budget analysis ( figures 9 and 10 ) , cooling temperature and brightness maps ( figures 11 - 12 ) , and a cooling time stability analysis ( figures 13 - 15 ) . it is also important to point out that these results do not contradict stamatellos et al . ( 2007 ) or krumholz et al . ( 2007 ) , who find fragmentation in massive , extended disks ( @xmath170 au ) . it is also pertinent to demonstrate that the chymera code can detect fragmentation when cooling rates are high and @xmath6 is low . figure 17 shows a snapshot for a simulation similar to the 512 @xmath98 simulation , but with the divergence of the fluxes artificially increased by a factor of two . in the normal simulation , kinks in the spiral waves form during the onset of the burst . as discussed above , the disk is very close to the fragmentation limit , but the knots do not break from the spiral wave . figure 15 suggests , although for a later time , that increasing the cooling rates by a factor of two would drop @xmath168 well below unity in low-@xmath6 regions . figure 17 confirms that the disk fragments with such enhanced cooling . however , we remind the reader that the @xmath98 simulation has the optimal optical depth for cooling ( @xmath101 ) . we can not imagine any physical process that could cause a factor of two enhancement in cooling . three clumps form , one for each spiral wave , between 4 and 5 au . the location of clump formation is consistent with the prediction by durisen et al . ( 2008 ) that a spiral wave is most susceptible to fragmentation near corotation . one of the clumps survives for several orbits and eventually passes through the inner disk boundary . resolution is always a concern for simulations . because these results are consistent with analytic fragmentation limits and numerical fragmentation experiments , we conclude that chymera can detect fragmentation at the resolutions employed for this study . as discussed above , when the opacity is abruptly decreased by a factor of 10@xmath171 , the disk does approach fragmentation - like behavior . a similar effect is reported by mayer et al . ( 2007 ) , who find that their disk only fragments when the mean molecular weight is suddenly switched from @xmath172 to @xmath173 . however , the simulation presented by boley et al . ( 2006 ) was accidentally run with @xmath173 , and cai et al . ( 2006 , 2008 ) purposefully ran their simulations at the high @xmath174 for comparison with the boley et al . none of the studies reported disk fragmentation . this may indicate that when a disk approaches fragmentation shortly after a sudden switch in a numerical parameter , e.g. , opacity here and @xmath174 in mayer et al . , the fragmentation may be numerically driven rather than physically , especially because sudden changes in cooling rates make disks more susceptible to fragmentation ( clarke et al . regardless , the results do indicate that disk fragmentation by gis may be possible under very extreme conditions . whether such conditions are physically possible or realistic remains to be shown . based on our results here and in earlier papers , disk fragmentation for @xmath175s of au appears to be a yet - unproven exception rather than the norm . in these simulations , the strong bursts of gi activity provide high mass fluxes ( @xmath176 yr@xmath76 , see fig . 16 ) throughout each disk . even though corotation is at @xmath0 au for the major spiral arms , the 2 au region of the disk is strongly heated . fluid elements approach peak temperatures of @xmath177 k in all simulations ( fig . it is not difficult to speculate that if a larger extent of the disk were modeled , the temperatures due to shocks would be large enough to ionize alkalis thermally and possibly sublimate dust ( 1400 - 1700 k ) . if such a condition is met , then an mri could activate and rapidly carry mass into the innermost regions of the disk . from these simulations it appears to be plausible , but by no means proven , that a burst of gi activity as far out as 4 au can drive mass into the inner regions of the disk and create strong temperature fluctuations , which may then be responsible for a thermal instablility . although we are simulating very massive disks , we note that such systems may exist during the class i yso phase . additional studies need to be conducted , preferably with a self - consistent buildup of a dead zone , to address the efficiency of this mechanism in low mass disks . there are at least three observable signatures for this mechanism . first , if a gi - bursting mass concentration at a few au ultimately results in an fu ori phenomenon , then one would expect to see an infrared precursor from the gi burst , with a rise time of approximately tens of years . second , one would also expect for a large abundance of molecular species , which would normally be frozen on dust grains , to be present in the gas phase during the infrared burst due to shock heating ( hartquist 2007 , private communication ) . third , approximately the first ten au of the disk should have large mass flows if the burst takes place near @xmath0 to 5 au . we speculate based on these results that if the burst were to take place at 1 au , then high outward mass fluxes should be observable out to a few au . each simulation shows that a large fraction of material goes through shocks with @xmath154 . although these shocks are weak , their abundant numbers may result in the processing of dust to some degree everywhere in the 2 - 10 au region . in fact , such processing may be necessary for prepping chondritic precursors for strong - shock survival ( ciesla 2007 , private communication ) . based on the arguments presented in 2 , the intent was to produce spirals with large pitch angles by constructing a disk biased toward a strong , sudden gi - activation near 5 au . even with this bias , the pitch angles of the spirals remain small , with @xmath178 . why are the pitch angles so small ? according to the wkb approximation , @xmath179 ( binney & tremaine 1987 ) , where @xmath180 is the radial wavenumber for some @xmath84-arm spiral . the most unstable wavelength for axisymmetric waves ( binney & tremaine 1987 ) is roughly @xmath181 for disk scale height @xmath59 . for a disk unstable to nonaxisymmetric modes , @xmath182 corresponds to some @xmath84-arm spiral ( e.g. , durisen et al . 2008 ) . by relating @xmath183 , @xmath184 in the linear wkb limit ( @xmath185 ) , where @xmath186 is a factor of order unity that relates @xmath182 to the @xmath84-arm spiral . because @xmath187 in gravitationally unstable disks , the linear wkb analysis may be marginally applicable . equation ( 7 ) predicts that linear spiral waves in these disks should have pitch angles @xmath188 to 11@xmath48 for @xmath186 between 1 and 1/2 , respectively . it appears that this estimate for @xmath35 extends accurately to the nonlinear regime , a result we did not expect . the gi spirals are efficient at heating the disk and transporting angular momentum , but _ not _ at producing chondrules . only for the @xmath95 @xmath5 simulation are multiple candidate chondrule - producing shocks detected , and there is only one chondrule - producing shock in the 1024 @xmath94 @xmath5 simulation . these low opacity disks have relatively flocculent spiral morphologies , including kinks in spiral waves . the @xmath98 simulations are on the verge of fragmentation . when interpreting these results , it is important to remember that the opacities were lowered abruptly and in a quite unphysical way . in addition , these detections are based on the fractional pressure change @xmath189 . if the temperature change is used , no chondrule - producing shocks are detected . this discrepancy may be due to efficient radiative cooling , which may make the assumption of adiabatic shock conditions incorrect , and/or to confusion with additional waves induced by shock bores . for the rest of this section , we assume that the pressure difference is the reliable shock identifier in order to discuss the implications of detecting chondrule - forming shocks in these disks . a thermal and spatial history for a fluid element that experiences a possible chondrule - forming shock is shown in figure 19 . to estimate the occurrence of a chondrule - forming shock , we employ a generous @xmath34-@xmath31 criterion ( fig . we do not take into account the optical depth criterion of miura & nakamoto ( 2006 ) on grounds that the large scale over which these shocks take place can allow for chondrules to equilibrate with their surroundings ( cuzzi & alexander 2006 ) . however , one should be aware that the optical depth criterion of miura & nakamoto will likely exclude all chondrule - producing shocks inside 4 au in the @xmath34-@xmath31 plane for these simulations . all candidate chondrule - forming shocks occur between @xmath190 and 5 au and at altitudes that are roughly less than a third of the gas scale height . because a large degree of settling is assumed , these shocks are located in regions that may be consistent with the dusty environments in which chondrules formed ( wood 1963 ; krot et al . 2005 ) . we estimate that @xmath191 of dust is processed through chondrule - forming shocks . because these shocks are limited to the onset of the gi burst , a few@xmath192 of dust would be processed in the @xmath95 @xmath5 disks if they went through about ten outbursts . if more outbursts take place than those that lead to an fu orionis event , then more chondritic material could be produced . in order to produce these shocks , the disk was pushed toward fragmentation by suddenly dropping the opacity by a factor of @xmath128 . we conclude that for bursts of gis near 4 or 5 au to produce chondrules , the disk must be close to fragmentation . the hypothesis behind the work summarized here is that bursts of gi activity , dead zones , the fu ori phenomenon , and chondrule formation are linked . the general picture is that mass builds up in a dead zone due to layered accretion and that gis erupt once @xmath6 becomes low . this activation of the instability causes a sudden rise in the mass accretion rate that heats up the disk inside 1 au to temperatures that can sustain thermal ionization and an mri . the mri shortly thereafter activates the thermal instability ( bell & lin . 1994 ; armitage et al . 2001 ; zhu et al . in addition , the strong shocks process dust and form chondritic material in the asteroid belt and at comet distances . given the results of our simulations , we think the boss & durisen conjecture that dead zones bursting at @xmath193 4 - 5 au can produce chondrules is _ unlikely _ , unless the disks are on the verge of fragmentation . because the conditions for fragmentation inside @xmath194s of au are very difficulty to achieve , we conclude that this is not viable as a general chondrule - formation mechanism . as a result , we return to figure 1 . based on our simple , analytic argument in equation ( 1 ) , chondrule formation can take place in the asteroid belt and at cometary distances for gi - bursting dead zones between about 1 and 3 au , even with @xmath178 . as suggested by wood ( 2005 ) , chondritic parent bodies may be representative of temporal as well as spatial formation differences . bursts that occur roughly between 1 and 3 au seem to be able to accommodate this scenario , and could drive the fu ori phenomenon ( armitage et al . 2001 ) . as the disk evolves , the location of the dead zone can vary , with multiple bursts occurring at a few au . accretion rates between @xmath82 and @xmath195 could build up a 0.01 @xmath74 dead zone every @xmath196 to @xmath197 years , respectively . some of these bursts may drive the fu ori phenomenon and produce chondrules , but others may only be responsible for chondrule - formation events . in this scenario , the disk need not fragment , and there should be between about 10 and 100 separate chondrule - formation events . future work is required to ascertain the plausibility of this modified version of the unified theory . we would like to thank f. ciesla , s. desch , and l. hartmann for fruitful discussions . we would also like to thank the anonymous referee for comments and suggestions that helped improve this manuscript . a.c.b.s contribution was supported by a nasa gsrp fellowship , and r.h.d.s contribution was supported by nasa grant nng05gn11 g . this work was supported in part by the iu astronomy department it facilities , by systems made available by the nasa advanced supercomputing division at nasa ames , by systems obtained by indiana university through shared university research grants through ibm , inc . , to indiana university , and by dedicated workstations provided to the astronomy department by iu s university information technology services . .simulation information . the name of the simulation indicates the fraction of the standard opacity that is used during the evolution of the disk . @xmath79 is the average @xmath6 between 3 and 6 au for a snapshot near 70/80 yr , depending on the simulation ( see text ) . the @xmath140 measurement is the sum of the time - averaged fourier components between roughly 55 and 80 yr . [ cols="^,^,^,^,^",options="header " , ] based on equation ( 1 ) . in the legend , @xmath198 for corotation radius of the spiral pattern @xmath37 and pitch angle @xmath35 . the colored regions highlight where chondrule formation is expected based on the desch & connolly ( 2002 ) shock calculations ( see text ) , with the yellow region appropriate for a mmsn density distribution and the orange for the same density distribution but with 10@xmath49 the mass . shocks that occur inside corotaton for @xmath199 will not produce chondrules between 1 and 5 au . however , chondrules can be produced by these low pitch angle spirals in shocks outside corotation . if the pitch angle is fairly open , such as @xmath200 , then a spiral wave with a corotation near 5 au can produce chondrules in the asteroid belt and at comet distances.,width=604 ] profiles for the the 512 simulations ( crosses ) and the 1024 simulations ( squares ) . the standard simulation does not have a 1024 counterpart because it was stopped at about 55 yr ( see text ) . these profiles demonstrate that the low - order structure dominates in the disk , with the fourier components dropping off quickly near @xmath125 . the lower opacity simulations have stronger @xmath124s , but there do seem to be some resolution effects present at large @xmath84 . the @xmath97 simulation is not shown for readability ; it falls between the @xmath93 and @xmath98 simulations and the 512 and 1024 resolutions follow each other closely.,width=604 ] simulations . the standard simulations are stopped early because the temperatures in the disk become too high to evolve due to inefficient cooling . the shock heating in the adiabatic ( no cooling ) simulation is indicated by the dark , dot - dash line . it is difficult to see because it follows the standard curves closely . for clarification , a @xmath201 is shown adjacent to the cooling curves and a @xmath18 is shown next to the heating curves.,width=604 ] and minimum , average , and maximum temperature plots for fluid elements in the 512 standard and the 1024 reduced opacity simulations . the temperature variations are quite large , approaching maximum temperatures of 1000 k. the values on the abscissa are due to the fluid elements that were lost from the simulated disk.,title="fig:",width=288 ] and minimum , average , and maximum temperature plots for fluid elements in the 512 standard and the 1024 reduced opacity simulations . the temperature variations are quite large , approaching maximum temperatures of 1000 k. the values on the abscissa are due to the fluid elements that were lost from the simulated disk.,title="fig:",width=288 ] and minimum , average , and maximum temperature plots for fluid elements in the 512 standard and the 1024 reduced opacity simulations . the temperature variations are quite large , approaching maximum temperatures of 1000 k. the values on the abscissa are due to the fluid elements that were lost from the simulated disk.,title="fig:",width=288 ] and minimum , average , and maximum temperature plots for fluid elements in the 512 standard and the 1024 reduced opacity simulations . the temperature variations are quite large , approaching maximum temperatures of 1000 k. the values on the abscissa are due to the fluid elements that were lost from the simulated disk.,title="fig:",width=288 ] simulation . the fluid element shows strong variations in radius and height as it passes through shocks , with a net inflow due to gravitational torques . strong variations in the pressure and temperature are also present . the strong shock near 30 yr is a potential chondrue - forming shock based on the pressure profile . however , the temperature peak is not as high as one would expect for the same pressure jump in a 1d adiabatic shock . this discrepancy may be due to efficient radiative cooling or to additional wave effects caused by shock bores . , width=571 ] -@xmath31 plane for the 1024 @xmath98 simulation , as determined from the fractional pressure change . the green area shows the chondrule - forming region , roughly based on the results of desch & connolly ( 2002 ) . the green strip should not be thought of as definitive , and slight changes in the strip s location could identify more chondrule - forming shocks , particularly at low @xmath34 . if the pre- and post - shock temperature ratio is used to determine @xmath34 , no chondrule - forming shocks are detected and most shocks have a @xmath206 4 km s@xmath76.,width=571 ]

the origin of a recently discovered close - in neptune - mass planet around gj436 poses a challenge to the current theories of planet formation . based on the sequential accretion hypothesis and the standard theory of gap formation and orbital migration , we show that around m dwarf stars , close - in neptune - mass ice - giant planets may be relatively common , while close - in jupiter - mass gas - giant planets are relatively rare . the mass distribution of close - in planets generally has two peaks at about neptune mass and jupiter mass . the lower - mass peak takes the maximum frequency for m dwarfs . around more massive solar - type stars ( g dwarfs ) , the higher - mass peak is much more pronounced . these are because planets tend to undergo type ii migration after fully accreting gas around g dwarfs while they tend to migrate faster than gas accretion around m stars . close - in neptune - mass planets may also exist around g dwarfs , though they tend to be mostly composed of silicates and iron cores and their frequency is expected to be much smaller than that of neptune - mass planets around m dwarfs and that of gas giants around g dwarfs . we also show that the conditions for planets migration due to their tidal interaction with the disk and the stellar - mass dependence in the disk - mass distribution can be calibrated by the mass distribution of short - period planets around host stars with various masses .

astro-ph0502566

in an attempt to place quantitative constraints on models of planet formation , we developed an algorithm to simulate the kinematic properties of gas giants formed in isolation ( ida & lin 2004a , hereafter paper i ) . this prescription is based on the sequential accretion model in which we assume that jupiter - mass gas - giant planets formed through 1 ) grain condensation , 2 ) runaway planetesimal coagulation @xcite , 3 ) oligarchic growth of protoplanetary embryos @xcite , and 4 ) gas accretion onto solid cores ( embryos ) @xcite . based on a distribution of 1 ) dust - disk masses ( @xmath0 ) inferred from mm data @xcite , 2 ) a range ( 1 - 10 myr ) of disk depletion time scale ( @xmath1 ) inferred from the observed decline in the ir @xcite and mm data @xcite , and 3 ) three different growth - termination criteria @xcite , we simulated a distribution of protoplanetary masses . in addition , we considered the effect of post formation type - ii orbital migration due to planet - disk interaction @xcite which is an important process in relocating protoplanets away from their birth places . the last process has been invoked @xcite to account for the origin of a population of jupiter - mass short - period planets such as 51 peg b @xcite . we compared the results of our simulation with the available data of extrasolar planets . we suggested that around solar - type stars , 1 ) there may be a deficit of intermediate - mass ( @xmath2 ) and intermediate - period ( @xmath3yr ) in their mass - period distribution ( paper i ) , 2 ) the frequency of gas giant planets may be an increasing function of their host stars metallicity [ fe / h]@xmath4 ( ida & lin 2004b , hereafter paper ii ) , and 3 ) a large fraction of the planets migrated to the proximity of their host stars may have perished ( paper i , ii ) . the first conclusion results directly from the expectation that a ) the growth of protoplanetary cores is limited by dynamical isolation in the inner regions of planetary systems and slow coagulation rate in the outer regions , b ) even under the most favorable locations , @xmath5 myr is needed for the formation of protoplanetary embryos ( cores ) , anywhere in the disk , with masses @xmath6 , c ) cores with mass larger than @xmath7 several @xmath8 can undergo runaway gas accretion , and d ) orbital migration occurs on a similar time scale ( @xmath9 ) to gas depletion in the disk ( @xmath1 ) . the model we have presented so far provides the first step in the construction of deterministic properties of planetary formation . its simplistic assumptions must be re - examined in a wider context of protoplanetary environment . for example , planets are assumed to form independently and in isolation in the model . dynamical isolation in this context means that there are no other major planets around the same host stars . most importantly , post - formation dynamical interaction between the planets have not yet been considered . while these effects will be examined in a future investigation , we consider here the mass function of dynamically - isolated planets and its dependence on the mass of the host star . on the observational side , most of the known planets are found around solar - type stars ( g dwarf stars ) because the most successful radial velocity surveys have been conducted for these target stars which have cool and well defined atmospheric spectroscopic features . however , the search window is rapidly expanding to lower - mass stars as detection techniques are being refined in both radial velocity and transit searches . figure [ fig : obs ] shows the distributions of semimajor axis ( @xmath10 ) and mass ( @xmath11 ) of discovered extrasolar planets around m , k , g , and f stars . the planets around subgiants are excluded , because the relation between stellar spectral - type and its mass is different from that for main sequence stars . the planets discovered by transit survey are also excluded , because the transit survey has different observational bias in planetary periods from that of doppler survey . although much more planets have been discovered around g stars than around f and k stars , the detection probability ( after correction of metallicity dependence ) is similar among these stars ( fischer & valenti 2005 ) . however , the detection probability may be by order of magnitude lower around m stars . since their signals are the most conspicuous , close - in planets are expected to be the first to be uncovered ( narayan _ et al . however , until recently , the only planets discovered around m stars are two jupiter mass planets with moderate semimajor axis ( 0.13au , 0.21au ) around gliese 876 . no jupiter - mass close - in planets have been discovered around m stars . recently , a short - period ( 2.6 days ) neptune - mass ( @xmath12 ) planet is found to be orbiting around an m - dwarf ( @xmath13 ) gj436 @xcite . with a mass between those of gas giants and the earth , this finding signifies a transition in the quest to search for terrestrial planets . in the solar system , two ice giants , uranus and neptune , have masses in this range . these planets are primarily composed of icy cores with a modest gaseous envelope . in accordance with the core accretion scenario , e.g. , @xcite , the cores in the outer solar system take a long time to emerge and when they finally acquired @xmath14 , the solar nebula was already so severely depleted that they can only accreted a small amount of gaseous envelope @xcite . it is natural to extrapolate that gj436b may also have attained @xmath14 but failed to accrete much gas . the main challenge to such a scenario is to account for the origin of both its low mass and short period . the period of the first extrasolar gas giant planet ( 51 peg b ) discovered @xcite around a main sequence star outside the solar system is comparable to that of gj436b . that planet and dozens others like it are thought to have formed through sequential accretion beyond the ice boundary and migrated to their present locations @xcite as a consequence of their tidal interaction with their nascent gaseous disks @xcite . but , the much lower mass @xmath15 of gj436b implies that gas accretion onto it may have been greatly suppressed prior to , during , and after its migration . two other short - period neptune - mass planets have been found around g dwarfs ( a planet with 2.8 day period and @xmath16 mass around 55 cnc and that with 9.6 day period and @xmath17 mass around hd 160691 ) @xcite . since 55cnc and hd160691 have three and two other giant planets ( maybe gas giants ) , these neptune - mass planets could form _ in situ _ by accumulation of rocky materials caused by sweeping mean motion resonance associated with migration of a giant planet , e.g. , @xcite or sweeping secular resonance associated with disk gas depletion , e.g. , @xcite . since we do not include interaction from other major planets , formation of these neptune - mass planets is beyond scope of the present paper . on the other hand , no gas giant planets has been found around gj436 . gj436b is dynamically - isolated , so that its formation must be considered without help of other giant planet(s ) and our calculation in the present paper can address its formation . although in the present paper , we will propose a scenario that type ii migration of a neptune - mass planet occurs without significant gas accretion onto the planet around an m star , there are several other potential scenarios for the origins of dynamically - isolated , close - in neptune - mass planets . @xcite suggested that , along with jupiter and saturn , uranus and neptune were formed through gravitational instability , in a massive disk . heavy elements settled to form the cores and the gas envelope ( clump ) was greatly depleted by the photo - evaporation due to the uv flux from nearby ob stars before it contracts to planetary size . since gravitational instability is unlikely at the present location ( 0.028au ) of gj436b , the planet must have migrated from outer region by tidal interaction of the disk . however , the gravitational potential of disks around m stars is shallower than that around g stars such that the evaporation of the envelope of collapsing clumps would also eliminate all the residual gas in the disk . it would be difficult for the collapsing fragment to lose a large fraction of their mass and migrate extensively to form a gj436b - like planet . on the other hand , under some extreme circumstance , the sporadic uv and x - ray irradiation from its host star could evaporate envelope of a gas giant planet that has migrated from outer region during m star s main sequence lifetime . but the overall impact of the photoevaporation process on planet s envelope and mass has not been determined . under the general concept of sequential accretion scenario , dynamically - isolated close - in planets may also form _ in situ _ @xcite . this scenario requires a concentration of planetesimals in the stellar proximity . one possible mechanism which may lead to such a situation is through embryo - disk interaction , or commonly known as type - i migration @xcite . the accumulation of building blocks for close - in planets also requires the termination of their migration process and the interaction between the embryos with their host stars as well as residual planetesimals . through a case study , we consider in this paper the origin of dynamically - isolated close - in planets of @xmath18 neptune mass around stars with various masses . with the model developed in paper i and ii , we show that neptune - mass close - in planets may be abundant around m stars . in 2 , we briefly recapitulate the sequential accretion hypothesis . using our prescription , we first compare the mass function of dynamically - isolated close - in versus modest to long - period planets around solar - type stars ( g dwarfs ) in 3 . this comparison is useful because we have most observational data and constraints for planets around solar - type stars at the present moment . based on the best available observed properties of pre - main sequence evolutionary tracks of different - mass stars and accretion rates onto their host stars , we construct disk models around main sequence stars with various masses . in 4 , we construct a conventional model of giant planet formation around lowest - mass stars and show that icy planets with 10-@xmath19 accrete from planetesimals at @xmath20au without any significant gas accretion onto the planets . these neptune - mass planets can also readily migrate to the proximity of the stellar surface , where gas accretion is quenched at neptune - mass . through a series of simulations , we show , in 5 , that the disk mass dependence on their host stars mass and the criteria of tidally induced migration in the disk may be quantitatively constrained by the mass distribution of short - period planets . we also place constraints , in 6 , on the dependence of disk mass on the stellar mass . finally , in 7 , we summarize our results and discuss their implications . the detailed description of sequential accretion scenario and our prescription to simulate the formation of planets are given in paper i. we briefly recapitulate the central features of our approach and define various quantities which are used in the discussions of our results . in the sequential accretion scenario , planetesimals grow into protoplanetary embryos ( cores ) which affects the velocity dispersion @xmath21 of nearby planetesimals and modifies their own growth @xcite . but , @xmath21 is also affected by gas drag @xcite . in a disk with a surface density of dust ( @xmath22 ) and gas ( @xmath23 ) around a host star with mass @xmath24 , protoplanetary embryos mass at any location @xmath10 and time @xmath25 is @xmath26 where @xmath27 is the typical mass of the planetesimals accreted by the embryos . since accretion time @xmath28 increases with @xmath29 , @xmath30 does not depend on its initial value @xmath31 as long as @xmath32 . in the limit of small @xmath21 , the full width of embryos feeding zone ( @xmath33 ) is limited to @xmath34 @xcite where @xmath35 is the embryos hill s radius @xmath35 @xmath36 . when all the residual planetesimals in an embryo s feeding zone have coagulated with it , the embryo attains an isolation mass @xmath37 equations ( [ eq : m_grow ] ) and ( [ eq : m_iso ] ) indicate that both @xmath30 and @xmath38 are determined by the distribution of @xmath22 and @xmath23 . in paper i and ii , we introduced a multiplicative factor ( @xmath39 and @xmath40 ) to globally scale disks with the minimum mass nebula model for the solar system @xcite such that @xmath41 , \\ \sigma_{\rm g } \label{eq : sigma_dust } & = 2.4 \times 10 ^ 3 f_{\rm g } h_{\rm g } ( a/ { \rm 1 au})^{-3/2 } \;\ ; [ { \rm g~cm}^{-2 } ] , \label{eq : sigma_gas } \end{array } \right.\ ] ] where the step function @xmath42 inside the ice boundary at @xmath43 and 4.2 for @xmath44 . [ note that the latter can be slightly smaller ( @xmath45 ) @xcite . ] the minimum mass model corresponds to @xmath46 . here we introduced a new scaling factor , @xmath47 and @xmath48 , representing the dependence on stellar mass ( see below ) . if the disk is optically thin and heated by stellar irradiation only @xcite , @xmath49 the stars luminosity @xmath50 is generally a function of their mass @xmath24 and age @xmath25 . additional heating due to viscous dissipation enlarges this boundary @xcite . ad hoc _ phenomenological prescription provides a useful working hypothesis for comparative analysis between solar system architecture and extrasolar planetary systems . since @xmath51 is generally an increasing function of @xmath52 , @xmath43 is small in disks around low - mass stars . around solar - type t tauri stars , the observationally inferred total mass of dust , @xmath0 , in the protostellar disks ranges from @xmath53 to @xmath54 @xcite , which corresponds to a range of @xmath55 . the disk - mass determinations are somewhat uncertain due to the poorly known radiative properties of the grains . nevertheless , their divergent dust content provides a reasonable evidence for a greater than an orders of magnitude dispersion in @xmath0 . in addition , the radio image of the disks is not well resolved in many cases . a rough magnitude of @xmath39 can be inferred from the the total mass of the disk under the assumption that all disks have similar sizes ( a few tens au to a hundred au ) . in this paper we follow paper ii and generate a set of @xmath39 with a unit variant of gaussian logarithmic distributions and a range between 0.1 - 10 with cut - off of high @xmath56 tails at @xmath57 ( see figure [ fig : f_d_dist ] ) . if @xmath58 , the disk is gravitationally unstable at @xmath59 a few au ( see discussion in section 7 ) , and the total mass of heavy elements in the disk would be significant fraction of that in the host star . in eq . ( [ eq : sigma_dust ] ) , the dependence of @xmath22 on @xmath24 is incorporated in a mass scaling function @xmath47 . in general , a relatively weak ir excess is associated with low - mass stars , which suggests @xmath47 may be an increasing function of @xmath24 , at least in the inner regions of the disk . for host stars with @xmath60 , @xmath61 . equation ( [ eq : m_grow ] ) indicates that @xmath29 is also a function of @xmath23 through damping of @xmath21 due to gas drag . the planetary migration rate that we mention later is also dependent on @xmath23 . around other stars , very little information is available on @xmath23 . since there is no indication of divergent depletion pattern between molecular hydrogen and mm - size dust emission @xcite , we follow the same prescription for @xmath22 with a disk mass scaling parameter @xmath40 and a stellar - mass dependence function @xmath48 ( eq . [ [ eq : sigma_dust ] ] ) . similar to paper i and ii , we adopt the conjecture that @xmath39 does not change except in those regions where they have been totally accreted by the cores and @xmath62 where @xmath1 is disk depletion time scale discussed below . the assumption of the uniform exponential decay is for simplicity . the effects of detailed decay pattern as a result of viscous evolution of disks will be discussed in a separate paper . for computational simplicity , we assume [ fe / h]=0 , _ i.e. _ , a solar composition for all stars so that @xmath63 and @xmath64 . the dependence of planetary @xmath65 distribution on [ fe / h ] for solar - type stars has already been discussed in paper ii . we now consider a prescription for the stellar mass dependence . the parameters are @xmath47 ( @xmath66 ) and @xmath1 . in young clusters , the fraction of stars with detectable ir @xcite and mm continuum @xcite from circumstellar disks around t tauri stars declines on the time scales of 1 - 10 myr . although this decline may be due to dust growth and planetesimal formation rather than the depletion of heavy elements @xcite , the correlation between the intensity of mm dust continuum with the gas decline emission @xcite and the uv veiling for ongoing gas accretion suggest that gas is depleted as the dust signature fades . the dust signature maintains up to 10myr also for disks around brown dwarfs @xcite . so , in this paper , we assume @xmath1 in the range of 1 - 10 myr for all stellar masses . although direct estimates of disk mass both in gas and dust are difficult to obtain for the inadequate sensitivity of existing observational instruments , the disk accretion rate @xmath67 can be inferred from the h@xmath68 line profiles @xcite such that @xmath69 with a large dispersion . if the angular momentum transfer and mass diffusion time scale is insensitive to @xmath24 , we could infer @xmath48 @xmath70 @xmath71 . in view of the large uncertainty in the data , we consider three possible dependence on @xmath24 , @xmath72 with @xmath73 as a standard case . with these prescriptions for disk parameters , we find that @xmath74 by assuming @xmath75 g . from eq . ( [ eq : sigma_dust ] ) , we estimate the core masses in protoplanetary systems with eq . ( [ eq : m_iso ] ) such that @xmath76 we have already pointed out in paper i and ii that the growth of planetesimals is limited by isolation at small @xmath10 and the slow coagulation rate at large @xmath10 . since for smaller @xmath52 , @xmath35 is larger , @xmath77 is larger for the same mass disks . however , both @xmath47 and @xmath48 are increasing functions of @xmath24 , so that the growth of @xmath78 and the isolation mass @xmath38 at any given @xmath10 are actually smaller for lower - mass stars . we do not explicitly include type i migration @xcite of cores , but take into account its effects in some runs ( see paper i and discussion in 5 ) . as the cores grow beyond a mass @xmath79 their planetary atmosphere is no longer in hydrodynamic equilibrium and they begin to accrete gas @xcite . in the above equation , we neglected the dependence on opacity ( see paper i ) . in regions where they have already attained isolation , the cores accretion @xmath80 is much diminished and @xmath81 can be comparable to an earth mass . but the gas accretion rate is still regulated by the efficiency of radiative transfer such that @xmath82 where @xmath15 is the planet mass including gas envelope and the kelvin helmholtz contraction time scale is ( for details , see paper ii ) @xmath83 gas accretion onto the core is quenched when the disk is depleted either locally or globally . a protoplanet induces the opening of a gap when its rate of tidally induced angular momentum exchange with the disk exceeds that of the disk s intrinsic viscous transport @xcite , that is , when the planet mass @xmath15 exceeds @xmath84 where we used an equilibrium temperature in optically thin disks @xcite and @xmath68-prescription for the effective viscosity @xmath85 ( shakura & sunyaev 1973 ) , in which @xmath86 k and @xmath87 where @xmath68 is a dimensionless parameter , @xmath88 and @xmath89 are the disk scale height and kepler frequency . since @xmath90 where @xmath91 is the sound velocity , @xmath92 , so that @xmath93 . since @xmath51 is an increasing function of @xmath52 , eq . ( [ eq : m_gas_vis ] ) indicates that @xmath94 is smaller around lower - mass stars . the orbital evolution of planets is locked to the viscous evolution of the disk gas ( type ii migration ) when their @xmath95 where @xmath96 is the dimensionless factor @xmath97 for a laminar disk @xcite . the type ii migration rate @xmath98 is given by @xmath99 with @xmath100 where @xmath101 is jupiter mass . we set a lower limit on @xmath9 at @xmath102 yrs ( paper i ) . we found that the formula for @xmath103 ( angular momentum flux at the radius of maximum viscous couple , @xmath104 ) given in paper i ( eq . [ 63 ] in the paper ) must be multiplied by a factor @xmath105 . if we use @xmath103 to evaluate evolution of planetary orbital radius @xmath10 , eq . ( [ eq : tau_mig ] ) is reduced by @xmath105 as well ( paper i ) . however , since the planetary migration may be caused by a fraction of @xmath103 and the fraction is uncertain , we use eq . ( [ eq : tau_mig ] ) in the present paper too . we adjust the migration time scale by the value of @xmath68 , which is also uncertain , comparing with observational data . as shown in paper ii , in order to reproduce period distribution of gas giant planets similar to that of observed extrasolar planets around solar - type stars , the disk viscous diffusion time scale @xmath106 yrs must be comparable to disk lifetime @xmath1 . hence , the @xmath68 viscosity in our model must be @xmath107 . although the regions at 1 - 10au could be `` dead zone '' for mhd turbulence @xcite , resulting in a very small @xmath68 in these regions , the best fit value @xmath108 may not necessarily reflect a realistic value because of rather simple assumptions for @xmath109 distribution and its exponential decay in our model . we will carry out more detailed calculations coupled with disk viscous evolution in a separate paper . the planets migration is terminated either when the disk is severely depleted ( @xmath110 ) or when they reach @xmath111 , which is set to be 0.04au in our calculations . there are potential mechanisms to stop migration at @xmath112au @xcite . however , paper ii suggests that only a small faction ( @xmath113% ) of migrating planets can survive in the vicinity of their host stars . this fact should be kept in mind when population of close - in planets is discussed with our model . the gap becomes locally severely depleted when the planets hill s radius ( @xmath35 ) exceeds the disk thickness @xmath88 @xcite , that is , when @xmath15 exceeds @xmath114 growth through gas accretion is quenched for planets with @xmath115 . numerical simulations show some uncertainties in the dimensionless parameter @xmath116 @xcite . planets with @xmath117 migrate with the disk while continue to accrete gas , albeit at a reduced rate , e.g. , @xcite . we use eq . ( [ eq : mgsdot ] ) without a reduction factor for simplicity , because the reduction factor is quite uncertain and introduction of the factor does not affect the results significantly . since gas accretion for @xmath118 is already very rapid ( [ eq : tau_kh ] ) , the reduction does not change total gas accretion time scale . following paper i , we adopt in this paper @xmath119 , that is , the truncation condition is @xmath120 . since @xmath51 rapidly increases with @xmath52 , @xmath121 is also smaller around lower - mass stars . even for planets with @xmath122 , gas accretion may be ultimately limited by the diminishing amount of residual gas in the entire disk . for our disk models , the maximum available mass is @xmath123 when @xmath124 becomes smaller than @xmath15 , gas accretion is terminated . a similar global limit @xmath125 ( see eq . [ [ eq : m_noiso_sp ] ] ) is also imposed if the formula ( [ eq : m_iso ] ) exceeds it . as the gas is severely depleted , the velocity dispersion @xmath21 of the embryos and residual planetesimals grows until they cross each other s orbits @xcite . eventually a few surviving embryos acquire most of the residual planetesimals and less massive cores during the late oligarchic - growth stage . the asymptotic embryos masses are given by eq . ( [ eq : m_iso ] ) with @xmath126 , @xmath127 where @xmath128 and @xmath129 are the surface escape speed and internal density of the embryo . we use this enlarge asymptotic mass when @xmath130 . in paper i and ii , we put all of these processes into a numerical scheme to simulate the formation and migration probabilities of planets around solar - type stars . cores with @xmath131 emerge on time scale shorter than @xmath1 in disks with modest - to - large values of @xmath39 around solar - type stars . in this limit , gas accretion and orbital migration lead to the formation of gas giants with kinematic properties similar to those observed we now apply our numerical methods to study the formation of planets around stars with various masses . our objective is to simulate the mass function of close - in planets and assess the influence of formation and migration on it . we show here that this quantity can provide clues on the dominant processes which regulate planet formation and it also can be used to distinguish between competing theories of planet formation . in all models , we choose @xmath132 based on assumption that the viscous evolution time scale for the disks is comparable to @xmath1 ( see 2.3 ) . using these models , we carry out monte carlo simulations . for simplicity , we generate a set of initial @xmath10 s of the protoplanets and @xmath1 with uniform distributions in log scale in the ranges of 0.1@xmath133au and @xmath134@xmath135 yrs . the assumed @xmath10 distribution corresponds to orbital separations @xmath136 that is proportional to @xmath10 , which may be the simplest choice . the distribution of @xmath39 was discussed in 2.2 ( also see figure [ fig : f_d_dist ] ) . in paper i and ii , we also assumed a distribution of @xmath24 in a range of 0.7 - 1.4@xmath137 . in this paper , in order to make the @xmath24 dependence clear , the value of @xmath24 is fixed in each run . in the following sections , we consider several sets of model parameters . we first discuss a standard model with @xmath138 and @xmath73 around a solar - type star with @xmath139 ( model 1.0 ) . the evolution of totally 20,000 planets are calculated for each run . model 1.0 is similar to the results we have already presented in paper i. in figure [ fig : ma]a , we highlight the mass @xmath140 and semi major axis @xmath141 distribution of planets at @xmath142 yrs after they have attained their asymptotic mass and gone through the initial migration due to their tidal interaction with their nascent disks . the main features to notice in this panel are : 1 ) a deficit of planets with intermediate masses ( @xmath143 ) at intermediate semi major axis ( @xmath144 ) , and 2 ) a large population of close - in ( @xmath145au ) gas giants with @xmath146 , although only a small faction of them ( @xmath113% ) may be able to survive ( paper ii ) . we have already indicated in paper i that these properties are due to 1 ) the runaway nature of dynamical gas accretion and 2 ) type ii migration . in order to distinguish between these two dominant effects , we trace back , in figure [ fig : ma]b , the initial semi major axis @xmath147 where the cores of both close - in ( marked by black circles ) and intermediate or long - period planets ( marked by gray dots ) formed . these results clearly indicate an one - to - one mapping between the mass function of the close - in planets and the locations where they are formed . for illustrative purposes , we also mark the domain where some physical processes operate and dominate the evolution of protoplanets . for example , the thin solid lines indicate the upper limit of the isolation mass that cores can attain prior to gas depletion ( it is obtained with @xmath148 in @xmath38 and @xmath149 ) . the transition at 2.7 au in model 1.0 corresponds to the ice boundary . the thick solid lines indicate the critical mass for the onset of type ii migration , @xmath150 with @xmath151 according to eq . ( [ eq : m_gas_vis ] ) . we also highlighted the asymptotic growth limit @xmath152 with @xmath153 ( eq . [ eq : m_gas_th ] ) by broken lines . at any given @xmath154 , a fraction of terrestrial planets can form with @xmath15 above @xmath155 through 1 ) gas accretion and 2 ) merger of residual planetesimals and other embryos after disk gas depletion . when the embryos reach their isolation mass , @xmath156 declines with the vanishing @xmath80 ( eq . [ [ eq : crit_core_mass ] ] ) . although gas accretion is initiated , planets can not grow significantly prior to severe gas depletion unless the planets isolation mass @xmath157 several @xmath8 . the gas accretion time scale given by eq . ( [ eq : tau_kh ] ) can be comparable to or shorter than @xmath1 only for cores with @xmath158 several @xmath8 . mergers of residual planetesimals and other embryos occur during and after the gas depletion ( see eq . [ [ eq : m_e_iso ] ] for asymptotic @xmath159 ) ( kominami & ida 2002 ) . if their @xmath160 and there is adequate residual gas in the disk ( @xmath161 is still large enough ) , they would migrate to the vicinity of their host stars ( eq . [ [ eq : tau_mig ] ] with eq . [ [ eq : fg_tdep ] ] ) . the magnitude of @xmath162 is an increasing function of @xmath10 . for @xmath163 au , @xmath164 for embryos which can migrate is longer than their @xmath9 . although their @xmath15 may be @xmath165 , their inward migration is sufficiently rapid that they do not acquire a significant amount of gas along the way @xcite . termination of migration allows low - mass planets ( formed at @xmath163au in relatively massive disks ) to grow through gas accretion to their asymptotic mass @xmath166 at the stalling location which is set to be @xmath167 au . thus , the final mass of close - in planets with @xmath168au is @xmath169 regardless of truncation mass @xmath121 at the original locations @xmath154 . these planets are formed interior to the ice boundary and they are likely to be mostly composed of silicates and iron , in contrast to the ice giants in the solar system . also , they can not accrete large amount of gas because the aspect ratio of their nascent disk at 0.04 au is so small that they induce clear gap formation when their mass reaches that of neptune . planets formed at slightly larger @xmath10 s ( @xmath20au ) must attain @xmath170 before they acquire the mass to start migration , @xmath171 ( eq . [ [ eq : m_gas_vis ] ] ) . with this critical mass , embryos which can initiate migration there can also accrete gas efficiently with @xmath172 provided the gas in the disk is not severely depleted . since @xmath121 decreases with decrease in @xmath10 , when they arrive close to their host stars , their @xmath15 may be @xmath173 so that they would not acquire any additional mass . note that cores of these planets are also mostly made of silicates and iron rather than ice and they should not be referred to as hot neptunes . at even larger radius , @xmath174 . prior to reaching isolation , embryos gas accretion is suppressed by the bombardment of residual planetesimals , _ @xmath175 for moderate @xmath176 . after isolation is reached , @xmath156 becomes smaller than @xmath29 and their @xmath164 due to gas accretion is reduced below @xmath9 and they quickly evolve into gas giants . thus , final @xmath15 in figure [ fig : ma]b coincides with @xmath121 at initial locations . beyond @xmath177 au , the time scale for the emergence of cores with @xmath131 or isolation is comparable to or longer than both @xmath9 or @xmath1 . although they may acquire @xmath15 in the range of @xmath178 through merges of residual embryos after the gas depletion , these planets generally do not migrate extensively . interior to the ice boundary , nearly all the planets with sufficient mass to initiate efficient gas accretion have migrated to the vicinity of the host star ( see figure [ fig : ma]b ) . but , a majority of the planets which migrated to the vicinity of solar - type stars were formed beyond the ice boundary , as gas giants , prior to their migration . there is a narrow window in the range of @xmath10 where the seed of intermediate - mass planets may form and migrate to the proximity of their host stars . in figure [ fig : histo ] , theoretically predicted mass distributions are plotted . since it is expected that most of close - in planets may fall onto their host stars , we plot the distributions of close - in planets , reducing the amplitude @xmath179 by a factor 10 . for comparison of the amplitude between close - in and distant planets , uncertainty in this calibration is noted . we also plotted observed distributions . since the number of runs in each model does not reflect the number of targets for current doppler survey , we can not compare the amplitude @xmath179 between the observed and the theoretically predicted distributions . only the shape of the distributions should be compared . also note that observed distributions do not exactly correspond to host stars mass of each model and numbers of observed planets are not large enough for statistical arguments for stars other than g stars ( model 1.0 ) . we show that , around stars with @xmath180 , the mass distribution for the close - in ( with @xmath145 au ) planets is skewed toward @xmath181 ( model 1.0 in figure [ fig : histo]a ) . this distribution is more enhanced near @xmath181 than that observed . the effect of post formation star - planet tidal interaction , which has not been taken into account in our model , may have caused the demise of a majority of the close - in planets , in particular massive planets ( gu _ et al . _ 2003 , paper ii ) . in disks with modest masses , planets form in the advance stages of evolution when gas depletion is well underway . some of these planets may migrate interior to the ice boundary and become stalled while others remain close to their place of birth . a majority of gas giant planets with @xmath182 au have migrated but not extensively . in contrast to the close - in planets , model 1.0 in figure [ fig : histo]b clearly shows a paucity of longer - period ( @xmath182 au ) planets at @xmath183 . this distribution reflects the stringent prerequisite that gas accretion into gas giants must be preceded by the rapid formation of sufficient mass cores whereas the build up of terrestrial planets can continue well after the severe depletion of the disk . in a generalization of the solar nebula model , nakano ( 1988 ) showed that the temperature distribution throughout the disk increases with @xmath24 . however , he did not consider the dependence of @xmath184 and @xmath23 on @xmath24 . in a recent paper , @xcite , considered a model in which the disk mass increases with the stellar mass . their objective is to demonstrate the difficulties to form gas giants around m dwarf stars . but , the effects of planetary migration , truncation of gas accretion due to gap opening and the gradual depletion of the disk gas are neglected . in this section , we consider the variation of 3 model parameters : 1 ) the stellar mass @xmath24 , 2 ) the dependence of disk mass on the stellar masses , @xmath47 , and 3 ) the condition for the onset of type ii migration . a standard series are model @xmath185 where @xmath186 represents models with @xmath187 . in the standard series , we set @xmath188 and @xmath189 . model @xmath185b ( series b ) and @xmath185c ( series c ) correspond to @xmath190 and 100 with @xmath73 . @xmath185 in @xmath185b and @xmath185c expresses @xmath191 as well . model @xmath185d ( series d ) and @xmath185e ( series e ) correspond to @xmath192 and 1 with @xmath138 . in figure [ fig : f_d_dist ] , distributions of @xmath193 we used for models @xmath185 and @xmath185b are shown . they represent the relative mass distribution of disks around stars with various masses . the mean value of @xmath22 is an increasing function of @xmath52 ( see eq . [ [ eq : sigma_dust ] ] ) . in order to limit additional model parameters , we assume zams mass - luminosity relationship , @xmath194 . since the time scale of pre - main sequence stage of lower - mass stars is long , planet formation around these stars may proceed during their pre - main sequence stage in which @xmath50 is rather large . in that case , the dependence of @xmath50 on @xmath24 is weaker , but it may still have a positive power - law dependence , so that the trend of the @xmath24-dependence of planetary systems shown below does not change . in the calculations in this section , @xmath195 @xmath196 @xmath197 @xmath198 @xmath199 @xmath200 and @xmath201 in figures [ fig : ma ] and [ fig : histo ] , we also included the results of the simulations for models 0.2 - 0.6 and 1.5 . we first present a low - mass model 0.2 since it is in strong contrast to model 1.0 . stars with @xmath202 corresponds to relatively light m stars . these stars are not only most numerous but they also contribute most to the initial stellar mass function . according to our prescription and model parameters , @xmath23 and @xmath22 around the host star in model 0.2 are 25 times smaller than those in model 1.0 . in the result of model 0.2 on the top panels of figures [ fig : ma]a and b , we find that jupiter - mass planets rarely formed around low - mass stars . this paucity is due to the slow growth rate of embryos and their low isolation mass such that little gas can be attained by them prior to its depletion . this result confirms the conclusion reached earlier by @xcite . the upper limit of the @xmath203 distribution is determined by the @xmath38 . this correlation arises because @xmath50 is a rapidly rising function of @xmath52 . in our prescription , the ice boundary is located at @xmath204au ( eq . [ [ eq : aice_sp ] ] ) which is 25 times closer to a host star with @xmath202 than in model 1.0 with solar - type stars . nearly all the cores formed around these low - mass stars are mostly composed of ice . planets can acquire masses @xmath205 through gas accretion . but with a relatively small @xmath38 , the accretion is inefficient ( eq . [ [ eq : tau_kh ] ] ) . they can also gain mass after the gas depletion through collisions and mergers of residual embryos . but the asymptotic mass @xmath206 is also relatively small for low - mass stars ( eq . [ [ eq : m_e_iso_sp ] ] ) . with the _ ad hoc _ @xmath68 prescription we have adopted , the necessary condition for the onset of type ii migration is satisfied for relatively low - mass planets ( eq . [ [ eq : m_gas_vis_sp ] ] ) . when their @xmath207 , these cores undergo orbital decay . similar to model 1.0 , there is a population of close - in planets with @xmath208 . the results on the top panel of figure [ fig : ma]b show that they indeed originated from a region between @xmath43 and @xmath209 au . they also indicate that planets with mass down to @xmath210 may migrate to the proximity of a @xmath211 star . at the arbitrary @xmath212 au around such a host star , the equilibrium temperature of these close - in neptune - mass planets is similar to that of the earth . the composition and structure of such planets have been already been discussed by @xcite . although a few planets formed with @xmath213 , they are the exceptional cases resulting from the tails of the @xmath39 distribution . consequently , the mass function of planets with modest or large @xmath214 ( @xmath215 au ) show a sharp decline at @xmath216 ( figure [ fig : histo]b ) . it also extends well into the low - mass range . in contrast , the mass function for the close - in planets ( with @xmath145 au ) shows a peak near @xmath217 ( figure [ fig : histo]a ) . in model 0.4 , we consider a host star with @xmath218 which corresponds to a relatively massive m dwarf star . it is an analog of gj436 , around which a single close - in neptune - mass planet has been discovered @xcite . figure [ fig : histo]a and b indicate that the frequency of planets observable with current doppler survey is significantly smaller for m stars than for k , g , and f stars ( also see figure [ fig : freq ] ) . this is consistent with the observation and the finding of @xcite . figure [ fig : mean_mp ] indicates the mean mass and characteristic mass associated with the highest peak of the mass distribution for a ) the close - in planets and b ) planets with @xmath219au . these masses increase with @xmath24 , in particular for the close - in planets . this accounts for the recent planetary finding around gj436 . we find _ the frequency of neptune - mass close - in planet peaks at @xmath220 _ ( figures [ fig : histo]a and [ fig : mean_mp ] ) . the formation of these planets depends sensitively on the environment and their frequency provides constraints on the sequential accretion scenario . most of these planets formed slightly beyond the ice boundary . for this stellar mass , the ice boundary is located at @xmath221 au . outside it , the upper limit of @xmath38 exceeds @xmath222 . in figure [ fig : tevol_mstar04 ] , we illustrate the evolution of a typical embryo that forms near the ice boundary in a disk with @xmath223 and a depletion time scale @xmath224 myr . the disk surface density @xmath223 is close to the tail of the distribution for @xmath225 stars ( figure [ fig : f_d_dist ] ) . during the initial @xmath226 yrs , an embryo grows to @xmath227 through coagulation . the rapid coagulation enhances @xmath81 , which prevents gas accretion ( see eq . [ [ eq : crit_core_mass ] ] and figure [ fig : tevol_mstar04 ] ) . since this embryo is located beyond the ice boundary and @xmath223 , @xmath77 is as large as @xmath228 . but , before it acquires @xmath77 , it opens up a gap and undergoes type ii migration at @xmath229 . during the migration , gravitational perturbations from the embryo prevents additional planetesimals from reaching the embryo and its growth is quenched @xcite . on a time scale of @xmath230 yr , the embryo migrates to 0.04 au where it is stalled . the termination of bombardment by residual planetesimals makes @xmath81 lower to start gas accretion onto the embryo . however , the gas accretion time scale @xmath164 is longer than @xmath9 ( figure 4 ) . gas accretion onto the planet actually proceeds after it reaches 0.04au . the planet s growth is quenched at @xmath231 as a clean gap is formed because @xmath232 at 0.04au . figures [ fig : ma ] and [ fig : histo ] show that many planets evolve in a similar way to acquire @xmath233 . the composition of these planets is similar to that of uranus and neptune and its day - side surface temperature may be @xmath234k . in the lowest - mass model 0.2 , the upper limit of @xmath38 does not exceed @xmath222 even outside the ice boundary . the migration of neptune - mass planets requires a delicate balance between @xmath38 and @xmath222 . thus , the frequency of close - in neptune - mass planets in model 0.4 is larger than that in model 0.2 . this frequency in model 0.4 is also larger than that in model 1.0 . although the upper limit of @xmath38 is larger than @xmath222 outside the ice boundary around solar - type stars , @xmath235 at the onset of migration , so that gas accretion is much more efficient onto such large cores ( eq . [ [ eq : tau_kh ] ] ) . thus , most isolated planets tend to arrive at the proximity of a solar - type star as gas giants and only a small fraction arrives as silicate and iron cores with limited gaseous envelopes . recently neptune - mass planets have been detected around 55cnc and hd160691 @xcite . since these stars are g - type stars , if the result of figure [ fig : histo ] is applied , the probability of formation of such planets is very low although it is not zero . unlike gj436 , there are three additional jupiter - mass planets around 55cnc and two additional ones around hd160691 . in the system of 55cnc , sweeping mean motion resonance associated with migration of the giant planet @xcite presently at 0.1au could bring rocky embryos / planetesimals to the vicinity of the host star , so that a neptune - mass rocky planet could accrete _ in situ_. on the other hand , in the system hd160691 , sweeping secular resonance associated with disk gas depletion @xcite , could bring rocky embryos / planetesimals to inner regions @xcite . in the present paper , we do not include such interactions . we will present elsewhere details . in the system of gj436 , however , no additional jupiter - mass planet has been found . our model accounts for formation of the isolated neptune - mass planet around an m star . figure [ fig : histo]b shows that in model 0.6 where @xmath236 , the intermediate - mass ( @xmath237 ) and intermediate-@xmath10 ( @xmath5au ) planets are more abundant than in either model 1.0 or model 0.4 . the mass function of close - in planets also appears to be smoother with some @xmath15 in the range of @xmath238 ( figure [ fig : histo]a ) . these intermediate - mass planets formed just outside the ice boundary in disks with modest @xmath39 where the upper limit of @xmath239 while that for typical disks ( with @xmath240 ) is @xmath241 . the critical mass for starting planetary migration at the ice boundary is @xmath242 in model 0.6 . a small fraction of emerged cores may accrete modest amount of gas as it starts to migrate . the formation of the intermediate - mass and intermediate-@xmath10 planets , which tends to smooth the mass distribution , may be one of characteristics of planets around k stars , compared with those around g and f stars , although it is much less pronounced than the characteristics of planets around m stars . for illustration , we present , in figure [ fig : tevol_mstar06 ] , the formation of a typical planet which formed with an intermediate mass and attain an intermediate @xmath10 during its migration . in this case , a seed embryo is formed at @xmath243au in a slightly massive @xmath244 disk with @xmath245myr . through planetesimal coagulation , this embryo attains a mass @xmath246 and become isolated in 1.5 myr . the cessation of the planetesimal bombardment enables the embryo to grow through gas accretion . when its mass reaches @xmath247 at @xmath248 myr , gas accretion is quenched by the severe depletion of gas near its orbit . the newly formed planet undergoes migration while gas is globally depleted . the orbital migration is eventually halted at an intermediate location , 0.4au . the condition for growth to be quenched between @xmath178 requires @xmath249 au . the above example shows that to halt migration at an intermediate location , both time scales of migration and growth due to gas accretion are required to be comparable to the gas depletion time scale . in general , such special circumstances are satisfied with small probability . however , they are more likely around k dwarfs than other type stars because cores with @xmath250 are more abundant at @xmath249au around k dwarfs as a result of the @xmath52 dependence of @xmath43 and @xmath251 . in model 0.6 , gas accretion may be quenched when the cores attain a mass @xmath252 while they migrated to @xmath253 au . in comparison with model 1.0 , the isolation mass @xmath38 at the ice boundary increases with @xmath24 . around a solar - type star , the isolation mass near the ice boundary is sufficiently large for efficient gas accretion to be initiated . eventually runaway gas accretion leads to the emergence of the intermediate - mass deficit in the mass distribution of planets around relatively high - mass stars . figure [ fig : obs ] shows that some fraction of planets discovered around k stars may have the intermediate mass . however , the number of the detected planets may be insufficient for statistical discussion . these stellar mass dependence in the extrapolated planetary characteristics can be tested with future observation . in model 1.5 where @xmath254 , the range of @xmath10 where gas giants are formed is more extended ( the bottom panel in figures [ fig : ma ] ) . this arises primarily because the disks around more massive stars have relatively large @xmath22 . in this case , @xmath255 ( the core mass required for rapid gas accretion ) can be attained before gas depletion even at large @xmath10 . however , larger @xmath43 in this case leads to slow core growth beyond the ice boundary . thus , most gas giants have cores composed of silicates and iron but not icy cores , in contrast with gas giants around lower mass stars . the very large @xmath43 also leads to less efficiency of formation of gas giants in the range of a few au to 10au than that around g stars , although gas giants form in broader range of @xmath10 . as a result , the fraction of f stars harboring giant planets with periods smaller than several years which are currently detectable with doppler survey is rather smaller than that of g stars ( figure [ fig : freq ] ) , although the difference is within a factor 1.5 . the similar fraction within a factor 2 among k , g , and f stars shown in figure [ fig : freq ] is consistent with the observation ( fischer & valenti 2005 ) . the predicted mass distributions for m , k , g , and f stars are not inconsistent with the observed distributions . since the numbers of observed planets are not enough for statistical discussion , in particular for stars other than g stars ( figure [ fig : histo ] ) , we can not discuss the agreement between the predicted and observed distributions in more detail . strong winds and jets from further higher mass stars ( a , b stars ) may decrease @xmath1 , which reduces formation rate of gas giants . it is not clear how much the fraction of massive stars with gas giant planets is reduced . we will address planetary formation around massive stars elsewhere . the results of the standard models clearly indicate that the mass function of planets depends on the delicate balance between growth and migration time scales . there are some uncertainties concerning the migration process . in the standard series , we set @xmath188 ( eq . [ [ eq : m_gas_vis_sp ] ) in accordance with the results of previous numerical simulations @xcite . however , the additional contribution from a torque imbalance between the lindblad resonances of low - mass embedded cores may also lead to type i migration while their masses are relatively small @xcite , although turbulence in the disk @xcite and self induced secondary instability @xcite can also retard the rate of migration . in our simulations , we can partly take into account the effect of type i migration by lowering the value of @xmath96 ( to unity ) . we also carried out simulations with @xmath96 ( @xmath256 ) such that the onset of migration is delayed until the cores have attained relatively large masses . the sensitive dependence of the mass function of close - in planets on the migration condition makes it an ideal observable feature which can be used to calibrate the criteria and efficiency of migration . in a variation of the standard models , we consider two new series : 1 ) series b ( model @xmath185b ) with @xmath257 and 2 ) series c ( model @xmath185c ) with @xmath258 , which have identical @xmath24 as the standard series with @xmath188 . in figures [ fig : mabc]a and b , we show the final mass and semimajor axis distribution of planets in models @xmath185b and @xmath185c . for models @xmath185b , cores undergo migration before they attain sufficient mass to engage in efficient gas accretion . although the cores migration may terminate close to their host stars , the relatively small aspect ratio of the disk for small @xmath10 implies low @xmath121 . at small @xmath10 , gap formation prevents the cores from accreting gas . the accumulation of cores and planetesimals in the proximity of their host star may promote their coagulation @xcite . although we can not rule out the possibility of a highly - efficient migration on the basis of the mass function of the close - in planets , it does pose difficulties to account for the modest frequency of gas giants with periods longer than a few weeks . even around solar - type stars , cores rapidly migrate to the stellar proximity before they have acquired sufficient mass to efficiently accrete gas so that the probability of gas - giant formation is strongly suppressed . for the low - stellar - mass models 0.2b and 0.4b , the isolation mass is only a few times larger than that of the earth . nevertheless , it is larger than @xmath222 for this low-@xmath96 case . most cores migrate toward their host stars with @xmath15 less than a few @xmath8 . although these masses are @xmath259 , gas accretion is too slow ( eq . [ [ eq : tau_kh ] ] ) for them to acquire any significant amount of mass prior to gas depletion . the detection of close - in neptune - mass around m dwarf stars and the modest detection frequency of gas giants around solar - type stars are inconsistent with the results of series b in figure [ fig : mabc]a and therefore , we suggest that @xmath96 is substantially larger than unity . an upper limit on the magnitude of @xmath96 may be inferred from models @xmath185c . with @xmath260 , the condition for the onset of migration becomes much more stringent and most cores do not have sufficient mass to undergo migration . for solar - type stars , a few relatively massive cores can form rapidly in disks with very large @xmath39 . these systems can migrate to form close - in jupiter - mass gas giants . the mass distribution of close - in planets around solar type stars ( model 1.0c ) is skewed to @xmath261 with lower cut - off below @xmath262 , which is inconsistent with the observed mass - period distribution of extra solar planets ( figure [ fig : obs ] ) . for low - mass host stars ( models 0.2c and 0.4c ) , many intermediate - mass planets can form during and after gas depletion . but they retain their initial semi major axis . the mass distribution of the close - in planets peaks near the mass of saturn and hardly any planets have masses comparable to that of neptune . these simulation results are again inconsistent with the observed mass - period distribution of extra solar planets . furthermore , in series c , the deficit of planets with intermediate masses and periods is too pronounced to be consistent with observed one . therefore , we infer @xmath263 . the expressions in equations ( [ eq : m_grow ] ) and ( [ eq : m_iso ] ) indicate that the growth rate and asymptotic mass of cores are increasing function of @xmath22 . in the standard series of models , we set @xmath73 . with this prescription , @xmath22 of disks around low - mass stars is relatively small . consequently , the emergence of gas giants occurs preferentially around massive stars . the dependence of the disks @xmath22 on the @xmath24 of their host stars is poorly known . on the theoretical side , gravitational instability may limit the amount of mass which can be retained by the disks , especially those around low - mass stars . but , smaller @xmath50 s and lower intensity of ionizing photons may also reduce the influence of the magneto - rotational instability @xcite and the angular momentum transfer efficiency so that more mass may be stored in disks around low - mass stars . best available observational data suggest @xmath69 ( see 2.2 ) but the dependence of @xmath22 on @xmath24 is poorly known . in view of these uncertainties , we introduce another two series of models . the parameters of models @xmath185d and @xmath185e are identical to those of models @xmath185 , respectively . the only difference is that we set @xmath192 in series d and it is set to be unity for all @xmath24 in series e. in comparison with the standard models , disks around low - mass stars are less deficient in these new models while those around the solar type stars remain the same . the predicted distributions for series d and e are shown in figure [ fig : made ] . since @xmath222 and @xmath264 do not depend on @xmath22 nor @xmath23 , planets migration and gas accretion undergo the same paths as in standard models for the same mass planets . as a result , the mass distribution of close - in planets is self - similar among models @xmath185 , @xmath185d , and @xmath185e . since massive disks exist around lower mass stars more frequently in series d ( and even more frequently in series e ) , the amplitude of mass distributions around lower mass stars are enhanced in these models . in series d and e , inferred frequency of jupiter mass planets around m stars are comparable to those around g stars ( figure [ fig : made ] ) . the sparse detection of close - in jupiter - mass and neptune - mass planets around m stars suggests that the disk masses are rapidly increasing function of @xmath24 as we have assumed in the standard models with @xmath73 . observational discovery of extrasolar planets is advancing rapidly . we now have sufficient amount of data to carry out statistical characterization of planetary properties and to place constraints not only on the dominant mode of planet formation but also the range of physical quantities which determine their growth and migration rates . in this paper , we focused our discussion on the mass function of close - in planets around stars with various masses , in particular lower masses than the solar mass , because its origin is determined by the delicate balance of various processes and they are the most conspicuous companions of nearby stars . although many ( @xmath265 ) planets which once migrated to the proximity of their host stars may be eliminated ( paper ii ) , we can compare the mass distribution of close - in planets among around stars with various masses . if the elimination factor is taken into account , rough comparison is also possible between close - in and more distant planets . 1 . dynamically - isolated , close - in , neptune - mass planets with silicate and iron cores can form in relatively massive disks around solar - type stars . but their frequency is expected to be an order of magnitude smaller than that of close - in jupiter - mass planets . dynamically - isolated , close - in , neptune - mass ice giants can form in lower - mass stars . their frequency peaks around the m dwarfs . since the luminosity of m dwarfs is weak , the ice boundary is located well inside 1 au . these planets are formed at around 1au but outside the ice boundary and their cores are primarily composed of volatile ices . planets which migrated to the stellar proximity with masses in the range of @xmath266 may acquire , _ in situ _ , a limited amount of additional gas , but gas accretion is immediately quenched by gap formation because of small aspect ratio of the disk in the proximity of the host star . because these planets compose mostly of icy material and m stars luminosity is relatively weak , they may have water - vapor atmosphere and water ocean @xcite . 3 . embryos with mass lower than @xmath267 can not migrate to the proximity of their f , g , and k dwarf host stars through type ii migration . detection of dynamically - isolated earth - mass close - in planets may be attributed to type i migration of low - mass embryos @xcite , sweeping secular resonances @xcite , or sweeping mean motion resonances . around late m dwarfs , however , dynamically - isolated , a few earth - mass planets can form with temperature comparable to that of the earth . 4 . around m dwarfs , the formation probability of gas giants is much reduced . the relatively low @xmath22 prevents the emergence of sufficiently massive cores prior to the severe depletion of gas in their nascent disks , which is consistent with the results by @xcite . the mass function of close - in planets generally have two peaks at about neptune mass and at about jupiter mass . the lower - mass peak takes the maximum frequency for m stars , while the higher - mass peak is much more pronounced around higher - mass stars ( f , g , k dwarfs ) . these are because planets tend to undergo type ii migration after fully accreting gas around the higher - mass stars while they tend to migrate faster than gas accretion around m stars . unless the termination location of planetary migration is a decreasing function of @xmath24 , close - in neptune - mass planets around m dwarfs are easier to detect than those around g dwarfs . 6 . the mass function of dynamically - isolated close - in planets around stars with various masses can also be used to calibrate the sufficient condition for the onset of planetary migration and for the termination of gas accretion due to planet - disk tidal interaction . the metallicity dependence on frequency of extrasolar gas giant planets may not be easily accounted for by the gravitational instability scenario , e.g. , @xcite , while it is naturally accounted for by the sequential core accretion scenario that we are based on ( paper ii ) . the condition of the gravitational instability is @xmath268 , and its radial wavelength @xmath269 @xcite where @xmath270 . hence , the instability may be more limited and result in smaller clumps around lower - mass stars , which could account for the above features 4 and 5 , if @xmath271 is assumed as in the standard series in the present paper . however , if @xmath272 , the gravitational instability scenario can not account for the above features 4 and 5 , while the sequential core accretion scenario still shows the tendency for the features . more detailed study on the dependence of disk mass on stellar mass @xmath24 is needed . our work is primarily motivated by the discovery of gj436b @xcite . our theoretical extrapolations can be tested with the following statistical properties of close - in planets to be discovered by various techniques . 1 . as mentioned above , the mass function of dynamically - isolated close - in planets as a function of the spectral classes of their host stars is particularly useful in the determination of the growth , migration , and disk depletion time scales . 2 . a comparison between the frequencies of gas giants with close - in orbits and those with extended orbits can provide constraints on the migration condition , survival criteria , and disk mass as functions of the host stars mass . 3 . in systems with multiple giant planets , the dynamical architecture may provide clues on whether the migration of the close - in planets are driven by planet - disk tidal interaction or sweeping secular resonance . a comparison of atmospheric properties of close - in neptune - mass planets around g dwarfs to that around m dwarfs can verify the conjecture that the former have silicate and iron cores whereas the latter have ice cores . 1 . a spatially resolved image of disks can directly provide information of @xmath22 and the temperature distribution around any given host star . the dependence of @xmath22 on the mass of the host stars determines the functional form of @xmath47 . 3 . a relation between the disk mass and accretion rate onto the host stars places a constraint on the rate of type ii migration . 4 . a direct measurement of the gas distribution is particularly important in determining @xmath1 and @xmath48 . 1 . the possibility of radiative feedback on the termination of gas accretion . 2 . the stoppage of type ii migration and the survival of short - period planets . 3 . the rate ( and direction ) of type i migration of cores . the enhanced probability of multiple planet formation . effect of dynamical interaction between multiple planets during and after gas depletion the radial distributions of @xmath22 and @xmath23 . effects of different power - law index of @xmath10 dependence from @xmath273 that we used . more realistic time evolution of @xmath23 . 7 . the influence of a stellar companion on the emergence and survival of planets . we thank g. laughlin , m. nagasawa , g. ogilvie , and s. vogt for useful discussions , and the anoymous referee for useful comments . this work is supported by the nasa through nags5 - 11779 under its origins program , jpl 1228184 under its sim program , and nsf through ast-9987417 and by jsps .

we undertake a detailed study of the performance of maximum likelihood ( ml ) estimators of the density matrix of finite - dimensional quantum systems , in order to interrogate generic properties of frequentist quantum state estimation . existing literature on frequentist quantum estimation has not rigorously examined the finite sample performance of the estimators and associated methods of hypothesis testing . while ml is usually preferred on the basis of its asymptotic properties - it achieves the cramer - rao ( cr ) lower bound - the finite sample properties are often less than optimal . we compare the asymptotic and finite - sample properties of the ml estimators and test statistics for two different choices of measurement bases : the average case optimal or mutually unbiased bases ( mub ) and a representative set of suboptimal bases , for spin-1/2 and spin-1 systems . we show that , in both cases , the asymptotic standard errors of the ml estimators grossly underestimate the estimation error in finite samples , rendering inference based on the asymptotic properties of the ml unreliable and misleading for experimentally realistic sample sizes . the results indicate that in order to fully exploit the information geometry of quantum states and achieve smaller reconstruction errors , the use of bayesian state reconstruction methods - which , unlike frequentist methods , do not rely on asymptotic properties - is necessary , since the estimation error is typically lower due to the incorporation of prior knowledge .

0904.1628

perhaps the most fundamental problem in quantum statistical inference ( qsi ) is the reconstruction of the density matrix of a quantum system on the basis of a finite number of quantum observations . due to the rapidly growing interest in quantum computation and quantum control , the ability to retrieve the maximum amount of information about a quantum state based on the smallest number of measurements is a subject of paramount importance . the accuracy of all derivative forms of qsi , including process estimation , is ultimately determined by that of the underlying state estimation . methods for quantum state estimation can be formally subdivided into three categories . the first , tomographic inversion @xcite , is the least computationally expensive and most popular technique . however , tomographic inversion can not enforce the constraints on the density matrix during estimation , and hence the estimates produced are often not physically meaningful . the second class consists of frequentist techniques of inference based on a likelihood function , the most notable of which is maximum likelihood ( ml ) estimation @xcite . this class of methods avoids the problems associated with tomography and is asymptotically more efficient than the latter ( as explained in section iii e ) . however , it delivers distributional results for the estimators of the parameters of interest under the assumption of an infinite number of measurements . hence , for finite sample sizes , the estimated confidence intervals for the parameters may have actual coverages quite different from the corresponding asymptotic ones and , as is well known in the statistics literature , this divergence typically tends to become more pronounced as the number of parameters and nonlinearity of the model increases . all forms of frequentist inference , including tomographic inversion and ml estimation , require a complete observation level , i.e. @xmath0 linearly independent observable operators where @xmath1 is the hilbert space dimension , in order to estimate all the parameters . the third type , bayesian estimation @xcite , which is based on updating a prior plausibility distribution of the parameters based on observed data , lends itself readily to both incomplete observation levels and a finite number of measurements . moreover , the estimation error that arises in bayesian methods is typically lower , reflected in shorter lengths of bayesian confidence intervals compared to their frequentist counterparts , due to the use of such priors . in this paper , we investigate the finite sample properties of ml estimators of the density matrix of finite - dimensional , spin-1/2 ( one qubit ) and spin-1 quantum systems . among frequentist estimation techniques , ml is usually preferred on the basis of its asymptotic properties - ( i ) the ml estimator is asymptotically efficient in the sense that its asymptotic variance achieves the cramer - rao lower bound for consistent estimators , ( ii ) likelihood based testing approaches are optimal , in the sense of the neyman - pearson fundamental lemma and the large deviation principle , for a broad class of hypothesis testing problems . however , the finite sample properties of ml estimators and test statistics are often less than optimal . existing literature on quantum ml has not rigorously examined the finite sample performance of the estimators and associated methods of hypothesis testing . to our knowledge , there is only one extant study on the efficiency of frequentist quantum state estimation @xcite , and robust numerical techniques are lacking . minimizing finite sample estimation errors is essential for making optimal quantum decisions , which underlie emerging quantum feedback control and computation strategies @xcite . lack of rigorous understanding of the small sample estimation errors has inhibited the application of ml to practical problems in quantum information and control . we shed light on the following issues : * how are the small sample biases of the ml estimators affected by the sample size ? * how do the finite - sample standard errors , and hence the associated @xmath2 confidence intervals , compare to the corresponding asymptotic ones and how does the coverage of the intervals change with increase in the sample size ? * how does the small sample behavior of the test statistics for physical quantities of interest in quantum decision theory compare to their known asymptotic behavior for different sample sizes ? we show that the finite sample properties of ml differ significantly from the corresponding asymptotic ones for experimentally realistic sample sizes . in particular , the finite - sample standard errors are orders of magnitude bigger than the corresponding asymptotic ones . this feature holds for sample sizes approaching the limit of experimental feasibility . consequently , the asymptotic confidence intervals grossly undercover in finite samples and the test statistics exhibit severe size distortions . in addition , the assessment of the estimation error in qsi is complicated by the existence of multiple measurement strategies due to the noncommutativity of the probability space , and ambiguities regarding the optimal measurement strategy . we compare the relative efficiencies of average - case optimal ( mub ) and representative suboptimal measurement strategies . we show that the asymptotically predicted advantages of optimal measurements are diminished in finite samples , to the extent that measurement strategies that are experimentally simpler to implement may perform almost as well as asymptotically optimal measurements . the paper is organized as follows . section ii discusses the asymptotic properties of the ml and compares it to the alternative estimation approaches . section iii details the properties of ml estimators of the quantum density matrix . section iv describes the bloch vector parameterization , mutually unbiased measurement bases ( mub ) , and representative suboptimal measurement strategies considered in this paper . section v discusses the details of the globally convergent newton - raphson and quasi - newton algorithms used for constrained parameter optimization , along with methods for kernel density estimation of finite sample distributions . in section vi , we present the estimation results for general mixed spin-1/2 and spin-1 density matrices , comparing the finite sample versus asymptotic properties of estimators and test statistics for physical quantities of interest in quantum decision theory . finally , in the concluding section vii , we draw conclusions regarding the efficiency of frequentist quantum state estimation and discuss bayesian extensions . let @xmath3 be an @xmath4 sample of size @xmath5 from a population with probability density function @xmath6 , which depends on the unknown parameter vector @xmath7 whose true value is @xmath8 . the value of the parameter vector that maximizes the likelihood function - the joint density of the sample defined as a function of the unknown parameter vector @xmath7 - is called the ml estimator of @xmath7 : @xmath9where @xmath10 denotes the admissible parameter space . typically , the logarithm of the likelihood function , @xmath11 , is easier to maximize numerically because of its separability . by maximizing the log likelihood , the ml estimator minimizes the kullback - leibler distance between the estimated and true probability distributions . ml has several properties that make it an attractive frequentist estimation procedure : 1 . _ consistency : _ an estimator @xmath12 is _ consistent _ for the parameter @xmath7 ( written as plim @xmath13 ) if for every @xmath14 , @xmath15 . invariance : _ the ml estimator of @xmath16 is @xmath17 , for a continuous and continuously differentiable function @xmath18 . asymptotic normality . _ for a sequence of estimators @xmath12 , if @xmath19 as @xmath20 , where @xmath21 denotes convergence in distribution and @xmath22 is any function of @xmath5 , @xmath12 is said to be @xmath23-consistent for @xmath7 and has an asymptotic normal distribution with asymptotic covariance matrix @xmath24 . + the ml estimator is asymptotically normally distributed:@xmath25 & \rightarrow & \mathcal{n}[0,i^{-1}(\theta_0 ) ] \text { , } \\ \text{where } i(\theta_0 ) & = & -\mathrm{e}\left [ \frac{\partial ^{2}\ln l(\theta_0 |x)}{\partial \theta \partial \theta ^{\prime } } \right ] \text{.}\end{aligned}\ ] ] + @xmath26 is called the expected fisher information matrix . note that the asymptotic covariance matrix of the ml estimator is a function of the unknown parameters . two approaches exist for consistent estimation of the expected fisher information matrix thereby providing feasible versions of the observed fisher information matrix . the first estimator replaces the expected second derivatives matrix of the log likelihood function with its sample mean evaluated at the maximum likelihood estimates,@xmath27 \label{observed fisher}\ ] ] + the second estimator is based on the result that the expected second derivatives matrix is the covariance matrix of the first derivatives vector,@xmath28 \label{observed fisher 2}\ ] ] 4 . _ asymptotically efficient . _ a sequence of consistent estimators @xmath12 is asymptotically efficient if @xmath29\overset{d}{\rightarrow } \mathcal{n}[0,i^{-1}(\theta_0 ) ] $ ] where @xmath30 $ ] ; @xmath31^{-1}$ ] is called the cramer - rao lower bound ( crb ) for consistent estimators . property 4 is the subject of the following classic lemma of frequentist inference . [ crb ] the eigenvalues of the covariance matrix of parameter estimates of an asymptotically unbiased frequentist estimator are bounded from below by the eigenvalues of @xmath32 \right\}^{-1}$ ] . the maximum likelihood estimator asymptotically ( i.e. , in the limit of an infinite number of measurements ) achieves this lower bound . in addition , likelihood based testing approaches are optimal , in the sense of the neyman - pearson fundamental lemma and the large deviation principle , for a broad class of hypothesis testing problems . a hypothesis test @xmath33 , based on a test statistic @xmath34 - a function of the data - is a rule that specifies for which values of @xmath35 ( the acceptance region @xmath36 ) the null hypothesis @xmath37 is accepted , and for which values ( the rejection region @xmath38 ) it is rejected ( and the alternative hypothesis @xmath39 , where @xmath40 is the complement of @xmath41 , is accepted ) . the _ size _ @xmath42 of a hypothesis test @xmath33 is the probability of rejecting the null hypothesis given that it is true . the _ power _ @xmath43 of a hypothesis test @xmath33 is given by the probability of rejecting the null hypothesis given that it is false . typically , when defining the power of a test , one assigns the test to a _ class _ based on its size . for @xmath44 a test with power function @xmath45 is a size @xmath46 test if @xmath47 . given a class of hypothesis tests for testing @xmath37 versus @xmath39 , where @xmath48 , the admissible parameter space , a test in that class with power function @xmath45 is _ uniformly most powerful ( ump ) _ if @xmath49 for every @xmath50 in that class . an important type of hypothesis test based on ml is the likelihood ratio test . likelihood ratio test statistics take the form @xmath51where @xmath52 is the constrained ml estimator . it can be shown @xcite that likelihood ratio tests are ump _ _ _ _ in their respective classes . in our simulation analysis , we rely on an alternative testing procedure , namely the wald test , which inherits the optimality properties of the likelihood ratio test on account of their asymptotic equivalence @xcite . this choice is made primarily on the basis of the ease of computation . the likelihood ratio test requires calculation of both restricted and unrestricted estimators . the wald test , on the other hand , requires only the unrestricted estimator . since some of our hypothesis tests involve nonlinear constraints and estimation of the constrained model is cumbersome , we rely on the wald testing procedure . because of properties 1 - 4 and the fact that likelihood ratio tests are ump , the maximum likelihood estimation methodology is considered the most desirable among frequentist estimation techniques . an alternative method of frequentist inference , which may be used to estimate the parameters of the quantum density matrix , is the method of moments approach . suppose that although the probability density function @xmath53 is unknown , @xmath54 moments of the density function have analytical representations in terms of the parameter vector @xmath7 . let these @xmath54 population moments be denoted by @xmath55 = \mu ( \theta ) $ ] , where @xmath56 and @xmath57 . the _ method of moments _ ( mm ) estimator of @xmath7 , denoted @xmath58 , is the value of the parameter vector that equates the population moments with the corresponding sample moments : @xmath59 note that the method of moments approach involves solving a system of @xmath54 ( possibly nonlinear ) equations in @xmath54 unknown parameters . like the ml estimator , the mm estimator is consistent and asymptotically normally distributed . however , while the ml estimator exploits all the information contained in the likelihood of the data , the mm estimator only uses the information in a chosen set of moments of the data . hence , unlike the ml estimator , the mm estimator is not asymptotically efficient , i.e. its asymptotic variance does not attain the cramer - rao lower bound . the asymptotic distribution of the mm estimaor is:@xmath60 \rightarrow \mathcal{n}[0,v]\text{,}\ ] ] where @xmath61 \text { , \ } \\ \omega & = & \mathrm{e}\left [ \left ( f(x_{i})-\mu ( \theta _ { 0})\right ) \left ( f(x_{i})-\mu ( \theta _ { 0})\right ) ^{t}\right ] \text{.}\end{aligned}\ ] ] in the alternative paradigm of bayesian estimation , the estimation error that arises is typically lower than that in frequentist estimation . bayesian estimation differs fundamentally from frequentist methods in that the parameters @xmath62 are treated as random variables . the goal is not to estimate a unique probability distribution , which can only truly be known in the limit of an infinite number of measurements , but to update a so - called prior plausibility distribution to a posterior plausibility distribution based ( only ) on the observed data . the posterior plausibility distribution is given by @xmath63 where @xmath64 denotes the prior plausibility distribution , i.e. , the probability of the parameter vector taking on the value @xmath65 given our prior information @xmath66 regarding the parameter space , @xmath67 denotes the joint probability density , and @xmath10 denotes the space of admissible parameters @xmath65 . conditional simulation is required to retrieve quantities of interest , including the parameter estimates , which are given formally by the posterior means @xmath68 unlike frequentist estimators , the notion of a confidence or credible interval can be rigorously defined for finite samples only for bayesian estimators . this allows one to rigorously report finite sample uncertainties . the @xmath69 bayesian credible interval for the parameter @xmath70 is the interval @xmath71 $ ] such that @xmath72 in this section we apply and extend the classical estimation framework in section ii to estimation of the quantum density matrix . quantum statistical inference is based on the notion of a quantum probability space . consider a measurable space @xmath73 , where @xmath74 is the set of all possible measurement outcomes and @xmath75 is the @xmath76-algebra of subsets of @xmath74 . an _ operator - valued probability measure ( povm ) _ is a ( set ) function @xmath77 , where @xmath78 is the set of bounded positive semidefinite , hermitian linear operators on a hilbert space @xmath79 . a _ quantum probability space _ is a measurable space @xmath80 , together with an operator - valued probability measure @xmath81 , such that the outcome @xmath82 has probability density function @xmath83 , where @xmath84 and @xmath85 is a positive - semidefinite , unit trace , hermitian matrix ( parametrized by a vector @xmath65 of parameters ) called the density matrix . note that @xmath35 in this context denotes the outcome of a single measurement . the measure @xmath81 is explicitly defined in terms of the operators @xmath86 , and an associated scalar - valued probability measure @xmath87 satisfying @xmath88 , through the relation @xmath89 . for @xmath1-dimensional ( finite ) quantum systems , we write @xmath90 , and denote the outcome of the @xmath91-th measurement by @xmath92 . @xmath93 is then a @xmath94 positive - semidefinite hermitian matrix . the outcomes @xmath35 are indexed by the set of integers @xmath95 . in this case , we have @xmath96 . for simplicity of exposition , we collect all the distinct parameters of the density matrix , @xmath97 , into the @xmath98-dimensional vector , @xmath65 . the most convenient parameterization of @xmath99 differs based on the state estimation method ; various parameterizations are discussed in section iv . the likelihood function for quantum state estimation is then @xmath100 which may be interpreted as the probability of obtaining the set of observed outcomes for a given density matrix @xmath85 . the ml estimator of the density matrix seeks to identify the admissible parameter vector @xmath65 at which this likelihood is maximal . a _ resolution of the identity _ on a hilbert space @xmath79 of quantum states is a normalized operator - valued measure . generally , the resolution of the identity satisfies @xmath101 for finite - dimensional systems , to which we restrict our attention , @xmath102 where @xmath103 denotes the @xmath94 identity matrix . an important feature of quantum probability is that the operators @xmath93 do not all mutually commute . the subsets @xmath104 of the space of possible measurement outcomes @xmath74 may be chosen to be pairwise disjoint and associated with subsets @xmath105 of commuting observables whose members do not commute with those of any other subset . the @xmath106 are then said to constitute distinct measurement bases " . writing each @xmath93 as an @xmath107 hermitian matrix , it is convenient to represent each basis @xmath108 in terms of an @xmath107 matrix of common eigenvectors @xmath109 . given that the density matrix is a function of @xmath110 independent parameters , the minimal cardinality resolution of the identity must be composed of @xmath111 subsets @xmath112 . we note , however , that many resolutions of the identity are redundant in that they are associated with @xmath113 bases @xmath106 . in the current work , the data @xmath35 consist of @xmath114 measurement outcomes in each of @xmath115 measurement bases with @xmath116 . the measurement bases used are discussed further in section [ bases ] . among frequentist estimation techniques , ml has been employed most extensively for reconstruction of quantum states . in quantum ml estimation , we aim to identify the maximum of the likelihood function ( [ likelihood ] ) over the set of _ admissible _ density matrices . all parametrizations of the density matrix require the imposition of constraints on parameter vector @xmath65 ( see section iv ) ; these constraints are necessary for expression ( [ likelihood ] ) to be a well - defined likelihood . assuming the constraints on the parameter vector @xmath65 are of the general form @xmath117 , the problem can be formulated in terms of the lagrangian function @xmath118 + \sum_{j=1}^{n}\lambda_j\left(a_j(\theta ) - \gamma_j^2\right),\ ] ] where the first term is @xmath119 in the absence of constraints ( i.e. , @xmath85 need not be an admissible density matrix and @xmath120 need not be a well - defined likelihood ) , the @xmath121 denote slack variables ( @xmath122 in the case of an equality constraint ) and the @xmath123 denote lagrange multipliers . it is convenient to order the @xmath1 constraints such that the first constraint enforces the unit trace of @xmath124 , and the following @xmath125 constraints enforce its positive semidefiniteness . note that @xmath126 is a well - defined likelihood function only in the presence of these constraints . for parameterizations where positive semidefiniteness is implicit in the parametrization ( such as the cholesky parametrization @xcite ) , @xmath127 , and for parameterizations where the unit trace constraint is implicit in the parameterization ( such as the bloch vector parametrization , sect . iv ) , @xmath128 . we denote the vector of parameters @xmath129 . finding the optimum corresponding to this lagrangian entails searching for parameter vector @xmath130 that renders the gradient vectors @xmath131 and a linear combination of @xmath132 , @xmath133 parallel . there are two common approaches to solving this problem : 1 ) minimization of the sum of squares " ( of the first - order conditions ) function @xmath134 ; 2 ) finding the roots of the system of nonlinear equations @xmath135 using the newton - raphson ( nr ) method . in fact , methods 1 ) and 2 ) may be combined to produce a globally convergent nr algorithm . further details on solving the constrained optimization problem are provided in section [ numerical ] . note that the ml estimator obtained by maximizing the likelihood defined by the lagrangian ( [ lagrangian ] ) is consistent , asymptotically normally distributed , and has an asymptotic covariance matrix equal to the inverse of @xmath5 times the expected fisher information matrix , @xmath136 ( see section ii for details ) . we estimate the expected fisher information using equation ( [ observed fisher ] ) in section ii . in quantum statistics , there are multiple cramer - rao type inequalities , each with its own associated ( quantum ) fisher information . some of these correspond to particular measurement strategies , whereas others are in fact unachievable . work in quantum probability theory @xcite has indicated that @xmath137 for an arbitrary choice of measurement bases is generally not the tightest asymptotic lower bound achievable in quantum ml estimation . however , the measurements that maximize the fisher information depend on the true , unknown state of the quantum system , rendering the practical utility of the notion of the tightest possible cramer - rao bound questionable . although the choice of measurement bases that can achieve the tightest possible cramer - rao bound depends on the true @xmath124 , there exists an approach to optimal measurement that is agnostic to the true value of @xmath124 . wootters @xcite proposed a construction of measurement bases that maximizes the _ average _ information ( over the set of all possible density matrices ) obtained via a set of @xmath5 measurements . these so - called mutually unbiased measurement bases ( mub ) are maximally noncommutative " in the sense that a measurement in one basis provides no information as to the outcome of a measurement over a basis unbiased with respect to the current one . let @xmath138 denote the fisher information given true state @xmath139 . mub aims to maximize the average fisher information over all possible @xmath140 s : @xmath141 where @xmath10 again denotes the admissible parameter space and @xmath142 is the volume of @xmath10 , by an appropriate choice of measurement bases ( as discussed in section parameterization , @xmath143 , the @xmath110 dimensional bloch vector space of admissible @xmath94 dimensional density matrices ) . it can be shown that this is equivalent to maximizing the average _ kullback - leibler ( kl ) information gain _ @xmath144 upon updating the flat prior distribution to the asymptotic multivariate normal distribution . the kl information gain is given by @xmath145 where in the present case @xmath146 is the asymptotic multivariate normal distribution over the bloch vector space after measurements and estimation , and @xmath147 is the uniform distribution on the bloch vector space . the kl gain is defined in terms of the shannon information of a distribution , @xmath148 . a measurement in each basis restricts the variance in @xmath125 directions , leaving it infinitely broad in the other @xmath149 . for estimation of a single parameter , the shannon entropy of the asymptotic normal distribution of parameter estimates is @xmath150 . maximizing the information gain is equivalent to minimizing the uncertainty volume " in the parameter space ; in the absence of measurements , this is equal to the volume of the bloch vector space . the uncertainty distance for estimation of a single parameter is the standard deviation of the estimator ; the uncertainty volume is the product of the standard deviations of the estimators for each of the parameters . in terms of uncertainty volumes , the information gained by measurements is @xmath151 where @xmath152 is the uncertainty volume after measurements and estimation and @xmath153 is volume of the bloch vector space . denote by @xmath154 the @xmath155-dimensional subspace of @xmath156 or @xmath157 associated with measurement basis @xmath158 . the total uncertainty volume is diminished by overlaps between the subspaces @xmath154 . it can be shown wootters1989 that this total volume @xmath152 may be written @xmath159 where @xmath154 is the @xmath155-dimensional subspace of @xmath156 associated with measurement basis @xmath158 and @xmath160 denotes the @xmath98-dimensional parallelipiped whose edges are the @xmath111 sets of eigenvectors associated with each of the subspaces @xmath154 . thus the kullback - leibler information gain ( [ kullback - leibler ] ) in updating the flat prior distribution to the asymptotically normal distribution is then @xmath161+\ln(w_0)-\left(\frac{n^2 - 1}{2}\right)\ln(\pi e).$ ] @xmath162 , the log of the average uncertainty volume in subspace @xmath154 , does not depend on the choice of measurements since it is the log of the product of standard deviations of multinomial parameters in a single measurement basis , averaged over all possible multinomial parameters @xmath163 . thus the average kullback - leibler or fisher information is a function of only @xmath164 , which in turn is determined by the relative orientations of the bases ( and not the parameters ) . the total uncertainty volume is minimized when @xmath160 is a rectangular solid with all the unit vectors defining the edges being orthogonal ; this is equivalent to the condition that the subspaces @xmath165 are mutually orthogonal . wootters showed wootters1989 that this condition is equivalent to requiring that @xmath166 are column vectors in the bases @xmath167 respectively , and @xmath168 denotes the modulus of the hermitian inner product . whereas mutual nonorthogonality of the edges of the parallelipiped may decrease the asymptotic uncertainty volume in particular subspaces @xmath154 , the total asymptotic uncertainty volume is always increased by such nonorthogonality . explicit formulas for measurement bases that satisfy ( [ mub condition ] ) are known in the cases where the hilbert space dimension @xmath169 is the power of a prime , and are discussed in section ivb . an unresolved question in the literature is the magnitude of the information loss for finite sample sizes incurred due to not using mub or other approaches to optimal quantum measurement . in many experimental setups , it is not convenient to use these specialized bases . we aim to clarify the practical utility of optimal quantum measurements in ml estimation , and assess the extent to which frequentist quantum estimation can effectively make use of the associated optimal efficiencies . an alternative frequentist method of quantum state estimation that is not based on a likelihood function is tomographic inversion @xcite . this involves an application of the method of moments approach described in section ii . adopting the notation @xmath170 for the @xmath171 observation returning outcome @xmath172 , let the mm estimator functions @xmath173 be given by @xmath174 . we then have @xmath175 = p_{i}={{\rm tr}}(\rho ( \theta ) f(x_{i_k}))$ ] , where @xmath176 denotes the probability of observing outcome @xmath177 . the tomographic inversion method estimates the parameters by equating these population moments to the corresponding sample moments : the parameter estimates @xmath178 are obtained by inverting a system of equations of the form @xmath179where @xmath180 denotes the frequency with which outcome @xmath181 is observed in the sample . introducing the notation @xmath182 , we may solve for the estimated parameter vector as @xmath183 for any parameterization @xmath184 that is linear in @xmath7 ( see section [ parameterization ] ) . however , this method has two major drawbacks . first , since no parametrization @xmath99 guarantees satisfaction of each of the positive - semidefiniteness , unit trace , and hermiticity constraints on @xmath97 ( see section [ parameterization ] ) , direct inversion can yield unphysical density matrix estimates . second , while the ml estimator exploits all the information contained in the likelihood of the data , the mm estimator only uses the information in a chosen set of moments of the data . hence , unlike the ml estimator , the mm estimator is not asymptotically efficient , i.e. its asymptotic variance does not attain the cramer - rao lower bound . for these reasons , we do not consider tomographic inversion in our assessment of the performance of frequentist quantum estimation . in the alternative , bayesian approach to quantum state estimation , the posterior distribution @xmath185 ( [ genpost ] ) takes the form : @xmath186p(\theta)\mathrm{d } \theta}{\int_{\theta } \left[\prod_k\mathrm{tr}(f_{i_k}\rho(\theta))\right]p(\theta)~\mathrm{d } \theta}.\end{aligned}\ ] ] the density matrix can be estimated by the posterior mean of each of its elements ( corresponding to a quadratic loss function " ) , namely @xmath187 p(\theta ) ~\mathrm{d}\theta } { \int_{\theta } \left [ \prod_{k}\mathrm{tr}(f_{i_{k}}\rho ) \right ] p(\theta ) ~\mathrm{d}\theta}\end{aligned}\ ] ] alternative loss functions can be used to retrieve other estimable quantities of interest . bayesian credible intervals can be obtained according to expression ( bayescred ) by sampling from the posterior density ( [ posterior ] ) . these credible intervals do not rely on asymptotic results / fisher information . we do not compute bayesian integrals in this work ; our goal is rather to determine whether such a need exists given the finite sample performance of the computationally simpler frequentist estimators . in maximum likelihood estimation of the quantum density matrix , a constrained optimization must be carried out , where the constraints correspond to preservation of the unit trace and/or positive semidefiniteness properties of the density matrix . the dimension of the parameter space increases quadratically with the hilbert space dimension , necessitating the use of efficient parameterizations of the density matrix . the three most commonly used parameterizations are the bloch vector @xcite , euler angle @xcite and the cholesky @xcite parameterizations . within the last few years , considerable advancements have been made in extending the bloch and euler parameterizations to arbitrary @xmath1-dimensional hilbert spaces . the euler angle parameterization of the density matrix , which is based on the euler angle parameterization of the special unitary group @xmath188 , employs the generators in the lie algebra @xmath156 to parameterize @xmath124 in terms of the lie algebra exponential map . the constraint equations are linear in the parameters , but because the parameters appear as exponents , the likelihood takes on a complicated form . in the cholesky parameterization , @xmath189 with @xmath36 upper triangular has real elements on the diagonal . @xmath124 is then is automatically positive - semidefinite , but is associated with a nonlinear expression for the likelihood , and is not standard in other applications . here , we employ the so - called bloch vector parameterization , where the probability of an observable outcome according to the born rule , @xmath190 , is a simple linear function of the parameter vector @xmath65 . moreover , the bloch vector parameterization is perhaps the most commonly used in the statistical physics of finite - dimensional quantum systems ( especially in quantum information applications ) . most importantly , asymptotic standard errors in the bloch vector parameterization can be computed using the standard methods described in section ii because the positive - semidefiniteness constraints are inequality restrictions that are nonbinding at the optimum . in the bloch vector parameterization @xcite , the hermitian operator @xmath124 is parameterized in terms of an orthogonal basis @xmath191 for the vector space of traceless hermitian operators on an @xmath1-dimensional hilbert space . in two dimensions , these are the familiar pauli spin matrices , whereas in three dimensions they are the so - called gell - mann matrices . @xmath124 can then be written @xmath192 where the @xmath110 matrices @xmath123 satisfy the conditions a ) @xmath193 , b ) @xmath194 , c ) @xmath195 . these are the defining conditions of the generators of the lie group @xmath188 that generalize the pauli spin matrices . the @xmath196 are given by @xmath197 ( i.e. , are expectation values of the observable generators ) . the vector @xmath198 is called the bloch vector . @xmath157 is a compact convex subset of @xmath199 . let @xmath200 denote the coefficients of the characteristic polynomial of @xmath124 , @xmath201 , where @xmath124 takes the form above . the unit trace constraint is automatically satisfied in the bloch vector parameterization . it can be shown that the conditions of hermiticity and positive - semidefiniteness of @xmath124 correspond to the following definition of the bloch vector set " @xmath157 of admissible values of @xmath65 @xmath202 this follows from the standard result that the roots of a characteristic polynomial are positive semidefinite iff the coefficients of the polynomial are positive semidefinite @xcite . the @xmath203 in the above definition of @xmath204 are themselves polynomials in @xmath65 whose coefficients can be expressed in terms of the structure constants of the lie algebra @xmath156 of traceless hermitian matrices , and will be written explicitly below for @xmath205 . the structure constants , which characterize the generators of @xmath156 , are the elements of the completely antisymmetric and completely symmetric tensors @xmath206 and @xmath207 , respectively defined by the relations : @xmath208= 2\imath f_{ijk } \lambda_k\ ] ] @xmath209_+ = \frac{4}{n}\delta_{ij } i_n + 2g_{ijk } \lambda_k,\ ] ] which can be solved @xcite for @xmath210 , @xmath211 : @xmath212\lambda_k\ } \\ g_{ijk}&=&\frac{1}{4 } \mathrm{tr}\{([\lambda_i,\lambda_j]_+-\frac{4}{n}\delta_{ij})\lambda_k\}.\end{aligned}\ ] ] where @xmath213 $ ] denotes the antisymmetric commutator and @xmath213_+$ ] denotes the symmetric commutator , and where we have used the einstein ( implicit ) summation convention for repeated indices . it can be shown @xcite that the @xmath214 that satisfy these conditions can be expressed : @xmath215 where @xmath216 when @xmath217 , the conditions @xmath218 ( equation ( [ bloch space ] ) ) correspond to @xmath219 and @xmath220 the latter condition defines the familiar bloch sphere for spin-@xmath221 systems . the lagrangian ( [ lagrangian ] ) in this case becomes @xmath222+\lambda_2\left(\frac{n-1}{n}-\frac{1}{2}|\theta|^2-\gamma_2 ^ 2\right),\ ] ] where we have omitted the constant term originating from @xmath223 since it is independent of @xmath65 . for @xmath224 , in addition to the constraint ( [ 2d bounds ] ) , we have @xmath225 where the structure constants @xmath211 are components of the completely symmetric tensor of the lie algebra @xmath156 ( eq . [ structure constants ] ) . the lagrangian ( [ lagrangian ] ) for spin-1 systems then becomes @xmath226 + \\ \lambda_2\left(\frac{n-1}{n}-\frac{1}{2}|\theta|^2-\gamma_2 ^ 2\right)+\\ \lambda_3\left[\frac{(n-1)(n-2)}{n^2}-\frac{3(n-2)}{2n}|\theta|^2+\frac{1}{2}g_{ijk}\theta_i\theta_j\theta_k-\gamma_3 ^ 2\right],\end{gathered}\ ] ] where we have ( again ) used the einstein summation convention for repeated indices , i.e. , @xmath227 . note that while for @xmath217 , the bloch vector space is exactly a ball , the additional constraints starting with @xmath228 restrict the bloch vector space for @xmath224 and higher dimensions to a proper subset of a ball . since the structure constants of @xmath156 for @xmath229 have no rotational invariance , neither do these conditions . the bloch vector space has an asymmetric structure in @xmath199 for @xmath229 . in the bloch vector parameterization , the fisher information takes on a particularly simple analytical form . for @xmath217 , we have for the score vectors @xmath230 \\ \frac{\partial \ln l(\theta|x)}{\partial \theta_2}=\frac{1}{2}\sum_{k=1}^m \frac{k}{\mathrm{tr}(\rho f_{i_k})}\left[f_{i_k}(1,2)+f_{i_k}(2,1)\right ] \\ \frac{\partial \ln l(\theta|x)}{\partial \theta_3}=\frac{1}{2}\sum_{k=1}^m \frac{1}{\mathrm{tr}(\rho f_{i_k})}\left[f_{i_k}(1,1)+f_{i_k}(2,2)\right]\end{aligned}\ ] ] with @xmath231_{ij}=\frac{\partial \ln l(\hat \theta)}{\partial \theta_i}\left(\frac{\partial \ln l(\hat \theta)}{\partial \theta_j}\right)^t$ ] . the fisher information decomposes similarly to @xmath217 for @xmath224 due to the linearity of the born probability in @xmath65 . as discussed in section iiib , quantum measurement strategies capable of fully reconstructing the density matrix can be characterized by a set of @xmath111 measurement bases ( matrices of eigenvectors ) @xmath232 . each such choice yields a different asymptotic variance for the ml estimator , i.e. , a different fisher information matrix . in the present work , we focus on the use of mutually unbiased measurement bases . for 1-qubit systems ( @xmath217 ) , mub bases @xmath232 can be written @xmath233 the observables are then simply the standard pauli spin operators . for @xmath224 , or more generally when @xmath1 is the power of an odd prime , these bases @xmath158 are given by @xcite @xmath234,~~ & 1\leq r\leq n. \\ & \end{array}\right.\ ] ] the orthonormal observables can then be taken to be @xmath235 recall that the mub measurements maximize the average fisher information over the set of all true density matrices @xmath140 , and hence are average - case optimal " as discussed in section iiid . in order to interrogate the asymptotic and finite sample losses induced by using biased or suboptimal measurement bases , the mub bases are rotated , causing the associated parallelipiped @xmath160 in section iii d to no longer be a rectangular solid and the average fisher information to decrease . after rotation , the new bases can be written @xmath236 where @xmath237 , @xmath36 being a random hermitian matrix specifying a random axis of rotation in the @xmath1-dimensional hilbert space ; @xmath238 is a scalar parameter specifying the extent of rotation ( magnitude of the solid angle ) . to generate a set of measurement bases that is sufficiently different from the mub , @xmath1 measurement bases were rotated according to the above formula ; for basis @xmath239 , the parameter @xmath238 was incrementally increased until @xmath240 runs over all the other bases and @xmath241 is a chosen scalar , for at least one pair of eigenvectors @xmath242 from bases @xmath239 and @xmath243 , respectively . if necessary , this procedure was iterated self - consistently . we refer to the resulting bases as mutually biased measurement bases ( mbb ) . in the bloch vector parametrization , the maximum of the likelihood corresponding to lagrangian function ( lagrangian ) can be found by solving for roots of the nonlinear system of @xmath244 equations @xmath245 in @xmath244 unknowns @xmath246 . the number of constraints and hence unknowns will differ in other parametrizations ; for example , for parameterizations where positive semidefiniteness is implicit in the parametrization ( such as the cholesky parametrization ) , @xmath247 in equation ( [ lagrangian ] ) . the newton - raphson algorithm can be used to find the roots of this nonlinear system . writing @xmath248 , the newton step for @xmath249 is @xmath250 with @xmath251 , where @xmath252 is the jacobian matrix . denoting the rows of @xmath253 by @xmath254 , we have @xmath255 in order to faciliate global convergence of the newton - raphson algorithm , the sum - of - squares " function @xmath256 is evaluated after each iteration , and the step length progressively shortened until the value of this function is found to decrease ( the existence of such a step length is guaranteed ) @xcite . alternatively , the sum - of - squares `` function @xmath257 may be minimized directly to locate the constrained maximum of the likelihood . in general , direct minimization of this function may be prone to encountering local traps . in the present case , minimization using an optimization algorithm capable of escaping from traps was employed . the quasi - newton algorithm , in which the algorithic step @xmath258 is given by @xmath259 , where @xmath260 denotes the approximate inverse hessian computed with the broyden - fletcher - goldfarb - shanno ( bfgs ) update , was first used to search for a zero of @xmath261 until convergence slowed below a specified stepwise tolerance '' , again using an adaptive line search strategy to identify the optimal step size . traps were often encountered that could not be escaped from using the above technique . to surmount them , a fixed number of stochastic simulated annealing steps were applied . in order to have a scalar measure of the accuracy of the entire density matrix estimate , the josza fidelity ( generalized overlap ) @xmath262 , which is related to the statistical distance on the space of density matrices , was used . the convergence tolerance ( i.e. , objective function value below which the optimized parameter estimate was accepted as an estimator ) for each @xmath97 was chosen by running a set of ml optimizations with a very large sample size ( @xmath263 observations ) , and determining the objective function value below which @xmath264 for all cases . in order to make the convergence tolerance compatible across difference sample sizes , the log of the unconstrained likelihood function was scaled as @xmath265 . kernel density estimators ( kdes ) , nonparametric density estimators that avoid some of the deficiencies of histograms , were used to estimate finite sample probability density functions . unlike histograms , they are smooth . kdes center a kernel function at each data point ; the contribution of data point @xmath266 to the estimate at @xmath267 depends on @xmath268 . the estimated density takes the form @xmath269 with @xmath270 . bandwidth ( @xmath271 ) optimization , which avoids values of @xmath271 that lead to spiky estimates ( undersmoothing ) or oversmoothing , was based on minimization of the asymptotic mean integrated squared error ( amise ) vaneeden1985 , i.e. , @xmath272 . a gaussian kernel @xmath273 was used . codes implementing the above algorithms , including both optimization and hypothesis testing , are available upon request from the author . we consider several monte carlo simulation environments in order to assess the asymptotic and finite - sample properties of the maximum likelihood estimators of the density matrix and test statistics for various physical quantities of interest . for both spin-1/2 and spin-1 systems , we report results for full rank , nondegenerate mixed states . the true @xmath97 s ( see tables i , ii ) are randomly chosen and , hence , the results may be considered representative for any true underlying density matrix ] . for each @xmath97 , @xmath274 hypothetical samples of i.i.d quantum observations each of size @xmath275 were simulated with the observations evenly distributed between the @xmath111 measurement bases . for simulating quantum observations from a given basis @xmath158 , the multinomial distribution probabilities @xmath276 were computed as @xmath277 where @xmath278 denotes the observable associated with the multinomial outcome @xmath172 obtained in draw @xmath91 from basis @xmath239 . the parameters of @xmath97 were then estimated using the maximum likelihood approach for each sample . among the types of possible optimal measurement strategies , mutually unbiased measurement bases ( mub ) minimizes the ( asymptotic ) estimation error across the widest range of hilbert space dimensions and true @xmath97 s ( as discussed section iii d ) . mub s are therefore used for the simulations in section ii b , and the performance of suboptimal mbb bases are then compared to the former in section c. the primary goal in this subsection is to determine whether the limiting asymptotic normal distribution of the parameter estimates provides a good approximation to the finite - sample distribution when mub s are used to generate quantum observations , and how the approximation improves with increase in the sample size . note that frequentist inferential methods are based on the notion of hypothetical repeated samples . hence , an important practical consideration when assessing the finite - sample performance of these methods is the choice of the number of simulated samples . in particular , the number of samples should be large enough such that the finite - sample statistics adequately characterize the corresponding population quantities and the finite - sample distributions are sufficiently smooth . to obtain such a choice , figure 1 compares the finite - sample distributions for the ml estimator of @xmath279 for the mixed spin-1/2 system for 100 , 500 , 1000 , and 1500 simulated samples . figure @xmath280 reveals that the finite - sample performance of the ml estimator is quite insensitive to the choice of the number of simulated samples . note that the finite - sample pdf s for different numbers @xmath281 of simulated samples are barely distinguishable . therefore , in all subsequent analysis , we set the number of hypothetical samples to @xmath274 . , mixed spin-1/2 system , mub bases , for various numbers q of hypothetical simulated samples . ( a ) m=100 ; ( b ) m=400 . in each panel , the finite sample distributions for q=100 , 500 , 1000 , and 1500 simulations are shown.,title="fig:",width=144,height=144 ] , mixed spin-1/2 system , mub bases , for various numbers q of hypothetical simulated samples . ( a ) m=100 ; ( b ) m=400 . in each panel , the finite sample distributions for q=100 , 500 , 1000 , and 1500 simulations are shown.,title="fig:",width=144,height=144 ] we next turn our attention to mixed spin-1/2 systems ( @xmath217 ) . table ii reports statistics from the asymptotic and finite - sample distributions of the ml estimators of the parameters and diagonal elements of the density matrix . panels a , b , and c report results for @xmath282 @xmath283 and @xmath274 , respectively . in particular , we report the following statistics from the finite - sample distribution : bias , standard deviation ( std ) , root mean square errors ( rmse ) , and @xmath2 confidence intervals . these statistics broadly summarize a probability distribution and are defined as follows . let @xmath284 denote the estimator of @xmath285 in sample @xmath286 . we have @xmath287 , @xmath288 where @xmath289 , @xmath290^{2}+\left [ \textmd{bias}(\hat { \theta}_{i})\right ] ^{2}\right\}^{1/2}$ ] , and the @xmath2 confidence interval is the region bounded by the @xmath291 and @xmath292 quantiles of the finite - sample distribution of @xmath293 . note that the diagonal elements of the density matrix are smooth real functions of the parameter vector @xmath7 . let @xmath294 denote the @xmath295-th element of the density matrix . given the invariance property of the ml estimator ( see section ii ) , the maximum likelihood estimator of @xmath294 is @xmath296 . also , the asymptotic distribution of the estimators of the diagonal elements of @xmath97 can be obtained using the continuous mapping theorem : @xmath297 has an asymptotic normal distribution with asymptotic variance given by @xmath298 where @xmath299 is the asymptotic variance of @xmath300 . row 1 of panel a reports the true values of the parameters and the diagonal elements of the density matrix . the row labeled `` asymptotic '' in each panel reports the point estimates of the parameters and diagonal elements , along with the asymptotic standard errors in parentheses and asymptotic @xmath2 confidence intervals in square brackets . we estimate the asymptotic covariance matrix consistently using the observed fisher information given by equation ( [ observed fisher ] ) . as mentioned above , the asymptotic distribution of the estimators of the diagonal elements of @xmath97 are obtained using the continuous mapping theorem . for the computation of the asymptotic distribution , 1 out of the 1000 hypothetical samples was selected randomly . note that the asymptotic standard errors are very small , ranging from @xmath301 for @xmath279 to @xmath302 for @xmath303 and @xmath304 in panel a. consequently , the asymptotic @xmath2 confidence intervals are tightly centered around the point estimates . the asymptotic standard errors decrease with the increase in sample size to @xmath305 and @xmath306 in panels b and c , respectively . consequently , as predicted by asymptotic theory , the distributions get narrower with increase in the sample size . however , note that , with the exceptions of @xmath303 in panel a , none of the asymptotic confidence intervals contain the true value of the parameter . [ cols="<,^,^,^,^,^,^ " , ] an important application of quantum state estimation is optimal decision making and control , and by extension , quantum information processing . control logic is often boolean , in that decisions are made based on whether a hypothesis is true or false , rather than the precise state of the system . therefore , in this subsection , we examine the finite - sample performance of maximum likelihood - based hypothesis testing procedures that aid in making decisions . hypothesis testing problems may involve a single restriction or multiple restrictions on the parameters . we consider first the case of single restrictions . the appropriate hypothesis test for testing a single restriction is the @xmath307-test . the @xmath307-test statistic for the null hypothesis @xmath308 against the alternative hypothesis @xmath309 , where @xmath310 is a known constant , is @xmath311where @xmath312 is a consistent estimate of the asymptotic standard error of @xmath313 . asymptotically , under the null hypothesis , the @xmath307-statistic converges in distribution to the standard normal distribution ; thus , for a two - sided size @xmath314 test , the null hypothesis is rejected if @xmath315 . the finite - sample _ size _ and _ power _ of the @xmath307-test vis - a - vis their corresponding asymptotic values provide a means of interrogating the finite sample performance of quantum state ml inferential methods . recall from section ii a that the size of a hypothesis test is the probability of rejecting the null hypothesis given that it is true . therefore , the finite sample size of the two - sided @xmath307-test for the hypothesis testing problem - @xmath316 : @xmath317 against the alternative hypothesis @xmath318 : @xmath319 - is given by the fraction of times @xmath320 is greater than @xmath321 in absolute value ; the @xmath307-test has an asymptotic size of @xmath322 . similarly , as defined in section ii a , the power of a hypothesis test is the probability of rejecting the null hypothesis given that it is false . hence , the finite - sample power for the hypothesis testing problem - @xmath323 against the alternative hypothesis @xmath324 , where @xmath325 - is given by the fraction of times @xmath326 is greater than @xmath321 in absolute value ; the @xmath307-test has an asymptotic power of @xmath280 . it can be shown @xcite that the @xmath307-test based on ml estimates is ump ; it is therefore ideal for interrogating finite sample performance of frequentist hypothesis testing . we consider single hypothesis testing problems involving restrictions on the parameters @xmath65 of the density matrix . in particular , for the spin-1/2 system , we test the hypothesis @xmath316 : @xmath327 against the alternative @xmath318 : @xmath328 . for the spin-1 system , the hypothesis tested is @xmath316 : @xmath329 against the alternative @xmath318 : @xmath330 . these hypotheses provide information about the finite - sample size of the testing procedure . for the finite - sample power , we test the hypothesis @xmath316 : @xmath331 against the alternative @xmath318 : @xmath332 for the spin-1/2 system and the hypothesis @xmath316 : @xmath333 against the alternative @xmath318 : @xmath334 for the spin-1 system . note that the true values of both @xmath335 for the spin-1/2 system and @xmath336 for the spin-1 system are sufficiently different from @xmath337 to ensure a proper size and power comparison of the tests . tables iv and v report the finite - sample size and power of the @xmath307-tests for the above hypotheses for the mixed spin-1/2 and spin-1 systems , respectively . panels a , b , and c in each table report results for @xmath282 @xmath283 and @xmath274 , respectively . consider first table iv . the finite - sample size of the @xmath307-test is 0.87 for @xmath338 , i.e. the probability of rejecting the null , @xmath316 : @xmath327 , given that it is true is @xmath339 . this is more than @xmath340 times bigger than the theoretically predicted asymptotic value of @xmath322 . this is a reflection of the fact that the asymptotic standard errors grossly underestimate the estimation error in finite - samples . the asymptotic standard error for @xmath303 is about @xmath302 - an order of magnitude smaller than the finite - sample value @xmath341 ( see table ii , panel a ) . consequently , the @xmath291 and @xmath292 quantiles from the finite - sample distribution of the @xmath307-statistic are @xmath342 and @xmath343 , respectively , whereas the corresponding asymptotic values are only @xmath344 and @xmath321 , respectively . the increase in sample size to @xmath345 and @xmath274 does not improve the size of the test . in fact , the finite - sample distribution of the @xmath307-statistic widens relative to the asymptotic standard normal distribution . the @xmath291 and @xmath292 quantiles from the finite - sample distribution of the @xmath307-statistic are @xmath346 and @xmath347 , respectively , for @xmath305 and @xmath348 and @xmath349 , respectively , for @xmath306 . this is because , although the finite - sample standard errors decrease with the increase in the sample size , the rate of convergence is much slower than the theoretically predicted rate @xmath350 . the asymptotic standard errors for @xmath303 for @xmath305 and @xmath306 relative to @xmath338 are @xmath351 and @xmath352 , respectively . however , the finite - sample standard errors for @xmath303 for @xmath305 and @xmath306 relative to @xmath338 are much higher at @xmath337 and @xmath353 , respectively ( see table ii ) . this is also shown in panel a of figure 11 that plots the finite sample @xmath307-statistic distribution for different sample sizes , along with the asymptotic standard normal distribution . similar results are obtained for the spin-1 system in table v and panel a of figure 12 . the power of the t - test is @xmath354 for @xmath338 , close to the asymptotic value of @xmath280 . for higher sample sizes , both the finite sample and asymptotic powers coincide at @xmath280 . however , note that the finite sample size ( above ) is also @xmath355 in spite of @xmath337 being substantially different from @xmath356 , indicating that the testing procedure is incapable of distinguishing between the two hypotheses . similar results are obtained for the spin-1 system ( table v ) . in certain applications , such as quantum information processing , it is important to simultaneously test whether several elements or parameters of the density matrix have prescribed values . for these testing problems involving multiple parameter restrictions , we rely on the wald test . as an example , for the spin-1/2 system , we test the hypothesis that each of the parameters @xmath357 is equal to its true known value : @xmath358 against the alternative @xmath359 . the wald statistic is given by @xmath360where @xmath361 and @xmath362 is the estimated asymptotic covariance matrix of the parameter estimates . asymptotically , under the null , the wald statistic converges in distribution to a chi - squared random variable @xmath363 , with the number of degrees of freedom @xmath91 equal to the number of parameter restrictions ( in this case @xmath364 ) . a size @xmath365 wald test has a rejection region corresponding to the tail of the @xmath363 distribution beyond which the cumulative probability density is @xmath322 . for @xmath366 , the null hypothesis is rejected if @xmath367 . the wald test at significance level @xmath314 has asymptotic size @xmath322 and power @xmath280 . the finite sample size of the test for the hypothesis testing problem - @xmath368 against the alternative hypothesis @xmath369 - is given by the fraction of times @xmath370 where @xmath361 , is greater than 7.81 . the finite - sample power for the hypothesis testing problem - @xmath371 against the alternative hypothesis @xmath372 , where @xmath373 - is given by the fraction of times @xmath370 , where @xmath374 , is greater than 7.81 . the wald test based on ml estimates is asymptotically equivalent to a likelihood ratio test , and hence is ump ( see section ii for details ) . the finite - sample size and power of the wald test for the above hypothesis for different sample sizes is reported in table iv , for the spin-1/2 system . the finite - sample size of the wald test for @xmath338 is @xmath354 , almost 20 times bigger than the asymptotic value of @xmath322 . the @xmath2 quantile from the finite - sample distribution of the wald - statistic is @xmath375 , whereas the corresponding asymptotic value is only @xmath376 . as with the @xmath307-test , the increase in sample size to @xmath345 and @xmath274 does not improve the size of the wald test . the finite - sample distribution of the wald- statistic shifts further away from the asymptotic chi - squared distribution with increase in sample size . this is also revealed in panel b of figure 11 that plots the finite sample wald - statistic distribution for different sample sizes , along with the asymptotic chi - squared distribution . similar results are obtained for the spin-1 system in table v and panel b of figure 12 , although the power of the test improves for @xmath306 . the finite sample performance of the wald statistic is also poor for the spin-1 system as reported in table v and panel b of fig . finally , another important conjecture to test regarding the density matrix is whether an eigenvalue of @xmath124 has a given value . hence , we also test the hypothesis @xmath316 that an eigenvalue of @xmath97 equals its true ( known ) value . we denote the @xmath172-th eigenvalue of the estimated density matrix @xmath377 by @xmath378 . the hellmann - feynman theorem can be used to compute @xmath379 . denoting the eigenvectors of @xmath124 by @xmath380 , we have : @xmath381 since @xmath382 , @xmath383 . for spin-1/2 systems , the exact analytical solution for the eigenvalue derivatives is available directly since the characteristic polynomial is a quadratic : @xmath384 where @xmath385 . however , it is difficult to analytically compute asymptotic standard errors for the estimated eigenvalues of higher - dimensional quantum states by using the hellmann - feynman theorem since the characteristic polynomial is of higher order . here , eigenvalue test statistics , along with the size and power of the associated tests , were computed for the spin-1 system under consideration ( table v ) , because they are representative of the higher - order nonlinearity common in applied problems . note that hypothesis tests involving a single eigenvalue are t - tests . the finite sample performance of the eigenvalue tests are roughly identical to those of the above parameter tests , indicating that decisions ( such as control logic ) based on density matrix eigenvalues or state entropy estimated by frequentist methods are expected to perform poorly for such sample sizes . a hypothesis of practical interest is the purity of the state . in this case , parameter values in the null hypothesis lie on the boundary of the maintained hypothesis . hence , standard regularity conditions ( the parameter value in the null should be an interior point of a compact set ) that ensure asymptotic convergence of the null distributions of the t - statistic and the wald statistic to standard normal and chi - squared distributions , respectively , fail to hold . andrews @xcite provides general asymptotic results for testing problems of this sort . he derives the asymptotic null and local alternative distributions of the test statistics under a set of high level conditions . although the distributions are non - standard , the critical values can be obtained by simulation . the salient conclusion from the above examples is that the size and power of hypothesis tests fall substantially far from the asymptotically predicted values , and that the finite sample test statistic distributions are not normal and chi - squared , as would be expected from asymptotic theory . in finite samples of practical size , hypothesis testing performs rather poorly for parameters , @xmath124 elements or eigenvalues , underscoring the inadequacy of the frequentist quantum state estimation approach for such sample sizes . . the asymptotic t - statistic distribution is standard normal ( superimposed ) . ( b ) wald - statistic distribution for null hypothesis @xmath386 . the asymptotic wald statistic distribution is chi - squared with degrees of freedom equal to 3 , the number of restrictions ( superimposed).,width=326,height=384 ] . the asymptotic t - statistic distribution is standard normal ( superimposed ) . ( b ) wald - statistic distribution for null hypothesis @xmath387 . the asymptotic wald statistic distribution is chi - squared with degrees of freedom equal to 3 , the number of restrictions ( superimposed).,width=326,height=384 ] in this paper we have examined the performance of frequentist estimators of the density matrix of a quantum system using quantum observations simulated using different measurement strategies . we provided numerical techniques for likelihood optimization under multiple constraints that are robust across arbitrary spin-1/2 and spin-1 system density matrices . in addition , we have presented methodologies and prescriptions for hypothesis testing in the frequentist quantum framework , an essential requirement for optimal decision making and control . performing inference in the frequentist framework , we find that the finite sample variances are signficantly larger than the asymptotic bounds ( predicted by the cramer - rao theorem ) for typical experimental sample sizes , and that these bounds are not approached at the rate predicted by asymptotic theory . this finding is robust to the choice of the density matrix and is more pronounced for small datasets . a prior study @xcite showed close correspondence between asymptotic and finite sample performance in a single example , but did not consider the rate of convergence to the crb , or higher - dimensional systems . a unique feature of quantum statistics is the quantum cramer - rao bound , a generalization of the classical cramer - rao bound that originates due to the dependence of the quantum fisher information on the mode of measurement . in prior work , considerable attention has been devoted to understanding the asymptotic relative efficiencies of different quantum measurement strategies . our results warrant a careful re - examination of the relative efficiencies of these measurement strategies in finite samples . recent studies @xcite have aimed to assess the resource requirements of various quantum tomography implementations employing different measurement strategies ; future efforts along these lines would benefit from attention to finite sample losses and the rate at which asymptotic predictions are approached . given that the finite sample variances are order(s ) of magnitude bigger than the corresponding asymptotic ones , we conclude that in order to improve parameter estimates in finite samples , it is important to incorporate additional information that exploits the information geometry of quantum states into the estimation procedure . an ideal approach in this regard is bayesian estimation . since it is based on updating a prior plausibility distribution about the parameters based on observed data , bayesian methods are generally more reliable than standard frequentist methods away from the asymptotic limit . the prior plausibility distribution permits the introduction of auxiliary information about the parameter space that is not contained within the likelihood . by contrast , such information is impossible to incorporate in frequentist estimation techniques . consequently , due to the incorporation of prior knowledge , the credible intervals produced by bayesian inference are generally shorter than their asymptotic frequentist counterparts . moreover , as we have seen , bayesian credible intervals do not refer to the asymptotic limit of an infinite number of measurements . the following complete , nonoptimal measurement bases were used in section [ measure ] in order to assess the finite sample losses incurred due to not using mutually unbiased measurements for spin-1 systems : @xmath388 these bases were generated through rotation of mub measurement bases according to the method described in section iv c , with @xmath389 in equation ( [ nonmub condition ] ) .

generating functions for plane overpartitions are obtained using various methods such as nonintersecting paths , rsk type algorithms and symmetric functions . we extend some of the generating functions to cylindric partitions . also , we show that plane overpartitions correspond to certain domino tilings and we give some basic properties of this correspondence .

0903.2039

the goal of the first part of this paper is to introduce a new object called plane overpartitions , and to give several enumeration formulas for these plane overpartitions . plane overpartition _ is a plane partition where ( 1 ) in each row the last occurrence of an integer can be overlined or not and all the other occurrences of this integer are not overlined and ( 2 ) in each column the first occurrence of an integer can be overlined or not and all the other occurrences of this integer are overlined . an example of a plane overpartition is @xmath0 this paper takes its place in the series of papers on overpartitions started by corteel and lovejoy @xcite . the motivation is to show that the generating function for plane overpartitions is : @xmath1 in this paper , we give several proofs of this result and several refinements and generalizations . namely , we prove the following results . content formula for the generating function for plane overpartitions of a given shape , see theorem [ t1 ] . the hook formula for the generating function for reverse plane overpartitions , see theorem [ reverse ] . the generating function formula for plane overpartitions with bounded parts , see theorem [ boundedparts ] . the goal of the second part of this paper is to extend the generating function formula for cylindric partitions due to borodin @xcite and the following 1-parameter generalized macmahon s formula due to the third author of this paper @xcite : @xmath2 where the weight @xmath3 is a polynomial in @xmath4 that we describe below . given a plane partition @xmath5 ( a ferrers diagram filled with positive integers that form nonincreasing rows and columns ) , a connected component of @xmath5 is the set of all connected boxes of its ferrers diagram that are filled with a same number . if a box @xmath6 belongs to a connected component @xmath7 then we define its level @xmath8 as the smallest positive integer such that @xmath9 does not belong to @xmath7 . a border component of level @xmath10 is a connected subset of a connected component whose all boxes have level @xmath10 , see figure [ fig11 ] . we associate to each border component of level @xmath10 , the weight @xmath11 . the polynomial @xmath3 is the product of the weights of its border components . for the plane partition from figure [ fig11 ] it is @xmath12 . [ htp ! ] we give a new proof of the 1-parameter generalized macmahon s formula . we also extend this formula to two more general objects : skew plane partitions and cylindric partitions . namely , we prove the following results . 1-parameter generalized formula for the generating function for skew plane partitions , see theorem [ skew ] . 1-parameter generalized formula for the generating function for cylindric partitions , see theorem [ uvodcyl ] . in the rest of this section we give definitions and explain our results in more detail . a partition @xmath13 is a nonincreasing sequence of positive integers @xmath14 . each @xmath15 is a part of the partition and the number of parts is denoted by @xmath16 . the weight @xmath17 of @xmath13 is the sum of its parts . a partition @xmath13 can be graphically represented by the ferrers diagram that is a diagram formed of @xmath16 left justified rows , where the @xmath18 row consists of @xmath15 cells ( or boxes ) . the conjugate of a partition @xmath13 , denoted by @xmath19 , is a partition that has the ferrers diagram equal to the transpose of the ferrers diagram of @xmath13 . for a cell @xmath6 of the ferrers diagram of @xmath13 the _ hook length _ of this cell is @xmath20 and the _ content _ is @xmath21 . it is well known that the generating function for partitions that have at most @xmath22 parts is @xmath23 , where @xmath24 . more definitions on partitions can be found , for example , in @xcite or @xcite . an _ overpartition _ is a partition where the last occurrence of an integer can be overlined @xcite . last occurrences in an overpartition are in one to one correspondence with corners of the ferrers diagram and overlined parts can be represented by marking the corresponding corners . the generating function for overpartitions that have at most @xmath22 parts is @xmath25 . let @xmath13 be a partition . a _ plane partition _ of shape @xmath13 is a filling of cells of the ferrers diagram of @xmath13 with positive integers that form a nonincreasing sequence along each row and each column . we denote the shape of a plane partition @xmath5 by @xmath26 and the sum of all entries by @xmath27 , called the weight of @xmath5 . it is well known , under the name of macmahon s formula , that the generating function for plane partitions is @xmath28 one way to prove this is to construct a bijection between plane partitions and pairs of semi - standard young tableaux of the same shape and to use the rsk correspondence between these young tableaux and certain matrices @xcite . recall that a plane overpartition is a plane partition where in each row the last occurrence of an integer can be overlined or not and in each column the first occurrence of an integer can be overlined or not and all the others are overlined . this definition implies that the entries strictly decrease along diagonals , i.e. all connected components are also border components . therefore , a plane overpartition is a diagonally strict plane partition where some entries are overlined . more precisely , it is easy to check that inside a border component only one entry can be chosen to be overlined or not and this entry is the upper right entry . plane overpartitions are therefore in bijection with diagonally strict plane partitions where each border component can be overlined or not ( or weighted by 2 ) . recently , those weighted diagonally strict plane partitions were studied in @xcite . the first result obtained was the shifted macmahon s formula that says that the generating function for plane overpartitions is indeed equation . this was obtained as a limiting case of the generating function formula for plane overpartitions which fit into an @xmath29 box , i.e. whose shapes are contained in the rectangular shape with @xmath30 rows and @xmath31 columns . @xcite[osnovna ] the generating function for plane overpartitions which fit in an @xmath29 box is @xmath32 this theorem was proved in @xcite using schur @xmath33 and @xmath34 symmetric functions and a suitable fock space . in @xcite the theorem was proved in a bijective way where an rsk type algorithm ( due to sagan @xcite , see also chapter xiii of @xcite ) was used to construct a bijection between plane overpartitions and matrices of nonnegative integers where positive entries can be overlined . in section 2 , we give a mostly combinatorial proof of the generalized macmahon formula @xcite . namely , we prove : @xcite @xmath35 [ mirjana ] in the above formula , @xmath36 is the set of plane partitions with at most @xmath30 rows and @xmath31 columns . when we set @xmath37 , only the border components of level @xmath38 have a non zero weight and we get back theorem [ osnovna ] . the main result of section 3 is a hook content formula for the generating function for plane overpartitions of a given shape . more generally , we give a weighted generating function where overlined parts are weighted by some parameter @xmath39 . let @xmath40 be the set of all plane overpartitions of shape @xmath13 . the number of overlined parts of a plane overpartition @xmath5 is denoted by @xmath41 . for example , @xmath42 is a plane overpartition of shape @xmath43 , with @xmath44 and @xmath45 . [ t1 ] let @xmath13 be a partition . the weighted generating function for plane overpartitions of shape @xmath13 is @xmath46 we prove this theorem using using a correspondence between plane overpartitions and sets of nonintersecting paths that use three kinds of steps . ( the work of brenti used similar paths to compute super schur functions @xcite . ) another way to prove this result is to show that plane overpartitions of shape @xmath13 are in bijection with super semistandard tableaux of shape @xmath13 . this is presented in a remark in section [ nonint ] . we also give the weighted generating function formula for plane overpartitions `` bounded '' by @xmath13 , where by that we mean plane overpartitions such that the @xmath18 row of the plane overpartition is an overpartition that has at most @xmath47 parts and at least @xmath48 parts . let @xmath49 be the set of all such plane overpartitions . [ t2 ] let @xmath13 be a partition . the weighted generating function for plane overpartitions such that the @xmath18 row of the plane overpartition is an overpartition that has at most @xmath47 parts and at least @xmath48 parts is @xmath50 note that it is enough to assign weights to overlined ( or nonoverlined ) parts only because generating functions where overlined and nonoverlined parts are weighted by @xmath39 and @xmath51 , respectively , follow trivially from the above formulas . the number of nonintersecting paths is given by a determinantal formula ( lemma 1 of @xcite ) . this result was anticipated by lindstrm @xcite and karlin and mcgregor @xcite , but gessel and viennot were first to use it for enumerative purpose of various classes of plane partitions @xcite . applying the result and evaluating the determinants we obtain hook content formulas . we use a simple involution to show that the stanley hook content formula ( theorem 7.21.2 of @xcite ) follows from our formula . from the symmetric function point of view , these formulas are given by schur functions in a difference of two alphabets , as explained in section 3 . the end of section 3 is devoted to _ reverse _ plane overpartitions . a reverse plane partition of shape @xmath13 is a filling of cells of the ferrers diagram of @xmath13 with nonnegative integers that form a nondecreasing sequence along each row and each column . a reverse plane overpartition is a reverse plane partition where ( 1 ) only positive entries can be overlined , ( 2 ) in each row the last occurrence of an integer can be overlined or not and ( 3 ) in each column the first occurrence of a positive integer can be overlined or not and all others ( if positive ) are overlined . an example of a reverse plane overpartition is @xmath52 it was proved by gansner @xcite that the generating function for reverse plane partitions of a given shape @xmath13 is @xmath53 let @xmath54 be the set of all reverse plane overpartitions of shape @xmath13 . the generating function for reverse plane overpartitions is given by the following hook formula . let @xmath13 be a partition . the generating function for reverse plane overpartitions of shape @xmath13 is @xmath55 [ reverse ] we construct a bijection between reverse plane overpartitions of a given shape and sets of nonintersecting paths whose endpoints are not fixed . using results of @xcite we obtain a pfaffian formula for the generating function for reverse plane partitions of a given shape . subsequently , we evaluate the pfaffian and obtain a proof of the hook formula . when @xmath13 is the partition with @xmath30 parts equal to @xmath31 , this result is the generating function formula for plane overpartitions fitting in an @xmath29 , box namely theorem [ osnovna ] . in section 4 we make a connection between plane overpartitions and domino tilings . we give some basic properties of this correspondence , such as how a removal of a box or an overline changes the corresponding tiling . this correspondence connects a measure on strict plane partitions studied in @xcite to a measure on domino tilings . this connection was expected by similarities in correlation kernels , limit shapes and some other features of these measures , but the connection was not established before . in section 5 we propose a bijection between matrices and pairs of plane overpartitions based on ideas of berele and remmel @xcite . we give another stronger version of the shifted macmahon s formula , as we give a weighted generating function for plane overpartitions with bounded entries . let @xmath56 be the set of all plane overpartitions with the largest entry at most @xmath22 . [ boundedparts ] the weighted generating functions for plane overpartitions where the largest entry is at most @xmath22 is @xmath57 in section 6 we study interlacing sequences and cylindric partitions . we say that a sequence of partitions @xmath58 is _ interlacing _ if @xmath59 or @xmath60 is a horizontal strip , i.e. a skew shape having at most one cell in each column . let @xmath61 be a sequence of 0 s and 1 s . we say that an interlacing sequence @xmath58 has _ profile _ @xmath62 if when @xmath63 , respectively @xmath64 , then @xmath59 , respectively @xmath65 is a horizontal strip . interlacing sequences are generalizations of plane partitions . indeed plane partitions are interlacing sequences with profile @xmath66 . [ htp ! ] we now define the _ diagram _ of an interlacing sequence . see figure [ fig1 ] . we start with a square grid ; we denote the two directions defined by the grid lines with 0 and 1 . a profile @xmath61 is represented by a path of length @xmath67 on this grid where the path consists of grid edges whose directions are given by @xmath68 . this path forms the ( upper ) border of the diagram . excluding the endpoints of the path we draw the diagonal rays ( which form @xmath69 angles with grid lines ) starting at the vertices of the path and we index them ( from left to right ) with integers from @xmath38 to @xmath70 . a diagram is a connected subset of boxes of a square grid whose ( upper ) border is given by the profile path and along the @xmath18 diagonal ray there are @xmath71 boxes . the filling numbers on the @xmath18 diagonal ray are parts of @xmath72 with the largest part at the top . observe that by the definition of interlacing sequences we obtain monotone sequences of numbers in the direction of grid lines . a ( skew ) plane partition and cylindric partition are examples of interlacing sequences . a plane partition can be written as @xmath73 with profile @xmath74 and @xmath72s are diagonals of the plane partition . a skew plane partition is an interlacing sequence @xmath75 with a profile @xmath76 . a cylindric partition is an interlacing sequence @xmath77 where @xmath78 , and @xmath70 is called the period of @xmath79 . a cylindric partition can be represented by the _ cylindric diagram _ that is obtained from the ordinary diagram by identification of the first and last diagonal . a _ connected component _ of an interlacing sequence @xmath79 is the set of rookwise connected boxes of its diagram that are filled with a same number . we denote the number of connected components of @xmath79 with @xmath80 . for the example from figure [ fig1 ] we have @xmath81 and its connected components are shown in figure [ fig1 ] ( bold lines represent boundaries of these components ) . if a box @xmath6 belongs to a connected component @xmath7 then we define its _ level _ @xmath82 as the smallest positive integer such that @xmath83 does not belong to @xmath7 . in other words , a level represents the distance from the `` rim '' , distance being measured diagonally . a _ border component _ is a rookwise connected subset of a connected component where all boxes have the same level . we also say that this border component is of this level . for the example from figure [ fig1 ] , border components and their levels are shown in figure [ fig2 ] ( different levels are represented by different colors ) . [ htp ! ] let @xmath84 be a sequence of nonnegative integers where @xmath85 is the number of @xmath10level border components of @xmath79 . we set @xmath86 for the example above @xmath87 . for a cylindric partition @xmath5 , we define _ cylindric connected components _ and _ cylindric border components _ in the same way but connectedness is understood on the cylinder , i.e. boxes are connected if they are rookwise connected in the cylindric diagram . we define @xmath88 where @xmath89 is the number of cylindric border components of level @xmath10 . in section 6 we give a generating function formula for skew plane partitions . let @xmath90 be the set of all skew plane partitions @xmath91 with profile @xmath92 , where @xmath93 and @xmath94 . ( generalized macmahon s formula for skew plane partitions ; hall - littlewood case ) @xmath95 [ skew ] note that as profiles are words in @xmath96 , a profile @xmath97 encodes the border of a ferrers diagram @xmath13 . skew plane partitions of profile @xmath62 are in one to one correspondence with reverse plane partitions of shape @xmath13 . moreover , one can check that @xmath98 therefore the theorem of gansner ( equation ) is theorem [ skew ] with @xmath99 and our theorem [ reverse ] on reverse plane overpartitions is theorem [ skew ] with @xmath37 . this theorem is also a generalization of results of vuleti @xcite . in @xcite a 2parameter generalization of macmahon s formula related to macdonald symmetric functions was given and the formula is especially simple in the hall - littlewood case . in the hall - littlewood case , this is a generating function formula for plane partitions weighted by @xmath100 . theorem [ skew ] can be naturally generalized to the macdonald case , but we do not pursue this here . let @xmath101 be the set of all cylindric partitions with period @xmath70 and profile @xmath102 . the main result of section 6 is : ( generalized macmahon s formula for cylindric partitions ; hall - littlewood case)[uvodcyl ] @xmath103 where @xmath104 is the smallest positive integer such that @xmath105 . the case @xmath99 is due to borodin and represents a generating function formula for cylindric partitions . cylindric partitions were introduced and enumerated by gessel and krattenthaler @xcite . the result of borodin could be also proven using theorem 5 of @xcite and the @xmath106-extension of bailey s @xmath107 summation due to gustafson ( equation ( 7.9 ) in @xcite ) @xcite . again theorem [ uvodcyl ] can be naturally generalized to the macdonald case . the trace generating function of those cylindric partitions could also be easily derived from our proof , as done by okada @xcite for the reverse plane partitions case . + the paper is organized as follows . in section 2 we give a mostly combinatorial proof of the generalized macmahon formula . in section 3 we use nonintersecting paths and obtain the hook length formulas for plane overpartitions and reverse plane partitions of a given shape . in section 4 we make the connection between tilings and plane overpartitions . in section 5 we construct a bijection between matrices and pairs of plane overpartitions and obtain a generating function formula for plane overpartitions with bounded part size . in section 6 we give the hook formula for reverse plane partitions contained in a given shape and the 1parameter generalization of the generating function formula for cylindric partitions . section 7 contains some concluding remarks . + * acknowledgment . * the authors want to thank alexei borodin , the advisor of the third author , for help and guidance and cdric boutillier , dominique gouyou - beauchamps , richard kenyon and jeremy lovejoy for useful discussions . the authors also want to thank one of the anonymous referee for his long list of constructive comments . in this section , we give an alternative proof of the generalization of macmahon s formula due to the third author @xcite . our proof is mostly combinatorial as it uses a bijection between plane partitions and pairs of strict plane partitions of the same shape and the combinatorial description of hall littlewood functions ( @xcite , chapter iii , equation ( 5.11 ) ) . let @xmath36 be the set of plane partitions with at most @xmath30 rows and @xmath31 columns . given a plane partition @xmath5 , let @xmath3 be the polynomial defined in , as @xmath108 . recall that theorem [ mirjana ] states that @xmath35 any plane partition @xmath5 is in bijection with a sequence of partitions @xmath109 . this sequence is such that @xmath110 is the shape of the entries greater than or equal to @xmath10 in @xmath5 for all @xmath10 . for example if @xmath111 the corresponding sequence is @xmath112 . note that the plane partition @xmath5 is column strict if and only if @xmath113 is a horizontal strip for all @xmath10 . we use a bijection between pairs of column strict plane partitions @xmath114 and plane partitions @xmath5 due to bender and knuth @xcite . we suppose that @xmath114 are of the same shape @xmath13 and that the corresponding sequences are @xmath115 and @xmath116 given a plane partition @xmath117 , we define the entries of diagonal @xmath118 to be the partition @xmath119 with @xmath120 and @xmath121 . the bijection is such that the entries of diagonal @xmath118 of @xmath5 are @xmath122 if @xmath123 and @xmath124 otherwise . note that as @xmath79 and @xmath125 have the same shape , the entries of the main diagonal ( @xmath126 ) are @xmath127 . for example , start with @xmath128 whose sequences are @xmath129 and @xmath130 , respectively and get @xmath131 this construction implies that : @xmath132 here we have @xmath133 and @xmath134 moreover @xmath135 where @xmath136 is a horizontal strip @xmath137,@xmath138 is the multiplicity of @xmath10 in @xmath13 and @xmath139 is the set of integers such that @xmath140 and @xmath141 . see @xcite , chapter iii sections 2 and 5 . indeed , the following statements are true . * each factor @xmath11 in @xmath142 is in one - to - one correspondence with a border component of level @xmath10 that goes through the main diagonal of @xmath5 . * each factor @xmath11 in @xmath143 is in one - to - one correspondence with a border component of level @xmath10 that ends in a non - negative diagonal . * each factor @xmath11 in @xmath144 is in one - to - one correspondence with a border component of level @xmath10 that starts in a non - positive diagonal . continuing with our example , we have @xmath145 and @xmath146 we recall the combinatorial definition of the hall littlewood functions following macdonald @xcite . the hall - littlewood function @xmath147 can be defined as @xmath148 where @xmath149 and @xmath150 is the number of entries equal to @xmath10 in @xmath79 . see @xcite chapter iii , equation ( 5.11 ) . a direct consequence of the preceding bijection is that the entries of @xmath125 are less than or equal to @xmath31 , and the entries of @xmath79 are less than or equal to @xmath30 if and only if @xmath5 is in @xmath151 . therefore : @xmath152 finally , we need equation ( 4.4 ) in chapter iii of @xcite . @xmath153 with the substitutions @xmath154 for @xmath155 and @xmath156 otherwise and @xmath157 for @xmath158 and @xmath156 otherwise , we get the result . in this section we represent plane overpartitions as nonintersecting paths . we use the determinantal formula for the number of nonintersecting paths ( see @xcite ) . evaluating these determinants we obtain the hook content formulas from theorems [ t1 ] and [ t2 ] . a similar approach was used for example in @xcite to compute super schur functions . we construct a bijection between the set of paths from @xmath159 to @xmath160 and the set of overpartitions with at most @xmath161 parts and the largest part at most @xmath118 . given an overpartition the corresponding path consists of north and east edges that form the border of the ferrers diagram of the overpartition except for corners containing an overlined entry where we substitute a pair of north and east edges with an north east edge . for example , the path corresponding to the overpartition @xmath162 is shown in figure [ path ] . note that this construction appears also in proposition 2.2 of @xcite . [ htp ! ] to each overpartition @xmath13 we associate a weight equal to @xmath163 , where @xmath164 is the number of overlined parts . to have the same weight on the corresponding path we introduce the following weights on edges . we assign weight 1 to east edges , @xmath165 to north edges on ( vertical ) level @xmath10 and weight @xmath166 to north - east edges joining vertical levels @xmath10 and @xmath167 . the weight of the path is equal to the product of weights of its edges . we will need the following lemma . @xcite [ lemaop]the generating function for overpartitions with at most @xmath161 parts is given by @xmath168 and the generating function for overpartitions with exactly @xmath161 parts is given by @xmath169 for a plane overpartition @xmath5 of shape @xmath13 we construct a set of nonintersecting paths using paths from row overpartitions where the starting point of the path corresponding to the @xmath10th row is shifted upwards by @xmath170 so that the starting point is @xmath171 . in that way , we obtain a bijection between the set of nonintersecting paths from @xmath171 to @xmath172 where @xmath10 runs from 1 to @xmath16 , and the set of plane overpartitions whose @xmath10th row has at most @xmath15 parts and at least @xmath173 parts with @xmath118 greater or equal to the largest part . the weights of this set of nonintersecting paths ( the product of weights of its paths ) is equal to @xmath174 . figure [ fig4 ] ( see also figure [ fig6 ] ) shows the corresponding set of nonintersecting paths for @xmath175 and the plane overpartition @xmath176 [ htp ! ] [ defm ] for a partition @xmath13 we define @xmath177 to be the @xmath178 matrix whose @xmath6th entry is given by @xmath179 for a partition @xmath13 let @xmath49 and @xmath40 be as in the introduction , i.e. the sets of all plane overpartitions bounded by shape @xmath13 and of the shape @xmath13 , respectively . [ bounded ] let @xmath13 be a partition . the weighted generating function for plane overpartitions whose @xmath18 row is an overpartition that has at most @xmath47 parts and at least @xmath48 parts is given by @xmath180 from ( [ genoverpartitions ] ) we have that the limit when @xmath118 runs to infinity of the number of paths from @xmath159 to @xmath160 is @xmath181 . using lemma 1 of @xcite we have that @xmath182 is the limit when @xmath118 goes to infinity of the generating function for @xmath16 nonintersecting paths going from @xmath183 to @xmath184 . thanks to the bijection between paths and overpartitions this is also the generating function for overpartitions whose @xmath10th row overpartition has at most @xmath47 and at least @xmath48 parts . [ gen ] let @xmath13 be a partition . the weighted generating function for plane overpartitions of shape @xmath13 is given by @xmath185 the proof is the same as the proof of proposition [ bounded ] . here we just use ( [ genoverpartitionsprim ] ) instead of ( [ genoverpartitions ] ) . the determinant of @xmath186 is given by the following formula : @xmath187 for the proof see for example pages 1617 of @xcite or theorem 26 ( 3.11 ) of @xcite . using ( [ m ] ) we obtain the product formulas for the generating functions from propositions [ bounded ] and [ gen ] . those are the hook content formulas given in theorems [ t1 ] and [ t2 ] . * remark 1 . * we give a bijective proof of formula ( [ m ] ) . indeed krattenthaler @xcite showed that the weighted generating function of super semistandard young tableaux of shape @xmath13 is @xmath188 where each tableau is weighted by @xmath189 . a super semistandard tableau @xcite of shape @xmath13 is a filling of cells of the ferrers diagram of @xmath13 with entries from the ordered alphabet @xmath190 such that * the nonoverlined entries form a column - strict reverse plane partition of some shape @xmath191 , where @xmath191 is a partition contained in @xmath13 , * the overlined entries form a row - strict reverse plane partition of shape @xmath192 . we give here a bijection from super semistandard young tableaux to plane overpartitions . we starting with a super semistandard young tableau where the nonoverlined ( resp . overlined ) entries are less than or equal to @xmath161 ( resp . @xmath193 ) . first , change all the entries equal to @xmath39 by @xmath194 and all the entries equal to @xmath195 to @xmath196 . for example , starting with the super semistandard young tableau with @xmath197 and @xmath198 @xmath199 now we use the order @xmath200 and we suppose that the cells outside of the shape @xmath13 are filled with 0 . at first all the nonoverlined entries are active . while there is an active part , we choose the smallest active part . if there are more than one , we choose the rightmost one . we swap it with its east or south neighbor , choosing the larger of the two . if they are equal then we choose the east one . we proceed in this way until this part is greater than or equal to both of its neighbors . when we reach this , we declare the part inactive . continuing with the preceding example and applying the algorithm to the smallest ( and rightmost ) active part until it becomes inactive we obtain : @xmath201 then we move the rest of the active parts . @xmath202 @xmath203 * remark 2 . * stanley s hook content formula states that the generating function of semistandard young tableaux of shape @xmath13 where the entries are less than or equal to @xmath22 is @xmath204 see theorem 7.21.2 of @xcite . formula ( [ m ] ) is equivalent to stanley s hook formula by examples i.2.5 and i.3.3 of @xcite . again we can give a bijective argument . formula ( [ shc ] ) is the weighted generating function of super semistandard tableaux of shape @xmath13 with @xmath205 . start with a super semistandard young tableau @xmath70 of shape @xmath13 and add @xmath22 to all the overlined entries . now transform the tableau into a reverse plane overpartition with the above algorithm using the order @xmath206 . this shows that super semistandard young tableaux @xmath70 of shape @xmath13 with weight @xmath207 are in bijection with reverse plane overpartitions @xmath5 of the same shape where the overlined entries are greater than @xmath22 with weight @xmath208 . now we define a sign reversing involution on these reverse plane overpartitions . given such a reverse plane overpartition , if there is at least one part greater than @xmath22 , we choose the uppermost and rightmost part greater than @xmath22 . if this part is overlined , then we take off the overline , otherwise we overline the part . note that in this case the parity of the number of overlined parts is changed and therefore the weight of the reverse plane overpartition is multiplied by @xmath209 . if no such part exists , the given reverse plane partition is a semistandard young tableau of shape @xmath13 where the entries are less than or equal to @xmath22 . + now , we give a generating formula for plane overpartitions with at most @xmath30 rows and @xmath31 columns . [ suma ] the weighted generating function for plane overpartitions with at most @xmath30 rows and @xmath31 columns is given by @xmath210 if @xmath30 is odd , and by @xmath211 if @xmath30 is even . in particular , the weighted generating function for all overpartitions is : @xmath212 this is a direct consequence of proposition [ bounded ] . we will use this result to get another `` symmetric function '' proof of the shifted macmahon s formula ( @xcite ) : @xmath213 in this section we give the weighted generalization of the shifted macmahon formula . we use a symmetric function identity to compute the last sum in proposition [ suma ] since @xmath214 , as we will soon see , has an interpretation in terms of symmetric functions . a symmetric function of an alphabet ( set of indeterminates , also called letters ) @xmath215 is a function of letters which is invariant under any permutation of @xmath215 . recall three standard bases for the algebra of symmetric functions : schur functions @xmath216 , complete symmetric functions @xmath217 and elementary symmetric functions @xmath218 ( see @xcite ) . the latter two are conveniently given by their generating functions : @xmath219 algebraic operations on alphabets , such as addition , subtraction and multiplication , can be defined using @xmath13-rings framework , see chapters i and ii of @xcite . in this framework symmetric functions are seen as ring operators . the addition of two alphabets @xmath215 and @xmath220 is defined naturally as their disjoint union and is denoted by @xmath221 . obviously , @xmath222 subtraction and multiplication are defined by @xmath223 analogous relations hold for elementary symmetric functions since @xmath224 . a specialization is an algebra homomorphism between the algebra of symmetric functions and @xmath225 . if @xmath226 is a specialization and @xmath227 is a symmetric function we denote its image by @xmath228 . let @xmath229 be a specialization given by @xmath230 since , for @xmath231 @xmath232 we have from definition [ defm ] that @xmath233 in the @xmath13ring framework , the @xmath234-binomial theorem ( see ( 2.21 ) of @xcite ) @xmath235 shows that the specialization @xmath229 is equivalent to considering symmetric functions in the difference of two alphabets @xmath236 and @xmath237 . thus , @xmath238 the weighted shifted macmahon formula can be obtained from ( [ sym ] ) and proposition [ suma ] . we have @xmath239 where @xmath19 is the transpose of @xmath13 and a partition is even if it has even parts . by ex . 10 b ) on p. 79 of @xcite we have @xmath240 if instead of @xmath215 we insert a difference of alphabets @xmath241 and @xmath242 then we obtain the following product formula : @xmath243 this gives us the weighted shifted macmahon formula : the weighted generating formula for plane overpartitions is @xmath244 [ propweighted ] substituting @xmath245 and @xmath246 in the preceding product formula , we get @xmath247 * remark . * this proof was suggested to the authors by one of the anonymous referees . in this section , we construct a bijection between the set of all reverse plane overpartitions and sets of nonintersecting paths whose endpoints are not fixed . we use this bijection and stembridge s results @xcite to obtain a pfaffian formula for the generating function for reverse plane overpartitions of a given shape . evaluating the pfaffian we obtain the hook formula for reverse plane overpartitions due to okada @xcite . let @xmath54 be the set of all reverse plane partitions of shape @xmath13 . [ okada ] the generating function for reverse plane overpartitions of shape @xmath13 is @xmath55 we construct a weight preserving bijection between reverse plane overpartitions and sets of nonintersecting paths on a triangular lattice in a similar fashion as in section [ 3.1 ] . the lattice consists of east , north and north east edges . east edges have weight @xmath38 , north edges on ( vertical ) level @xmath10 have weight @xmath248 and north east edges joining vertical levels @xmath10 and @xmath167 have weight @xmath248 . the weight of a set of nonintersecting paths @xmath249 is the product of the weights of their edges and is denoted by @xmath250 . let @xmath5 be a reverse plane overpartition whose positive entries form a skew shape @xmath251 and let @xmath252 . then @xmath5 can be represented by a set of @xmath22 nonintersecting lattice paths such that * the departure points are @xmath253 and * the arrivals points are @xmath254 , for a large enough @xmath118 and @xmath255 . for example let @xmath175 , @xmath256 and @xmath257 . figure [ fig44 ] shows the corresponding set of nonintersecting paths for the reverse plane overpartition of shape @xmath137 @xmath258 [ htp ! ] this implies that all reverse plane overpartitions of shape @xmath259 can be represented by nonintersecting lattice paths such that * the departure points are an @xmath193element subset of @xmath260 and * the arrivals points are @xmath261 with @xmath262 . + now , for @xmath263 we define @xmath264 where @xmath265 is the set of all nonintersecting paths joining an @xmath193-element subset of @xmath266 with @xmath267 . note that for @xmath268 we have @xmath269 by stembridge s pfaffian formula for the sum of the weights of nonintersecting paths where departure points are not fixed ( theorem 3.1 of @xcite ) we obtain @xmath270 where if @xmath193 is even @xmath271 is the @xmath272 skew symmetric matrix defined by @xmath273 for @xmath274 and if @xmath193 is odd @xmath271 is the @xmath275 skew symmetric matrix defined by @xmath273 for @xmath276 and @xmath277 for @xmath278 . let @xmath279 . then @xmath280 from lemma [ lemaop ] we have that the generating function for overpartitions with at most @xmath22 parts is @xmath281 and the generating function for overpartitions with exactly @xmath22 parts is @xmath282 . this implies that @xmath283 moreover , @xmath284 we prove ( [ forzaf2 ] ) by induction on @xmath216 . the formula for the base case @xmath285 holds by the above . so , we assume @xmath286 . by lindstrm s determinantal formula ( lemma 1 of @xcite ) we have that @xmath287 summing over @xmath288 and using ( [ pz ] ) we obtain @xmath289 then @xmath290 it is enough to prove that @xmath291 and ( [ forzaf2 ] ) follows by induction . now , @xmath292 using the inductive hypothesis for @xmath293 we obtain ( [ kolic ] ) . let @xmath294 be the generating function for reverse plane overpartitions of shape @xmath13 . then , using the bijection we have constructed , we obtain @xmath295 which , after applying stembridge s result , gives us a pfaffian formula . this pfaffian formula can be expressed as a product after the following observations . let m be @xmath296 skew symmetric matrix . one of definitions of the pfaffian is the following : @xmath297 where the sum is over all of perfect matchings ( or fixed point free involutions ) of @xmath298 $ ] . also , for @xmath299 , @xmath300 then @xmath301 where @xmath302 if @xmath303 and @xmath304 if @xmath305 ( see ( [ rekw ] ) in this case ) . inductively we obtain theorem [ okada ] . in @xcite a measure on ( diagonally ) strict plane partitions was studied . strict plane partitions are plane partitions were all diagonals are strict partitions , i.e. strictly decreasing sequences . they can also be seen as plane overpartitions where all overlines are deleted . there are @xmath306 different plane overpartitions corresponding to the same strict plane partition @xmath5 , where @xmath307 is the number of connected components of @xmath5 . alternatively , a strict plane partition can be seen as a subset of @xmath308 consisting of points @xmath309 where @xmath118 is a part of the diagonal partition indexed by @xmath4 . see figure [ fig5 ] . we call this set the 2-dimensional diagram of that strict plane partition . the connected components are connected sets ( no holes ) on the same horizontal line . the 2-dimensional diagram of a plane overpartition is the 2-dimensional diagram of its corresponding strict plane partition . [ htp ! ] the measure studied in @xcite assigns to each strict plane partition a weight equal to @xmath310 . the limit shape of this measure is given in terms of the ronkin function of the polynomial @xmath311 and it is parameterized on the domain representing half of the amoeba of this polynomial . this polynomial is also related to plane tilings with dominoes . this , as well as some other features like similarities in correlation kernels @xcite suggested that a connection between this measure and domino tilings is likely to exist . alternatively , one can see this measure as a uniform measure on plane overpartitions , i.e. each plane overpartition @xmath5 has a probability proportional to @xmath312 . in section [ nonint ] we have constructed a bijection between plane overpartitions and sets of nonintersecting paths . see figure [ fig6 ] . this figure is obtained from figure [ fig4 ] by a rotation . note that the paths are incident with all points in the corresponding 2-dimensional diagram . the paths consist of edges of three different kinds : horizontal ( joining @xmath309 and @xmath313 ) , vertical ( joining @xmath314 and @xmath309 ) and diagonal ( joining @xmath314 and @xmath313 ) . there is a standard way to construct a tiling with dominoes using these paths(see for example @xcite ) . we explain the process below . an example is given on figure [ fig6 ] . [ htp ! ] we start from @xmath315 and color it in a chessboard fashion such that @xmath316 , @xmath317 , @xmath318 and @xmath319 are vertices of a white square . so , the axes of this infinite chessboard form angles of @xmath320 and @xmath321 degrees with the axes of @xmath315 . a domino placed on this infinite chessboard can be one of the four types : @xmath322 , @xmath323 , @xmath324 or @xmath325 , where we say that a domino is of type @xmath326 if @xmath326 is the vector parallel to the vector whose starting , respectively end point is the center of the white , respectively black square of that domino . now , take a plane overpartition and represent it on this chessboard by its corresponding set of nonintersecting paths . we cover each edge by a domino that satisfies that the endpoints of that edge are midpoints of sides of the black and white square of that domino . in that way , we obtain a tiling of a part of the plane with dominoes of three types : @xmath322 , @xmath323 and @xmath325 . more precisely , horizontal edges correspond to @xmath322 dominoes , vertical to @xmath323 and diagonal to @xmath325 . to tile the whole plane we fill the rest of it by dominoes of the fourth , @xmath324 type . see figure [ fig6 ] for an illustration . in this way , we have established a correspondence between plane overpartitions and plane tilings with dominoes . we now give some of the properties of this correspondence . first , we describe how a tiling changes when we add or remove an overline or we add or remove a box from a plane overpartition . we require that when we add / remove an overline or a box we obtain a plane overpartition again . in terms of the 2-dimensional diagram of a plane partition , adding an overline can occur at all places where @xmath309 is in the diagram and @xmath313 is not . adding an overline at @xmath309 means that a pair of horizontal and vertical edges , @xmath327 and @xmath328 , is replaced by one diagonal edge , @xmath329 . this means that the new tiling differs from the old one by replacing a pair of @xmath322 and @xmath323 dominoes by a pair of @xmath325 and @xmath324 dominoes . see figure [ fig7 ] . [ htp ! ] removing an overline is the inverse of adding an overline . we now explain the operation of removing a box . observe that if a box can be removed from a plane overpartition then the corresponding part is overlined or it can be overlined to obtain a plane overpartition again . so , it is enough to consider how the tiling changes when we remove an overlined box since we have already considered the case of adding or removing an overline . if we remove an overlined box we change a diagonal edge to a pair of vertical and horizontal edges . if the box was represented by @xmath309 in the 2-dimensional diagram then the edge @xmath329 is replaced by the pair of @xmath330 and @xmath331 . this means that the new tiling differs from the old one by replacing a pair of @xmath325 and @xmath324 dominoes by a pair of @xmath322 and @xmath323 dominoes . see figure [ fig8 ] . [ htp ! ] all four operations are described by a swap of a pair of adjacent @xmath322 and @xmath323 dominoes and a pair of adjacent of @xmath325 and @xmath324 dominoes . we conclude this section by the observation that plane overpartitions of a given shape @xmath13 and whose parts are bounded by @xmath22 are in bijection with domino tilings of the rectangle @xmath332\times [ 0,n]$ ] with certain boundary conditions . these conditions are imposed by the fact that outside of this rectangle nonintersecting paths are just straight lines . we describe the boundary conditions precisely in the proposition below . the set @xmath333 of all plane overpartitions of shape @xmath13 and whose largest part is at most n is in bijection with plane tilings with dominoes where a point @xmath334 is covered by a domino of type @xmath323 if * @xmath335 * @xmath336 and @xmath337 * @xmath338 for some @xmath10 and @xmath339 and a domino of type @xmath324 if * @xmath340 * @xmath341 and @xmath342 * @xmath343 for all @xmath10 and @xmath344 the boundary conditions for @xmath345 and @xmath346 are shown in figure [ fig9 ] . the example from figure [ fig6 ] satisfies these boundary conditions . in this section we are going to give a bijection between certain matrices and pairs of plane overpartitions of the same shape . this bijection is obtained by an algorithm similar to the algorithm rs2 of berele and remmel @xcite which gives a bijection between matrices and pairs of @xmath347-semistandard tableaux . these tableaux are super semistandard tableaux , defined in section 3 , where the nonoverlined ( resp . overlined ) parts are less than or equal to @xmath161 ( resp . @xmath193 ) . we then apply properties of this algorithm to enumerate plane overpartitions , as done by bender and knuth @xcite for plane partitions . let @xmath348 be a @xmath349 matrix , made of four @xmath350 blocks @xmath351 and @xmath271 . the blocks @xmath62 and @xmath271 are nonnegative integer matrices , and @xmath352 and @xmath7 are @xmath353 matrices . we denote the set of all such matrices with @xmath354 . we represent a matrix in @xmath354 by a sequence of pairs of numbers @xmath355 , @xmath356 , @xmath357 and @xmath358 . the encoding of @xmath359 into pairs is made using the following rules : * for each nonzero entry @xmath360 of @xmath62 , we create @xmath360 pairs @xmath355 , * for each nonzero entry @xmath361 of @xmath352 , we create one pair @xmath356 , * for each nonzero entry @xmath362 of @xmath7 , we create one pair @xmath357 , * for each nonzero entry @xmath363 of @xmath271 , we create @xmath363 pairs @xmath358 . for example let @xmath364 . after encoding @xmath359 , we obtain @xmath365 0.5 cm from now on , we fix the order @xmath366 . we sort the pairs to create a two line array @xmath367 such that * the first line is a nonincreasing sequence , * if two entries of the first line are equal and overlined ( resp . nonoverlined ) then the corresponding entries in the second line are in weakly increasing ( resp . decreasing ) order . for the example above , after sorting , we obtain the two line array @xmath368 we now describe the _ insertion algorithm_. it is based on an algorithm proposed by knuth in @xcite and quite similar to the algorithm rs2 of @xcite . we first explain how to insert a number @xmath288 into an overpartition @xmath13 of length @xmath193 , so that the sequence @xmath13 is still an overpartition . if @xmath288 can be inserted at the end of @xmath13 , then insert and stop otherwise , find the smallest part that can be replaced by @xmath288 . bump this part and insert @xmath288 . at the end we obtain a new overpartition that contains @xmath288 as a part and whose length is @xmath193 if we bumped a part or @xmath369 if @xmath288 was added at the end and no part was bumped . for example if @xmath370 then * insert 5 , obtain @xmath371 and bump 4 , * insert 3 , obtain @xmath372 and bump @xmath373 , * insert @xmath373 , obtain @xmath374 and bump @xmath373 , * insert 1 , obtain @xmath375 and bump nothing . to insert @xmath288 into a plane overpartition @xmath33 insert @xmath288 in the first row . if nothing is bumped then stop . otherwise , insert the bumped part in the second row . if something is bumped from the second row , insert it in the third and so on . stop when nothing is bumped . for example when @xmath373 is inserted in @xmath376 we define how to insert a pair @xmath355 into a pair of plane overpartitions of the same shape @xmath377 . we first insert @xmath288 in @xmath33 with the insertion algorithm . if the insertion ends in column @xmath31 and row @xmath30 of @xmath33 , then insert @xmath10 in column @xmath31 and row @xmath30 in @xmath34 . finally going from the two line array @xmath367 to pairs of plane overpartitions of the same shape works as follows : start with two empty plane overpartitions and insert each pair of @xmath367 going from left to right . this is identical to the classical rsk algorithm @xcite . continuing with the previous example and applying the insertion algorithm we get @xmath378 let @xmath56 be the set of all plane overpartitions with the largest entry at most @xmath22 . there is a one - to - one correspondence between matrices @xmath379 and pairs of plane overpartitions of the same shape @xmath380 . this correspondence is such that : * @xmath161 appears in @xmath33 exactly @xmath381 times , * @xmath382 appears in @xmath33 exactly @xmath383 times , * @xmath161 appears in @xmath34 exactly @xmath384 times , * @xmath382 appears in @xmath34 exactly @xmath385 times . the proof is identical to the proof in the case of the rsk algorithm @xcite or the rs2 algorithm @xcite . details are given in @xcite . if the insertion algorithm produces @xmath377 with input matrix @xmath359 , then the insertion algorithm produces @xmath386 with input matrix @xmath387 . [ insalg ] the proof is again analogous to the one in @xcite . given a two line array @xmath388 , we partition the pairs @xmath389 in classes such that @xmath390 and @xmath391 are in the same class if and only if : * @xmath392 and if @xmath393 , then @xmath394 is overlined and * @xmath395 and if @xmath396 , then @xmath397 is overlined . then one can sort each class so that the first entries of each pair appear in nonincreasing order and then sort the classes so that the first entries of the first pair of each class are in nonincreasing order . for example if the two line array is @xmath398 , we get the classes @xmath399 @xmath400 if the classes are @xmath401 with @xmath402 then the first row of @xmath33 is @xmath403 and the first row of @xmath34 is @xmath404 moreover one constructs the rest of @xmath33 and @xmath34 using the pairs : @xmath405 one can adapt the proof of lemma 1 of @xcite for a complete proof . this is done in the master thesis of the second author @xcite . as the two line array corresponding to @xmath387 is obtained by interchanging the two lines of the array and rearranging the columns , the theorem follows . this implies that @xmath406 if and only if @xmath407 . there is a one - to - one correspondence between symmetric matrices @xmath408 and plane overpartitions @xmath409 . in this correspondence : * @xmath161 appears in @xmath33 exactly @xmath381 times , and * @xmath382 appears in @xmath33 exactly @xmath383 times . [ symmetric ] we can get the generating function for plane overpartitions whose largest entry is at most @xmath22 from theorem [ symmetric ] . by this bijection , if @xmath359 is a symmetric matrix of size @xmath349 with blocks @xmath351 and @xmath271 , each of size @xmath350 corresponding to a plane partition @xmath410 , we have that @xmath411 as @xmath359 is symmetric , i.e. @xmath412 , @xmath413 and @xmath414 , we can express the above formulas as @xmath415 and @xmath416 let @xmath417 . we are now ready to prove theorem [ boundedparts ] which states that : @xmath418 indeed , @xmath419 when @xmath22 tends to infinity we get back the weighted generating function of proposition [ propweighted ] . we can also get this result with the method of the proof of proposition [ propweighted ] with the substitution @xmath245 for @xmath420 and 0 otherwise and @xmath246 for @xmath421 and 0 otherwise . we can also get some more general results , as in @xcite . the generating function for plane overpartitions whose parts lie in a set @xmath422 of positive integers is given by : @xmath423 for example , as a corollary we get the generating function for plane overpartition with odd parts . one can get a similar formula for even parts . the generating function for plane overpartitions with odd parts is @xmath424 we want to combine results of @xcite and @xcite to obtain a 1parameter generalization of the formula for the generating function for cylindric partitions related to hall - littlewood symmetric functions . we use the definitions of interlacing sequences , profiles , cylindric partitions , polynomials @xmath3 and @xmath425 given in the introduction . for an ordinary partition @xmath13 we define a polynomial @xmath426 where @xmath138 denotes the number of times @xmath10 occurs as a part of @xmath13 and @xmath427 . for a horizontal strip @xmath428 we define @xmath429 let @xmath430 then @xmath431 for an interlacing sequence @xmath58 with profile @xmath61 we define @xmath432 : @xmath433 where @xmath434 for @xmath58 and @xmath435 such that @xmath436 we define @xmath437 and @xmath438 then @xmath439 for an interlacing sequence @xmath58 with profile @xmath440 we define the reverse @xmath441 with profile @xmath442 . then @xmath443 for an ordinary partition @xmath13 we construct an interlacing sequence @xmath444 of length @xmath445 , where @xmath72 is obtained from @xmath13 by truncating the last @xmath446 parts . then @xmath447 in @xcite ( propositions 2.4 and 2.6 ) it was shown shown that for a plane partition @xmath5 @xmath448 the following proposition is an analogue of ( [ aphiplane ] ) for interlacing sequences . if @xmath58 is an interlacing sequence then @xmath449 if we show that the statement is true for sequences with constant profiles then inductively using ( [ aphiproiz ] ) we can show that it is true for sequences with arbitrary profile . it is enough to show that the statement is true for sequences with @xmath450 profile because of ( [ aphirev ] ) . so , let @xmath58 be a sequence with @xmath451 profile . then @xmath452 is a plane partition and from ( [ aphiproiz ] ) , ( [ aphirev ] ) and ( [ aphiseq ] ) we obtain that @xmath453 and @xmath454 . then from ( [ aphiplane ] ) it follows that @xmath455 . for skew plane partitions and cylindric partitions we obtain the following two corollaries . [ corskew ] for a skew plane partition @xmath5 we have @xmath456 [ corcyl ] if @xmath5 is a cylindric partition given by @xmath457 then @xmath458 . the last corollary comes from the observation that if a cylindric partition @xmath5 is given by a sequence @xmath457 then @xmath459 in the rest of this section we prove generalized macmahon s formulas for skew plane partitions and cylindric partitions that are stated in theorems [ skew ] and [ uvodcyl ] . the proofs of these theorems were inspired by @xcite , @xcite and @xcite . we use a special class of symmetric functions called hall - littlewood functions . in this subsection we introduce weights on sequences of ordinary partitions . for that we use hall - littlewood symmetric functions @xmath33 and @xmath34 . we recall some of the facts about these functions , but for more details see chapters iii and vi of @xcite . we follow the notation used there . recall that hall - littlewood symmetric functions @xmath460 and @xmath461 depend on a parameter @xmath4 and are indexed by pairs of ordinary partitions @xmath13 and @xmath462 . in the case when @xmath99 they are equal to ordinary schur functions and in the case when @xmath37 to schur @xmath33 and @xmath34 functions . the relationship between @xmath33 and @xmath34 functions is given by ( see ( 5.4 ) of ( * ? ? ? * chapter iii ) ) @xmath463 where @xmath51 is given by ( [ formulab ] ) . recall that ( by ( 5.3 ) of ( * ? ? ? * chapter iii ) and ( [ relqp ] ) ) @xmath464 we set @xmath465 and @xmath466 . recall that ( ( 4.4 ) of ( * ? ? ? * chapter iii ) ) @xmath467 a specialization of an algebra @xmath468 is an algebra homomorphism @xmath469 . if @xmath226 and @xmath470 are specializations of the algebra of symmetric functions then we write @xmath471 , @xmath472 and @xmath473 for the images of @xmath460 , @xmath461 and @xmath474 under @xmath226 , respectively @xmath475 . every map @xmath476 where @xmath477 and only finitely many @xmath478 s are nonzero defines a specialization . these specializations are called evaluations . a multiplication of a specialization @xmath226 by a scalar @xmath479 is defined by its images on power sums : @xmath480 if @xmath226 is the specialization of @xmath79 where @xmath481 then by ( 5.14 ) and ( 5.14 ) of ( * ? ? ? * chapter vi ) @xmath482 similarly , @xmath483 let @xmath484 be an integer and @xmath485 be finite sequences of specializations . for a sequence of partitions @xmath486 we set the weight function @xmath487 to be @xmath488 where @xmath234 and @xmath4 are parameters and the sum ranges over all sequences of partitions @xmath489 . we can also define the weights using another set of specializations @xmath490 where @xmath491 . then @xmath492 where @xmath493 we will focus on two special sums , namely @xmath494 and @xmath495 let @xmath496 and @xmath497 be sequences of 0 s and 1 s such that @xmath498 . if the specializations @xmath499 are evaluations given by @xmath500 then by ( [ spec ] ) and ( [ specp ] ) the weight @xmath487 vanishes unless @xmath79 is an interlacing sequence of profile @xmath501 , and in that case @xmath502 then , from corollaries [ corskew ] and [ corcyl ] , we have @xmath503 and @xmath504 where @xmath501 in both formulas is given by the fixed profile @xmath62 of skew plane partitions , respectively cylindric partitions . if @xmath226 is @xmath505 and @xmath470 is @xmath506 then @xmath507 thus , for the specializations @xmath508 , where @xmath509 are given by ( [ evaluations ] ) , we have @xmath510 we now use proposition 2.2 of @xcite ) : @xmath511 this proposition together with ( [ h ] ) implies theorem [ skew ] . the generating function formula for skew plane partitions can also be seen as the generating function formula for reverse plane partitions as explained in the introduction . each skew plane partition can be represented as an infinite sequence of ordinary partitions by adding infinitely many empty partitions to the left and right side . in that way , the profiles become infinite sequences of 0 s and 1 s . theorem [ skew ] also gives the generating function formula for skew plane partitions of infinite profiles @xmath512 : @xmath513 similarly for cylindric partitions , using ( [ simpleh ] ) together with the following proposition we obtain theorem [ uvodcyl ] . [ z ] @xmath514 where @xmath104 is the smallest positive integer such that @xmath105 . we use @xmath515 the proof of this is analogous to the proof of proposition 5.1 ( ( [ zamena ] ) for @xmath37 ) that appeared in @xcite . also , see example 26 of chapter i , section 5 of @xcite . the proof of ( [ formulaforz ] ) uses the same idea as in the proof of proposition 1.1 of @xcite . we start with the definition of @xmath516 : @xmath517 if @xmath518 is a vector we define the shift as @xmath519 . we set @xmath520 , @xmath521 and @xmath522 . if using the formula ( [ zamena ] ) we substitute the sums over the @xmath523 s by the sums over the @xmath524 s we obtain @xmath525 since @xmath526 , if we apply the same trick @xmath70 times , we obtain @xmath527 where @xmath528 . thus , @xmath529 from @xmath530 and @xmath531\\ & = & \prod_{k,\,l=1}^th(q^{(k - l)_{(t)}}r^-_k , r^+_l),\end{aligned}\ ] ] we conclude that ( [ formulaforz ] ) holds . observe that if in theorem [ uvodcyl ] we let @xmath532 , i.e. the circumference of the cylinder tends to infinity then we recover ( [ infiniteskew ] ) . in this paper , we determine the generating functions for plane overpartitions with several types of constraints . in particular , we can compute the generating function for plane overpartitions with at most @xmath30 rows and @xmath31 columns and the generating function for plane overpartitions with entries at most @xmath22 . all the proofs can be done with symmetric functions , but we also highlight combinatorial proofs when these are simple . the natural question is therefore to put those constraints together and to compute the generating function of plane overpartitions with at most @xmath30 rows , @xmath31 columns and entries at most @xmath22 . unfortunately , this generating function can not be written in terms of a product in the style of theorems 18 . for example , when @xmath533 , this generating function can be written as @xmath534 computer experiments show that there is a large irreducible factor in the product . let @xmath215 be the alphabet @xmath535 and @xmath220 be the alphabet @xmath536 . we can write this generating function as @xmath537 where the sum is taken on partitions @xmath13 such that all the parts of @xmath19 have the same parity as @xmath30 . in section 6 we compute generating functions for cylindric partitions . their generating functions are elegant products . in the case @xmath99 ( cylindric partitions and @xmath37 cylindric overpartitions , one could certainly adapt the ideas of gansner @xcite to give a constructive proof of the result . it should be interesting to generalize the combinatorial techniques used in this paper in sections 3 and 5 to understand the combinatorics of plane partitions for general @xmath4 . n. hoffman and j. f. humphreys , _ projective representations of the symmetric groups . @xmath34-functions and shifted tableaux _ , oxford mathematical monographs . oxford science publications . the clarendon press , oxford university press , new york , ( 1992 ) , xiv+304 pp . c. krattenthaler , _ a bijective proof of the hook - content formula for super schur functions and a modified jeu de taquin _ , the foata festschrift . j. combin . 3 ( 1996 ) , no . 2 , research paper 14 , 24 pp . ( electronic ) a. lascoux , _ symmetric functions and combinatorial operators on polynomials _ , cbms regional conference series in mathematics , 99 . published for the conference board of the mathematical sciences , washington , dc ; by the american mathematical society , providence , ri , ( 2003 ) . xii+268 pp . i. g. macdonald , _ symmetric functions and hall polynomials _ , second edition . with contributions by a. zelevinsky . oxford mathematical monographs . oxford science publications . the clarendon press , oxford university press , new york , ( 1995 ) . x+475 s. okada , _ @xmath538-deformations of multivariate hook product formulae _ , journal of algebraic combinatorics ( march 2010 ) 118 ( electronic ) a. okounkov and n. reshetikhin , _ correlation function of schur process with application to local geometry of a random 3-dimensional young diagram _ , j. amer . math . ( 2003 ) , no . 3 , 581603 . r. stanley , _ enumerative combinatorics _ , 2 . with a foreword by gian - carlo rota and appendix 1 by sergey fomin . cambridge studies in advanced mathematics , 62 . cambridge university press , cambridge , ( 1999 ) . xii+581 pp . j. r. stembridge , _ nonintersecting paths , pfaffians , and plane partitions _ , adv . math . 83 ( 1990 ) , 96131 . m. vuleti , _ the shifted schur process and asymptotics of large random strict plane partitions _ , res . not . imrn 2007 , no . i d rnm043 , 53 pp .

following the detection of strong tev-ray flares from the bl lac object 1es 1959 + 650 with the whipple 10 m cherenkov telescope on may 16 and 17 , 2002 , we performed intensive target of opportunity ( too ) radio , optical , x - ray and tev-ray observations from may 18 , 2002 to august 14 , 2002 . observations with the x - ray telescope _ rxte _ _ ( rossi x - ray timing explorer ) _ and the whipple and hegra _ ( high energy gamma ray astronomy ) _ -ray telescopes revealed several strong flares , enabling us to sensitively test the x - ray/-ray flux correlation properties . although the x - ray and-ray fluxes seemed to be correlated in general , we found an `` orphan ''-ray flare that was not accompanied by an x - ray flare . while we detected optical flux variability with the boltwood and abastumani observatories , the data did not give evidence for a correlation between the optical flux variability with the observed x - ray and-ray flares . within statistical errors of about 0.03 jy at 14.5 ghz and 0.05 jy at 4.8 ghz , the radio fluxes measured with the university of michigan radio astrophysical observatory ( umrao ) stayed constant throughout the campaign ; the mean values agreed well with the values measured on may 7 and june 7 , 2002 at 4.9 ghz and 15 ghz with the very large array ( vla ) , and , at 4.8 ghz with archival flux measurements . after describing in detail the radio , optical , x - ray and-ray light curves and spectral energy distributions ( seds ) we present initial modeling of the sed with a simple synchrotron self - compton ( ssc ) model . with the addition of another tev blazar with good broadband data , we consider the set of all tev blazars to begin to look for a connection of the jet properties to the properties of the central accreting black hole thought to drive the jet . remarkably , the temporal and spectral x - ray and-ray emission characteristics of tev blazars are very similar , even though the masses estimates of their central black holes differ by up to one order of magnitude .

astro-ph0310158

the egret _ ( energetic gamma ray experiment telescope ) _ detector on board of the _ compton gamma - ray observatory _ discovered 100 mev@xmath11 gev @xmath0-ray emission from 66 blazars , mainly from flat spectrum radio quasars and unidentified flat spectrum radio sources @xcite . ground - based cherenkov telescopes discovered tev @xmath0-ray emission from 6 blazars , 4 of which are not egret sources . the electromagnetic emission of these active galactic nuclei ( agns ) is dominated by a non - thermal continuum with a low - energy synchrotron component and a high - energy inverse compton component ( see coppi ( 1999 ) , sikora & madejski ( 2001 ) , krawczynski ( 2003a ) for recent reviews ) . the tev sources all belong to the class of bl lac objects , blazars with relatively low luminosity but with spectral energy distributions ( seds ) that peak at extremely high energies . in the case of tev blazars , the large detection area of cherenkov telescopes of several times 10@xmath2 m@xmath3 makes it possible to assess @xmath0-ray flux variations on time scales of minutes . as the kev x - ray and tev @xmath0-ray emission from these sources is probably produced by electrons of overlapping energy ranges as synchrotron and inverse compton emission , respectively , observations of rapid flux and spectral variability in both bands complement each other ideally . the observations can thus be used to constrain and , in principal , even over - constrain models . more specifically , the x - ray and tev @xmath0-ray observations yield a measurement of the jet doppler factor @xmath4 and the jet magnetic field @xmath5 at the jet base . observations of tev blazars can thus reveal key information about the astrophysics of mass - accretion onto supermassive black holes and the formation of agn jets . unfortunately , the interpretation of the tev @xmath0-ray data is not unambiguous owing to the highly uncertain extent of extragalactic absorption of tev @xmath0-rays in pair - production processes with photons of the cosmic infrared background ( cib ) and the cosmic optical background ( cib ) . although x - ray and @xmath0-ray observations of tev - blazars might ultimately be used to measure the cib / cob , a considerable number of sources is needed as it is difficult to disentangle source physics and cib / cob absorption for individual sources @xcite . + owing to its hard x - ray synchrotron emission and low redshift ( @xmath6 0.047 ) , the bl lac object 1es 1959 + 650 had long been considered a prime - candidate tev @xmath0-ray source ( e.g. stecker , de jager & salamon 1996 , costamante & ghisellini 2002 ) . the `` utah seven telescope array '' collaboration reported the detection of tev @xmath0-ray emission from the source with a total statistical significance of 3.9 @xmath7 @xcite . the average flux measured during the 1998 observations was about that from the crab nebula . motivated by the x - ray properties , the telescope array detection , and a tentative detection of the source by the hegra cherenkov telescopes in 2000 and 2001 , we proposed pre - approved pointed _ rxte _ target of opportunity observations . these observations were to take place immediately after a predefined increase in the x - ray or gamma - ray activity was detected with the _ rxte _ all sky monitor ( asm ) or the whipple 10 m cherenkov telescope . following the detection of a spectacular tev @xmath0-ray flare on may 17 , 2002 with the whipple 10 m telescope by the veritas _ ( very energetic radiation imaging telescope array system ) _ collaboration we invoked the pointed _ rxte _ observations as well as simultaneous observations in the radio , optical , and tev @xmath0-ray bands . the whipple @xcite and hegra @xcite data showed that the @xmath0-ray flux was strongest during the first 20 days of observations with peak fluxes of between 4 and 5 crab units ; subsequently , the flare amplitude decreased slowly . following mrk 421 ( @xmath6 0.031 ) and mrk 501 ( @xmath6 0.034 ) , 1es 1959 + 650 is now the third tev @xmath0-ray blazar with a high - state flux much stronger than that from the crab nebula , allowing us to measure the @xmath0-ray lightcurve on a time scale of a couple of minutes , and to take energy spectra with good photon statistics on a nightly basis . since the discovery of the first tev blazar mrk 421 in 1992 @xcite , the number of well established blazars has now grown to 6 ( see table [ blazars ] ) . [ asm ] shows the 2 - 12 kev flux from these 6 sources as measured in the years 1996 to 2003 with the _ rxte _ asm . for mrk 421 , mrk 501 , 1es 1959 + 650 , and pks 2155 - 304 long flaring phases extending over several weeks can be recognized . while mrk 421 , 1es 1959 + 650 , and pks 2155 - 304 flare frequently , mrk 501 flared in 1997 , but showed only modest fluxes thereafter . the prolonged flaring phases offer ideal opportunities to study these objects with high photon statistics . + in this paper , we discuss the results of the 2002 multiwavelength campaign on 1es 1959 + 650 . we present new radio , optical and _ rxte _ x - ray data taken between may 16 , 2002 and august 14 , 2002 , and combine these data with the already published whipple and hegra tev @xmath0-ray data . in sect . [ data ] we present the data sets and the data reduction methods . in sect . [ overview ] we give an overview of the combined light curves , and in sect . [ lightcurves ] we scrutinize certain episodes of the light curves in more detail . after discussing the flux correlations in different energy bands in sect . [ sed ] , we present the radio to @xmath0-ray seds of 1es 1959 + 650 and show results of initial modeling with the data in sect . [ seds ] . with the addition of another tev blazar with good broadband data , we consider the set of all tev blazars to begin to look for a connection of the jet properties to the properties of the central engine in sect . we discuss the implications of our observations in sect . [ disc ] . we use the following cosmological parameters @xmath8 @xmath9 100 km s@xmath10 mpc@xmath10 with @xmath11 , @xmath12 0.3 , and @xmath13 0.7 . the redshift of 1es 1959 + 650 translates into a luminosity distance of 229.5 mpc . errors on the best - fit results of @xmath14-fits to the _ rxte _ data are given on the 90% confidence level . all other errors are quoted on the 1 @xmath7 confidence level . we used the university of michigan 26-meter paraboloid to monitor 1es 1959 + 650 at 4.8 ghz and 14.5 ghz between may 5 and august 9 , 2002 . each observation consisted of a series of on - off measurements taken over a 30 - 40 minute time period . all observations were made within a total hour angle range of about 5 hours centered on the meridian . the calibration and reduction procedures have been described in @xcite . some daily observations were averaged to improve the signal - to - noise ratio . additional flux density measurements were made with the very large array ( vla ) of the national radio astronomy observatory at frequencies of 43.315 ghz , 22.485 ghz , 14.965 ghz , 8.435 ghz , and 4.885 ghz on may 7 and june 7 , 2002 , in snap - shot mode ( single scans ) . observations were made in a - array configuration on may 7 and b - array configuration june 7 . pointing checks were incorporated in an effort to keep the program sources near the centers of the primary antenna beams . the data were reduced within the aips software package supplied by nrao , following the standard procedures outlined in the aips cookbook . flux density calibration was accomplished through observations of the source 3c 286 . some of the flux density measurements at the highest three frequencies were discarded owing to erratic variations among antenna pairs . several secondary calibration sources were used for checks on the final flux density scale . we present two optical data sets . one was taken with the 0.4 m telescope at boltwood observatory ( stittsville , ontario ) between may 18 and august 14 , 2002 , using v , r , and i broadband filters . the aperture photometry was performed with custom software and used the comparison star 4 from villata et al . data points were obtained from averaging over between 4 and 6 2-minute exposures . relative v and r band magnitudes were converted to absolute magnitudes using the published absolute magnitudes from villata et al . we are not aware of a published measurement of the absolute i magnitude of star 4 and we give the results only as relative magnitudes mag(1es 1959 + 650)-mag(star 4 ) . the typical statistical error on the relative photometry of each data point is 0.02 mag . the absolute photometry has an additional error of 0.03 mag . the other data set was taken with the 0.7 m telescope at the abastumani observatory in georgia from may 19 to july 12 , 2002 , using an r filter for all observations . the frames were reduced using ` daophot ii ` . the absolute magnitude of 1es 1959 + 650 was determined by comparison with the standard stars 4 , 6 , and 7 from villata et al . ( 1998 ) . in 20 nights , 192 measurements of 5 minutes exposure time were taken . the statistical error on the relative photometry is 0.1 mag . the absolute photometry has an additional error of 0.05 mag . for both data sets we did not attempt to remove the light contribution from the host galaxy . the x - ray analysis was based on the 3 - 25 kev data from the proportional counter array ( pca ; jahoda et al . 1996 ) on board the _ rxte _ satellite . standard-2 mode pca data gathered with the top layer of the operational proportional counter units ( pcus ) were analyzed . the number of pcus operational during a pointing varied between 2 and 4 . we did not use the 15250 kev data from the high - energy x - ray timing experiment hexte ( rothschild et al . 1998 ) owing to their poor signal to noise ratio . after applying the standard screening criteria and removing by hand abnormal data spikes , the net exposure in each good time interval ranged from 160 secs to 4.43 ksecs ( see table [ xflux ] ) . spectra and lightcurves were extracted with ` ftools ` v5.1`a ` . background models were generated with the tool ` pcabackest ` , based on the _ rxte _ gof calibration files for a `` bright '' source with more than 40 counts / sec . comparison of the background models and the data at energies above 30 kev showed that the model underestimated the background by 10% . we corrected for this shortcoming by scaling the background model with a correction factor of 1.1 . response matrices for the pca data were created with the script ` pcarsp ` v.7.11 . the spectral analysis was performed with the ` sherpa ` v.2.2.1 package . a galactic neutral hydrogen column density of 1.027 @xmath15 was used for all observations . since the analysis is restricted to the energy region above 3 kev the hydrogen column density has only a very minor influence on the estimated model parameters . single power - law models resulted in statistically acceptable fits for all data sets . 1es 1959 + 650 was monitored on a regular basis as part of the bl lac program at the whipple cerenkov telescope during the 2001/2002 observing season ; it was during these observations that 1es 1959 + 650 was seen to go into an active state . following the detection of strong flares on may 16 and 17 , 2002 , we coordinated simultaneous observations of 1es 1959 + 650 with the whipple and hegra cherenkov telescopes . the observations with the whipple 10 m cherenkov telescope began on may 16 , 2002 and ended on july 8 , 2002 @xcite . the total data set consists of 39.3 hrs of on - source data , together with 7.6 hrs of off - source data for background comparison . the whipple telescope is located in southern arizona , usa , on mt . hopkins and is part of the whipple observatory . at this latitude , 1es 1959 + 650 culminates at a zenith angle of 33.5@xmath16 , and so the data were necessarily taken at large zenith angles , between 33.5@xmath16 and 53.5@xmath16 . the data were corrected for large zenith angles and for a temporary reduction of the telescope detection efficiency using the method of lebohec and holder ( 2003 ) which involves measuring the response of the telescope to cosmic rays . while correcting the @xmath0-ray detection rates for the reduced telescope sensitivity is straightforward , energy spectra can not be determined with the standard tools and further studies of the whipple energy spectra are underway . the peak energy - ray detection rate peaks , assuming a source with the same @xmath0-ray spectrum as the crab nebula . ] lies at about 600 gev for the majority of observations . motivated by the hegra detection of the source in 2000 and 2001 , as well as by the strong flaring activity in may , 2002 , the hegra system of five cherenkov telescopes ( canary island la palma ) regularly monitored 1es 1959 + 650 in 2002 . a total of 89.6 hrs of data were taken during moonless nights from may 18 to september 11 , 2002 ( aharonian et al . typically , each night comprises about 1 hour of observation time around the object s culmination . owing to the declination of 1es 1959 + 650 , the object could only be observed at zenith angles above 35.9@xmath16 leading to a mean peak energy of 1.4 tev . all observations were carried out in the so - called wobble mode allowing for a simultaneous measurement of the background rate induced by charged cosmic rays . the hegra collaboration determined the differential 1.3 tev - 12.6 tev energy spectrum @xmath17 @xmath18 @xmath19 of 1es 1959 + 650 for a high - flux data set and a low - flux data set . the high - flux data set used all 2002 data for which the diurnal integral flux above 2 tev surpassed that from the crab and gave @xmath20 ( [email protected]@[email protected]@xmath23 ) 10@xmath24 photons @xmath25 s@xmath10 tev@xmath10 , and @xmath26 ( [email protected]@[email protected]@xmath23 ) . the low - flux data set used all 2000 - 2002 data for which the diurnal integral flux above 2 tev was less than 0.5 crab units and gave @xmath20 ( [email protected]@[email protected]@xmath23 ) 10@xmath27 photons @xmath25 s@xmath10 tev@xmath10 , and @xmath26 ( [email protected]@[email protected]@xmath23 ) . in the following , we quote integral @xmath0-ray flux in crab units above energy thresholds of 600 gev and 2 tev , for the whipple and hegra data points , respectively . in the case of hegra , the analysis threshold has been chosen well above the peak energy to minimize systematic uncertainties in the region of the trigger threshold . the normalization of the fluxes in crab units renders the results largely independent of monte carlo simulations . the drawback of the method is that different energy thresholds can introduce flux offsets if the source energy spectrum deviates from the crab energy spectrum . based on the hegra results on the correlation of the @xmath0-ray flux level and @xmath0-ray photon index , we estimate that these offsets are smaller than 20% for @xmath28 gev flux levels on the order of 1 crab and higher , and smaller than a factor of 2 for flux levels well below 1 crab . based on the whipple measurement of the energy spectrum from the crab nebula @xcite , a flux of 1 crab corresponds to a differential 1 tev flux of ( [email protected]@xmath22@xmath21@xmath29 ) @xmath30 photons @xmath25 s@xmath10 tev@xmath10 and to a @xmath31 flux of ( [email protected]@xmath22@xmath21@xmath32 ) @xmath30 ergs @xmath25 s@xmath10 . figure [ all ] shows from top to bottom the integral tev flux , the x - ray flux at 10 kev , the 3 - 20 kev x - ray photon index , the v , r , and i band optical data , and the 14.5 ghz and 4.8 ghz radio data . the tev @xmath0-ray data ( fig . [ all]a ) show several strong flares during the first 20 days of the campaign with a flux surpassing 2 crab units on may 1720 ( mjd 5241152414 ) and again roughly two weeks later on june 4 ( mjd 52429 ) . subsequently , the flux leveled off to about 0.3 crab units with the exception of two flares on july 1112 ( mjd 52466 - 52467 ) and july 1415 ( mjd 52469 - 52470 ) with a flux between 1 and 1.5 crab units . holder et al . ( 2002 ) studied the fastest @xmath0-ray flux variability time scales based on the whipple data and found a rapid flux increase with an @xmath33-folding time of 10 hrs . the large `` gaps '' in the @xmath0-ray lightcurves originate from the fact that the cherenkov telescopes are operated during moonless nights only . the 10 kev x - ray flux ( fig . [ all]b , table [ xflux ] ) was strongest on may 18 - 20 ( mjd 52412 - 52414 ) . it slowly decreased by a factor of 18.7 from the maximum on may 20 to a minimum on june 17 ( mjd 52442 ) . as we will discuss in more detail in the next section , the tev @xmath0-ray and x - ray fluxes seem to be correlated , with the notable exception of an `` orphan '' tev @xmath0-ray flare on june 4 , 2002 ( mjd 52429 ) that is not associated with increased x - ray activity . from july 17 ( mjd 52469 ) until the end of the campaign , the x - ray flux stayed at a consistently high level : a factor of 1.7 below the maximum flux observed at the beginning of the campaign , and a factor of 11.5 above the minimum flux measured on june 17 ( mjd 52442 ) . the tev emission level during this `` x - ray plateau state '' is about a factor of 2 lower than at similar x - ray flux levels earlier in the campaign . we analyzed the x - ray flux variability time scale by computing the @xmath33-folding times from the flux changes between observations : @xmath34 @xmath35 @xmath36 with @xmath37 being the time difference between two observations and @xmath38 is the difference of the logarithms of the 10 kev fluxes . the shortest @xmath33-folding times are given in table [ inc ] . we detected faster flux increases than flux decreases : the fastest flux increase has an @xmath33-folding time of @xmath395.9 hrs ; the fastest flux decrease has an @xmath33-folding time of @xmath3915.2 hrs . the 3 - 25 kev photon index @xmath40 ( @xmath41 ) ( fig . [ all]c , table [ xflux ] ) varies between 1.6 and 2.4 . the x - ray photon index and the x - ray flux are clearly correlated , higher flux corresponding to harder energy spectrum . values well below and well above the value of @xmath26 2 show that the low - energy ( presumably synchrotron ) component sometimes peaked above 10 kev , and sometimes below 10 kev . we searched for rapid spectral changes by analyzing photon index variations between _ rxte _ observations , see table [ hard ] . the photon index @xmath40 decreased ( spectral hardening ) by up to 0.09 hr@xmath10 and increased ( spectral softening ) by up to 0.04 hr@xmath10 . as a consequence of synchrotron cooling which is more efficient at higher energies , leptonic models predict that the x - ray emission is harder during the rising phase of a flare than during its decaying phase @xcite . careful inspection of the x - ray lightcurve and the photon indices does not show evidence for such a behavior . while the detection of this effect would impose a constraint on the jet magnetic field and the doppler factor , the non - detection allows large regions in the @xmath4 - @xmath5 plane @xcite . the v , r , and i band optical data ( fig . [ all]d - f ) show flux variations of about 0.1 mag on typical time scales of about 10 days . remarkably , the mean optical brightness increased from the first 4 weeks to the last 2 weeks of the campaign by about 0.1 mag in all three optical bands . both , the optical and the x - ray fluxes increased slowly during the campaign . apart from this joint slow flux increase , we did not find any evidence for a correlation between the optical and the x - ray or the tev @xmath0-ray fluxes . we searched for optical intra - day flux variability by fitting models to the data of individual days . although we performed very long observations of up to @xmath397 hrs per night with small statistical errors of 0.02 mag per 10 minute exposure time , the reduced chi - square values did not show any evidence for statistically significant intra - day flux variability . based on the diurnal brightness averages in the three bands , we computed fastest rise and decay @xmath33-folding times of 0.07 mag day@xmath10 and 0.03 mag day@xmath10 , respectively . within the statistical errors , the v - r and v - i colors stay constant throughout the full campaign . the 14.5 ghz and 4.8 ghz radio data taken with the umrao ( fig . [ all]g - h ) do not show significant flux variations . a fit of a constant flux level to the 14.5 ghz data gives a mean of [email protected] jy with a chi - square value of 24.5 for 21 degrees of freedom ( chance probability of 27% ) . the mean flux is consistent with the 14.965 ghz flux of [email protected] jy measured with the vla on may 7 . a fit of a constant flux level to the 4.8 ghz umrao data gives a mean of [email protected] jy with a chi - square value of 7.1 for 7 degrees of freedom ( chance probability of 42% ) . the mean 4.8 ghz flux is compatible with the 4.885 ghz flux measured on june 7 with the vla of [email protected] jy , and with 4.85 ghz values of [email protected] jy and [email protected] jy reported by gregory & condon ( 1991 ) and becker & white ( 1991 ) , respectively . in this section we discuss the lightcurves in more detail by dividing the data into 4 epochs ( epoch 1 : mjd 5241052419 , epoch 2 : mjd 5242052445 , epoch 3 : mjd 5246052474 , epoch 4 : mjd 52486 - 52500 ) . in the following figures , we show only the observational bands where a substantial number of data points were recorded . [ epoch1 ] shows the data from epoch 1 ( may 1625 , mjd 5241052419 ) . note that the x - ray observations started on may 18 , utc 3:26 , within less than 24 hrs of the initial detection of strong @xmath0-ray flaring activity from 1es 1959 + 650 with the whipple 10 m telescope . the @xmath0-ray and x - ray fluxes seem to be correlated , both showing a strong flux increase on may 18 ( mjd 52412 ) and a strong flux decrease on may 21 ( mjd 52415 ) . from may 19 to may 20 , the x - ray flux increases by 20% without a similar increase in the @xmath0-ray band . while the source brightened in the time interval may 1925 ( mjd 52413.2752419.40 ) by about 0.1 mag in all three optical bands , the x - ray flux decreased by a factor of @xmath13 over the same time interval . the data from epoch 2 ( may 26june 21 , mjd 5242052446 ) are shown in fig . [ epoch2 ] . except for one strong @xmath0-ray flare , the @xmath0-ray flux stayed well below 2 crab units . the x - ray flux decayed slowly , and the optical brightnesses in the three bands meandered around their mean values by 0.05 mag . the most interesting feature of the full observation campaign is the strong `` orphan '' @xmath0-ray flare on june 4 ( mjd 52429.30852429.362 ) which fig . [ orphan ] shows in more detail . while hegra measured a low flux of [email protected] crab units on mjd 52429.106 , the whipple observation 5 hrs later revealed a high flux of 4 crab units . the x - ray flux ( measured at the same time as the @xmath0-ray data ) did not show any sign of an increased activity : the 10 kev flux stayed constant and later even decreased compared to the observation taken 5 hrs earlier . similarly , the x - ray photon index and the optical magnitudes do not show any irregularity during the @xmath0-ray flare . the results from epoch 3 ( july 519 , mjd 5246052474 ) are presented in fig . [ epoch3 ] . the @xmath0-ray and x - ray fluxes show a very similar development with joint flux minimums on july 13 ( mjd 52468 ) and july 18 ( mjd 52473 ) , and a joint flux maximum on july 15 ( mjd 52470 ) . as during the full campaign , the x - ray flux and x - ray photon index are tightly correlated . the data from epoch 4 ( july 31august 14 , mjd 5248652500 ) are shown in fig . [ epoch4 ] . while statistical errors hamper the interpretation of the @xmath0-ray data , the x - ray flux varied by 50% and the optical flux by 0.1 mag . the correlation between simultaneously measured @xmath0-ray and x - ray fluxes during the full campaign is shown in fig . [ xgcorr ] ; even though the fluxes seem to be correlated in general , the orphan flare clearly deteriorates the quality of the correlation . during the observation campaign , the 3 - 25 kev x - ray photon index and the 10 kev flux were tightly correlated ( fig . [ hardness ] ) . higher flux levels are accompanied by harder energy spectra , as is typical for bl lac objects . the photon - index / flux correlation shows some slow evolution during the multiwavelength campaign with some exceptionally hard energy spectra recorded during epoch 3 , around july 15 ( mjd 52470 ) . in the hardness intensity plane we did not detect clockwise or anti - clockwise loops during flares . such loops are expected to occur as a consequence of diffusive particle acceleration at strong shocks and synchrotron cooling of the radiating particles ( electrons or protons ) @xcite . the sparse observational sampling might be responsible for our non - detection . the x - ray and @xmath0-ray emission from tev - blazars are commonly attributed to the ssc mechanism in which a population of high - energy electrons emits synchrotron radiation , followed by inverse compton scattering of synchrotron photons to tev energies . in fig . [ sed01 ] we show the radio to @xmath0-ray sed of 1es 1959 + 650 together with a simple one - zone ssc model . the model ( krawczynski 2003b , see inoue & takahara ( 1996 ) , and kataoka et al . ( 1999 ) for similar codes ) assumes a spherical emission volume of radius @xmath42 , that moves with bulk lorentz factor @xmath40 toward the observer . the radiation is doppler shifted by the doppler factor @xmath43^{-1},\ ] ] with @xmath44 the bulk velocity of the plasma in units of the speed of light , and @xmath45 the angle between jet axis and the line of sight in the observer frame . the emission volume is filled with an isotropic electron population and a randomly oriented magnetic field @xmath5 . we assume that the energy spectrum of the electrons in the jet frame can be described by a broken power law with low - energy ( @xmath46 to @xmath47 ) and high - energy ( @xmath47 to @xmath48 ) indices @xmath49 2 and @xmath50 3 , respectively ( @xmath51 from @xmath52 @xmath53 , @xmath54 is the electron energy in the jet frame ) . motivated by the similar seds and flux variability time scales of mrk 501 in 1997 and 1es 1959 + 650 in 2002 , we chose parameter values similar to those inferred for mrk 501 from time dependent modeling of 1997 x - ray and @xmath0-ray data @xcite . the dotted line shows the model prediction before taking into account extragalactic extinction and the solid line shows the sed modified by intergalactic extinction as predicted by a cib / cob model with a reasonable shape . we choose the cib / cob model of kneiske et al . ( 2002 ) , see e.g. primack et al . ( 2001 ) and de jager & stecker ( 2002 ) for alternative detailed model calculations . the parameter values for all subsequent models are given in the respective figure captions . while the model shown in fig . [ sed01 ] gives a satisfactory fit to the x - ray to @xmath0-ray data , it under - predicts the radio and optical fluxes . the model thus suggests that the low - energy radio to optical radiation is dominated by emission from other regions than those that emit the bulk of the x - rays and @xmath0-rays . this finding is consistent with the fact that we found much less flux variability in the radio and optical bands than in the x - ray and @xmath0-ray bands . the inverse compton sed corrected for extragalactic absorption peaks in our model at 1.7 tev . between 100 gev and 400 gev the cib / cob model predicts a characteristic sharp turnover . the next generation cherenkov telescopes cangaroo iii , h.e.s.s . , magic , and whipple should be able to measure such sharp turnovers in blazar energy spectra . we explored several ways to produce the orphan @xmath0-ray flare in the framework of ssc models . given the observed _ rxte _ energy spectrum and our choice of model parameters , it is not possible to produce an orphan @xmath0-ray flare by moving the high - energy cutoff of accelerated electrons to higher energies ( fig . [ orphan1 ] ) . the reason for this behavior is that high - energy electrons that emit synchrotron radiation above the _ rxte _ energy range , emit inverse compton @xmath0-rays at energies above those sampled by the observations ( above @xmath110 tev ) . the additional photons show up at energies above @xmath110 tev . extragalactic extinction reduces the flux above 10 tev already by so much that it is not shown in the figure . adding a low - energy electron population ( fig . [ orphan2 ] , left panel ) succeeds in producing an orphan @xmath0-ray flare and predicts an extremely steep @xmath0-ray energy spectrum . however , the model needs careful fine - tuning , as the density of low - energy electrons is constrained by the optical measurements . studies of the tev @xmath0-ray energy spectrum during the flare are underway to test the prediction of a steep spectrum . a more natural way to explain the flare is to postulate a second , dense electron population within a small emission region ( fig . [ orphan2 ] , right panel ) ; compared to the region where the quiescent emission comes from , the 1200 times larger energy density of this electron population and the 5400 times smaller emission volume lead to a high inverse compton to synchrotron luminosity ratio , and thus to a @xmath0-ray flare without a strong x - ray flare . note that this model does not suffer from a `` compton catastrophe '' . the optical thickness for internal absorption in @xmath55 @xmath56 pair - production processes is well below 1 over the full range of gamma - ray energies covered by the tev observations . in fig . [ sed02 ] we compare the x - ray and @xmath0-ray energy spectra of 1es 1959 + 650 with those of the three other tev blazars with measured tev energy spectra . the x - ray and @xmath0-ray energy spectra of 1es 1959 + 650 are very similar to those of mrk 501 . in comparison to these two sources , the x - ray spectra of mrk 421 are softer , while the @xmath0-ray energy spectra are similar . the x - ray energy spectrum of h1426 + 428 seems to be relatively hard . a meaningful comparison of the high - energy tev @xmath0-ray energy spectra of h 1426 + 428 and the other 3 sources is hampered by the highly uncertain extent of extragalactic absorption for the high - redshift source h 1426 + 428 . with the addition of another tev blazar with good broadband data , we consider the set of all tev blazars to begin to look for a connection of the jet properties to the properties of the central accreting black hole thought to drive the jet . ferrarese & merrit ( 2000 ) and gebhardt ( 2000 ) discovered a close correlation between the mass of the central black holes , @xmath57 , and the host galaxy s stellar velocity dispersion , @xmath58 . the present data on nearby galaxies do not show evidence for an intrinsic scatter of the correlation and the upper limit on the width of the correlation is @xmath59 . the correlation is significantly tighter than that of @xmath57 and the galactic bulge luminosity , @xmath60 . based on both correlations falomo , kotilainen & treves ( 2002 ) and barth et al . ( 2003 ) estimated the black hole masses of several bl lac objects , including 5 of the 6 established tev - blazars . for 1es 1959 + 650 , falomo et al . estimated @xmath61 [email protected] , using the @xmath57-@xmath58 correlation . as can be seen from the black hole masses given in table [ blazars ] , the black hole of 1es 1959 + 650 seems to be the least massive of all tev blazars and is separated by one order of magnitude from the most massive one , mrk 501 . the black hole mass estimates allow us to explore the correlation between mass and parameters describing the jet emission and therefore with the jet properties . in fig . [ mb1]a - f we show the correlation between the black hole mass and 6 parameters that characterize the jet continuum emission : ( i ) the luminosity at the peak of the low - energy ( synchrotron ) emission component ; ( ii ) the frequency at which the low - energy sed peaks ; ( iii ) the range of observed luminosities at ( 1+@xmath62 ) tev ; ( iv ) the 15 tev photon index ; ( v ) the x - ray `` flare duty cycle '' , @xmath63 , defined as the fraction of time during which the _ rxte _ asm flux exceeds 50% of the time averaged flux ( see more detailed description below ) ; and , ( vi ) the range of @xmath0-ray @xmath33-folding time observed so far . the first two parameters describe the sed of the low - energy ( synchrotron ) component ; the second two parameters describe the sed of the high - energy ( inverse compton ) component ; the last two parameters describe temporal properties of the x - ray and the @xmath0-ray emission . the @xmath0-ray parameters were corrected for extragalactic extinction based on the cib / cob model of kneiske et al . the flare duty cycle was computed from the _ rxte _ asm data taken between 1996 and mid 2003 . binning the data into 28-day bins , we determined the fraction of bins where the flux surpassed the mean flux from that source by 50% . we determined error bars on these duty cycles with a monte - carlo simulation , by re - calculating the values 1000 times , modifying each time the flux values according to a gaussian distribution with a width given by the experimental flux errors . the choice of the time binning changes the values of the flare duty cycle , but does not change the results qualitatively . the symbols differentiate the sources ; the solid and dashed error bars show the black hole mass estimates based on stellar velocity dispersion measurements from falomo et al . ( 2002 ) and barth et al . ( 2003 ) , respectively . the dotted error bars show the black hole mass estimates from bulge luminosity measurements ( also from falomo et al . horizontal error bars show the statistical uncertainty on the @xmath64-estimates , and the vertical error bars show the ranges of observed values . differences between parameter ranges can be highly significant from a statistical point of view even if the vertical `` error bars '' span various orders of magnitudes and exhibit a substantial overlap . the figures do not show clear correlations . the only quantity that shows an indication for a correlation is the x - ray flare duty cycle @xmath63 . early ssc modeling of mrk 421 and mrk 501 data indicated that simple one - zone ssc models were capable of describing a wealth of data satisfactorily ( inoue & takahara 1996 , takahashi et al . 2000 , krawczynski et al . 2001 ) . for mrk 501 however , detailed time dependent modeling showed that the very simplest ssc models failed to account for the combined broadband x - ray ( bepposax , _ rxte _ ) and tev @xmath0-ray data @xcite . in order to consistently fit the data from several flares , the authors had to introduce a second emission zone , as well as a poorly justified `` minimum lorentz factor of accelerated electrons '' on the order of @xmath65 10@xmath2 and higher . in this paper we presented evidence for an `` orphan '' @xmath0-ray flare without x - ray counterpart . also this finding contradicts the most simple 1-zone ssc models . there are several ways to explain the orphan flare : * * multiple - component ssc models : * a high density electron population confined to a small emission volume can account for an orphan @xmath0-ray flare ( see sect . [ sed ] ) . `` low duty - cycle '' fast variability has been observed for a number of sources . a prime example is the detection of a strong x - ray flare from mrk 501 with a doubling time of 6 minutes @xcite . such observations strongly suggest that indeed small regions with high electron densities produce strong and rapid flares . alternatively , a second electron population with a low high - energy cutoff might produce an orphan @xmath0-ray flare , as mentioned above . however , the corresponding ssc model requires fine - tuning of the model parameters . as a third possibility a population of electrons with a very hard energy spectrum might produce the gamma - ray flare while it emits synchrotron radiation at energies above those sampled with the _ rxte_. * * external compton models : * in external compton models , the @xmath0-ray flux originates from inverse compton processes of high - energy electrons with radiation external to the jet . variations of the external photon intensity in the jet frame can cause @xmath0-ray flares without lower - energy counterparts . such variations could have different origins : the external photon flux , e.g. from the accretion disk , could be intrinsically variable . alternatively , the motion of the emission region relative to an external photon reflector could result in a time - variable photon flux in the jet frame @xcite . in external compton models , the external photon field is highly anisotropic in the jet frame , owing to the highly relativistic motion of the jet plasma ( @xmath66 ) . as a consequence , the inverse compton emission has a narrower beaming angle than the synchrotron emission and a slight precession of the jet could cause a large change in the tev flux accompanied by a small change of the x - ray flux . * * magnetic field aligned along jet axis : * if the magnetic field in the emission region of the orphan flare is aligned with the jet axis and thus with the line of sight , the observer would not see the synchrotron flare . the electrons however would scatter ssc gamma - rays into our direction and we would thus be able to see the inverse compton flare . * * proton models : * in proton models the low - energy radiation is produced by a population of non - thermal electrons and high - energy radiation by accelerated protons , either directly as synchrotron radiation @xcite , or via a proton induced cascade ( pic ) @xcite . as electron and proton injection rates and high - energy cutoffs may vary in a different way with the plasma conditions , proton models naturally account for orphan flares . in pic models , the tev @xmath0-ray emission originates from a thin surface layer of an optically thick pair plasma , while the x - ray emission originates from the full emission volume . the model naturally accounts for orphan @xmath0-ray flares , as the thin surface layer can produce more rapid flares than the larger x - ray emission region . we consider it unlikely that this latter explanation applies to the observation of the orphan flare from 1es 1959 + 650 , as the x - ray and @xmath0-ray fluxes varied on comparable time scales throughout the rest of the observation campaign . the black hole mass estimates from stellar velocity dispersion measurements allow us to study the connection between the jet emission parameters and the central black hole mass . we expect to find correlations as the characteristic length and time scales of the accretion system scale with @xmath64 ( see e.g. , mirabel et al . our data however , did not reveal any correlations . it is remarkable that 1es 1959 + 650 and mrk 501 show very similar x - ray and @xmath0-ray energy spectra and flux variation time scales while their black hole masses differ by about one order of magnitude . variations of parameters like jet viewing angle , jet magnetic field , or intensity and energy spectrum of the ambient photon field may mask the correlations . furthermore , our correlation plots suffer from the limitations of the observations : flux threshold selection effects , limited energy coverage of the observations , and short time over which the data were acquired ( relative to the lifetime of the jet ) . alternatively , the @xmath64@xmath58 and @xmath64@xmath60 correlations found for nearby galaxies may not hold for blazars , rendering the black hole mass estimates used in our analysis inaccurate ( barth et al . + _ acknowledgements : _ we thank jean swank , david smith and the _ rxte _ gof for their excellent collaboration in scheduling the _ rxte _ observations . we thank the veritas and hegra collaborations for the tev @xmath0-ray light curves and energy spectra . and sh gratefully acknowledge support by nasa through the grant nasa nag5 - 12974 . am acknowledges support by the national science foundation grant ast-0098579 . the university of michigan radio astrophysical observatory ( umrao ) is operated by funds from the university of michigan department of astronomy . we acknowledge helpful comments by an anonymous referee . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + krawczynski , h. , 2003a , new astronomy reviews , in press , astro - ph/0309443 + + + + + + + + + + + + + + + + + + + + + + + + + p1.6cmp0.7cmp2cmp1.5cmp1.5cmp1.5cmp1.3cmp1cmc mrk 421 & 0.031 & punch et al . 1992 , petry et al . 1996 & 17.218.0@xmath67 & 43.243.9@xmath67 & 44.2 ( 450)@xmath68 & 2.12.9 ( 2.53.3)@xmath69 & [email protected]@xmath70 [email protected]@xmath71 & 8.69@xmath71 + mrk 501 & 0.034 & quinn et al . 1996 , bradbury et al . 1997 & 17.319.7@xmath72 & 42.744.2@xmath72 & 44.2 ( 365)@xmath73 & 1.92.4 ( 2.32.8)@xmath74 & [email protected]@xmath70 [email protected]@xmath71 & 9.07@xmath71 + 1es 2344 + 514 & 0.044 & catanese et al . 1998 , tluczykont et al . 2003 & 17.218.7@xmath75 & 42.643.2@xmath75 & 43.4 ( 30)@xmath76 & & [email protected]@xmath70 & + 1es 1959 + 650 & 0.047 & nishiyama et al . 1999 , holder et al . 2003 , aharonian et al . 2003b & 17.019.7@xmath77 & 43.044.5@xmath77 & 44.4 ( 215)@xmath78 & 2.22.6 ( 2.8 - 3.2)@xmath79 & [email protected]@xmath71 & 8.64@xmath71 + pks 2155 - 304 & 0.116 & chadwick et al . 1999 , hinton et al . 2003 & 16.017.0@xmath75 & 44.444.8@xmath75 & 45.1 ( 45)@xmath80 & & & + h 1426 + 428 & 0.129 & horan et al . 2002 , aharonian et al . 2003a & 18.219.2@xmath75 & 43.344.5@xmath75 & 45.1 ( 30)@xmath81 & 0.6 ( 2.2)@xmath82 & & 8.77@xmath71 + cccccc 52412.1406 & 0.45 & 1.550 @xmath21 0.011 & 1.831 @xmath21 0.010 & 0.68 / 45 & 0.95 + 52412.3477 & 0.36 & 2.229 @xmath21 0.014 & 1.772 @xmath21 0.009 & 0.61 / 45 & 0.98 + 52412.9922 & 0.16 & 1.028 @xmath21 0.014 & 2.006 @xmath21 0.018 & 1.46 / 45 & 0.02 + 52413.1406 & 0.12 & 1.865 @xmath21 0.023 & 1.717 @xmath21 0.018 & 0.53 / 45 & 1.00 + 52413.3516 & 0.24 & 2.277 @xmath21 0.018 & 1.710 @xmath21 0.011 & 0.67 / 45 & 0.96 + 52413.7734 & 0.43 & 2.285 @xmath21 0.011 & 1.717 @xmath21 0.007 & 0.93 / 45 & 0.60 + 52414.4609 & 0.40 & 2.937 @xmath21 0.016 & 1.660 @xmath21 0.008 & 1.28 / 45 & 0.10 + 52415.1367 & 0.14 & 1.292 @xmath21 0.017 & 1.878 @xmath21 0.019 & 0.82 / 45 & 0.80 + 52415.3828 & 0.42 & 1.202 @xmath21 0.010 & 1.862 @xmath21 0.011 & 0.51 / 45 & 1.00 + 52416.1211 & 0.16 & 1.151 @xmath21 0.015 & 1.891 @xmath21 0.018 & 0.70 / 45 & 0.93 + 52416.4375 & 0.43 & 1.183 @xmath21 0.010 & 1.797 @xmath21 0.012 & 0.59 / 45 & 0.99 + 52417.1758 & 0.23 & 0.824 @xmath21 0.011 & 1.995 @xmath21 0.018 & 0.45 / 45 & 1.00 + 52417.8672 & 0.24 & 1.155 @xmath21 0.013 & 1.755 @xmath21 0.016 & 0.89 / 45 & 0.68 + 52417.3594 & 0.43 & 0.986 @xmath21 0.008 & 1.909 @xmath21 0.010 & 0.85 / 45 & 0.75 + 52417.4258 & 0.44 & 0.892 @xmath21 0.007 & 1.952 @xmath21 0.010 & 0.49 / 45 & 1.00 + 52418.4141 & 0.44 & 1.485 @xmath21 0.008 & 1.678 @xmath21 0.008 & 0.54 / 45 & 0.99 + 52419.4062 & 0.17 & 0.927 @xmath21 0.012 & 1.860 @xmath21 0.017 & 0.67 / 45 & 0.95 + 52420.4648 & 0.33 & 1.339 @xmath21 0.010 & 1.795 @xmath21 0.010 & 0.56 / 45 & 0.99 + 52421.8477 & 0.08 & 1.166 @xmath21 0.019 & 1.821 @xmath21 0.023 & 0.95 / 45 & 0.57 + 52422.9102 & 0.16 & 1.142 @xmath21 0.015 & 1.891 @xmath21 0.016 & 0.76 / 45 & 0.88 + 52423.3672 & 0.24 & 0.995 @xmath21 0.010 & 1.867 @xmath21 0.013 & 0.75 / 45 & 0.89 + 52424.7891 & 0.19 & 0.801 @xmath21 0.011 & 2.001 @xmath21 0.017 & 0.31 / 45 & 1.00 + 52425.2188 & 0.17 & 0.658 @xmath21 0.011 & 2.041 @xmath21 0.023 & 0.90 / 45 & 0.67 + 52426.2070 & 0.17 & 1.179 @xmath21 0.015 & 1.860 @xmath21 0.018 & 0.64 / 45 & 0.97 + 52427.3047 & 0.46 & 1.147 @xmath21 0.009 & 1.817 @xmath21 0.011 & 1.02 / 45 & 0.43 + 52428.3047 & 0.47 & 1.156 @xmath21 0.009 & 1.856 @xmath21 0.011 & 0.85 / 45 & 0.75 + 52429.0859 & 0.33 & 0.793 @xmath21 0.009 & 1.874 @xmath21 0.016 & 0.77 / 45 & 0.86 + 52429.3008 & 0.27 & 0.793 @xmath21 0.011 & 1.854 @xmath21 0.018 & 0.58 / 45 & 0.99 + 52429.3398 & 0.65 & 0.750 @xmath21 0.007 & 1.913 @xmath21 0.012 & 1.06 / 45 & 0.36 + 52429.6016 & 0.17 & 0.503 @xmath21 0.010 & 2.051 @xmath21 0.026 & 0.92 / 45 & 0.63 + 52430.3477 & 0.49 & 0.577 @xmath21 0.007 & 1.928 @xmath21 0.016 & 0.53 / 45 & 1.00 + 52430.2969 & 0.10 & 0.563 @xmath21 0.014 & 2.033 @xmath21 0.033 & 0.72 / 45 & 0.92 + 52431.0820 & 0.16 & 0.751 @xmath21 0.013 & 1.931 @xmath21 0.023 & 0.49 / 45 & 1.00 + 52431.3164 & 0.93 & 0.789 @xmath21 0.007 & 1.773 @xmath21 0.012 & 1.15 / 45 & 0.22 + 52432.3047 & 0.61 & 0.498 @xmath21 0.007 & 2.041 @xmath21 0.018 & 0.74 / 45 & 0.91 + 52432.1172 & 0.75 & 0.654 @xmath21 0.006 & 1.938 @xmath21 0.012 & 0.46 / 45 & 1.00 + 52432.8359 & 0.26 & 0.521 @xmath21 0.009 & 1.953 @xmath21 0.023 & 0.76 / 45 & 0.88 + 52433.9141 & 0.28 & 0.371 @xmath21 0.007 & 2.119 @xmath21 0.024 & 0.64 / 45 & 0.97 + 52433.2930 & 2.52 & 0.583 @xmath21 0.005 & 1.937 @xmath21 0.011 & 0.52 / 45 & 1.00 + 52434.1562 & 0.85 & 0.477 @xmath21 0.005 & 2.017 @xmath21 0.013 & 0.44 / 45 & 1.00 + 52434.3477 & 0.94 & 0.399 @xmath21 0.003 & 2.087 @xmath21 0.011 & 1.39 / 45 & 0.04 + 52434.6797 & 0.24 & 0.489 @xmath21 0.009 & 2.040 @xmath21 0.023 & 0.72 / 45 & 0.92 + 52435.3398 & 0.94 & 0.430 @xmath21 0.005 & 2.045 @xmath21 0.016 & 0.80 / 45 & 0.82 + 52436.3281 & 0.72 & 0.428 @xmath21 0.006 & 2.105 @xmath21 0.020 & 1.29 / 45 & 0.09 + 52437.3164 & 0.70 & 0.291 @xmath21 0.005 & 2.281 @xmath21 0.024 & 0.99 / 45 & 0.50 + 52438.3047 & 0.68 & 0.226 @xmath21 0.005 & 2.169 @xmath21 0.029 & 0.64 / 45 & 0.97 + 52439.0938 & 0.54 & 0.180 @xmath21 0.004 & 2.345 @xmath21 0.025 & 0.91 / 45 & 0.64 + 52439.3008 & 0.47 & 0.182 @xmath21 0.006 & 2.340 @xmath21 0.041 & 0.62 / 45 & 0.98 + 52439.3672 & 0.47 & 0.170 @xmath21 0.007 & 2.384 @xmath21 0.051 & 0.93 / 45 & 0.61 + 52440.0859 & 0.95 & 0.248 @xmath21 0.003 & 2.183 @xmath21 0.017 & 0.81 / 45 & 0.81 + 52440.2891 & 0.48 & 0.384 @xmath21 0.009 & 2.058 @xmath21 0.030 & 0.73 / 45 & 0.91 + 52440.3516 & 0.56 & 0.426 @xmath21 0.008 & 2.002 @xmath21 0.026 & 0.60 / 45 & 0.98 + 52440.4219 & 0.60 & 0.490 @xmath21 0.007 & 1.942 @xmath21 0.021 & 1.19 / 45 & 0.18 + 52441.0742 & 2.53 & 0.221 @xmath21 0.003 & 2.200 @xmath21 0.015 & 0.79 / 45 & 0.84 + 52441.3359 & 0.30 & 0.219 @xmath21 0.006 & 2.107 @xmath21 0.034 & 0.56 / 45 & 0.99 + 52442.1289 & 0.70 & 0.157 @xmath21 0.003 & 2.335 @xmath21 0.022 & 0.77 / 45 & 0.87 + 52442.3320 & 0.54 & 0.196 @xmath21 0.006 & 2.251 @xmath21 0.042 & 0.98 / 45 & 0.51 + 52442.3984 & 0.58 & 0.212 @xmath21 0.005 & 2.218 @xmath21 0.029 & 0.91 / 45 & 0.65 + 52443.0781 & 1.91 & 0.239 @xmath21 0.003 & 2.148 @xmath21 0.016 & 0.84 / 45 & 0.77 + 52443.3789 & 0.72 & 0.248 @xmath21 0.004 & 2.142 @xmath21 0.020 & 0.95 / 45 & 0.58 + 52444.1055 & 0.96 & 0.479 @xmath21 0.005 & 1.911 @xmath21 0.015 & 0.75 / 45 & 0.89 + 52444.3750 & 0.56 & 0.333 @xmath21 0.005 & 2.134 @xmath21 0.022 & 0.80 / 45 & 0.83 + 52445.4219 & 0.47 & 0.249 @xmath21 0.005 & 2.127 @xmath21 0.024 & 0.94 / 45 & 0.58 + 52467.2266 & 0.39 & 0.961 @xmath21 0.009 & 1.678 @xmath21 0.015 & 0.60 / 45 & 0.98 + 52468.2148 & 0.44 & 0.598 @xmath21 0.007 & 1.885 @xmath21 0.017 & 0.51 / 45 & 1.00 + 52469.2031 & 0.33 & 1.249 @xmath21 0.012 & 1.625 @xmath21 0.014 & 0.68 / 45 & 0.95 + 52470.1406 & 0.20 & 1.883 @xmath21 0.018 & 1.555 @xmath21 0.014 & 0.60 / 45 & 0.99 + 52471.1797 & 0.25 & 1.028 @xmath21 0.012 & 1.734 @xmath21 0.017 & 1.45 / 45 & 0.03 + 52472.1680 & 0.25 & 1.012 @xmath21 0.012 & 1.805 @xmath21 0.017 & 0.68 / 45 & 0.95 + 52473.1562 & 0.28 & 0.963 @xmath21 0.008 & 1.773 @xmath21 0.012 & 0.81 / 45 & 0.82 + 52485.9297 & 0.54 & 1.182 @xmath21 0.005 & 1.812 @xmath21 0.007 & 0.99 / 45 & 0.49 + 52486.9219 & 0.56 & 0.989 @xmath21 0.005 & 1.875 @xmath21 0.008 & 0.98 / 45 & 0.50 + 52487.9688 & 0.76 & 1.089 @xmath21 0.006 & 1.820 @xmath21 0.007 & 0.62 / 45 & 0.98 + 52488.9570 & 0.93 & 1.094 @xmath21 0.006 & 1.824 @xmath21 0.008 & 0.82 / 45 & 0.80 + 52489.9453 & 0.94 & 0.925 @xmath21 0.006 & 1.762 @xmath21 0.009 & 1.09 / 45 & 0.31 + 52490.9336 & 0.94 & 1.226 @xmath21 0.007 & 1.746 @xmath21 0.008 & 0.87 / 45 & 0.71 + 52491.9414 & 0.46 & 1.415 @xmath21 0.010 & 1.759 @xmath21 0.010 & 0.48 / 45 & 1.00 + 52493.9180 & 0.50 & 1.003 @xmath21 0.006 & 1.988 @xmath21 0.008 & 0.87 / 45 & 0.72 + 52494.9727 & 0.50 & 1.017 @xmath21 0.008 & 1.951 @xmath21 0.011 & 0.72 / 45 & 0.92 + 52495.9609 & 0.53 & 1.079 @xmath21 0.008 & 1.973 @xmath21 0.010 & 0.79 / 45 & 0.84 + 52496.9297 & 0.51 & 1.362 @xmath21 0.009 & 1.844 @xmath21 0.010 & 0.82 / 45 & 0.80 + 52497.9297 & 0.60 & 1.185 @xmath21 0.008 & 1.961 @xmath21 0.009 & 0.76 / 45 & 0.88 + 52498.9727 & 0.56 & 1.167 @xmath21 0.009 & 1.981 @xmath21 0.011 & 0.72 / 45 & 0.92 + 52412.15 & 52412.36 & 4.89 & 13.5 @xmath21 0.4 + 52412.36 & 52413.00 & 15.40 & -19.9 @xmath21 -0.4 + 52413.00 & 52413.14 & 3.49 & 5.9 @xmath21 0.2 + 52414.47 & 52415.14 & 16.06 & -19.6 @xmath21 -0.3 + 52417.37 & 52417.43 & 1.57 & -15.6 @xmath21 -1.4 + 52429.30 & 52429.35 & 1.10 & -19.6 @xmath21 -5.6 + 52429.35 & 52429.61 & 6.09 & -15.2 @xmath21 -0.8 + 52432.32 & 52432.13 & 4.49 & -16.5 @xmath21 -1.1 + 52440.11 & 52440.30 & 4.68 & 10.8 @xmath21 0.7 + 52440.30 & 52440.37 & 1.58 & 15.3 @xmath21 4.4 + 52440.37 & 52440.43 & 1.61 & 11.4 @xmath21 2.0 + 52444.12 & 52444.39 & 6.26 & -17.2 @xmath21 -0.9 + cccc 52412.36 & 52413.00 & 15.40 & 0.015 @xmath21 0.001 + 52413.00 & 52413.14 & 3.49 & -0.083 @xmath21 0.007 + 52431.09 & 52431.34 & 5.98 & -0.026 @xmath21 0.004 + 52440.43 & 52441.12 & 16.65 & 0.015 @xmath21 0.002 + 52444.12 & 52444.39 & 6.26 & 0.036 @xmath21 0.004 +

in fluid dynamical simulations in astrophysics , large deformations are common and surface tracking is sometimes necessary . smoothed particle hydrodynamics ( sph ) method has been used in many of such simulations . recently , however , it has been shown that sph can not handle contact discontinuities or free surfaces accurately . there are several reasons for this problem . the first one is that sph requires that the density is continuous and differentiable . the second one is that sph does not have the consistency , and thus the accuracy is zeroth order in space . in addition , we can not express accurate boundary conditions with sph . in this paper , we propose a novel , high - order scheme for particle - based hydrodynamics of compressible fluid . our method is based on kernel - weighted high - order fitting polynomial for intensive variables . with this approach , we can construct a scheme which solves all of the three problems described above . for shock capturing , we use a tensor form of von - neumann - richtmyer artificial viscosity . we have applied our method to many test problems and obtained excellent result . our method is not conservative , since particles do not have mass or energy , but only their densities . however , because of the lagrangian nature of our scheme , the violation of the conservation laws turned out to be small . we name this method consistent particle hydrodynamics in strong form ( cphsf ) .

1701.05316

in fluid dynamical simulations in astrophysics , large deformations are common and surface tracking is sometimes necessary . mesh - free methods , in which particles move following the motion of fluid , are very useful for such simulations . in particular , smoothed particle hydrodynamics ( sph ; @xcite , @xcite ) has been widely used in astrophysics and also in computer - aided engineering . sph is one of lagrangian methods . in sph , we assume that the fluid equation can be expressed by interactions between fluid particles . therefore , sph is not only suitable for simulation of large deformations , but also can satisfy the conservation laws . recently , however , it has become known that sph has several difficulties . for example , it can not handle contact discontinuities ( e.g. @xcite , @xcite ) or fluid surfaces . in our opinion , the standard formation of sph has the following three problems , 1 ) the density distribution must be differentiable ( e.g. @xcite ) . hence , sph can not handle the contact discontinuity properly . 2 ) since the approximation of quantities in sph is of zeroth order in space , sph does not have the consistency to the original partial differential equation ( e.g. @xcite ) . according to the lax equivalence theorem ( @xcite ) , a method , which does not have the consistency , does not converge to the original partial differential equation in the limit of the infinite resolution . 3 ) there is no mathematically sound way to specify boundary conditions in sph , except for the mirror boundary condition . traditionally , fixed particles have been used to express reflecting boundaries such as walls and bottom of a well . they are necessary because sph can not express a sharp cutoff in the density distribution . however , there is no way to let these fixed particles change their physical quantities correctly . thus , smoothed estimate of physical quantities of particles near the boundary contains large errors . to solve problem one , @xcite , @xcite , @xcite and @xcite proposed modified formulation of sph in which the differentiability of the density is not required . their methods can handle the density discontinuity better than standard sph ( hereafter ssph ) does . in previous studies , high - order scheme have been proposed as the solution to problem two . @xcite formulated moving least squares particles hydrodynamics ( mlsph ) based on the moving least square ( mls ) technique . in reproducing kernel particle method ( rkpm ; @xcite ) , the formulation is not mls , but is similar . these methods were applied to inviscid fluid dynamics simulations . however , they have not been applied to large deformations . corrective sph method ( cspm ; @xcite , @xcite , @xcite , @xcite ) is based on the taylor expansion and method of weighted residuals ( mwr ) . they calculated burgers equation , conduction of heat , linear elastodynamics and others with cphm . in addition , @xcite modified this scheme and they called it modified sph ( msph ) . msph was applied to the elastic wave and the diffusion equation . for non - compressible fluid , @xcite introduced high - order formulation into moving particle semi - implicit method ( @xcite ) . in addition , finite particle method ( fpm ; @xcite ) was developed using mwr to handle viscous fluid . they calculated dam break test with fpm . corrected sph ( csph ) in @xcite successfully handled the dam break test of inviscid fluid by re - fitting the density frequently . the frequent re - fitting generates numerical viscosity , and particles move to reduce the number - density deviation . thus , large viscosity occurs in their simulation . in other word , if fluid particles do not move following fluid line to prevent the number - density deviation , large numerical viscosity is induced . there seems to be no high - order scheme without large numerical viscosity for inviscid fluid tested with the calculation of large deformations . @xcite argued that it is difficult to handle large deformations with a high - order scheme . figure [ fig : khi_noise ] shows the result of a simulation of the kelvin - helmholtz instability performed using the high - order mesh - free method presented in section [ sec : form_norm ] . the simulation time is @xmath0 , where @xmath1 is the time scale of the kelvin - helmholtz instability . we can see large deviations of the number - density of particles . these deviations are the result of the fact that each particle move following the fluid motion at its location accurately . small - scale vortex can easily generate highly disordered distribution of particles . this problem is mathematically same as the generation of large density fluctuation in cold keplerian disk ( e.g. @xcite ) . . the right panel is the enlarge image of low - density region of the left.,width=453 ] from the above , it is necessary to rearrange particles , when the large number - density deviation occurs . one potential problem of high - order method is that they do not completely satisfy the conservation law because particles do not have extensive variables ( @xcite , @xcite ) . to construct a high - order mesh - free method which satisfies the conservation laws , we must define fluid particles which have extensive variables . hence , each particle must also have its volume . this volume has to be represented by some physical shape of the particle . if the mass of a particle is constant , its shape of the particle has to change following the fluid deformation . consider an initially spherical particle in velocity field with a uniform shear . it will become elongated very soon , and thus numerical scheme would break down . therefore , no high - order method can satisfy the conservation law completely . however , we found that the error of the conservation law is very small for high - order methods . concerning problem three , we can apply boundary conditions in mathematically meaningful and well - defined way because we approximate the partial differential equation using fitting polynomial for intensive variables in high - order mesh - free method . we call our method consistent particle hydrodynamics in strong form ( cphsf ) . we performed several numerical tests , and results were excellent . in the rest of this paper , we present cphsf ( section [ sec : form ] ) , and report the results of numerical tests ( section [ sec : test ] ) . finally , we discuss and summarize our study ( section [ sec : sumanddis ] ) . as described in introduction , many different high - order approximation methods have been proposed in previous studies . in the following , we call these approximation methods `` shape functions '' , since they can be written as @xmath2 where @xmath3 is the approximation of a spacial partial derivative of function @xmath4 with respect to variables @xmath5 at position @xmath6 , @xmath7 is the value of function @xmath4 at the location of particle @xmath8 , @xmath9 , and @xmath10 is the `` shape function '' of particle @xmath8 . these shape function can be classified by whether or not they are defined as the minimum of the l2 norm of the residual of the fitting . we call shape functions based on the minimization of l2 norm type 1 and others type 2 . in this section , we consider an approximation of function @xmath4 , @xmath11 , which minimizes @xmath12 ^ 2 . \end{aligned}\ ] ] we assume @xmath13 is of @xmath14 class , i.e. @xmath13 is @xmath15-times differentiable . here , @xmath16 is the weight of particle @xmath8 for approximation at particle @xmath17 . to derive shape functions in the way similar to those in @xcite and @xcite , we define @xmath18 where @xmath19 are the approximations of @xmath20 and have the error @xmath21 , and @xmath22 is the dimension of the space . therefore , the right hand side of equation ( [ eq : taylor , normal ] ) matches the taylor expansion around @xmath23 of @xmath24 up to for ( @xmath25)-th order for @xmath26-th order spatial derivatives . if we do not need the interpolation formula for @xmath27 , it is possible to set @xmath28 ( e.g. @xcite ) . however , in this paper , we need the interpolation formula for @xmath27 in section [ sec : form_ref ] . therefore , we do not use this form . the optimal @xmath29 for which @xmath30 takes the extreme , is given by @xmath31 where @xmath32 is a regular matrix . note that if @xmath32 becomes a rank deficient , it means that the set of @xmath7 does not have enough information to derive a unique @xmath29 . therefore , in such a case , we widen the non - zero region of @xmath16 to increase the number of particles in the shape function . in this paper , we set @xmath33 , where @xmath34 is a kernel function that depends on @xmath35 and @xmath36 . here , @xmath36 is a kernel length that denotes the width of the kernel function . in sph , it is usually given by @xmath37 where @xmath38 is a constant coefficient . the parameters @xmath26 and @xmath39 are the mass and the mass - density . from the above , the contribution of @xmath7 to @xmath40 is given by @xmath41_{0 \alpha } { w}_{ij } p_{\alpha , ij } , \end{aligned}\ ] ] and that to derivatives @xmath42 $ ] is @xmath43_{\beta \alpha } { w}_{ij } p_{\alpha , ij } , \end{aligned}\ ] ] where @xmath44_{\alpha \beta}$ ] is element @xmath45 of matrix @xmath32 and @xmath46 is element @xmath47 of vector @xmath48 . note that index @xmath49 takes @xmath50 in equation ( [ eq : my , sf1 ] ) . alternatively , in mlsph , the derivative of equation ( [ eq : my , sf0 ] ) is used as @xmath51 $ ] . in this case , all components of @xmath52 are freely changed to minimize the residual . therefore , the error is smaller than that of equation ( [ eq : my , sf1 ] ) . however , we can get only zeroth order differential shape function , and thus it takes calculation cost to differentiate shape function which consists of a matrix and others . therefore , we use equation ( [ eq : my , sf1 ] ) as a differential shape function . in this paper , we define that the scheme is @xmath15-th order in space , if the first - order spatial derivative is @xmath15-th order . we derived the shape functions using mls in section [ sec : form_norm ] . however , the shape functions of cspm , msph and fpm can not be expressed in terms of mls because these functions are derived from the idea different from mls . in this section , we derive them using mwr . in mwr , the weighted approximation error is set zero . @xmath53 \widehat{w}_{ij } d\boldsymbol{r}_j = 0 . \end{aligned}\ ] ] in this section , we show that the shape functions of cphm , msph and fpm can be derived using the trial function of the form in equation ( [ eq : taylor , normal ] ) . here , since there are @xmath54 unknown parameters , we need to give @xmath54 differential forms of @xmath55 . in these methods , these functions are given by @xmath56 thus , we have @xmath57 \boldsymbol{q}_{ij } d\boldsymbol{r}_j = 0 . \end{aligned}\ ] ] note that the @xmath15-th order differential shape functions in cspm are derived recursively by using zero to @xmath58 order derivatives . therefore , @xmath40 is given by setting @xmath59 , and then the first derivatives are produced by using this @xmath40 and setting @xmath60 . finally , the integral @xmath61 is approximated by summation @xmath62 , and we have @xmath63 where @xmath64 is the `` volume '' of particle @xmath8 . from the above , the contribution of @xmath7 to @xmath40 is given by @xmath65_{0\alpha } v_j p_{\alpha , ij } , \end{aligned}\ ] ] and that of derivatives @xmath66 is @xmath67_{\beta\alpha } v_j p_{\alpha , ij}. \end{aligned}\ ] ] in the following , we use the shape functions of equations ( [ eq : my , sf0 ] ) and ( [ eq : my , sf1 ] ) . many forms of the artificial viscosity have been proposed for sph to capture shocks ( e.g. @xcite , @xcite ) . however , most of them cause unwanted shear viscosity ( e.g. @xcite , @xcite ) . to reduce unwanted shear viscosity , @xcite and @xcite introduced shear switches that reduce viscosity when the shear exists . alternatively , @xcite and @xcite proposed the use of riemann solvers in order not to use the artificial viscosity . @xcite introduced a form of the artificial viscosity different from that of @xcite . this artificial viscosity ( hereafter nrav ) was derived by adding the bulk viscosity to the artificial viscosity formulated by @xcite . this viscosity is applied only when @xmath68 , where @xmath69 is the velocity . @xcite compared many different forms of artificial viscosities , including the usual @xcite type , and nr type , both with and without different forms of shear switches and time - dependent switches . they found the nr type viscosity is the best , when the estimate of @xmath70 is of high order . when a low - order estimate is used , it causes unwanted shear viscosity . these proposed forms of the artificial viscosity can handle fluid with the velocity shear better than the standard artificial viscosity of sph can . the standard sph artificial viscosity is defined for pairs of particles , and thus there is no easy way to apply it to high - order schemes in which particles do not have extensive quantities . moreover , @xcite demonstrated that even for sph schemes , nrav is better than the standard sph artificial viscosity . therefore , we adopted nrav . initially , we used the usual form of nrav extended to multi - dimensional space @xmath71 where @xmath72 is a constant coefficient and @xmath73 is the internal energy . the function @xmath74 is the theta function . we , however , found that this form leads to numerical instability . the reason why the instability takes place is that the pressure - like quantity in nrav in equations ( [ eq : avmom , mdim , org ] ) and ( [ eq : aveng , mdim , org ] ) is isotropic . therefore , these `` pressure '' can and does operate to directions perpendicular to the direction of compression , resulting in the increase of the kinetic energy . therefore , we extend the nrav to multi - dimensions so that the artificial viscosity operates only in the direction of the maximum compression . note that we assume that the number of shock waves at one position is only one . let @xmath75 the coordinate of the direction of a shock wave . @xcite introduced the following the artificial viscosity to equations of moment and energy . @xmath76 to extend this artificial viscosity to multi - dimensions , first , we determine the direction of @xmath75 axis along which the fluid is maximally compressed . second , the artificial viscosity for this @xmath75-direction is calculated . finally , we transform the calculated the artificial viscosity into the original system of coordinates . the strain rate tensor is given by @xmath77 using eigenvalues @xmath78 and eigenvectors @xmath79 , @xmath80 is diagonalized as @xmath81 we assume eigenvalues are ordered , @xmath82 . thus , if @xmath83 , we apply the artificial viscosity in the direction of @xmath84 . here , the transformed coordinate is given by @xmath85 first , we derive the artificial viscosity in the equation of motion . in the transformed coordinate , the artificial viscosity term is given by @xmath86 therefore , in the original coordinate , the artificial viscosity term in the equation of motion is @xmath87 here , by using the quantities in the original coordinate , @xmath88 can be expressed by @xmath89 from the above , the artificial viscosity term in the equation of motion in the original coordinate is given by @xmath90 where @xmath91 is @xmath92 . here , @xmath93 denotes the first element of @xmath94 . next , we derive the artificial viscosity in the equation of energy . in this case , it is the same artificial viscosity for both the original coordinate and the transformed one because the equation of energy is scalar . therefore , @xmath95 in this paper , we also use the bulk viscosity introduced by @xcite . therefore , @xmath96 changes to @xmath97\lambda_1 \theta(-\boldsymbol{\nabla}\cdot\boldsymbol{v } ) , \end{aligned}\ ] ] in this paper , we use @xmath98 and @xmath99 where @xmath100 is a constant coefficient , and @xmath101 is the sound velocity . the introduction of the linear bulk viscosity of equation ( [ eq : avgm , mlutdim , org ] ) implies that the viscosity is active even for infinitesimal compression . in other words , the fluid is viscous even if there is no shock . in order to reduce the viscosity in the absence of the shock , so - called `` bulk switches '' have been proposed ( @xcite , ) . we use the switch @xmath102 multiplied to @xmath96 . the time evolution of @xmath102 is given by @xmath103 where @xmath104 is a parameter which determines the decay time scale of the artificial viscosity after shock . in this paper , we use @xmath105 and @xmath106 , following and @xcite . we discuss the choice of @xmath107 in section [ sec : test_shock ] . finally , we introduce a new term which weakens the artificial viscosity when fluid is compressed in one direction , but is expanding in other directions , @xmath108 where @xmath109 is a positive coefficient . consider the case that @xmath110 and @xmath111 . obviously , @xmath112 , and there is no compression . however , numerically determined @xmath70 can be negative due to truncation errors and in that case equation ( [ eq : avgm , mlutdim , org ] ) can result in the strong artificial viscosity . we can reduce this errortic activation of the artificial viscosity using this term . in this paper , we use @xmath113 . we multiply @xmath96 by @xmath102 and @xmath114 . thus , in our study , we use @xmath115 instead of @xmath96 . as we have discussed in section [ sec : intro ] , the rearrangement of particles is necessary to handle large deformations of inviscid fluid , if we use high - order schemes . therefore , in this section , we describe how to rearrange particles . in this paper , we consider two dimensional cases . first , we consider periodic boundary . in this case , particles are rearranged to the initial pattern . of cause , the initial pattern does not have large deviations of the number density of particles . for example , consider the case in which initial pattern is a grid . if the distribution of particles has become distorted as in the left - hand side panel in figure [ fig : ref , grid ] , we rearrange particles in a grid pattern ( see the right in figure [ fig : ref , grid ] ) . next , we consider the fluid with the surface . in this paper , particles are rearranged to the original grid pattern . consider the case where particles are distributed as in the left - hand side panel of figure [ fig : ref , surf ] . first , we connect surface particles with lines which are expressed as black lines in the central panel of figure [ fig : ref , surf ] . then , particles are placed on the intersections of the black line and the grid line , for example , the black dots in the central panel of figure [ fig : ref , surf ] . if the interval between particles is smaller than @xmath116 , where @xmath117 is the width of a grid , one particle is removed . finally , we put particles on grid points as in the right - hand side panel of figure [ fig : ref , surf ] . the physical quantity , @xmath27 , of a new particle @xmath17 is calculated by using @xmath118 , where @xmath118 is the quantity of particle @xmath8 before rearrangement . @xmath119 finally , we describe the criterion for rearrangement . typically , rearrangement is required when local distribution of particles has become highly anisotropic . consider a simple velocity field with linear shear , such as @xmath120 , applied to initial particle distribution of a tilted cartesian grid ( see the left side panel of figure [ fig : ref , shear ] ) . particles move through shear velocity and the distribution transforms to that given by the central panel of figure [ fig : ref , shear ] . very soon , large difference between distance of particles in one direction and that in the orthogonal direction develops as shown in the right panel in figure [ fig : ref , shear ] . in order to detect this kind of anisotropy , we use an approximate kernel weighted moment tensor defined as @xmath121 where @xmath122 is a sign function , and @xmath123 and @xmath124 are @xmath125 and @xmath126 . the parameter @xmath127 equal to the width of the kernel function @xmath34 . therefore , if @xmath128 is larger than @xmath127 , the value of @xmath34 is zero . the criterion of the rearrangement is given by @xmath129 where @xmath130 and @xmath131 are maximum and minimum eigenvalues of @xmath132 , and @xmath133 is a constant coefficient . note that it is difficult to apply this criterion to the fluid surface , since it lacks particles near the surface . hence , we apply rearrangement at a constant time interval , when we need to follow the surface . in this section , we derive our high - order mesh - free fully lagrangian discretization of equations of continuity , moment and energy . the original set of partial differential equations are given by @xmath134 where @xmath135 is pressure . these equations are discretized using equation ( [ eq : my , sf1 ] ) , and the artificial viscosity is added . hence , we have @xmath136,\\ \label{eq : fluid_eoe } \frac{du_i}{dt } & = & -\frac{1}{\rho_i}\sum_j\left(p_i\boldsymbol{v}_{j}\cdot \boldsymbol{\psi_{ij } } + q_j^{{\rm av}}\lambda_1\right ) . \end{aligned}\ ] ] if the fluid is an ideal gas , the equation of state is given by @xmath137 where @xmath138 is the ratio of specific heat . following @xcite , we use the equation for the water given by @xmath139 + p_{\rm air } , \label{eq : eos_water } \end{aligned}\ ] ] where @xmath140 and @xmath141 are density and pressure at the surface . the constant coefficient @xmath142 is @xmath143 here @xmath144 is the gravitational acceleration and @xmath145 is the height of a fluid . for a weakly compressible fluid , we use the linearized equation of state @xmath146 where @xmath147 . following @xcite and @xcite , the timestep @xmath148 is given by @xmath149}. \end{aligned}\ ] ] we set the constant coefficient @xmath150 to @xmath151 unless we state otherwise . the kernel length @xmath36 is calculated as @xmath152 where @xmath153 and @xmath154 are an initial or rearranged particle spacing and density . in this paper , we set @xmath38 to @xmath155 and @xmath156 for first- , second- , third- and fourth - order fitting formulae in space . in our method , we can express boundary conditions directly , since our method is based on fitting polynomials for intensive variables . first , we consider a free surface . at the free surface , the pressure of the fluid is same as that of thin air or vacuum . thus , the boundary condition is given by @xmath157 where @xmath158 is a constant pressure of the assumed air . now let us consider a fixed slipping boundary . the boundary condition is @xmath159 here , @xmath160 is the velocity perpendicular to the boundary . in the case of the non - slipping boundary , a fluid element should stay at its initial point . therefore , for the non - slip fixed boundary , the boundary condition is @xmath161 finally , at the slipping contact discontinuity , pressure and velocity perpendicular to the boundary are continuous . thus , the boundary conditions of contact discontinuity of fluid 1 and 2 are @xmath162 where @xmath163 and @xmath164 are pressure and velocity perpendicular to the discontinuity of the fluid @xmath17 . the boundary conditions at the non - slipping contact discontinuity are given by @xmath165 where @xmath166 is velocity of the fluid @xmath17 . in this section , we show the results of several numerical tests . first , we compare capabilities of cphsf and ssph to handle the fluid surface by calculating linearize sound modes in section [ sec : test_fs ] . in section [ sec : test_shock ] , the result of the sod shock tube test is presented . here we investigate the errors of the conservation laws . in section [ sec : rotcone ] , we show the result of the rotating cone test . kelvin - helmholtz instability ( section [ sec : test_khi ] ) and rayleigh - taylor instability ( section [ sec : test_rti ] ) are also calculated . they are suitable to survey the capability to handle large deformations and fluid instability . in addition , we investigate if cphsf can handle free surfaces by gravity wave test ( section [ sec : test_gw ] ) and dam break test ( section [ sec : test_dam ] ) . finally , a cold keplerian disk is calculated in section [ sec : test_disk ] . in these tests , we used the backward euler integrator for tests with boundary conditions , and we used a third - order runge - kutta method for other tests . for shape functions , we used the first - order shape function in space unless stated otherwise . the kernel function is the fourth - order wendland function ( @xcite ) . in this section , we investigate how ssph and cphsf handle the fluid surface by analyzing one - dimensional sound wave modes . we assume the fluid is water with equation of state given by equation ( [ eq : eos_water ] ) with @xmath167 and @xmath168 . the position , velocity , density and pressure of particle @xmath17 in the equilibrium are given by @xmath169 , @xmath170 , @xmath171 and @xmath158 . in addition , perturbations are @xmath172 , @xmath173 , @xmath174 and @xmath175 . the computational domain is @xmath176 . the number of particles is 101 . the values of @xmath170 , @xmath171 and @xmath158 are @xmath177 , @xmath178 and @xmath177 for all @xmath17 . first , we derive the linearized equation of the density perturbation in ssph . the fluid equation of ssph is given by @xmath179 note that the value of @xmath180 given by equation ( [ eq : dens , ssph ] ) does not become @xmath178 because the approximation has zeroth - order error . therefore , we derive @xmath181 , which satisfies @xmath182 for all @xmath17 , by implicitly calculating the following equation , @xmath183 alternatively , we could adjust the locations of particles so that equation ( [ eq : unit , ssph ] ) is satisfied for equal - mass particles , but we chose to change mass for simplicity . in this test , mass distribution becomes as figure [ fig : fs_ssphmass ] . ) with the equal particle spacing . , width=226 ] the perturbation equation for equations ( [ eq : x , ssph ] ) to ( [ eq : acc , ssph ] ) is given by @xmath184 where @xmath185 is the sound velocity of the equilibrium state . we derive the linearized equation of the density perturbation in cphsf . the fluid equation of cphsf is given by equations ( [ eq : fluid_eoc ] ) , ( [ eq : fluid_eom ] ) and ( [ eq : fluid_eoe ] ) . the linearized perturbation equation is given by @xmath186 equation ( [ eq : boundary , freesurface ] ) is used for the boundary condition . therefore the value of @xmath174 is set zero at @xmath187 , @xmath188 , where @xmath188 is the width of the fluid and equal @xmath189 in this test . we used cphsf of first- and third - order in space . the perturbation equation at the continuous limit is the wave equation , @xmath190 the boundary condition is @xmath191 at @xmath192 and @xmath193 as that of cphsf . consequently , the general solution is given by @xmath194 where @xmath195 is the amplitude of mode with wave number @xmath196 . figure [ fig : fsana_vl ] shows the eigenvalues of the modes as function of wave number @xmath196 . we can see that the third - order cphsf gives very accurate angular frequency even for large wave numbers . the first - order cphsf and standard sph give similar errors . figure [ fig : fsana_vc ] shows the eigenfunctions of @xmath197 , and figure [ fig : fsana_vc_error ] gives the error of eigenfunctions . from these figures , we can see that the error of eigenfunction calculated with cphsf is much smaller than that with ssph , even when the spatial order of cphsf is one . we can conclude that cphsf is more accurate than ssph , in particular near the boundary , even when the spatial order is low . plotted against the number of the wave @xmath196 for ssph and first and third space - order cphsf.,width=226 ] . the left and central panels are for cphsf with first- and third - order in space . the right is for ssph.,width=453 ] but show the errors of eigenfunctions.,width=453 ] in this section , we present the result of the sod shock tube test ( @xcite ) . first , we determine the value of @xmath107 so that after - shock numerical oscillation is suppressed . second , we investigate the errors of conserved quantities . we assume that fluid is ideal gas with @xmath198 . the computational domain is @xmath199 with a periodic boundary condition , and the initial boundary of two fluids is at @xmath200 . in this test , we use equal - mass particles . the numbers of particles are @xmath201 ( high density region ) and 200 ( low density region ) , and the total number of particles is @xmath178 . initial velocity is given by @xmath202 . the density is smoothed by a polynomial , and it is given by @xmath203 + \frac{\rho_h + \rho_l}{2 } , & { \rm otherwise } \end{array } \right . \end{aligned}\ ] ] where @xmath204 and @xmath205 are the initial density of the high- and low - density regions . we used @xmath206 and @xmath207 . the parameter @xmath208 represents the width of the smoothing region , and the value is given by @xmath209 , where @xmath210 and @xmath211 are the kernel length in the high- and low - density regions . the position of particle @xmath17 in the smoothing region is determined so that they satisfy @xmath212 the smoothed pressure is given by @xmath213 + \frac{p_h + p_l}{2 } , & { \rm otherwise}\\ \end{array } \right.\end{aligned}\ ] ] where @xmath214 and @xmath215 are the initial pressure of the high- and low - density regions . we used @xmath216 and @xmath217 . figure [ fig : sodshock101av ] shows the numerical solution at @xmath218 with first - order cphsf with @xmath219 . the post shock oscillation is rather strong . figure [ fig : sodshock105av ] shows the numerical solution for @xmath220 . in this figure , the post shock oscillation is suppressed . we use @xmath220 for other tests in this paper unless stated otherwise . , but for @xmath220.,width=302 ] , but for @xmath220.,width=302 ] we now investigate the errors in conserved quantities . we used first- and third - order schemes . we set @xmath221 in these tests . in the following , we report the result of two series of test calculations . we varied the number of particles from @xmath222 to @xmath223 for one test and from @xmath222 to @xmath224 for the other . the kernel length is calculated using equations ( [ eq : kernellength , cphsf ] ) and ( [ eq : mass , kernellength , cphsf ] ) . in the first series , we used the nrav of the form described in section [ sec : form_av ] . this means that the strength of av is weaker for higher resolution , so that the number of particles used to resolve shock is approximately constant . in the second series , we fixed the value of @xmath225 for equation ( [ eq : avgm , mlutdim , org ] ) and initial smoothing parameter @xmath208 in equations ( [ eq : dens , shock ] ) and ( [ eq : pres , shock ] ) , so that the physical thickness of the shock is independent of @xmath226 . we used the second series to test the convergence of our scheme . figures [ fig : sodshock1s1n_valh ] to [ fig : sodshock3s1n_conh ] give the calculation result for @xmath222 and @xmath227 , for tests 1 and 2 and first- and third- order schemes . we can see that all four results are good in capturing shocks . the shock is broader for test 1 than for test 2 , since the coefficient for the artificial viscosity is larger for test 1 . on the other hand , weak oscillation is visible in the left - hand - side region of the contact discontinuity , in particular for the third - order scheme . , but the results with third - order cphsf.,width=302 ] , but the results with third - order cphsf.,width=302 ] , but the results with third - order cphsf.,width=302 ] , but the results with third - order cphsf.,width=302 ] in test 1 , we compare results with @xmath228 , @xmath229 , @xmath230 , @xmath223 to the analytical solution . since it is difficult to derive the analytical solution for test 2 , we compare the results with @xmath228 , @xmath229 , @xmath230 , @xmath223 , @xmath224 to that with @xmath231 . we calculate the errors of the total energy and the total momentum in the region @xmath232 as the measure of the conservative quantities . these errors are given by @xmath233 where @xmath234 and @xmath235 are numerical solutions of energy and the momentum , and @xmath236 and @xmath237 are analytical solutions of energy and the momentum ( test 1 ) or the results of @xmath231 calculation ( test 2 ) . the parameters @xmath238 and @xmath239 are the values of @xmath236 and @xmath237 at @xmath218 . energy and the momentum of the numerical solution are calculated by @xmath240 where @xmath241 , @xmath242 and @xmath243 are internal energy , density and velocity of the numerical solution . using the shape functions , this integral is expressed as @xmath244}{2^{m+1}(m+1 ) } \sum_j \left(u_j + \frac{\rho_j}{2}v_{x , j}^2\right ) \psi_{m , ij},\\ p & \simeq & \sum_m^n \sum_{i , |x_i| \leq 0.25 } \frac{\left[\left({x_{i+1 } - x_i}\right)^{m+1 } - \left({x_{i } - x_{i-1}}\right)^{m+1}\right]}{2^{m+1}(m+1 ) } \sum_j \rho_j v_{x , j } \psi_{m , ij } , \end{aligned}\ ] ] where @xmath15 is the spatial order of the scheme . the parameter @xmath245 is the @xmath26-th differential - order shape function . figure [ fig : shockconsvalh ] shows the time evolutions of @xmath246 and @xmath247 with first- and third - order schemes and @xmath228 , @xmath229 , @xmath230 , @xmath223 for test 1 . we can see that the error is proportional to @xmath248 . in test 1 , the initial smoothing length @xmath208 and the strength of the artificial viscosity depend on @xmath248 in a linear way , and thus the errors of the fluid equation and artificial viscosity are of zeroth order . the width of region , where these error occurs , is proportional to @xmath248 . therefore , the total errors at the shock become @xmath249 . and the lower panels show @xmath246 . the left - hand side is the result of first - order cphsf and the right is that of third - order one . , width=453 ] figure [ fig : shockconsconh ] show the time evolutions of @xmath246 and @xmath247 with first- and third - order schemes and @xmath228 , @xmath229 , @xmath230 , @xmath223 @xmath224 for test 2 . we can see that the error is much smaller than those in figure [ fig : shockconsvalh ] . if the result converge to an exact solution following the order of the scheme , the error relation to @xmath250 result should be given by @xmath251 where @xmath15 is the order of the scheme . in figure [ fig : shockconsnn ] , we plot the error at @xmath218 as a function of @xmath226 . we can see that the result of test 2 with third - order scheme actually shows the @xmath252 error , demonstrating that the spatial order of our cphsf scheme is consistent with the numerical order with of the fitting polynomial . , but for test 2.,width=453 ] ( left ) and @xmath247 ( right ) at @xmath227 plotted against @xmath248 . crosses and `` x ' 's show the results with first- and third - order cpshsf for test 1 , and squares and circles those for test 2 . solid and dashed curves show the theoretical models for the error of first- and third - order schemes for test 2 . , width=453 ] in this section , we present the result of the rotating cone test ( e.g. @xcite , @xcite , @xcite ) to discuss the effect of the numerical diffusion caused by the rearrangement . in this test , the rigid rotation a cone - shaped object is followed for two complete circles . since cphsf is fully lagrangian scheme , without the rearrangement of particles it will perfectly conserve the initial shape of the object . we forced the periodic rearrangement of particles to see its effect . we used the initial condition the same as in @xcite . the computational domain is @xmath253 . the initial density distribution is given by @xmath254 where @xmath255 , @xmath256 and @xmath257 where @xmath226 is the number of particles . initial velocity is the rigid body rotation given by @xmath258 where @xmath259 is the angular velocity and we set @xmath260 . we integrate position and velocity with respect to time analytically . the time step is given by @xmath261 in this test , we use first- , third- and fourth - order schemes . the numbers of particles are @xmath262 and @xmath263 . we rearrange particles to the grid pattern @xmath264 times within two rotations . we compare the results with @xmath265 , @xmath266 , @xmath267 , @xmath189 and @xmath177 . we rearrange at @xmath268-th , @xmath269-th , @xmath270-th and @xmath271-th steps when the rotating angle is @xmath272 ( @xmath273 ) for @xmath265 , @xmath266 , @xmath267 and @xmath189 . figures [ fig : rotcone11]-[fig : rotcone44 ] show cones after two rotations for the number of particles @xmath262 and @xmath274 . the orders of schemes are first and fourth . table [ tab : rotcone ] shows the errors of the height of the cone after two rotations defined as @xmath275 we derive @xmath276 using equation ( [ eq : my , sf0 ] ) when @xmath264 is not zero since , due to the rearrangement , there is no particle at @xmath277 . we can see that the errors are smaller for larger @xmath264 , higher order schemes , and higher resolution ( larger number of particles ) . the error for the case of @xmath278 and the fourth - order scheme are better or similar to those of mesh and mesh - free schemes discussed in @xcite or @xcite , except for those of extremely high accuracy schemes based on fourier transform and one of finite - volume scheme which the interpolynomial is a cubic spline ( @xcite ) . in our initial model , we used only one particle to express the peak , while in the traditional rotating cone test , four grid points are used to express the peak . thus , with our initial condition it is much harder to keep the height of the peak . thus , we can conclude that even when we forced frequent rearrangement of particles , our scheme can achieve the accuracy comparable to those of high - order eulerian schemes . note that the required frequency of the rearrangement is generally quite low . for example , the total number of rearrangement in the kelvin - helmholtz instability test with the first - order scheme discussed in section [ sec : test_khi_abel ] is 9 in 5291 timesteps . [ tab : rotcone ] the kelvin - helmholtz instability ( hereafter khi ) test has been used to investigate the ability of numerical schemes to handle large deformations and hydrodynamical instabilities ( @xcite , @xcite , @xcite , @xcite ) . we performed two - dimensional khi from two different initial conditions . one is used in @xcite , and another is in @xcite . the difference between two initial conditions is that density and shear velocity are discontinuous in @xcite , while smoothed in @xcite . we present the results of calculations with the initial condition in @xcite in section [ sec : test_khi_abel ] and that in @xcite in section [ sec : test_khi_price ] . we performed two - dimensional calculations using a computational domain of @xmath279 , with a periodic boundary condition . we made the smoothed contact discontinuity by setting the initial conditions as @xmath280 where we used @xmath281 , @xmath282 and @xmath283 . the parameter @xmath284 is given by @xmath285 . the smoothed velocity of the @xmath286-direction ( shear velocity ) is @xmath287 where @xmath288 and @xmath289 are the reference values of the @xmath286-directional velocity in the high- and low - density regions , respectively . we used @xmath290 and @xmath291 . we assume that fluid is an ideal gas with @xmath292 and set @xmath293 and @xmath294 . the number of particles is 256 @xmath295 256 . the velocity perturbation in the @xmath296-direction is as follows : @xmath297 , \end{aligned}\ ] ] where @xmath298 and @xmath299 . the growth timescale of khi is @xmath300 for our setup , @xmath301 . we rearranged particles with @xmath302 for the first - order cphsf and @xmath303 for the third - order one . figures [ fig : khi_mc_1s ] and [ fig : khi_mc_3s ] show the results obtained using first- and third - order schemes . the first - order result looks similar to the highest resolution pencil code result of @xcite . in figure [ fig : khi_mc_3s ] , we can clearly see the development of the second - level khi , which does not exist in the first - order result . thus , we can conclude that the effective resolution of the third - order scheme is significantly higher than that of the first - order scheme with the same number of particles . , @xmath304 from left to right . the order of the scheme is first . , width=453 ] , but for the third - order scheme.,width=453 ] the initial condition is same as that in section [ sec : test_khi_abel ] , except for density , shear velocity and the perturbation . in this simulation , initial density is given by @xmath305 where we used @xmath206 and @xmath306 . velocity of the @xmath286-direction is @xmath307 where we used @xmath290 and @xmath291 . the velocity perturbation in the @xmath296-direction is @xmath308 , & 0 \leq |y-0.25| \leq 0.025\\ a \sin [ 2\pi ( x + 0.5 ) / \lambda ] , & 0 \leq |y-0.75| \leq 0.025 \end{array } \right.\end{aligned}\ ] ] where @xmath309 and @xmath310 . for our setup , @xmath311 . we rearranged particles with @xmath302 for the first - order cphsf and @xmath312 for the third - order one . figures [ fig : khi_pr_1s ] and [ fig : khi_pr_3s ] show the results obtained using first- and third - order schemes . the vortexes in our result are clearer than that of @xcite in which sph with the artificial conductivity is used . we conclude that cphsf can handle khi , even if the initial density and shear velocity at the contact discontinuity are really discontinuous . , @xmath304 from left to right . the order of the scheme is first . , width=453 ] , but for the third - order scheme.,width=453 ] the rayleigh - taylor instability ( hereafter rti ) test is one of popular tests to investigate the capability of the scheme to handle fluid instability . we used the initial condition the same as in @xcite . the computational domain is @xmath176 , @xmath313 , and the boundary of two fluids is at @xmath314 . we applied a periodic boundary condition at @xmath192 , @xmath189 and equation ( [ eq : boundary , fixedslipping ] ) at @xmath315 , @xmath189 as boundary conditions . the number of particles is @xmath316 @xmath295 @xmath317 . the gravitational acceleration is @xmath318 . initial velocity is @xmath202 and @xmath294 . initial density is @xmath319 , & 0 \leq y \leq 0.5 \\ \rho_h\left[1+\frac{\gamma-1}{\gamma}\frac{\rho_hg(y-0.5)}{p_0}\right ] , & 0.5 \leq y \leq 1 \end{array } \right.\end{aligned}\ ] ] where we used @xmath282 and @xmath283 . since the fluid is in equilibrium , initial pressure is given by @xmath320 where @xmath158 is the pressure at the boundary of two fluids , and we used @xmath321 in this test . the velocity perturbation in the @xmath296-direction for @xmath322 is @xmath323 \{1+\cos[5\pi(y-0.5)]\},\end{aligned}\ ] ] where we used @xmath324 . we used equation ( [ eq : rea , criterion ] ) with @xmath302 for the rearrangement of particles . since equation ( [ eq : rea , criterion ] ) can not be applied to particles near the surface , we excluded particles with @xmath325 or @xmath326 when evaluating equation ( [ eq : rea , criterion ] ) . figure [ fig : rti ] shows the result . we can see that cphsf can handle rti . we can see small - scale khi features develop near the bottom of the heavier fluid , and we can also see small - scale rti features near the `` root '' of sinking heavier fluid . these fine features indicate that cphsf has high resolution and small dissipation . , @xmath267 , @xmath327 and @xmath266 from left to right . the order of the scheme is first . , width=604 ] the gravity wave test is useful to investigate the capability of numerical schemes to handle the free surface . note that standard sph can not handle the gravity wave well , because it can not correctly evaluate the density of particles near the surface . sph schemes specially designed to handle free surface exist ( e.g. @xcite , @xcite ) . however , the most sophisticated schemes require local high - order diffusion of velocity to stabilize the wave ( e.g. @xcite ) . the computational domain is @xmath328 , @xmath329 . we applied a periodic boundary at @xmath192 , equation ( [ eq : boundary , freesurface ] ) at @xmath315 and equation ( [ eq : boundary , fixedslipping ] ) for particles initially at @xmath330 as boundary conditions . the number of particles is @xmath331 @xmath295 @xmath332 . the equation of state is given by equation ( [ eq : eos_water ] ) with @xmath333 , @xmath168 and @xmath334 , and initial density is @xmath335 initial velocity is the same as in @xcite , @xmath336 where @xmath337 , @xmath196 and @xmath338 are the amplitude , the number of wave and the frequency . in this test , we set @xmath299 , @xmath339 and @xmath340 . we used @xmath341 . figure [ fig : gw ] shows the time evolution up to @xmath342 . figure [ fig : gwe ] and figure [ fig : gwe2 ] show the error of velocity at @xmath343 for runs with the number of particles @xmath226 , @xmath344 , @xmath345 and @xmath346 . the error is given by @xmath347 where @xmath348 and @xmath349 are analytical solutions . we can see that the error becomes smaller as we increase @xmath226 , showing the first - order convergence , as we used the first - order scheme . , @xmath350 , @xmath351 and @xmath352 are shown . , width=453 ] - and @xmath296-directional velocity at @xmath343 in the gravity wave test with the numbers of particles @xmath226 @xmath353 , @xmath354 , @xmath355 . , width=453 ] , but the error with @xmath356 is multiplied by two , while that with @xmath357 is multiplied by three.,width=453 ] figure [ fig : gwy ] shows @xmath296 of the particle initially at @xmath358 . in an initial stage , the phases of results for @xmath359 , @xmath354 and @xmath355 agree well with each other . however , the wave for @xmath360 has slightly longer period than those of @xmath361 and @xmath357 , and the difference of @xmath296 between the result of @xmath359 and those of @xmath361 and @xmath357 grows in time . in addition , the phase of @xmath361 deviates from that of @xmath357 slightly at the later stage . -coordinate of the particle initially at @xmath358 in the gravity wave test in which the numbers of particles @xmath226 are @xmath353 , @xmath354 , @xmath355 . , width=604 ] the dam break test is the most widely used test for numerical schemes for the fluid with the free surface . the initial condition of the dam break test is the same as that used by @xcite . the computational domain is @xmath362 , @xmath363 . we applied equation ( [ eq : boundary , freesurface ] ) at @xmath364 and @xmath365 and equation ( [ eq : boundary , fixedslipping ] ) at @xmath192 and @xmath315 as boundary conditions . the number of particles is 60 @xmath295 61 . we set that the equation of state is given by equation ( [ eq : eos_water ] ) with @xmath366 , @xmath168 and @xmath334 , and initial density is @xmath367^{1/6}. \end{aligned}\ ] ] we set that density and pressure of the particles at @xmath364 is @xmath368 initial velocity is @xmath369 . in dam break simulation , the rearrangement of particles is necessary . we rearrange particles every @xmath370 time unit . we used @xmath371 . figure [ fig : dam ] shows the time evolution up to @xmath372 . the result looks similar to those in the previous dam break test ( @xcite ) . figure [ fig : damx ] gives the position of the forefront of water plotted against @xmath373 , where @xmath373 is the dimensionless time defined as @xmath374 . the results are for first - order cphsf , an analytical solution ( @xcite ) and the experimental data of @xcite , by @xcite distributed . it was compared to numerical solutions in previous studies ( e.g. @xcite , @xcite ) . according to @xcite , the forefront velocity @xmath375 becomes @xmath376 in static state , if we use on the theory of shallow water waves . therefore , the position of the forefront is @xmath377 analytically . note that the analytical solution is not exact . clearly , in the limit of @xmath378 , the solution should be quadratic and not linear in time . the experimental result is not free from the real viscosity . therefore , the discrepancy between our numerical result and experimental result does not imply the problem in the side of our scheme . @xcite calculated the dam break using his csph and standard sph schemes and compared the result with the experimental data . the agreement between numerical and experimental data was actually pretty good , better than that in our case . this result probably imply both of experimental result and @xcite s result suffer the effect of viscosity on other dispational effect . though @xcite did not explicitly use av , he applied the re - evaluation of the density , which would cause significant dissipation . therefore , we believe our scheme is less dissipative than his csph . , @xmath350 , @xmath351 and @xmath352 from top to bottom . , width=453 ] . filled circles , crosses show the result of first - order cpshf and experimental result ( see text ) , respectively . the solid line indicate the analytic estimate based on the shallow water equation . , width=226 ] the cold keplerian disk is important in astronomy . we prepared a two - dimensional computational domain of @xmath379 . we applied equation ( [ eq : boundary , freesurface ] ) at the inner edge and equation ( [ eq : boundary , fixedslipping ] ) at the outer edge as boundary conditions . the number of particles is 46368 , and we place the particles in concentric rings with the same intervals for @xmath380 axis and @xmath381 axis . we assumed that the fluid is ideal gas with @xmath198 and set @xmath382 and @xmath383 . initial velocity is the pure keplarian rotation given by @xmath384 where @xmath385 is the gravitational constant and @xmath386 is the mass of the central star . we set @xmath387 . the time step is smaller one between the time step given by the courant conditions for the pressure gradient term and the central force term . therefore , we compare equation ( [ eq : dt_cfl ] ) with @xmath388 and then smaller one is taken . the parameter @xmath389 is a constant coefficient , and the value of @xmath390 denotes how many steps is calculated in a orbit . we set @xmath391 . in this test , we used fourth - order cphsf , because the accuracy of the approximation of the artificial viscosity is important to prevent unphysical angular momentum transfer ( e.g. @xcite ) . in addition , we use fifth - order radau method for the time integration . figure [ fig : disk ] shows the results . we can see that the disk does not break till @xmath392 orbits . note that this lifetime is longer than the lifetime of any lagrangian scheme , reported in the literature ( @xcite , @xcite ) by more than a factor three . figure [ fig : diske ] shows the error of the density plotted against time . the error is given by @xmath393 it is clear that the error grows rapidly after @xmath394 orbits . from figure [ fig : disk ] , we can see that the error becomes large at the outer edge after @xmath395 orbits . therefore , if we can improve the treatment of the outer edge , we may be able to reduce the error . we conclude that the cold keplerian disk integrated using cphsf can survive for @xmath396 orbits . cphsf can handle rotating disks better than the lagrangian scheme variants proposed in previous studies . @xmath177 , @xmath178 , @xmath397 and @xmath229 orbits from left to right . , width=604 ] one limitation of our current cphsf scheme is that there is no easy way to handle topological changes of fluid , for example the collision of two droplets . in principle , we can handle collisions by detecting the moment of the collision and change the surface particles ( usually with special boundary conditions ) to bulk particles . however , finding the exact moment of collision for each of surface particles involved can be very expensive . therefore approximate treatment is practically necessary . another problem is that we currently need implicit time integration in order to apply boundary conditions . an explicit or semi - implicit scheme would be better , and should be applicable at least to the bulk of the fluid . standard sph scheme and its variations are not high order and can not handle free surfaces . we formulate cphs , a high - order mesh - free method . cphsf can handle kelvin - helmholtz instability and rayleigh - taylor instability better than previous sph methods do . we also simulate the gravity wave and dam break . the results are excellent . additionally , the cold keplerian disk , which is very important in astrophysics , can be calculated for much larger time than possible with previous mesh - free methods . cphsf does not exactly satisfy the conservation laws . however we showed that the violation of the conservation laws is acceptable , because of the lagrangian nature and high - order accuracy of cphsf . therefore we conclude that cphsf is more useful than previous mesh - free method . we thanks takayuki saitoh for discussion about mesh - free method and others , natsuki hosono , masaki iwasawa and many others for providing useful comments about numerical simulations , keigo nitadori for discussion about formulation . we thank the referee for insightful and helpful comments . this work was supported by riken junior research associate program and mext spire and jicfus . part of the research covered in this paper was funded by mext s program for the development and improvement for the next generation ultra high - speed computer system , under its subsidies for operating the specific advanced large research facilities . agertz , o. , moore , b. , stadel , j. , et al . 2007 , , 380 , 963 antuono , m. , colagrossi , a. , marrone , s. , & lugni , c. 2011 , computer physics communications , 182 , 866 balsara , d. s. 1995 , journal of computational physics , 121 , 357 bilotta , g. , russo , g. , hrault , a. , & del negro , c. 2011 , annals of geophysics , 54 chen , j. , beraun , j. , & jih , c. 1999a computational mechanics , 23 , 279 chen , j. , beraun , j. , & jih , c. 1999b computational mechanics , 24 , 273 chen , j. , beraun , j. , & carney , t. 1999c international journal for numerical methods in engineering , 46 , 231 . chen , j. , & beraun , j. 2000 computer methods in applied mechanics and engineering , 190 , 225 chock , david p. 1991 , atmospheric environment . part a. general topics , 25 , 853 crowley , wp . 1968 , monthly weather review , 96 , 1 cullen , l. , & dehnen , w. 2010 , , 408 , 669 dilts , g. a. 1999 , international journal for numerical methods in engineering , 44 , 1115 fang , j. , parriaux , a. , rentschler , m. , & ancey , c. , 2009 applied numerical mathematics 59 , 251 . frontiere , n. , raskin , c. d. , & owen , j. m. 2016 , arxiv:1605.00725 gingold , r. a. , & monaghan , j. j. 1977 , , 181 , 375 hernquist , l. , & katz , n. 1989 , , 70 , 419 hopkins , p. f. 2015 , , 450 , 53 hosono , n. , saitoh , t. r. , & makino , j. 2016 , , 224 , 32 imaeda , y. , & inutsuka , s .- i . 2002 , , 569 , 501 inutsuka , s .- 2002 , journal of computational physics , 179 , 238 koshizuka , s. , & oka , y. 1996 , nuclear science engineering , 123 lattanzio , j. c. , monaghan , j. j. , pongracic , h. , & schwarz , m. p. 1985 , , 215 , 125 lax , p. d. , & richtmyer , r. d. 1956 communications on pure and applied mathematics , 9 , 267 . liu , w. k. , jun , s. , & zhang , y. f. 1995 international journal for numerical methods in fluids , 20 , 1081 . liu , m. , xie , w. , & liu , g. 2005 applied mathematical modelling , 29 , 1252 . lobovsk , l. , botia - vera , e. , castellana , f. , mas - soler , j. , & souto - iglesias , a. 2014 , journal of fluids and structures , 48 , 407 lucy , l. b. 1977 , , 82 , 1013 martin , j. c. , & moyce , w. j. 1952 , philosophical transactions of the royal society of london series a , 244 , 312 mcnally , c. p. , lyra , w. , & passy , j .- c . 2012 , , 201 , 18 monaghan , j. j. 1994 , journal of computational physics , 110 , 399 monaghan , j. j. 1997 , journal of computational physics , 136 , 298 monaghan , j. j. , & gingold , r. a. 1983 , journal of computational physics , 52 , 374 morris , j. p. , & monaghan , j. j. 1997 , journal of computational physics , 136 , 41 okamoto , t. , jenkins , a. , eke , v. r. , quilis , v. , & frenk , c. s. 2003 , , 345 , 429 ott , f. , & schnetter , e. 2003 , arxiv : physics/0303112 price , d. j. 2008 , journal of computational physics , 227 , 10040 ritchie , b. w. , & thomas , p. a. 2001 , , 323 , 743 rosswog , s. , davies , m. b. , thielemann , f .- k . , & piran , t. 2000 , , 360 , 171 saitoh , t. r. , & makino , j. 2013 , , 768 , 44 sod , g.a . 1978 , journal of computational physics , 27 , 1 staroszczyk , r. 2010 , archives of hydro - engineering and environmental mechanics , 57 , 61 . tamai , t. , shibata , k. , & koshizuka , s. 2013 trans . jsces vijay , k. 1998 , applied computational fluid dynamics , ( united states : crc press ) , 1 von neumann , j. , & richtmyer , r. d. 1950 , journal of applied physics , 21 , 232 wendland , h. 1995 , advances in computational mathematics , 4 , 389 . whitham , g. 1999 , linear and nonlinear waves , ed . allen iii , m. b. , cox , d.a . , lax , p. , toland , j. , ( state of new jersey : wiley - interscience ) , 1 yamamoto , s. , saitoh , t. r. , & makino , j. 2015 , , 67 , 37 yamartino , rj . 1993 , monthly weather review , 121 , 753 zhang , g. , & batra , r. 2004 , computational mechanics , 34 , 137 .

in this work we have employed bethe - salpeter equation ( bse ) under covariant instantaneous ansatz ( cia ) to study electromagnetic decays of ground state equal mass vector mesons : , , , and through the process . we employ the generalized structure of hadron - quark vertex function which incorporates various dirac structures from their complete set order - by - order in powers of inverse of meson mass . the electromagnetic decay constants for the above mesons are calculated using the leading order ( lo ) and the next - to - leading order ( nlo ) dirac structures . the relevance of various dirac structures in this calculation is studied . corresponding author . [email protected] + [email protected]

1307.4044

meson decays provide an important tool for exploring the structures of these simplest bound states in qcd , and for studies on non - perturbative behavior of strong interactions . these studies has become a hot topic in recent years . flavourless vector mesons play an important role in hadron physics due to their direct coupling to photons and thus provide an invaluable insight into the phenomenology of electromagnetic couplings to hadrons . thus , a realistic description of vector mesons at the quark level of compositeness would be an important element in our understanding of hadron dynamics and reaction processes at scales where qcd degrees of freedom are relevant . there have been a number of studies@xcite on processes involving strong , radiative and leptonic decays of vector mesons . such studies offer a direct probe of hadron structure and help in revealing some aspects of the underlying quark - gluon dynamics . in this work we study electromagnetic decays of ground state equal mass vector mesons : @xmath10 and @xmath4 ( each comprising of equal mass quarks ) through the process @xmath11 which proceeds through the coupling of quark - anti quark loop to the electromagnetic current in the framework of bethe - salpeter equation ( bse ) , which is a conventional non - perturbative approach in dealing with relativistic bound state problems in qcd and is firmly established in the framework of field theory . from the solutions , we obtain useful information about the inner structure of hadrons which is also crucial in high energy hadronic scattering and production processes . despite its drawback of having to input model dependent kernel , these studies have become an interesting topic in recent years since calculations have satisfactory results as more and more data is being accumulated . we get useful insight about the treatment of various processes using bse due to the unambiguous definition of the 4d bs wave function which provides exact effective coupling vertex ( hadron - quark vertex ) of the hadron with all its constituents ( quarks ) . we have employed qcd motivated bethe - salpeter equation ( bse ) under covariant instantaneous ansatz ( cia)@xcite to calculate this process . cia is a lorentz - invariant generalization of instantaneous ansatz ( ia ) . what distinguishes cia from other 3d reductions of bse is its capacity for a two - way interconnection : an exact 3d bse reduction for a @xmath12 system ( for calculation of mass spectrum ) , and an equally exact reconstruction of original 4d bse ( for calculation of transition amplitudes as 4d quark loop integrals ) . in these studies , the main ingredient is the 4d hadron - quark vertex function @xmath6 which plays the role of an exact effective coupling vertex of the hadron with all its constituents ( quarks ) . the complete 4d bs wave function @xmath13 for a hadron of momentum @xmath14 and internal momentum @xmath15 comprises of the two quark propagators ( corresponding to two constituent quarks ) bounding the hadron - quark vertex @xmath6 . this 4d bs wave function is considered to sum up all the non - perturbative qcd effects in the hadron . now one of the main ingredients in 4d bs wave function ( bsw ) is its dirac structure . the copius dirac structure of bsw was already studied by c.h.l . smith@xcite much earlier . recent studies@xcite have revealed that various mesons have many different dirac structures in their bs wave functions , whose inclusion is necessary to obtain quantitatively accurate observables . it was further noticed that all structures do not contribute equally for calculation of various meson observables@xcite . further , it was amply noted in @xcite that many hadronic processes are particularly sensitive to higher order dirac structures in bs amplitudes . it was further noted in @xcite that inclusion of higher order dirac structures is also important to obtain simultaneous agreement with experimental decay widths for a range of processes such as : @xmath16 , @xmath17 , @xmath18 , etc . for a given choice of parameters . towards this end , to ensure a systematic procedure of incorporating various dirac covariants from their complete set in the bsws of various hadrons ( pseudoscalar , vector etc . ) , we developed a naive power counting rule in ref.@xcite , by which we incorporate various dirac structures in bsw , order - by - order in powers of inverse of meson mass . using this power counting rule we calculated electromagnetic decay constants of vector mesons ( @xmath19 ) using only the leading order ( lo ) dirac structures [ @xmath20 and @xmath21 . however in ref.@xcite , we rigorously studied leptonic decays of unequal mass pseudoscalar mesons @xmath22 and calculated the leptonic decay constants @xmath23 for these mesons employing both the leading order ( lo ) and the next - to - leading order ( nlo ) dirac structures . the contributions of both lo and nlo dirac structures to @xmath23 was worked out . we further studied the relevance of both the lo and the nlo dirac structures to this calculation . in the present paper , we extended these studies to vector mesons and have employed both lo and nlo dirac structures identified according to our power counting rule , to calculate @xmath24 for ground state vector mesons , @xmath25 and in the process we studied the relevance of various dirac structures to calculation of decay constants @xmath24 for vector mesons in the process @xmath16 . we found that contributions from nlo dirac structures are smaller than those of lo dirac structures for all vector mesons . in what follows , we give a detailed discussion of the fit and calculation up to nlo level after a brief review of our framework . the paper is organized as follows : in section 2 we discuss the structure of bsw for vector mesons under cia using the power counting rule we proposed earlier . in section 3 we give the calculation of @xmath24 for vector mesons . a detailed presentation of results and the numerical calculation is given in section 4 . section 5 is relegated to discussion . we briefly outline the bse framework under cia . for simplicity , lets consider a @xmath26 system comprising of scalar quarks with an effective kernel @xmath27 , 4d wave function @xmath28 , and with the 4d bse , @xmath29 where @xmath30are the inverse propagators , and @xmath31 are ( effective ) constituent masses of quarks . the 4-momenta of the quark and anti - quark , @xmath32 , are related to the internal 4-momentum @xmath33 and total momentum @xmath34 of hadron of mass @xmath35 as @xmath36 where @xmath37/2 $ ] are the wightman - garding ( wg ) definitions of masses of individual quarks . now it is convenient to express the internal momentum of the hadron @xmath15 as the sum of two parts , the transverse component , @xmath38 which is orthogonal to total hadron momentum @xmath14 ( ie . @xmath39 regardless of whether the individual quarks are on - shell or off - shell ) , and the longitudinal component , @xmath40 , which is parallel to p. thus we can decompose @xmath33 as , @xmath41 , where the transverse component , @xmath42 is an effective 3d vector , while the longitudinal component , @xmath43 plays the role of the fourth component and is like the time component . to obtain the 3d bse and the hadron - quark vertex , use an ansatz on the bs kernel @xmath27 in eq . ( [ eq:2.1 ] ) which is assumed to depend on the 3d variables @xmath44 , @xmath45 @xcite i.e. @xmath46 hence , the longitudinal component , @xmath47 of @xmath33 , does not appear in the form @xmath48 of the kernel . defining @xmath49 as the 3d wave function , @xmath50 integrating eq.(1 ) , and making use of eqs.(2 - 3 ) , we obtain the 3d salpeter equation , @xmath51 where @xmath52 is the 3d denominator function defined as @xcite , @xmath53 whose value given above is obtained by evaluating the contour integration over inverse quark propagators in the complex @xmath54-plane by noting their corresponding pole positions ( for details see@xcite ) . which is used for making contact with mass spectra of @xmath12 mesons . further , making use of eq.(2 ) and ( 3 ) on rhs of eq.(1 ) , we get , @xmath55 from equality of rhs of eq.(4 ) and ( 6 ) , we see that an exact interconnection between 3d and 4d bs wave functions is thus brought out . the 4d hadron - quark vertex function for scalar quarks under cia can be identified as : @xmath56 now for fermionic quarks , the 4d bse under gluonic which is akin to vector type interaction kernel with a 3d support can be written as : @xmath57 here , @xmath58 is the color factor , @xmath59 and the potential @xmath60 involves the scalar structure of the gluon propagator in the perturbative ( o.g.e . ) as well as the non - perturbative ( confinement ) regimes . the full structure of @xmath60 is @xcite : @xmath61 which is taken as one - gluon - exchange like as regards color [ @xmath62 and spin ( @xmath63 ) dependence . the scalar function @xmath64 is a sum of one - gluon exchange @xmath65 and a confining term @xmath66 . this confining term simulates an effect of an almost linear confinement ( @xmath67 ) for heavy quark ( @xmath68 ) sector , while retaining harmonic form ( @xmath69 ) for light quark ( @xmath70 ) sector as is believed to be true for qcd . @xmath71 the values of basic constants are : @xmath72 gev , @xmath73 gev , @xmath74 gev , @xmath75 gev , @xmath76 gev and @xmath77 gev @xcite . however the form of bse in eq.(8 ) is not convenient to use in practice since dirac matrices lead to several coupled integral equations . however a considerable simplification is effected by expressing them in gordon - reduced form which is permissible on the mass shells of quarks ( ie . on the surface @xmath78 ) . the gordon reduced bse form of the fermionic bse can be written as @xcite : @xmath79 where the connection between @xmath80 and @xmath81 ( whose structure is identical as @xmath81 in case of scalar quarks in eq.(1 ) ) is , @xmath82 now to reduce the above bse to the 3d form , all time - like components @xmath83 of momenta in @xmath84 on rhs of eq.(9 ) are replaced by their on - shell values giving us the 3d form , @xmath85 . thus @xmath86 the 3d form of bse then works out as @xcite : @xmath87 this is reducible to the equation for a 3d harmonic oscillator with coefficients depending on the mass @xmath35 and total quantum number @xmath88 . the ground state wave functions @xcite deducible from this equation have gaussian structure and are expressed as : @xmath89 and is appropriate for making contact with o(3)-like mass spectrum ( for details see @xcite ) . it is to be noted that this 3d bse ( in eq.(14 ) ) which is responsible for determination of mass spectra of mesons in cia is formally equivalent ( see @xcite ) to the corresponding mass spectral equation deduced earlier using null - plane approximation ( npa)@xcite . thus the mass spectral predications for @xmath90 systems in bse under cia are identical to the corresponding mass spectral predictions for these systems in bse under npa @xcite ( see @xcite for details ) . we further wish to mention that a similar form for ground state wave function in harmonic oscillator basis using variational arguments has been used in @xcite . in ground state wave function @xmath91 in eq.(15 ) , @xmath92 is the inverse range parameter which incorporates the content of bs dynamics and is dependent on the input kernel @xmath93 ( for details see @xcite ) the structure of inverse range parameter @xmath92 in wave functions @xmath91 is given as @xcite : @xmath94 thus , for fermionic quarks , the full 4d bs wave function can be written as @xmath95 where the 4d hadron - quark vertex function is @xcite i.e. @xmath96 the 4d hadron - quark vertex @xmath6 in the above equation , satisfies a 4d bse with a natural off - shell extension over the entire 4d space ( due to the positive definiteness of the quantity @xmath97 throughout the entire 4d space ) and thus provides a fully lorentz - invariant basis for evaluation of various transition amplitudes through various quark loop diagrams . @xmath98 in the above equation is the 4d bs normalizer . in the hadron - quark vertex function , @xmath6 above , ( @xmath99 ) contains the relevant dirac structures @xcite which makes @xmath6 a @xmath100 matrix in the spinor space . in our model , the relevant dirac structures in @xmath6 are incorporated in accordance with our recently proposed power counting rule @xcite which identifies the leading order ( lo ) dirac structures from the next - to - leading order ( nlo ) dirac structures and for a vector meson is expressed as : @xmath101\frac{a_2}{m}\\ & & \nonumber+i\frac{a_3}{m^2}[(\gamma\cdot\varepsilon)(\gamma\cdot p)(\gamma\cdot q ) - ( \gamma\cdot\varepsilon)(\gamma\cdot q)(\gamma\cdot p)+2i(q\cdot\varepsilon)(\gamma\cdot p)]\\ & & + ( q\cdot\varepsilon)\frac{a_4}{m}+i(q\cdot\varepsilon)(\gamma\cdot p)\frac{a_5}{m^2 } - i(q\cdot\varepsilon)(\gamma\cdot q)\frac{a_6}{m^2 } + ( q\cdot\varepsilon)[(\gamma\cdot p)(\gamma\cdot q)-(\gamma\cdot q)(\gamma\cdot p)]\frac{a_7}{m^3},\end{aligned}\ ] ] where @xmath102 are taken as the eight dimensionless constant coefficients to be determined . but since we take constituent quark masses , where the quark mass , @xmath35 is approximately half the hadron mass , @xmath35 , we can use the ansatz , @xmath103 in the rest frame of the hadron . then each of the eight terms in the above equation receives suppression by different powers of @xmath104 . thus we can arrange these terms as an expansion in powers of @xmath105 . we can then see that in the expansion of @xmath106 that the structures associated with the coefficients , @xmath107 have magnitudes , @xmath108 and are of leading order ( lo ) . those with @xmath109 are @xmath110 and are next - to - leading order ( nlo ) , while those with @xmath111 are @xmath112 and are nnlo . this naive power counting rule suggests that the maximum contribution to calculation of any vector meson observable should come from dirac structures @xmath113 and @xmath114 associated with coefficients , @xmath115 and @xmath116 respectively , followed by the higher order dirac structures associated with the other four coefficients , @xmath117 and then by dirac structures associated with coefficients , @xmath111 . in this work we try to calculate the decay constants using lo and nlo dirac structures since we wish to calculate the decay constants up to next - to - leading orders . thus we take the form of the hadron - quark vertex function as , @xmath118\frac{a_2}{m}\\ & & \nonumber+i\left[(\gamma\cdot\varepsilon)(\gamma\cdot p)(\gamma\cdot q ) - ( \gamma\cdot\varepsilon)(\gamma\cdot q)(\gamma\cdot p)+2i(q\cdot\varepsilon)(\gamma\cdot p)\right ] \frac{a_3}{m^2}\\ & & \left.+(q\cdot\varepsilon)\frac{a_4}{m}+i(q\cdot\varepsilon)(\gamma\cdot p ) \frac{a_5}{m^2}\right\}\frac{1}{2\pi i}n_vd(\widehat{q})\phi(\widehat{q}).\end{aligned}\ ] ] in the above equation , @xmath98 is the 4d bs normalizer for ground state vector meson with internal momenta @xmath15 , and is worked out in the framework of covariant instantaneous ansatz ( cia ) to give explicit covariance to the full fledged 4d bs wave functions , @xmath13 and hence to the hadron - quark vertex function , @xmath6 , employed for calculation of decay constants . in the structure of @xmath6 above , @xmath91 is the ground state 3d bs wave function for vector meson with internal momenta @xmath15 and is given in eq.(15 ) . the vector meson decay proceeds through the quark - loop diagram shown below ( see fig.1 ) . showing the coupling of electromagnetic current to the quark loop . ] the coupling of a vector meson of momentum @xmath14 and polarization @xmath120 to the photon is expressed via dimensionless coupling constant @xmath121 which can be described by the matrix element , @xmath122 ( where @xmath123 is the flavour multiplet of quark field and @xmath124 is the quark electromagnetic charge operator ) which can in turn be expressed as a loop integral , @xmath125.\ ] ] here @xmath126 arises from the flavour configuration of individual vector mesons and has values : @xmath127 and @xmath128 for @xmath129 and @xmath4 respectively , and the polarization vector @xmath130 of v - meson satisfies , @xmath131 . defining leptonic decay constant,@xmath24 as , @xmath132 @xcite , we can express @xmath133\ ] ] plugging @xmath134 , which involves the structure of hadron - quark vertex function in eq . ( 2.8 ) flanked by the dirac propagators of the two quarks as in eq . ( 2.3 ) , into the above equation , evaluating trace over the gamma matrices and noting that only the components of terms on the right hand side in the direction of @xmath130 will contribute to the integral , we multiply both sides of the above integral by @xmath135 , we can then express the leptonic decay constant @xmath136 as , @xmath137 where @xmath138 are contributions to @xmath136 from the six dirac structures associated with @xmath139 in the expression for hadron - quark vertex function @xmath140 and are expressed analytically in terms of @xmath141 integrations over the poles @xmath142 of the quark propagators as : @xmath143\\ \nonumber\ f_v^1&=&\sqrt{3}n_v\frac{a_1}{m}\int d^3\widehat{q}d(\widehat{q})\phi(\widehat{q})\int\frac{md\sigma}{2\pi i\delta_1\delta_2}(-4mm)\\ \nonumber\ f_v^2&=&\sqrt{3}n_v\frac{a_2}{m}\int d^3\widehat{q}d(\widehat{q})\phi(\widehat{q})\int\frac{md\sigma}{2\pi i\delta_1\delta_2}\left[-\frac{4}{3}m(\delta_1-\delta_2)\right]\\ \nonumber\ f_v^3&=&\sqrt{3}n_v\frac{a_3}{m}\int d^3\widehat{q}d(\widehat{q})\phi(\widehat{q})\int\frac{md\sigma}{2\pi i \delta_1\delta_2}\left[-\frac{8}{3}(\delta_1+\delta_2)+\left(\frac{16}{3}m^2-\frac{4}{3}m^2\right)\right]\\ \nonumber\ f_v^4&=&\sqrt{3}n_v\frac{a_4}{m}\int d^3\widehat{q}d(\widehat{q)}\phi(\widehat{q})\int\frac{md\sigma}{2\pi i \delta_1\delta_2}\left[-\frac{2m}{3m}(\delta_1+\delta_2)+\left(\frac{4m^3}{3m}-\frac{1}{3}mm\right)\right]\\ \ f_v^5&=&\sqrt{3}n_v\frac{a_5}{m}\int d^3\widehat{q}d(\widehat{q)}\phi(\widehat{q})\int\frac{md\sigma}{2\pi i \delta_1\delta_2}\left[\left(\frac{4}{3}m^2-\frac{2}{3}m^2\right)(\delta_1-\delta_2)\right].\end{aligned}\ ] ] in deriving the above expressions , we have made use of the following relation showing the normalization over the polarization vector , @xmath144 for @xmath60- meson of 4-momentum , @xmath14 as , @xmath145 we made use of the above relation to express the quantities involving dot products of @xmath146 with various momenta like , @xmath147 , @xmath148 and @xmath149 in terms of dot products of momenta as , @xmath150 these dot products of momenta were in turn expressible in terms of the inverse propagators , @xmath142 as : @xmath151 thus all expressions for @xmath152 above were expressible in terms of @xmath142 . then carrying out integrations over the off - shell variable , @xmath141 by the method of contour integrations by noting the pole positions in the complex @xmath153 plane : @xmath154 we get , the following integrals : @xmath155 where , @xmath156 thus we can express the various components @xmath157 @xmath158 of @xmath24 in eq . ( 3.5 ) as , @xmath159\\ & & \nonumber\ f_v^1=\sqrt{3}n_v\frac{a_1}{m}\int d^3\widehat{q}\phi(\widehat{q})(-4mm)\\ & & \nonumber\ f_v^2=0\\ & & \nonumber\ f_v^3=\sqrt{3}n_v\frac{a_3}{m}\int d^3\widehat{q}\phi(\widehat{q})\left[-\frac{8}{3}d_{0}(\widehat{q})+\frac{16}{3}m^2-\frac{4}{3}m^2\right]\\ & & \nonumber\ f_v^4=\sqrt{3}n_v\frac{a_4}{m}\int d^3\widehat{q}\phi(\widehat{q})\left[-\frac{2m}{3m}d_0(\widehat{q})+\frac{4m^3}{3m}-\frac{1}{3}mm\right]\\ & & \ f_v^5=0\end{aligned}\ ] ] note that the components , @xmath160 and @xmath161 are @xmath162 on account of equal mass kinematics . note that each of the @xmath157 involves the bs normalizer @xmath163 . this is evaluated using the current conservation condition:@xcite , @xmath164\psi(p , q)s_{f}^{-1}(-p_2)\right\ } + ( 1\rightleftharpoons2)\ ] ] putting bs wave function @xmath13 for a vector meson in eqs . ( 2.3 ) and ( 2.8 ) in the above equation , carrying out derivatives of inverse quark propagators of constituent quarks with respect to total hadron momentum @xmath165 , evaluating trace over products of gamma matrices , following usual steps , and multiplying both sides of the equation by @xmath166 to extract out the normalizer @xmath163 from the above equation , we then express the above equation in terms of the integration variables @xmath42 and @xmath54 . noting that the 4d volume element @xmath167 , we then perform the contour integration in the complex @xmath54- plane by making use of the corresponding pole positions . for details of these mathematical steps involved in the calculations of bs normalizers for vector and pseudoscalar mesons see @xcite , where in the present calculation , we take both , the leading order ( lo ) as well as the next - to - leading order ( nlo ) dirac structures for vector mesons in their respective 4d bs wave functions @xmath13 . then integration over the variable @xmath42 is finally performed to extract out the numerical results for @xmath163 for different vector mesons . the calculation of @xmath163 is quite complex due to the @xmath168 dirac structures involved in the calculation . the structure of the normalizer is of the form , @xmath169 here , @xmath170 are functions of parameters @xmath171 , @xmath172 @xmath173 , where the @xmath174 are the results of integrations over off - shell parameter @xmath54 whose results are given in eq . ( [ eq : thesi ] ) . the @xmath175 are involved functions of @xmath176 and @xmath177 . explicit expressions are listed below in eq . ( [ eq : i20 ] ) : @xmath178\\ i_{11}&=&\frac{2}{3}\left[-\frac{4d_{11}}{m^2}+\frac{3d_{12}}{m^2}-\frac{4d_{22}}{m^2}-\frac{8m^2d_1}{m^2}+\frac{(8m^2+m^2)d_2}{m^2}+2(4m^2+m^2)r\right]\\ i_{22}&=&\frac{2}{3}\left[\frac{d_{11}}{m^2}+\frac{3d_{12}}{m^2}-\frac{4d_{22}}{m^2}-\frac{8m^2d_1}{m^2}+\frac{(8m^2+m^2)d_2}{m^2}+2(4m^2+m^2)r\right]\\ i_{33}&=&\frac{1}{3}\left[\frac{d_{11}}{m^2}\left(\frac{-6m^4}{m^4}+\frac{9m^2}{m^2}-\frac{22m^3}{m^3}-\frac{55m^2}{2m^2}+13\right)\right.\\ & \phantom{= } & + \frac{d_{12}}{m^2}\left(\frac{18m^4}{m^4}-\frac{7m^2}{m^2}+\frac{70m^2}{m^2}+\frac{12m^3}{m^3}-4\right)\\ & \phantom{= } & + \frac{d_{22}}{m^2}\left(\frac{9m^4}{m^4}-\frac{67m^2}{m^2}+\frac{10m^3}{m^3}+35-\frac{9m^2}{m^2}\right)\\ & \phantom{= } & + \frac{d_1}{m^2}\left(\frac{29}{2}m^2 - 13m^2+\frac{12m^4}{m^2}-\frac{3m^3}{m}-\frac{48m^4}{m^2}+\frac{16m^6}{m^4}-\frac{36m^5}{m^3}\right)\\ & \phantom{= } & + \frac{d_2}{m^2}\left(22m^2 - 7m^2-\frac{12m^4}{m^2}+\frac{3m^3}{m}+\frac{48m^4}{m^2}-\frac{16m^6}{m^4}+\frac{4m^5}{m^3}\right)\\ & \phantom{=}&\left.+\frac{r}{m^2}\left(-m^2m^2 - 14m^3m+76m^4 + 16mm^3 + 6m^4+\frac{5}{2}m^4\right)\right]\\ i_{44}&=&\frac{1}{3}\left[\frac{d_{11}}{m^2}\left(\frac{4m^2}{m^2}+1\right)+\frac{7d_{12}}{2m^2}+\frac{d_{22}}{m^2}\left(\frac{1}{6}-\frac{4m^2}{m^2}\right)\right.\\ & \phantom{= } & + \left(2-\frac{6m^4}{m^4}-\frac{m^5}{m^5}-\frac{6m^2}{m^2}-\frac{m^3}{m^3}\right)d_1+\left(\frac{1}{6}+\frac{m^5}{m^5}+\frac{6m^4}{m^4}-\frac{8m^2}{m^2}\right)d_2\\ & \phantom{=}&\left . + \left(-4m^2+\frac{8m^4}{m^2}+\frac{m^5}{m^3}+\frac{m^2}{3}\right)r\right]\\ i_{55}&=&\frac{1}{3}\left[-\frac{3}{2m^2}\left(\frac{m^4}{3m^4}+\frac{11m^2}{3m^2}+\frac{3}{2}\right)d_{12}+\frac{d_1}{m^2}\left(-m^2+\frac{3m^4}{2m^2}+\frac{7m^5}{2m^3}+\frac{17m^2}{4}\right)\right.\\ & \phantom{=}&\left.+\frac{d_2}{m^2}\left(-\frac{m^2}{4}-\frac{3m^5}{2m^3}-\frac{7m^4}{2m^2}+\frac{15m^2}{4}\right)+\frac{r}{m^2}\left(-\frac{3m^4}{2}+\frac{9m^2m^2}{4}+\frac{7m^4}{2}\right)r\right]\\ i_{01}&=&\left[-\frac{4md_{11}}{m^3}-\frac{8md_{22}}{m^3}+\frac{4md_{12}}{m^3}-\frac{4d_1}{3m}\left(\frac{4m^3}{m^2}+m\right)\right.\\ & & \left.+\frac{4d_2}{m}\left(\frac{4m^3}{3m^2}-m\right)+\left(\frac{16m^3}{3m}-\frac{4mm}{3}r\right)\right]\\ i_{02}&=&\left[\frac{8m}{m}(d_1-d_2)-8mmr\right];\\ i_{03}&=&\left[-\frac{2m}{m^3}d_{11}+\frac{4m}{3m^3}d_{12}+\frac{2m}{m^3}d_{22}+\frac{8d_1}{3m}(\frac{m^3}{m^2}-m)+\frac{3m^3}{3m^3}d_2\right.\\ & & \left.+(\frac{8m^3}{3m}-\frac{2mm}{3})r\right]\\ i_{04}&=&\left[-\frac{7m^2d_{11}}{3m^4}+\frac{d_{22}}{3m^2}\left(-\frac{2m^4}{m^4}-\frac{7m^2}{m^2}+8\right)+\frac{d_{12}}{3m^2}\left(\frac{2m^2}{m^2}-8\right)\right.\\ & & \left.+\frac{d_1}{m^2}\left(-m^2-\frac{2m^4}{3m^2}+\frac{8m^2}{3}\right)+\frac{d_2}{m^2}\left(\frac{4m^2}{3}+\frac{2m^4}{3m^2}-\frac{8m^2}{3})+\frac{r}{m^2}(-m^2m^2 - 2m^4\right)\right]\\ i_{05}&=&\left[\frac{d_{11}}{m^2}\left(\frac{6m^2}{m^2}-\frac{34}{3}\right)+\frac{d_{22}}{3m^2}\left(\frac{26m^2}{m^2}-14\right)-\frac{d_{12}}{3m^2}\left(\frac{44m^2}{m^2}+20\right)\right.\\ & & + \frac{d_1}{m^2}\left(-6m^2+\frac{8m^4}{3m^2}-12m^2\right)d_1+\frac{d_2}{m^2}\left(-\frac{22m^2}{3}-\frac{8m^4}{m^2}+\frac{44m^2}{3}\right)\\ & & + \frac{r}{m^2}\left(14m^2m^2-\frac{8m^4}{3}-\frac{10m^4}{3}\right)\}\\ i_{12}&=&\left[\frac{d_{11}}{m^2}\left(\frac{m^2}{m^2}+\frac{2}{3}\right)+\frac{d_{22}}{m^2}\left(\frac{m^2}{m^2}-4\right)+\frac{d_{12}}{m^2}\left(-\frac{2m^2}{m^2}+\frac{10}{3}\right)\right.\\ & & \left.+\frac{d_1}{m^2}\left(\frac{8m^2}{3}-\frac{2m^2}{3}\right)+\frac{d_2}{m^2}\left(-2m^2+\frac{2m^2}{3}\right)+\frac{5m^2r}{3}\right]\\ i_{13}&=&\left[\frac{d_{11}}{3m^3}\left(-\frac{4m^3}{m^2}-34m\right)+\frac{d_{22}}{3m^3}\left(-\frac{4m^3}{m^2}-26m\right)+\left(\frac{8m^2}{m^4}+\frac{20}{m^2}\right)d_{12}\right.\\ & & + \frac{d_{22}}{m^3}\left(-\frac{16m^5}{3m^4}+\frac{8m^3}{3m^2}+28m\right)+\frac{d_2}{m}\left(\frac{16m^5}{3m^4}-\frac{8m^3}{3m^2}-\frac{46m}{3}\right)\\ & & \left.+\left(-\frac{28m^4}{3m^2}-\frac{8m^3}{3m}-2mm\right)r\right]\\ i_{14}&=&\left[\frac{d_{11}}{m^2}\left(-\frac{m^4}{m^4}+\frac{4}{3}\right)+\frac{d_{22}}{m^2}\left(-\frac{m^2}{m^2}+2\right)+\frac{d_{12}}{m^2}\left(\frac{2m^2}{m^2}-\frac{10}{3}\right)\right.\\ & & \left.+\frac{d_1}{m^2}\frac{13m^2}{3}-\frac{d_2}{m^2}\frac{5m^2}{3}+\left(-\frac{8m^2}{3}+\frac{2m^2}{3}\right)r\right]\\ i_{15}&=&\left[\frac{5m}{3m^3}d_{11}-\frac{m}{3m^3}d_{22}-\frac{4m}{3m^3}d_{12}+\frac{d_2}{m}\left(-\frac{2m^3}{m^2}+\frac{2m}{3}\right)d_2+\left(-\frac{2m^3}{m}+mm\right)r\right]\\ i_{23}&=&\left[\left(\frac{7m^3}{3m^5}+\frac{3m^2}{4m^4}+\frac{5m}{12m^3}\right)d_{11}+\left(\frac{7m^3}{3m^5}-\frac{3m^2}{6m^4}\right)d_{22}\right.\\ & & + \left(-\frac{14m^3}{3m^5}+\frac{m}{3m^3}-\frac{3m^2}{4m^4}\right)d_{12}\\ & & + \left(-\frac{7m^3}{6m^3}+\frac{m}{6m}-\frac{3m^4}{8m^4}+\frac{3m^2}{8m^2}\right)d_1+d_2\left(\frac{7m^3}{6m^3}-\frac{m}{m}+\frac{3m^4}{8m^4}\right)\\ & & \left.+\left(\frac{5m^3}{6m}-\frac{mm}{4}+\frac{3m^4}{16m^2}\right)r\right];\\ i_{24}&=&\left[\frac{13m^3}{m}r+\frac{d_{11}}{m^3}\left(-\frac{3m^3}{2m^2}+\frac{7m}{3}\right)-\frac{d_{22}}{m^3}\left(\frac{m^3}{3m^2}+6m\right)+\frac{d_{12}}{m^3}\left(\frac{3m^3}{m^2}+\frac{11m}{3}\right)\right.\\ & & \left.+\frac{d_1}{3m}\left(-\frac{25m^3}{m^2}-5m\right)+\frac{d_2}{3m}\left(\frac{25m^3}{m^2}-19m\right)\right]\\ i_{25}&=&\left[\frac{d_{11}}{m^2}\left(-\frac{3m^2}{2m^2}+\frac{4m^4}{m^4}+1\right)+\frac{d_{22}}{m^2}\left(-\frac{7m^2}{2m^2}-\frac{4m^4}{3m^4}+\frac{10}{3}\right)\right.\\ & & + \frac{d_1}{m^2}\left(\frac{7m^2}{6}+\frac{m^4}{m^2}-10\frac{4m^2}{3}-\frac{8m^6}{3m^4}\right)+\frac{d_2}{m^2}\left(\frac{8m^2}{3}-\frac{m^4}{3m^2}+\frac{8m^6}{3m^4}-8m^2\right)\\ & & \left.+\frac{d_1d_2}{m^2}\left(\frac{5m^2}{m^2}-\frac{8m^4}{3m^4}+2\right)+\left(-\frac{23m^2}{6}+\frac{19m^4}{3m^2}+\frac{8m^6}{3m^4}+\frac{2m^2}{3}\right)r\right]\\ i_{34}&=&\left[-\frac{m^3}{m^5}\left(d_{11}+d_{22}\right)+\frac{4m^3}{m^5}d_{12}\right]\\ i_{35}&=&\left[\frac{d_{11}}{m^2}\left(\frac{m^4}{m^4}+\frac{28m^2}{m^2}-\frac{16m^4}{m^4}-\frac{14}{3}\right)+\frac{d_{22}}{m^2}\left(\frac{m^2}{m^2}-\frac{8m^4}{3m^4}+\frac{16m^3}{m^3}-\frac{2}{3}\right)\right.\\ & & + \frac{d_1}{m^2}\left(\frac{2m^2}{3}-\frac{10m^3}{m}-\frac{2m^4}{3m^2}+\frac{16m^6}{3m^4}+\frac{22m^2}{3}\right)+\frac{d_2}{m^2}\left(2m^2-\frac{12m^2}{3}+\frac{26m^4}{3m^2}-\frac{2m^4}{3m^2}-\frac{16m^6}{3m^4}\right)\\ & & \left.+\frac{d_{12}}{m^2}\left(-\frac{2m^4}{3m^4}+\frac{50m^2}{3m^2}+\frac{4}{3}\right)+\frac{r}{m^2}\left(-2m^4-\frac{8m^4}{3}-\frac{14m^6}{3m^2}-\frac{10m^2m^2}{3}\right)\right]\\ i_{45}&=&\left[\frac{d_{11}}{m^3}\left(-\frac{8m^3}{m^2}-\frac{8m}{3}\right)+\frac{d_{22}}{m^3}\left(-\frac{8m^3}{m^2}+\frac{16m}{3}\right)+\frac{d_{12}}{m^3}\left(\frac{16m^3}{m^2}-\frac{8m}{3}\right)\right.\\ & & \left.\frac{d_1}{m^2}\left(\frac{8m^3}{m^2}-\frac{4m}{3}\right)+\frac{d_2}{m}\left(-\frac{8m^3}{m^2}+\frac{4m}{3}\right)\right ] \label{eq : i20}\end{aligned}\ ] ] @xmath179 here @xmath180 are the analytic results of contour integrations over the off - shell parameter @xmath141 in the complex @xmath54-plane . the results of these integrals are given as : @xmath181 final result for the bs normalizer has the form , @xmath182k_0\left(\frac{m^2}{2\beta^2}\right)\right.\nonumber\\ & & + [ g_{13}(m , m , a ) + g_{11}(m , m , a)\beta^2 + g_9(m , m , a)\beta^4 + g_7(m , m , a)\beta^6]k_1\left(\frac{m^2}{2\beta^2}\right)\nonumber\\ & & + \left[h_5(m , m)\beta^6u\left(\frac{1}{2},-3,\frac{m^2}{\beta^2}\right)\right.\nonumber\\ & & + h_7(m , m)\beta^4u\left(\frac{1}{2},-2,\frac{m^2}{\beta^2}\right)\nonumber\\ & & \left.\left.+h_{11}(m , m)u\left(\frac{1}{2},0,\frac{m^2}{\beta^2}\right)\right]\beta^2a_3a_5\right\ } \label{eq : formnvm2}.\end{aligned}\ ] ] here , @xmath183 is the second class modified bessel function , @xmath184 is the confluent hypergeometric function , @xmath185 and @xmath186 are polynomials of @xmath187-th degree in @xmath188 and @xmath35 , and the @xmath185 are quadratic functions of the @xmath171 coefficients . in these expressions , @xmath49 is a decaying function of @xmath189 . thus despite the fact that the integrands contain growing factors like @xmath189 , the overall integrals converge , and can be analytically integrated . then , the @xmath190 @xmath158 can be expressed in the following analytic form , @xmath191\right\}\\ \nonumber f_v^4&=&-\frac{a_4}{m^3}n_v\frac{4\pi^{1/4}}{\sqrt{3}}\sqrt{\beta } \left\{m^2m^2 e^{m^2/(4\beta^2)}k_1\left(\frac{m^2}{4\beta^2}\right ) + 2\sqrt{\pi}\beta^3\left[3\sqrt{2}m-8\beta u\left(-\frac{3}{2},-2,\frac{m^2}{2\beta^2}\right ) \right]\right\}\\ f_v^5&=&0 \label{eq : formfv}\end{aligned}\ ] ] with @xmath98 given in the previous equation . eq . ( 3.18 ) which expresses decay constants @xmath136 of vector mesons in terms of the parameters @xmath192@xmath115 , @xmath116 , @xmath193 , @xmath194 , @xmath195 , @xmath196 is a non linear function of the @xmath171 s . then numerical methods must be applied to solve the problem of determining the best values of the @xmath171 s . we used a simple ` mathematica ' procedure for searching for accurate values of the @xmath171 ( @xmath197 ) . we defined the following auxiliary function @xmath198 , which is positive definite , as , @xmath199 ^ 2 . \label{eq:3.14}\ ] ] the summation in the above equation runs over the five ground state vector mesons @xmath0 , @xmath1 , @xmath2 , @xmath3 and @xmath4 studied in this work . @xmath200 are the central values of experimental data on decay constants used ( indicated in table 2 ) which are calculable from the data on the partial widths , @xmath201 for the five studied mesons in @xcite . their results are the following : @xmath0(770 ) : [email protected] kev , @xmath1(782 ) : [email protected] kev , @xmath2(1020 ) : [email protected] kev , @xmath3(1s ) : [email protected] kev , @xmath4(1s ) : [email protected] kev . they are related to our decay constants @xmath136 by formula , @xmath203 here , @xmath35 is the meson mass , @xmath204 is the qed coupling constant ( _ i.e. _ the fine structure constant ) , @xmath126 plays the role of effective electric charge of the meson and has values as listed after eq.(3.2)for different vector mesons . from formula ( [ eq : width2decay ] ) were obtained the data cited in table 1 . we see that the error bars of experimental data on decay widths given by pdg tables for the @xmath0 , @xmath1 , @xmath2 , @xmath3 and @xmath4 mesons represent , 0.8% , 3.3% , 3.1% , 1.8% , 1.5% , respectively . we can say that average error bars are 2.1% or 0.05 kev . the average error in decay constants derived from experimental data of the five mesons studied is 1% . as a general rule , while we are averaging data coming from different measurements , we should give larger weight to more precise data and lesser weight to less precise data . for instance , @xmath0 s experimental error bar of @xmath136 is 0.4% , while for @xmath1 meson error bar of @xmath136 is 1.6% ; _ i.e. _ @xmath0 has a higher precision than @xmath1 meson . however , by using the @xmath198 function , we have fit the data for all mesons from @xmath0 to @xmath4 at the same level . this means that we can not expect a fitting with an average error significantly less than @xmath205% . from the numerical point of view , the problem reduces to finding values of @xmath171 s such that @xmath198 has a minimum , and that at such a minimum it takes the value zero . we used the mathematica package which has some useful functions for minimizing . it is clear that the 6-dimensional hypersurface @xmath198 has many minima , but the only acceptable minima are those minima for which @xmath198 is very close to zero . the experimental values of the @xmath136 s are 0.220 , 0.195 , 0.228 , 0.410 and 0.708 gev . then , we adopt as criterium of `` sufficient closeness '' to zero the value of @xmath198 which is less than @xmath206 gev@xmath9 . besides , it is important to ensure a degree of robustness of the solution . it means , that if we change each of the experimental values @xmath200 according to their error bars , then the values of the @xmath171 which minimize @xmath198 remain near the corresponding values which minimize at the central values @xmath200 . this is the main criterion of stability which must be satisfied for the model to be physically acceptable . in other words , when point @xmath207 is within a box determined by the averages and the error bars of the @xmath171 , the function @xmath198 has a minimum very near to zero for @xmath136 in a box determined by the experimental data of the @xmath136 and their error bars . an additional check is done by evaluating the percent average of the absolute values of the differences between the predicted @xmath136 values from the experimental value @xmath200 . using this method , we found that the values of coefficients should respectively be : @xmath208 , @xmath209 , @xmath210 , @xmath211 , @xmath212 , @xmath213 to predict the decay constant values , @xmath214 gev , @xmath215 gev , @xmath216 gev , @xmath217 gev , and @xmath218 gev . these decay constant values have an average error with respect to the experimental data of 4% . the robustness of the model can be quantified in the following way . the set of @xmath171 which , when replaced in @xmath219 give a value `` near '' the experimental value of the decay constant for each of the studied vector mesons . we found that point @xmath207 is located within certain `` box '' of sides @xmath220 . the center of the box is at the values given in the last paragraph . all those results were obtained by randomly choosing 22 sets of the @xmath136 within their experimental error bars and finding in each case the point @xmath221 where @xmath198 has a minimum near to zero . 2 represents the @xmath136 amplitudes for all five studied mesons , sets of dots were obtained by varying the @xmath171 coefficients around their average values , procedure was done to show stability of results . the normalization factors @xmath163 were found to be : @xmath222 gev@xmath223 , @xmath224 gev@xmath223 , @xmath225 gev@xmath223 , @xmath226 gev@xmath223 , and @xmath227 gev@xmath223 . values of @xmath136 along with the contributions from various covariants and experimental results are listed in table 2 . comparison of our results with those of other models and data is presented in table 3 . formulas found in section [ sec : theory ] express decay constants @xmath136 of vector mesons in terms of the constant parameters @xmath115 , @xmath116 , @xmath193 , @xmath194 , @xmath195 , @xmath228 . our model should be capable of predicting the values of those parameters if one uses the known experimental values of the decay constants for the @xmath0 , @xmath1 , @xmath2 , @xmath3 , @xmath229 and @xmath4 mesons . expression for @xmath136 in eq . ( 3.12 ) and eq . ( 3.18 ) is a linear function of the @xmath171 s ( for i=0,1,2,3,4,5 ) . however @xmath24 involves the bs normalizer @xmath163 , which is evaluated by integrating with respect to @xmath230 , is a highly non - linear function of the @xmath171 s . analytical expressions for the @xmath171 s as functions of quark masses and other parameters corresponding to each of the vector mesons can not be obtained . however , numerical methods give acceptable solutions of the problem . expression for normalizer ( [ eq : formnvm2 ] ) has the form , @xmath231 where matrix elements @xmath232 were given in eq . ( [ eq : i20 ] ) . contributions @xmath157 to the decay constants , given by eq . ( @xmath233 ) , by definition are proportional to @xmath171 , @xmath234 our idea is that the @xmath171 can be obtained by fitting formulas to experimental results , @xmath235 ( see eq . ( [ eq : nvm2 ] ) and eq . ( [ eq : i20 ] ) ) . we must notice that [ eq : equationai ] is a homogenous function of the @xmath171 , fact that precludes finding solutions for the @xmath236 by using available data of the five considered mesons . however , a nonhomogenous system of equations can be constructed by putting @xmath208 and leaving as unknowns @xmath237 . with this procedure it is sufficient to consider the five mesons @xmath0 , @xmath1 , @xmath2 , @xmath3 , and @xmath4 . system of five equations is obtained by introducing in eq . ( [ eq : equationai ] ) the appropriate parameters of the five mesons . for the @xmath200 we used the results from pdg tables @xcite . we see that the `` error bars '' on data for @xmath24 for @xmath10 and @xmath4 mesons are 0.4% , 1.7% , 1.6% , 0.9% and 0.8% respectively , whose average is about 1.1% . system of five equations has many solutions , several of them complex . complex solutions appear in conjugate pairs . available algorithms allow finding all solutions with very high precision . in this way the percent average of the absolute values of the difference between the calculated @xmath136 and the @xmath238 experimentally found , is much lower than the error bars of experiments . however the values of the @xmath171 found were complex , which lead to complex @xmath136 s . we made different checks and selected the parameter set giving the results shown in table 2 which predicts the experimental values of all the @xmath136 s with less precision . the @xmath136 s predicted for the considered mesons match approximately with the experimental values . regarding meson @xmath229 , our theoretical value is 0.1798 gev , while value calculated from experiments is 0.2200 gev , with error 18% . [ cols="<,^,^,^,^,^,^",options="header " , ] in this paper we have calculated the decay constants @xmath136 of vector mesons @xmath0 , @xmath1 , @xmath2 , @xmath3 and @xmath4 in bse under covariant instantaneous ansatz ( cia ) using hadron - quark vertex function which incorporates various dirac covariants order - by - order in powers of inverse of meson mass within its structure in accordance with a recently proposed power counting rule from their complete set . this power counting rule suggests that the maximum contribution to any meson observable should come from dirac structures associated with leading order ( lo ) terms alone , followed by dirac structures associated with next - to - leading order ( nlo ) terms in the vertex function . incorporation of all these covariants is found to bring calculated @xmath136 values much closer to results of experimental data @xcite and some recent calculations @xcite for @xmath0 , @xmath1 , @xmath2 , @xmath3 and @xmath4 mesons . the @xmath136 predicted are within the error bars of experimental data for each one of these five mesons . the results for @xmath0 , @xmath1 , @xmath2 , @xmath3 and @xmath4 mesons with parameter set : @xmath239 , @xmath240 , @xmath241 , @xmath242 , @xmath243 , @xmath244 ( giving @xmath136 values with average error with respect to experimental data of 4% ) are presented in table 2 . comparison with experimental data and other models is shown in table 2 . in fig . 3 we are plotting vs @xmath230 the integrands of @xmath245 , @xmath246 and @xmath136 for each of studied mesons . those plots show that the contribution to @xmath136 from nlo covariants is smaller than the contribution from lo covariants for @xmath0 , @xmath1 , @xmath2 , @xmath3 and @xmath4 mesons . and for @xmath3 and @xmath4 mesons , nlo contribution is negligible in comparison to lo contribution . then , it is concluded from table 2 that as far as the various contributions to decay constants @xmath136 are concerned , for @xmath0 and @xmath1 mesons , the lo terms contribute only 55% , while nlo terms 45% . however as one goes to @xmath2 meson , the lo contribution increases to 62.8% , while nlo contribution is 37.2% . but as one goes to heavy ( @xmath247 and @xmath248 ) mesons , for @xmath3 meson , lo contribution is 84.8% , while nlo contribution is 15.2% , and for @xmath4 meson the lo contribution is 92.6% , while nlo contribution reduces to just 7.4% . thus the drop in contribution to decay constants from nlo covariants vis - a - vis lo covariants is more pronounced for heavy mesons @xmath3 and @xmath4 . and among the two lo covariants , it can be seen that the most leading covariant @xmath249 contributes the maximum for all vector mesons from @xmath0 to @xmath4 . these results on decay constants @xmath24 for vector mesons are completely in conformity with the corresponding results on decay constants @xmath23 for pseudoscalar mesons , @xmath250 and @xmath251 done recently@xcite where it was also noticed that the nlo contribution is much smaller than the lo contribution for hevier mesons like @xmath252 and @xmath251 , where the contribution drops from 10% to 4% , and the most leading covariant was found to be @xmath253 . this is in conformity with the power counting rule according to which the leading order covariants , @xmath254 and @xmath255 ( associated with coefficients @xmath115 and @xmath116 ) should contribute maximum to decay constants followed by the next - to - leading order covariants , @xmath256 and @xmath257(1/m^2)$ ] ( associated with coefficients @xmath258 and @xmath194 ) in the bs wave function . we observe in fig . 2 that though lo and nlo dirac covariants are sufficient to correctly predict amplitudes for @xmath2 and @xmath259 vector mesons , but only lo covariants are sufficient for @xmath4 meson . but for @xmath0 and @xmath1 mesons , the lo and nlo dirac covariants are not sufficient to predict accurately their amplitudes and is thus necessary to include even higher order nnlo dirac covariants in their hadron - quark vertex functions . this can also be seen from fig . however , the numerical results for @xmath136 for equal mass vector mesons , obtained in our framework with use of leading order dirac covariants along with the next to leading order dirac covariants , along with a similar calculation for @xmath23 done recently @xcite for pseudoscalar mesons demonstrates the validity of our power counting rule , which also provides a practical means of incorporating various dirac covariants in the bs wave function of a hadron . by this rule , we also get to understand the relative importance of various covariants to calculation of meson observables . this would in turn help in obtaining a better understanding of the hadron structure . one of the authors ( sb ) is thankful to prof . li , shandong university , china for discussions . jm thanks the support from programa de sostenibilidad university of antioquia . r.alkofer , p.watson , h.weigel , phys . rev . d*65 * , 094026 ( 2002 ) . m.a.ivanov , yu a. kalinovski , c.d . roberts , phys . rev . d*60 * , 034018 ( 1999 ) . f.e.close , a.donnachi , yu s. kalashnikova , phys . rev . d*65 * , 092003 ( 2002 ) . h-m.li , s-l.wan , chin . . lett . * 25 * , 1239 ( 2008 ) . g.cvetic et al . , phys . b*596 * , 84 ( 2004 ) . d.s.hwang , g.h.kim , phys . rev . d*55 * , 6944 ( 1997 ) . r.alkofer , l.w.smekel , phys . rep . * 353 * , 281 ( 2001 ) . g.l.wang , intl . j. mod . a*23 * , 3263 ( 2008 ) . a. n. mitra , b. m. sodermark , nucl . a*695 * , 328 ( 2001 ) and references therein . a.n.mitra , pinsa * 65 * , 527 ( 1999 ) . s. bhatnagar , d. s. kulshreshtha , a. n. mitra , phys . lett . b*263 * , 485 ( 1991 ) . a. n. mitra , s. bhatnagar , intl . j. a*7 * , 121 ( 1992 ) . s.bhatnagar , s - y . li , j. phys . g*32 * , 949 ( 2006 ) s. bhatnagar , s-y.li , j. mahecha , intl . j. mod e*20*,1437 ( 2011 ) . c.h.l.smith , ann . phys . * 53 * , 521 ( 1969 ) . f.t.hawes , m.a.pichowsky , arxiv : nucl - th/9806025 , 1998 . e.mengesha , s.bhatnagar , intl . j. phys . e*20 * , 2521 ( 2011 ) . s.bhatnagar , intl . j. e14 , 909 ( 2005 ) . k.k.gupta et al . d42 , 1604 ( 1990 ) . s.bhatnagar and s.y.li , _ in the proceedings of 9th workshop on non - perturbative quantum chromodynamics , paris , france , 4 - 8 jun 2007 , pp 12_. d.arndt , c-r.ji , _ light - cone quark model analysis of radially excited pseudoscalar and vector mesons_. arxiv : hep - ph/9905360v1 , 1999 . j.beringen et al . , ( particle data group ) , phys . rev . d*86 * , 010001 ( 2012 ) .

starting from the linearized fluctuating boussinesq equations we derive an expression for the structure factor of fluids in stationary convection - free thermal nonequilibrium states , taking into account both gravity and finite - size effects . it is demonstrated how the combined effects of gravity and finite size causes the structure factor to go through a maximum value as a function of the wave number . the appearance of this maximum is associated with a crossover from a dependence for larger to a dependence for very small . the relevance of this theoretical result for the interpretation of light scattering and shadowgraph experiments is elucidated . the relationship with studies on various aspects of the problem by other investigators is discussed . the paper thus provides a unified treatment for dealing with fluctuations in fluid layers subjected to a stationary temperature gradient regardless of the sign of the rayleigh number , provided that is smaller than the critical value associated with the appearance of rayleigh - bnard convection . boussinesq equations , light scattering , nonequilibrium fluctuations , rayleigh - bnard convection , shadowgraph experiments , swift - hohenberg equation 05.40 , 05.70.l , 44.25 , 47.20 , 78.35

cond-mat0109250

questions concerning the nature of thermal fluctuations in fluids in thermal nonequilibrium states have been the subject of many studies during the past two decades . specifically , in this paper we consider thermal fluctuations in a horizontal layer of a one - component fluid that is heated either from above or from below in such a way that the fluid is still in a stable convection - free state . a number of different and independent approaches for dealing with various aspects of this problem have appeared in the literature . most of the earlier studies focused on the nonequilibrium fluctuations close to the convective rayleigh - bnard instability because of their possible influence on the pattern - selection process when convection appears @xcite . the effects of both linear and nonlinear terms in the hydrodynamic fluctuations close to the convection threshold have been evaluated . this line of research turned out to be very important , mainly because it lead to the development of convection model equations for dealing with spatiotemporal - pattern formation above the threshold , as reviewed by cross and hohenberg @xcite . a second and independent line of research developed when investigators became interested in the nature of the fluctuations in nonequilibrium steady states far away from any hydrodynamic instability . the first important theoretical work along this line was done by kirkpatrick et al . , who calculated the correct expression for the structure factor of a fluid subjected to a stationary temperature gradient @xcite . althought their result was originally obtained on the basis of the same mode coupling theory that in equilibrium leads to long - time tail contributions to the correlation functions for the transport coefficients , it turned out that the same result could also be obtained from the simpler scheme of landau s fluctuating hydrodynamics @xcite . this line of research received an important momentum from the experimental side when the research group of sengers and coworkers verified from light - scattering experiments the main conclusion of the theoretical result , namely that the nonequilibrium contribution to the structure factor of a liquid is proportional to the square of the temperature gradient @xmath5 and inversely proportional to the fourth power of the wave number @xmath0 of the fluctuations @xcite . theory and experiments have subsequently been extended to also deal with nonequilibrium fluctuations in liquid mixtures and polymer solutions , as reviewed elsewhere @xcite . a third line of independent research has focused on the real - space behavior of the nonequilibrium correlations resulting from the heat - diffusion equation @xcite . a study of the correlations resulting from the heat - diffusion equation is equivalent to a study of the nonequilibrium correlations in a liquid in the direction co - incident with the temperature gradient . this line of research has shown how the nonequilibrium correlation function of the fluctuating variables encompasses the entire system not involving any intrinsic lenght scales . the real - space analysis has demonstrated the long - range nature of the nonequilibrium fluctuations in the direction of the temperature gradient , which has been further confirmed by numerical integration of the equations @xcite , by computer simulations @xcite and by a lattice - gas automaton approach @xcite . it is our present goal to further extend the second aforementioned research so as to describe nonequilibrium fluctuations far away from any convective instability , nonequilibrium fluctuations close to the convective instability and nonequilibrium fluctuations produced by the heat - diffusion equation from one unified point of view . the original work of kirkpatrick et al . and of others @xcite yielded an expression for the structure factor of a fluid in thermally nonequilibrium states without considering any gravity or boundary effects . segr at al . extended the theory to include gravity effects , first in a one - component liquid @xcite and then in a liquid mixture @xcite . they predicted that gravity will cause the singular @xmath1 dependence of the structure factor to saturate at small values of @xmath0 . this prediction was confirmed experimentally by vailati and giglio from light - scattering experiments at ultra - small scattering angles @xcite . more recently , ortiz de zrate et al . evaluated the nonequilibrium structure factor including finite - size effects but neglecting gravity @xcite . they concluded that deviations from the @xmath1 dependence due to finite - size effects will be just as important as deviations due to gravity . hence , for a quantitative interpretation of ultra - low - angle light scattering or shadowgraph experiments it is important to account for both gravity and finite - size effects . in this paper we extend the previous work so as to include gravity and finite - size effects simultaneously . starting from the linearized fluctuating boussinesq equations , we shall derive a complete expression for the structure factor of the nonequilibrium fluid making no other approximations than the ones incorporated in the boussinesq equations . we shall then show that the resulting expression is capable of describing nonequilibrium fluctuations both far away and close to the convective instability . close to the convective instability we shall recover the linear swift - hohenberg correlation function as an approximation to our more complete result . from the study of the behavior of the nonequilibrium correlation function in real space , we shall recover with appropriate approximations , expressions and plots previously obtained by other investigators for the case of the heat - diffusion equation . thus it will be shown how these various conclusions can be reached from one single theoretical result . attempts to arrive at a unified description of nonequilibrium fluctuations both far away and close to the convective instability has been made by kirkpatrick and cohen @xcite and by schmitz and cohen @xcite , but the consequences of their theoretical results for the interpretation of actual experiments are not clear . in contrast , we shall be able to make a connection with available experimental data for the thermal fluctuations near the convective rayleigh - bnard instability @xcite . there has been an extensive debate whether fluctuating hydrodynamics can properly account for the thermal fluctuations close to the onset of convection @xcite , but the most recent answer to this question appears to be positive @xcite . the present paper will provide additional evidence for the validity of fluctuating hydrodynamics to describe nonequilibrium fluctuations close to the onset of the classical rayleigh - bnard instability in simple fluids . we shall proceed as follows . we start from the linear fluctuating boussinesq equations which will be written in a form most suitable for our analysis in section [ sec:0 ] . in section [ sec:1 ] we review the derivation of the well - known expression for the structure factor of a fluid subjected to a stationary temperature gradient without taking into account the presence of boundaries , but including the effects of gravity . the relationship of the resulting expression for the effect of gravity with that obtained previously by segr et al . @xcite will be elucidated . in section [ sec:2 ] we then consider the modifications to the nonequilibrium structure factor due to the finite height of the fluid layer . in section [ sec : ls ] we present a detailed analysis of how the combined effects of gravity and finite - size will appear in low - angle light scattering and in shadowgraph experiments . sections [ sec:2 ] and [ sec : ls ] contain the most important new features of the present work . in section [ sec : ins ] we introduce an approximation valid close to the convective instability and show how one can recover the linear swift - hohenberg model for the fluctuations near the convective threshold . in section [ sec : rs ] we discuss the spatial behavior of the nonequilibrium structure factor and we shall make contact with the results obtained by previous investigators for the long - range nature of the nonequilibrium fluctuations in the direction of the temperature gradient . our results are summarized in section [ sec : con ] . we consider a fluid layer between two horizontal plates separated by a distance @xmath6 . the fluid layer is subjected to a temperature gradient in the vertical direction by maintaining the plates at two different temperatures . the size of the system in the two horizontal x- and y - directions is much larger than the size @xmath6 in the vertical z - direction . to determine the structure factor of the fluid we consider small fluctuations around the conductive solution . these small fluctuations shall be described by the linearized boussinesq equations supplemented with random noise terms , as first considered by zaitsev and shliomis and by swift and hohenberg for studying the influence of thermal noise close to the convective instability @xcite . use of the boussinesq approximation to the full hydrodynamics equations implies that we neglect the sound modes and consider only density fluctuations caused by temperature fluctuations . the temperature gradient is applied along the z - direction and is written in the form : @xmath7 where @xmath8 is the average local value of the temperature @xmath9 and where @xmath10 . hence , @xmath11 is positive when the fluid layer is heated from above and negative when heated from below . the gravitational force @xmath12 is directed in the negative z - direction : @xmath13 where @xmath14 is the gravitational acceleration constant . in this notation , the rayleigh number may be defined as : @xmath15 where @xmath16 is the thermal expansion coefficient , @xmath17 the kinematic viscosity and @xmath18 the thermal diffusivity of the fluid . we shall evaluate the structure factor of the fluid maintained in a convection - free thermal nonequilibrium state , where the average value @xmath19 of the fluid velocity @xmath20 will be zero . such states correspond to both negative and positive values of the rayleigh number @xmath3 as long as @xmath3 is smaller than the critical value @xmath4 . for this purpose we write the linearized fluctuating boussinesq equations in the form @xcite : [ rb ] @xmath21 @xmath22 where @xmath23 represents the local fluctuating temperature and @xmath24 is the fluctuating z - component of the fluid velocity @xmath25 . to eliminate the stationary pressure gradient from the equations we find it convenient to consider eq . ( [ rb:1 ] ) for @xmath26 , rather than an equation for the fluctuating fluid velocity @xmath20 itself @xcite . finally , @xmath27 and @xmath28 represent the contributions from rapidly varying short - range fluctuations and are related to landau s random stress tensor @xmath29 and random heat flow @xmath30 in such a way that @xcite : [ deff ] @xmath31\right\}_z \label{deff : a}\ ] ] @xmath32 where @xmath33 and @xmath34 are the density and the thermal conductivity of the fluid , while the subscript @xmath35 in eq . ( [ deff : a ] ) indicates that @xmath27 has to be identified with the z - component of the vector between the curly brackets . we note that we are here considering only the effects of additive noise represented by @xmath27 and @xmath28 in eqs . ( [ rb ] ) . multiplicative noise has been also considered in the literature @xcite , but that subject lies outside the scope of the present paper . since @xmath36 and @xmath37 everywhere in the fluid layer , we note that the solution of the fluctuating boussinesq equations will be independent of the sign adopted for @xmath38 or @xmath24 @xcite . the conditions for the validity of the boussinesq equations , also referred to as oberbeck - boussinesq equations , have been discussed extensively in the literature @xcite . specifically , in deriving the boussinesq equations from the more general hydrodynamic equations , one neglects the temperature dependence of the various thermophysical properties , except for the density which is taken to vary linearly with temperature . deviations of this so - called boussinesq approximation do affect the characteristic nature of pattern formation upon the appearance of convection @xcite . however , the intensity of the nonequilibrium fluctuations in convection - free states below the convective instability appears to be much less sensitive to deviations from the boussinesq approximation . segr et al . have analyzed and measured the temperature fluctuations in liquid toluene subjected to temperature gradients up to 220 k @xmath39 @xcite . at the higher value of the temperature gradient some of the thermophysical properties varied significantly over the height of the fluid layer nevertheless , the intensity of the observed nonequilibrium temperature fluctuations was equal to the intensity predicted when all thermophysical - property values are taken at their average value in the fluid layer . similar results were obtained by ahlers , measuring the critical rayleigh number for the onset of rayleigh - bnard convection and the heat transfer slightly above the threshold @xcite . we note that in the derivation of the boussinesq equations it is also assumed that the adiabatic temperature gradient @xmath40 , where @xmath41 is the isobaric specific heat capacity and @xmath42 the average temperature in the fluid layer , is small and can be neglected compared to the magnitude of the imposed temperature gradient . in practice , this is a very good approximation @xcite . in addition , in the boussinesq eq . ( [ rb:2 ] ) the coefficient multiplying @xmath43 is usually identified with the thermal diffusivity of the fluid @xmath18 . for consistency , we have also expressed the prefactor of the second random noise term in eq . ( [ deff : b ] ) in terms of the same diffusivity @xmath18 @xcite . starting from the fluctuating boussinesq equations , we first derive an expression for the structure factor of a fluid subjected to a stationary temperature gradient in the presence of gravity but in the absence of any boundary conditions . althought the resulting expression for such a bulk structure factor will only be meaningful for negative rayleigh numbers , it will enable us to develop a procedure most suited for subsequent incorporation of boundary effects . in the absence of any boundary conditions , a temporal and spatial fourier transformation can be applied to eqs . ( [ rb ] ) to obtain a set of equations for the fluctuations in the vertical component of the velocity @xmath44 and for the fluctuations in the temperature @xmath45 as a function of the frequency @xmath46 and the wavevector @xmath47 : @xmath48 where @xmath49 represents the magnitude of the component of the wavevector @xmath50 in the xy - plane , i.e. , the component of @xmath51 perpendicular to the temperature gradient . we note that in this paper the parallel and perpendicular directions are defined with respect to the horizontal plane , as done by most researchers in the field @xcite , but unlike the notation used by segr et al . the fourier transforms @xmath52 and @xmath53 of the random noise terms can be expressed as a function of the fourier transforms of the random stress tensor and the random heat flow : @xmath54_z - q_z { \mathbf q}\cdot\left[{\mathbf q}\cdot\delta{\mathsf t}(\omega,{\mathbf q})\right]\right\},\ ] ] @xmath55 the solution for @xmath44 and @xmath45 can be readily obtained by inverting the matrix in eq . ( [ bulk1 ] ) , so that : @xmath56~\left[\omega - \mathrm{i } \omega_-(\mathbf{q})\right ] } \begin{pmatrix } f_1 ( \omega,{\mathbf q } ) \\ f_2 ( \omega,{\mathbf q } ) \end{pmatrix } , \label{matrizinversa}\ ] ] where we have introduced the quantities : @xmath57 . \label{defomega}\ ] ] stricly speaking , in the expression ( [ defomega ] ) for the decay rates @xmath58 the temperature gradient @xmath11 should be identified with the effective temperature gradient @xmath59 , as shown by segr et al . however , as mentioned in section [ sec:0 ] , the contribution @xmath40 from the adiabatic temperature gradient is neglected in the boussinesq approximation . in the limit @xmath60 , we have @xmath61 and @xmath62 , which are the decay rates associated with the transverse - velocity fluctuations @xmath24 and the temperature fluctuations @xmath38 , respectively . in the boussinesq approximation , density fluctuations are caused only by temperature fluctuations , while pressure fluctuations are neglected . consequently , the structure factor @xmath63 will be related to the autocorrelation function of the temperature fluctuations : @xmath64 to deduce the autocorrelation function @xmath65 of the temperature fluctuations from eqs . ( [ matrizinversa ] ) , we need the correlation functions for the langevin noise terms @xmath66 and @xmath67 , defined by eqs . ( [ deff ] ) . in nonequilibrium fluctuating hydrodynamics it is assumed that the random stress tensor and the random heat flow retain their local - equilibrium values @xcite . thus , as in equilibrium , we shall consider both @xmath68 as @xmath69 as white " noise . their autocorrelation functions are short ranged in time and in space , being proportional to delta functions . fourier transforming the explicit expressions for the correlation functions of @xmath70 and @xmath71 as , for instance , given by eqs . ( 3.12 ) in ref . @xcite , and using the definition ( [ deff ] ) of @xmath72 and @xmath73 , we obtain : @xmath74 where @xmath75 represents boltzmann s constant . combining eqs . ( [ matrizinversa ] ) , ( [ defs ] ) and ( [ corrf ] ) we obtain for the dynamic structure factor @xmath76 of a nonequilibrium fluid in the boussinesq approximation the following expression : @xmath77~\left[\omega^2+\omega_-^2(\mathbf{q})\right ] } + \frac{2 \nu ( c_p / t)~\nabla t_0 ^ 2 q_\parallel^2 } { \left[\omega^2+\omega_+^2(\mathbf{q})\right]~\left[\omega^2+\omega_-^2(\mathbf{q})\right ] } \right\ } , \label{bulk2}\ ] ] where @xmath78 represents the intensity of the fluctuations in thermodynamic equilibrium : @xmath79 here @xmath80 denotes the heat - capacity ratio @xmath81 and @xmath82 is the isothermal compressibility . in deriving eq . ( [ bulk2 ] ) we have employed the thermodynamic relation @xmath83 \lambda \kappa_t / t$ ] . the equation for the rayleigh spectrum of a fluid subjected to a stationary temperature gradient including the contribution from gravity was first derived by segr et al . we recover the same result , but without the ( negligible ) contribution from the adiabatic temperature gradient . in the limit @xmath60 , eq . ( [ bulk2 ] ) reduces to the expression of the rayleigh spectrum of a fluid subjected to a stationary temperature gradient , first obtained by kirkpatrick et al . @xcite , and subsequently reproduced by other investigators @xcite . the nonequilibrium structure factor , as given by eq . ( [ bulk2 ] ) , is anisotropic , and it depends on the magnitude of the horizontal component @xmath84 of the scattering vector directly and through the decay rates @xmath85 . in this paper we shall focus our attention on the static structure factor : @xmath86 which determines the total intensity of rayleigh scattering @xcite . for subsequent use , we find it convenient to introduce dimensionless wave numbers @xmath87 and @xmath88 , where @xmath6 is the finite height of the fluid layer . upon substituting eq . ( [ defomega ] ) for @xmath85 into eq . ( [ bulk2 ] ) and integrating the resulting expression for @xmath89 , we obtain : @xmath90 in eq . ( [ bulk3 ] ) @xmath91 represents the strength of the nonequilibrium enhancement of the structure factor , which is given by : @xmath92 where @xmath93 is the prandtl number . since we have not yet considered any finite - size effects , eqs . ( [ bulk3 ] ) and ( [ strength ] ) do not depend explicitly on the height @xmath6 . equations ( [ bulk3 ] ) and ( [ strength ] ) are identical to eq . ( 2.35 ) in ref . the term in eq . ( [ strength ] ) proportional to @xmath3 is related to the adiabatic temperature gradient and is in practice negligibly small . hence , the intensity of nonequilibrium fluctuations continues to be proportional to @xmath94 in the presence of gravity . note that ( [ bulk3 ] ) can only be valid for negative rayleigh number @xmath3 , i.e. , when the fluid layer is heated from above so that @xmath95 . for any @xmath96 , the nonequilibrium contribution to the structure factor will always diverge at some finite value of the horizontal component @xmath84 of the scattering vector . the appearance of this divergence is a direct consequence of us having performed the calculation without taking into account boundary conditions . as will be demonstrated in the next section , when boundary conditions are taken into account , a different expression for @xmath97 is obtained . this new expression will be valid not only for negative rayleigh numbers , but also for a finite interval of positive rayleigh numbers , up to some critical rayleigh number @xmath4 . since in practice the fluid layer is confined between two horizontal plates separated by a ( small ) distance @xmath6 , the nonequilibrium structure factor will be affected by the presence of boundary conditions in the z - direction . in some previous publications , we have evaluated finite - size effects on the nonequilibrium structure factor neglecting gravity @xcite . a major conclusion of this previous work is that finite - size effects appear at wave numbers where in practice the nonequilibrium structure factor will also be affected by gravity . hence , for the interpretation of experimental measurements it becomes imperative to incorporate both finite - size effects and gravity effects simultaneously . this task is implemented in the present section . as in our previous publication @xcite , we again apply a fourier transformation of the fluctuating boussinesq equations ( [ rb ] ) in space and in time , but restricting the spatial fourier transformation to the xy - plane . we thus obtain the following set of linear stochastic differential equations : @xmath98 - \nu~\left[\cfrac{d^2}{dz^2}-q_\parallel^2\right]^2 & \alpha~g~q_\parallel^2 \\ \nabla t_0 & \mathrm{i}~\omega - d_t~\left[\cfrac{d^2}{dz^2}-q_\parallel^2\right ] \end{pmatrix } \begin{pmatrix } w(\omega,{\mathbf q_\parallel},z ) \\ \theta ( \omega,{\mathbf q_\parallel},z ) \end{pmatrix } = \begin{pmatrix } f_1 ( \omega,{\mathbf q_\parallel},z ) \\ f_2 ( \omega,{\mathbf q_\parallel},z ) \end{pmatrix } , \label{finite1}\ ] ] where @xmath99 is the wavevector in the xy - plane . the random noise terms @xmath100 and @xmath101 in eqs . ( [ finite1 ] ) are related to the partial fourier transforms @xmath102 of the random stress tensor and @xmath103 of the random heat flux . the actual expressions are a bit complicated and can be found elsewhere @xcite . as is often done in the literature @xcite , for the sake of simplicity , we adopt stress - free boundary conditions for the vertical velocity and perfectly conducting walls for the temperature , so that : @xmath104 note that these boundary conditions imply the absence of any possible fluctuations in the temperature and velocity of the fluid adjacent to the walls . as is well known , stress - free boundary conditions represent a fluid bounded by two free surfaces , which is a rather unrealistic case @xcite . for the realistic case of a fluid confined between two rigid surfaces , no - slip boundary conditions in the velocity are more appropriate . for the particular case of @xmath105 , we have evaluated the nonequilibrium structure factor using both stress - free @xcite and no - slip @xcite boundary conditions . while there are numerical differences between the two solutions , the dependence of the nonequilibrium structure factor on the wave number @xmath0 appears to be qualitatively the same for the two sets of boundary conditions . for mathematical simplicity , we evaluate here the nonequilibrium structure factor using stress - free boundary conditions in the expectation that also in the presence of gravity the more realistic no - slip boundary conditions will not modify the qualitative nature of the wave - number dependence of the structure factor . to search for a solution of eq . ( [ finite1 ] ) , we represent @xmath106 and @xmath107 as a series expansion in a complete set of eigenfunctions of the differential operator in eq . ( [ finite1 ] ) , satisfying the boundary conditions ( [ sf ] ) @xcite . because of the simplicity of the boundary conditions considered here , an appropriate set of eigenfunctions is the fourier sine basis in the @xmath108 $ ] interval @xcite . we thus assume : @xmath109 to deduce the coefficients @xmath110 and @xmath111 from eq . ( [ finite1 ] ) , we also have to represent the random noise terms @xmath112 and @xmath113 as a fourier sine series : @xmath114 where we have introduced the set of random functions @xmath115 and @xmath116 , which are the projections of the random noise terms over the eigenfunction basis . they are given by : @xmath117 representing the random noise terms by eq . ( [ p11 ] ) , one readily deduces from eq . ( [ finite1 ] ) expressions for the coefficients of the fourier series @xmath110 and @xmath111 . to calculate the structure factor in the boussinesq approximation , we only need the coefficients @xmath111 for the temperature fluctuations . substituting eqs . ( [ p10 ] ) and ( [ p11 ] ) into eq . ( [ finite1 ] ) we obtain : @xmath118 f_{2,n}(\omega,\mathbf{q}_\parallel ) } { \left(\frac{n^2\pi^2}{l^2}+q_\parallel^2\right ) \left(\omega - \mathrm{i}~\omega_+(n , q_\parallel)\right ) \left(\omega - \mathrm{i}~\omega_-(n , q_\parallel)\right)},\ ] ] where , similarly to eq . ( [ defomega ] ) , we have introduced decay rates @xmath119 , which are now given by : @xmath120.\ ] ] since the differential operator in eq . ( [ finite1 ] ) depends only on the modulus of @xmath99 , the problem has cylindrical symmetry and the solution will only depend on the magnitude @xmath84 of the two - dimensional wavevector @xmath99 . in analogy to eq . ( [ defs ] ) , the relationship between the dynamic structure factor @xmath121 and the autocorrelation function of the temperature fluctuations is now given by : @xmath122 to obtain the structure factor @xmath121 we need the correlation functions between the different projections of the random noise terms . these have been calculated in a previous publication with the result @xcite : @xmath123 as in the case of the absence of any boundary conditions , we continue to assume that the correlation functions between the different components of the random current tensor and the random heat flux retain their equilibrium values . this assumption remains valid as long as @xmath6 is a macroscopic distance , much larger than the molecular distances in the fluid . in this paper we are interested in the static structure factor @xmath124 . with the information presented above , this quantity can be readily calculated . in analogy to eq . ( [ bulk3 ] ) for @xmath125 in the absence of any boundary conditions , the modified nonequilibrium structure factor can be written in the form : @xmath126 , \label{g23}\ ] ] where @xmath78 and @xmath91 are again given by eqs . ( [ se ] ) and ( [ strength ] ) , respectively , while @xmath127 incorporates the finite - size effects and is given by : @xmath128 the first term in eq . ( [ g23 ] ) is again the static structure factor of a fluid in thermodynamic equilibrium , which is not affected by any finite - size effects @xcite . the second term in eq . ( [ g23 ] ) represents the nonequilibrium enhancement of the structure factor . this nonequilibrium enhancement is proportional to @xmath94 through the expression ( [ strength ] ) for @xmath129 , it depends on the gravitational acceleration constant @xmath14 through the appearance of the rayleigh number in eqs . ( [ strength ] ) and ( [ g23 ] ) , and it depends on the finite height @xmath6 of the fluid layer explicitly in eq . ( [ g23 ] ) and also through @xmath88 . in the limit @xmath130 we recover the expression for @xmath131 presented in a previous publication @xcite . it is important to note that eq . ( [ g24 ] ) is valid for both negative and positive rayleigh numbers , provided that @xmath132 for @xmath133 there always exist values of @xmath84 for which right - hand side of eq . ( [ g24 ] ) diverges . of course , the value @xmath134 equals the well - known value obtained form a linear stability analysis of the boussinesq equations with no - slip boundary conditions @xcite . as expected , for @xmath135 the present calculation breaks down because the fluctuations around the conductive state are no longer small . equation ( [ g24 ] ) for the normalized nonequilibrium enhancement @xmath136 represents our principal result for the combined effects of gravity and finite - size on the nonequilibrium structure factor of the fluid . the remaining part of this paper is concerned with an analysis of some of the physical consequences that follow from eq . ( [ g24 ] ) . the double fourier series contained in eq . ( [ g24 ] ) , when it is convergent , can be easily summed . to do so , we first find the roots of the denominator for @xmath137 . we then split the inverse of the product of three binomials as a sum of three terms , and we finally obtain for @xmath136 : @xmath138 } , \label{g25}\ ] ] where @xmath139 for @xmath140 are the three complex cubic roots of @xmath3 : @xmath141 in eq . ( [ defla ] ) @xmath142 represents the phase of @xmath3 when considered as a complex number : @xmath143 for @xmath96 and @xmath144 for @xmath145 . this procedure is only valid for @xmath146 . however , the finite - size effects on the structure factor for the particular case @xmath147 have already been considered in a previous publication @xcite , where the sum of the series in eq . ( [ g24 ] ) for this particular case has been performed by a different method . hence , we do not need to re - evaluate @xmath148 for the case @xmath147 . to implement the summation in eq . ( [ g25 ] ) we consider the relation : @xmath149 this equation can be obtained from formula 1.445 in ref . @xcite and it is valid for @xmath150 $ ] and for any complex number @xmath151 , provided that in the right - hand side of ( [ g26 ] ) the square root of @xmath151 with positive real part is chosen . now , the sum of the double fourier series contained in eq . ( [ g24 ] ) can be performed , but the result is a long expression , not particularly informative , which can be easily obtained by the reader . explicit expressions to be obtained upon further integration of @xmath152 will be presented in the following section . the nonequilibrium fluctuations can be obserbed in small - angle light - scattering experiments . for a discussion of such light - scattering experiments we consider an experimental arrangement like the ones employed at the university of maryland @xcite and by vailati and giglio at the university of milan @xcite . a schematic representation of such a light - scattering experiment is shown in fig . [ segunda ] . the scattering medium is a thin horizontal fluid layer bounded by two parallel plates whose temperatures can be controlled independently so as to establish a temperature gradient across the fluid layer . the temperature gradient can be parallel or antiparallel to the direction of gravity . the horizontal plates are furnished with windows allowing laser light to propagate through the fluid in the direction parallel to the gravity and to the temperature gradient . light scattered over an angle @xmath153 arises from fluctuations with a wave number such that @xcite : @xmath154 where @xmath155 is the wave number of the incident light inside the scattering medium . to observe any nonequilibrium fluctuations one needs to observe the scattered light at small wave numbers and , hence , at small scattering angles . from electromagnetic theory @xcite it follows that the scattering intensity @xmath125 is obtained from an integration of the structure factor over the scattering volume , so that @xcite : @xmath156 in eq . ( [ p31 ] ) we have assumed that the scattering volume extends over the full height of the fluid layer as is the case in small - angle light scattering from a thin fluid layer @xcite . in this situation scattered light received in the collecting pinhole of the detector indeed arises from all points illuminated by the laser beam inside the fluid layer . from eq . ( [ p30 ] ) and the geometrical arrangement shown in fig . [ segunda ] , we note that @xmath84 and @xmath157 in an actual light - scattering experiment are not independent variables , because they are related to the scattering angle , @xmath153 , by : @xmath158 equation ( [ p32 ] ) shows that for actual small - angle light - scattering experiments one may use the approximation @xmath159 , @xmath160 . an alternative promising experimental technique for measuring the intensity , though not the temporal spectrum , of nonequilibrium fluctuations is quantitative shadowgraph analysis @xcite . the experimental arrangement is similar to the one depicted in fig . [ segunda ] but , instead of a laser beam , an extended uniform monochromatic light source is employed to illuminate the sample . then many shadowgraph images of a plane perpendicular to the temperature gradient are obtained with a ccd detector , which measures the spatial distribution of intensity @xmath161 , where @xmath162 is a two - dimensional position vector in the imaging plane . for each image a shadowgraph signal is defined by : @xmath163 where @xmath164 is the blank intensity distribution , when there are no fluctuations in the index of refraction of the sample . in practice @xmath164 is calculated by averaging over many shadowgraph images , so that fluctuations cancel out and the resulting @xmath164 contains only contributions from nonuniform illumination of the sample . from physical and geometrical optics one can demonstrate that the modulus squared of the two - dimensional fourier transform of the shadowgraph signal , @xmath165 , after taking an azimuthal average , can be expressed as @xcite : @xmath166 where the overline indicates the azimuthal average , in which case the result depends only on the modulus @xmath84 of the two - dimensional fourier vector @xmath50 . the factor @xmath167 represents an optical transfer function , which can be derived from the optical arrangement used to produce the shadowgraph pictures ; it includes contributions from the response of the ccd detector and the density dependence of the refractive index @xcite . therefore , applying a two - dimensional fourier transform to the shadowgraph images , one can deduce the structure factor of the fluid as a function of the wave number @xmath0 at @xmath168 . there exist an equivalence between small - angle light scattering and shadowgraphy , in the sense that both methods give us @xmath169 . for light scattering , @xmath0 is the scattering wavevector as given by ( [ p30 ] ) , whereas for shadowgraph techniques @xmath0 is the modulus of the two - dimensional fourier vector in the imaging plane . as mentioned by bodenschatz , what it is measured in the experiments is a kind of vertical average of the fluctuations @xcite . to obtain an explicit expression for the structure factor @xmath170 , we substitute eq . ( [ g23 ] ) with @xmath127 given by eq . ( [ g24 ] ) into eq . ( [ p31 ] ) , implement the summation by using the relation ( [ g26 ] ) and perform the double integration in eq . ( [ p31 ] ) . introducing again dimensionless wave numbers @xmath88 and @xmath171 , we thus obtain : @xmath172 with @xmath173 ^ 2 } ~\frac{\cos{(\tilde{q}_\perp)}-\cos{\left(\tilde{q}_\parallel^{1/3 } \sqrt{\lambda_j -\tilde{q}_\parallel^{4/3}}\right ) } } { \sin{\left(\tilde{q}_\parallel^{1/3 } \sqrt{\lambda_j -\tilde{q}_\parallel^{4/3}}\right)}}. \nonumber\end{aligned}\ ] ] in eq . ( [ g29 ] ) , @xmath78 and @xmath174 are again given by eqs . ( [ se ] ) and ( [ strength ] ) , respectively , while @xmath139 in eq . ( [ g30 ] ) again represent the three complex cubic roots of @xmath3 defined by eq . ( [ defla ] ) . in thermodynamic equilibrium @xmath175 and @xmath147 , so that @xmath176 and we recover from eq . ( [ g29 ] ) the well - known isotropic equilibrium structure factor given by eq . ( [ se ] ) . if we retain only the first term on the right - hand side of eq . ( [ g30 ] ) and substitute it in eq . ( [ g29 ] ) we recover the expression ( [ bulk3 ] ) for the bulk structure factor of the fluid . as noted in section [ sec:1 ] , this expression for the bulk structure factor breaks down for any positive value of the rayleigh number because of a divergence at @xmath177 . the second term in eq . ( [ g30 ] ) for @xmath178 , containing a sum over the three complex cubic roots of @xmath3 , represents the finite - size effects on the nonequilibrium structure factor and depends on the height @xmath6 of the fluid layer . although perhaps not immediately transparent , it can be verified that this second term when multiplied by @xmath91 vanishes in the limit @xmath179 . moreover , for any finite value of @xmath6 , the divergence of the first term in eq . ( [ g30 ] ) at @xmath177 for positive @xmath3 is now exactly compensated by a similar divergence in the second term of eq . ( [ g30 ] ) for the same wave number , since @xmath180 for positive @xmath3 . thus @xmath178 , as given by eq . ( [ g30 ] ) , is continuous at @xmath177 for positive @xmath3 . however , due to the presence of a sine term in the denominator , new divergences appear when : @xmath181 for @xmath182 . for @xmath183 this new divergence is again compensated due to the presence of the term @xmath184 in the numerator . but for @xmath185 , there is no compensation and the contribution ( [ g30 ] ) to the nonequilibrium enhancement of the structure factor diverges . since condition ( [ g31 ] ) for @xmath185 is satisfied when @xmath186 , this divergence will appear for any rayleigh number larger than the critical rayleigh number @xmath187 . for @xmath188 , condition ( [ g31 ] ) is satisfied at a critical wave number @xmath189 . we thus recover from an analysis of the divergences in the nonequilibrium structure factor the well - known results of the linear stability theory for a fluid layer bounded by two free surfaces @xcite . a more detailed analysis of the behavior of our solution for the nonequilibrium structure factor for positive @xmath3 near @xmath4 will be presented in section [ sec : ins ] . as mentioned earlier , for the interpretation of low - angle scattering or shadowgraph experiments one may consider the approximation @xmath190 , @xmath191 . in the limit @xmath192 , ( [ g30 ] ) for @xmath178 can be slightly simplified ; in this approximation the nonequilibrium enhancement @xmath193 will depend only on @xmath194 , which is the experimentally relevant quantity . in the remainder of this paper the small - angle approximation will always be assumed . we note that all the divergent features discussed above remain the same in this approximation . it is interesting to look at the asymptotic behavior of the nonequilibrium enhancement for @xmath195 and for @xmath196 . from eq . ( [ g30 ] ) it can be readily shown that : @xmath197 and @xmath198 we note from eqs . ( [ tes5:7 ] ) and ( [ tes5:8 ] ) that the limiting behavior of @xmath193 for @xmath195 and @xmath196 does not depend on the value of the rayleigh number @xmath3 . in accordance with eq . ( [ g29 ] ) , the product @xmath199 represents the nonequilibrium enhancement of the structure factor with @xmath91 given by eq . ( [ strength ] ) and @xmath193 given by eq . ( [ g30 ] ) with @xmath200 . from eq . ( [ strength ] ) we see that the prefactor @xmath91 depends on various thermophysical properties of the fluid and on the magnitude of the applied temperature gradient , but @xmath91 is independent of the wave number @xmath194 . hence , to evaluate the wave - number dependence of the nonequilibrium enhancement of the structure factor , we only need to focus our attention on @xmath193 . assuming @xmath201 , we have evaluated @xmath193 from eq . ( [ g30 ] ) . in fig . [ tercera ] we show on a double - logarithmic scale @xmath193 as a function of @xmath194 for three different values of the rayleigh number , namely for a large negative value , @xmath202 , for a value @xmath203 which corresponds to the case investigated in a previous publication @xcite and for a value @xmath204 which is close to the critical value @xmath205 . at larger values of @xmath194 , @xmath193 varies as @xmath206 , independent of the rayleigh number , in agreement with eq . ( [ tes5:8 ] ) . upon decrease of @xmath194 , @xmath193 goes through a maximum and for very small values of @xmath194 , @xmath193 decreases as @xmath207 , in agreement with eq . ( [ tes5:7 ] ) . for positive @xmath3 , @xmath193 develops a prominent peak close to @xmath208 , which diverges as @xmath209 , as further discussed in section [ sec : ins1 ] . hence , a major effect of the additive noise terms in the fluctuating boussinesq equations is the appearance of ( fluctuating ) patterns in the fluid , even below the convective instability as discussed by some other investigators @xcite . sengers and coworkers have measured the nonequilibrium fluctuations in liquid toluene @xcite and in liquid @xmath210-hexane @xcite . these experiments correspond to rayleigh numbers from @xmath211 to @xmath212 at dimensionless wave numbers ranging from @xmath213 down to @xmath214 . the experiments have provided an accurate confirmation of the @xmath1 dependence of the intensity of nonequilibrium fluctuations in this range of wave numbers . giglio and coworkers have measured the intensity of nonequilibrium fluctuations for negative rayleigh numbers down to wave numbers @xmath194 of order unity , both with ultra - low angle light - scattering experiments @xcite and from quantitative analysis of shadowgraph images @xcite . they actually measured the intensity of nonequilibrium _ concentration _ fluctuations in a liquid mixture . however , due to the similar structure of the underlying hydrodynamic equations , the @xmath0 dependence of the contribution of nonequilibrium concentration fluctuations to the structure factor in a liquid mixture is expected to be similar to the @xmath0 dependence of the contribution of nonequilibrium temperature fluctuations to the structure factor of a one - component fluid @xcite . giglio and coworkers have not only confirmed the @xmath1 dependence of the nonequilibrium structure factor , but they have also observed the crossover to a region of @xmath194 close to unity where the nonequilibrium structure factor is independent of @xmath194 , in agreement with the flat region indicated in fig . [ tercera ] for @xmath193 at large negative rayleigh numbers . wu et al . have measured the nonequilibrium structure factor in a layer of fluid carbon dioxide , at a pressure of about 3 mpa , near the convective instability . for positive rayleigh numbers , @xmath193 strongly depends on the parameter @xmath215 which measures the distance from the rayleigh - bnard instability . in fig . [ tercerab ] we show @xmath193 as a function of @xmath194 close to @xmath208 as calculated from eq . ( [ g30 ] ) for @xmath216 , @xmath217 and @xmath218 . we did not attempt to make a quantitative comparison with the measurements of wu et al . , because the slip - free boundary conditions adopted in this paper do not correspond to the experimental boundary conditions . however , a comparison of fig . [ tercerab ] with fig . 3 in ref . @xcite shows that our solution yields a good qualitative representation of the @xmath0 dependence of the nonequilibrium structure factor as measured by wu et al . we conclude that our solution of the linearized fluctuating boussinesq equations for the nonequilibrium structure factor is consistent with the characteristic features of the nonequilibrium structure factor observed in experiments for both negative and positive rayleigh numbers . the nature of thermal noise near the convective instability has been the subject of studies by many investigators @xcite . hence , it is of interest to make a comparison of those results with our solution for the intensity of temperature fluctuations for thermal nonequilibrium states . it turns out that for an analysis of @xmath193 near the convective instability it is more convenient not to perform the summation of the series in eq . ( [ g24 ] ) . instead , substituting eq . ( [ g24 ] ) into eq . ( [ p31 ] ) and integrating the terms in the series individually , we can write @xmath193 as : @xmath219 which is equivalent to the expression ( [ g30 ] ) . zaitsev and shliomis @xcite were the first to compute thermal fluctuations in a fluid layer subjected to a stationary temperature gradient near the convective instability . using linear perturbation theory they found that the structure factor diverges as @xmath220 . the same divergence follows from our solution for @xmath193 . to reproduce the divergence we first calculate the value @xmath221 corresponding to the maximum of @xmath193 by expanding @xmath221 in powers of @xmath215 around @xmath222 . differentiating the expression ( [ g30b ] ) for @xmath193 and demanding the derivative to vanish , we find that the position of the maximum corresponds to : @xmath223 ^ 2 } + o(\epsilon^3 ) \nonumber \\ & \approx&\frac{\pi}{\sqrt{2}}+6.324\times 10^{-4}~\epsilon^2 . \label{div1}\end{aligned}\ ] ] substituting eq . ( [ div1 ] ) into eq . ( [ g30b ] ) we conclude that the structure factor , which is proportional to the intensity of the scattered light , diverges when the convective instability is approached such that : @xmath224}+o(\epsilon ) . \label{div2}\ ] ] we thus recover the linear divergence of @xmath148 as a function of @xmath225 obtained by zaitsev and shilomis @xcite and confirmed by swift and hohenberg @xcite . linear fluctuation theory amounts to a mean - field theory of fluctuations . extremely close to the instability nonlinear effects will cause a smearing out of the sharp mean - field transition @xcite , but this effect will only be noticeable for very small values of @xmath226 @xcite hence , observation of this linear divergence is possible in experiments @xcite . deviations from linear fluctuation theory have been observed by scherer et al . in the case of electroconvection @xcite . the approximation scheme used by zaitsev and shliomis and by swift and hohenberg is equivalent to retaining only the term @xmath185 in the series expansion ( [ g30b ] ) for @xmath193 . note that , for @xmath185 , when @xmath3 is close to @xmath4 and @xmath84 is close to @xmath227 , the denominator in the corresponding term in eq . ( [ g24 ] ) approaches zero . therefore , in this situation the term for @xmath185 is much larger than the terms for any other value of @xmath228 . consequently , when @xmath229 and @xmath230 , truncating the series ( [ g30b ] ) at @xmath185 yields a very good approximation . we thus deduce from our solution for the structure factor as measured in low - angle light scattering or shadowgraph experiments : @xmath231 where the equilibrium contribution to @xmath232 has been neglected ( cf . , eq . ( [ g29 ] ) ) . this is usually called the most - unstable - mode approximation in the literature @xcite . the denominator in eq . ( [ g33b ] ) is zero at @xmath233 and @xmath234 . expanding the denominator in powers of @xmath235 around @xmath236 , one obtains : @xmath237 where , following the notation of hohenberg and swift @xcite , we have introduced the parameter @xmath238 . equation ( [ g34 ] ) is proportional to the correlation function of the swift - hohenberg ( sh ) model _ below _ the convective instability , when nonlinear terms are negligible @xcite . the results of sh refer to the autocorrelation of a field @xmath239 , which is a generalized two - dimensional order parameter introduced by them to describe the convective instability as a nonequilibrium phase transition . if we compare the term @xmath185 in eq . ( [ g24 ] ) with eq . ( 2.10 ) in ref . @xcite , we identify our @xmath240 as proportional to the equal - time autocorrelation function of the field @xmath142 . therefore the sh model can be obtained as an approximation close to the convective instability from our exact results for @xmath241 , as given by eqs . ( [ g29 ] ) and ( [ g30 ] ) . as discussed by sh , eq . ( [ g34 ] ) is valid for both stress - free " and no - slip " boundary conditions , but the numerical values of the constants @xmath4 , @xmath242 and @xmath243 do depend on the boundary conditions ( see table i in ref . an interesting feature of the sh model is that is a lagrangian " model , because the autocorrelation function ( [ g34 ] ) can be obtained from a simple probability functional for the field @xmath244 , as is discussed in detail in the relevant literature @xcite . to compare our more complete solution of the fluctuating boussinesq equations for @xmath193 with the various approximations proposed for the structure factor close to the convective instability , we have plotted in fig . [ cuarta ] @xmath193 as a function of @xmath194 . the solid line represents @xmath193 as given by our exact solution ( [ g30 ] ) , the dotted curve ( hardly visible ) the most - unstable - mode approximation ( [ g33b ] ) and the dashed curve the sh - model approximation ( [ g34 ] ) . all three curves correspond to a fluid layer subjected to a temperature gradient such that @xmath245 . from fig . [ cuarta ] it is evident that eq . ( [ g33b ] ) yields a very good approximation although at large @xmath194 deviations appear . the reason is that @xmath193 for large @xmath194 should vary as @xmath206 , while eq . ( [ g33b ] ) yields @xmath246 for large @xmath194 . many authors have represented the @xmath0 dependence of the structure factor near the instability in terms of a lorentzian profile centered at @xmath247 with a width proportional to @xmath248 @xcite , and wu et al . have also analyzed their experimental data in terms of a lorentzian profile @xcite . however , our actual solution ( [ g30 ] ) does not yield a lorentzian profile , although it can be obtained by expanding the denominator in eq . ( [ g33b ] ) not in powers of @xmath207 around @xmath249 , but in powers of @xmath194 around @xmath247 . the problem is that a lorentzian does not lead to the proper asymptotic behavior for either small @xmath0 or large @xmath0 . it even leads to an apparent divergence when one tries to calculate the power of the fluctuations by integrating @xmath250 over all two - dimensional wave vectors @xmath0 @xcite . we conclude that it would be much better to represent the experimental data in terms of an equation like eq . ( [ g33b ] ) , which is not a lorentzian . another feature of nonequilibrium fluctuations close to the convective instability that has been studied theoretically and experimentally is the so - called power of the thermal fluctuations , @xmath251 . the mean - square amplitude @xmath251 of the temperature fluctuations is related to the integral of the structure factor measured in the experiments : @xmath252 which corresponds to a vertical average of the real - space temperature autocorrelation function . due to the symmetry of the problem this quantity does not depend on the point @xmath253 in the horizontal plane at which is evaluated . in eq . ( [ g35a ] ) the prefactor @xmath254 accounts for the conversion from density fluctuations to temperature fluctuations ( cf . ( [ defs ] ) and ( [ g23 ] ) ) and the factor @xmath255 results from the vertical average ( compare eq . ( [ p31 ] ) with eq . ( a14 ) in ref . @xcite ) . from eq . ( [ g30b ] ) we note that in terms of our solution of the fluctuating boussinesq equations ( neglecting the equilibrium contribution ) : @xmath256 the integral in ( [ noise2 ] ) can be performed analytically , but the result is long and not particularly interesting . asymptotically close to the convective instability we obtain : @xmath257 thus the mean - square amplitude @xmath251 calculated from our solution will diverge as @xmath258 , which is the kind of divergence observed experimentally by ahlers and coworkers @xcite . in eq . ( [ noise3 ] ) , @xmath91 has to be evaluated at the critical temperature gradient . it is interesting to note that the sh approximation gives for @xmath251 the same asymptotic behavior , eq . ( [ noise3 ] ) , than the exact solution . in this section we study the real - space behavior of the autocorrelation function of the nonequilibrium density fluctuations . we also shall make contact with the third approach dealing with nonequilibrium fluctuations that was mentioned in the introduction . to calculate the autocorrelation function in real space we apply an inverse fourier transform in the horizontal plane to the structure factor as given by eq . ( [ g23 ] ) . the equilibrium contribution gives a triple delta function . the nonequilibrium contribution gives a nontrivial contribution , which we denote as @xmath259 and which is equal to : @xmath260 where @xmath261 is the bessel function of the first kind and order zero . note that due to the cylindrical symmetry of the problem the correlation in real space depends only on the cylindrical radial coordinate ( @xmath262 ) . it is difficult to study the integral ( [ r1 ] ) in the general case , but there are a couple of particular cases for which integral ( [ r1 ] ) can be evaluated explicitly or where simple approximations can be made so that the real - space behavior of the correlation function can be studied in detail . the first interesting case for which the long - range part of the real - space correlation function @xmath259 can be studied is for @xmath263 , i.e. , along the vertical z - direction . in this particular case , the correlation function @xmath264 will only depend on the variables @xmath35 and @xmath265 and on the rayleigh number @xmath3 . close to the convective instability , the behavior of the real - space correlation function will be dominated by the most unstable mode . from eq . ( [ g24 ] ) we note that @xmath264 will be proportional to the product @xmath266 . therefore , the concavity of @xmath267 as a function of @xmath35 for fixed value of @xmath265 is negative . for the interesting case @xmath147 , @xmath267 was evaluated exactly in a previous publication @xcite . it turns out that the concavity of @xmath267 changes at @xmath147 , being negative for positive @xmath3 and positive for negative @xmath3 ( except for a small region near @xmath268 ) . in fig . [ figcorr1 ] we have plotted the dimensionless nonequilibrium correlation function , @xmath269 , as a function of @xmath35 , for @xmath270 , @xmath263 and for three values of the rayleigh number @xmath3 . the dotted lines correspond to @xmath147 , the solid curve corresponds to a positive @xmath3 and the dashed curve to a negative @xmath3 . the change of the concavity of the function at @xmath147 is evident . this feature can be investigated analytically since , for small values of the rayleigh number , @xmath267 can be readily expanded in powers of @xmath3 yielding : @xmath271 \right\ } , \label{rs1}\end{aligned}\ ] ] where @xmath272 is the bernoulli polynomial of the 6th degree ( actually , it is possible to obtain a systematic expansion of @xmath267 in terms of the bernoulli polynomials , but the resulting series is not very informative ) . for @xmath147 , we recover from eq . ( [ rs1 ] ) the expression obtained for the spatial dependence of the correlation function in a previous publication ( see eq . ( 25 ) and fig 2 in ref . the change of concavity of @xmath267 at @xmath147 is also evident from eq . ( [ rs1 ] ) . we note that for @xmath147 the function @xmath267 has a kink at @xmath268 , indicating the presence of a discontinuity in the derivative at this point . for any other value of @xmath273 the function and its derivative are continuous . from fig [ figcorr1 ] we also note that the nonequilibrium real - space equal - time correlation function is always positive , has a maximum at @xmath268 and decreases monotonically from this maximum reaching zero at both ends of the @xmath108 $ ] interval . thus , we conclude that the real - space correlation function in the vertical direction is nonlocal , long ranged and does not involve any intrinsic length scale , i.e. , the correlation encompasses the entire system only to be cut off by the finite size of the system itself . hence , the presence of the temperature gradient induces long - range correlations in the direction coincident with the temperature gradient , as has been elucidated by previous investigators , mainly on the basis of the nonequilibrium heat - diffusion equation @xcite . considering only the heat - diffusion equation means that one neglects any coupling between temperature and velocity fluctuations . this approach neglects also any gravity effects and is restricted to fluctuations with @xmath274 . the existence of long - range correlations in the direction of the temperature gradient has been verified by numerical integration @xcite , by computer simulation @xcite , by lattice - gas - automaton dynamics @xcite and by a master - equation approach @xcite . our result for the correlation function along the direction of the temperature gradient is in agreement with that obtained by malek mansour and coworkers @xcite , as is also evident form a comparison of our fig . [ figcorr1 ] for @xmath147 with fig . 1 in ref . @xcite . some of these authors have also considered a coupling between temperature and velocity fluctuations @xcite . but to simplify the hydrodynamic equations they made the assumption that the thermal expansion coefficient of the fluid vanishes . this is equivalent to assuming that density fluctuations are caused only by pressure fluctuations . hence , their calculations are concerned with brillouin scattering , while here we are considering rayleigh scattering , for which accurate experimental information on nonequilibrium fluctuations has been obtained . more recently , bena et al . have studied the velocity autocorrelation function adopting an hydrodynamic approximation known as kolmogorov flow @xcite ; again , due to the nature of the hydrodynamic simplifications performed , the results can not be compared with those obtained in the present paper . a second particular case for which the real - space behavior of the correlations in a liquid subjected to a stationary temperature gradient can be easily studied is in a horizontal plane , that is , in the plane perpendicular to the temperature gradient and far from the boundaries ( @xmath275 ) . we shall analyze here the spatial dependence of the correlations in the horizontal direction near the convective instability using the most - unstable - mode approximation , eq . ( [ g33b ] ) , which , as was shown in section [ sec : ins ] , yields an excellent representation of the structure factor near the convective instability . in this approximation : @xmath276 where @xmath277 . the integral ( [ r5 ] ) can be evaluated exactly , when @xmath273 , but the result is a rather long expression . instead we prefer to plot in fig . [ figcorr2 ] the resulting @xmath278 as a function of @xmath262 for two values of the rayleigh number . the solid curve represents the density autocorrelation function for @xmath217 , which is close to the convective instability ; the dotted curve represents the same function , but for @xmath279 . for @xmath3 well below @xmath4 , the correlation function is overdamped but , as the value of @xmath3 approaches @xmath4 , it starts to oscillate and the number of oscillations before the correlation function decays to zero increases the closer one approaches @xmath4 . to see this mathematically , we have made an asymptotic expansion of @xmath280 for large @xmath262 and , when @xmath3 is close to @xmath281 , we obtain : @xmath282}. \label{r6}\end{aligned}\ ] ] the approximation ( [ r6 ] ) for @xmath217 is shown in fig . [ figcorr2 ] as a dashed curve , to be compared with the solid curve representing the autocorrelation function as given by eq . ( [ r5 ] ) for the same value of @xmath248 . as can be inferred from fig . [ figcorr2 ] , the approximation is excellent for small @xmath248 and for large @xmath283 . from eq . ( [ r6 ] ) , the real - space behavior of the autocorrelation function in the horizontal plane as a kind of damped oscillation is evident . starting from the linearized fluctuating boussinesq equations we have derived the structure factor of a horizontal fluid layer taking into account both gravity and finite - size effcts . the resulting expression reproduces the @xmath1 dependence of the nonequilibrium structure factor predicted theoretically @xcite and confirmed experimentally @xcite for negative rayleigh numbers , it accounts for the saturation of the nonequilibrium enhancement of the intensity of the fluctuations at small wave numbers observed by giglio and coworkers @xcite and it is consistent with the experimental observation by ahlers and coworkers @xcite of the structure factor close to the convective instability . we have thus provided a unified approach for describing nonequilibrium fluctuations for both negative and positive values of the rayleigh number , provided that @xmath273 . our solution also enables us to evaluate the real - space behavior of the correlation function of the temperature fluctuations which agrees with the results obtained by other investigators for the spatial direction co - incident with the direction of gravity @xcite . there are still a number of issues that deserve further attention . first of all , for the sake of mathematical simplicity , we have adopted stress - free boundary conditions in the present paper . for a quantitative comparison with experimental data at very small wave numbers and close to the convective instability modifications due to no - slip boundary conditions should be considered @xcite . second , we want to emphasize that neither our solution nor that of swift and hohenberg ( sh ) close to the convective instability yield a lorentzian wave number dependence on @xmath0 as is often assumed in the literature since a lorentzian does not yield the proper asymptotic behavior for either small or large @xmath0 . moreover , even when the sh equation is not approximated by a lorentzian , it does not yield the expected asymptotic @xmath0 behavior either . we plan to study this issue in more detail in the future . finally , while the average local values @xmath284 and @xmath285 are zero , the average cross correlation @xmath286 outside equilibrium is not generally zero . as a result , the coupling between temperature and transverse - velocity fluctuations may cause a small convective heat transfer , even below the instability @xcite . a satisfactory solution of this effect has not yet been obtained @xcite . the authors acknowledge some valuable discussions with g. ahlers , d.s . cannell , t.r . kirkpatrick , r. prez cordn and m.g . velarde . the research at the university of maryland is supported by the chemical sciences , geosciences and biosciences division of the office of basic energy sciences of the us department of energy under grant no . de - fg-02 - 95er14509 . j.v.s . acknowledges the hospitality of the departamento de fsica aplicada i , universidad complutense de madrid , where the manuscript was completed .

we analyse mid - infrared ( mir ) spectroscopic properties for 19 ultra - luminous infrared quasars ( ir qsos ) in the local universe based on the spectra from the infrared spectrograph on board the _ spitzer space telescope_. the mir properties of ir qsos are compared with those of optically - selected palomar - green qsos ( pg qsos ) and ultra - luminous infrared galaxies ( ulirgs ) . the average mir spectral features from to 30 m , including the spectral slopes , 6.2 m pah emission strengths and [ neii ] 12.81 m luminosities of ir qsos , differ from those of pg qsos . in contrast , ir qsos and ulirgs have comparable pah and [ neii ] luminosities . these results are consistent with ir qsos being at a transitional stage from ulirgs to classical qsos . we also find that the colour index(30 , 15 ) is a good indicator of the relative contribution of starbursts to agns for all qsos . correlations between the [ neii ] 12.81 m and pah 6.2 m luminosities and those between the [ neii ] , pah with 60 m luminosities for ulirgs and ir qsos indicate that both [ neii ] and pah luminosities are approximate star formation rate indicators for ir qsos and starburst - dominated galaxies ; the scatters are , however , quite large ( 0.7 to 0.8 dex ) . finally the correlation between the ew ( pah 6.2 m ) and outflow velocities suggests that star formation activities are suppressed by feedback from agns and/or supernovae . [ firstpage ] galaxies : active galaxies : evolution galaxies : interactions quasars : general infrared : galaxies

0807.3653

since the discovery of ultra - luminous infrared galaxies ( ulirgs , l@xmath3@xmath410@xmath5 ) by the _ infrared astronomical satellite _ ( _ iras _ ) in the 1980 s ( e.g. , * ? ? ? * ) , it is widely accepted that ulirgs result from strong interactions / mergers between gas - rich disk galaxies . these mergers form elliptical galaxies and ulirgs are an important intermediate stage in the process during which at least a fraction of ulirgs manifest as dust - enshrouded qsos ( see , e.g. , * ? ? ? * ; * ? ? ? in addition , active galactic nuclei ( agns ) triggered by mergers tend to appear at the final merging stages ( e.g. , @xcite , @xcite , @xcite , @xcite and reference therein ) . there is mounting evidence that qsos with far - infrared ( fir ) excess have massive starbursts in their host galaxies . for example , @xcite investigated 9 qsos with fir excess and found that their host galaxies are tidally interacting or major merger systems with obvious recent star - forming activities . from the detections of mid - infrared ( mir)/firm , fir 35 - 350 @xmath1 m , and ir 8 - 1000@xmath1 m . ] emissions for felobals ( broad absorption line qsos with low - ionisation and iron absorption lines ) by multiband imaging photometer on _ spitzer _ ( mips ; * ? ? ? * ) on board the _ spitzer space telescope _ @xcite , @xcite find that all their 9 felobals are extremely infrared ( ir ) bright , and concluded that these qsos are in transition from ulirgs to classical qsos with ongoing or recent starbursts , because the iron absorption may be from iron ejected during starbursts . @xcite studied 31 qsos / seyfert1s selected from the local ulirg samples ( termed as ir qsos for simplicity ) . by comparing the fir spectral index of ir qsos with those of optically selected palomar - green qsos ( pg qsos ; @xcite ) , they argued that the fir excess of ir qsos relative to pg qsos is from massive starbursts and inferred star formation rate ( sfr ) in the host galaxies of ir qsos . recently , from studies of z@xmath06 qsos with strong sub - mm emissions , @xcite , @xcite and @xcite concluded that massive starbursts also exist in their host galaxies . the conclusion is consistent with the results of @xcite that high redshift ( sub)mm - loud qsos follow the same trend for fir to bolometric luminosities established by low redshift ir qsos @xcite . all these studies suggest that there exists a transitional stage , during which both the central black hole and the spheroidal component of qso hosts grow rapidly in a coeval fashion . however , there is still a debate about the origin of fir emission from qsos , because one can not firmly exclude the possibility that fir emissions are from dust tori heated by central agns ( for detailed discussions see * ? ? ? moreover , from the molecular gas properties of pg qsos with ir excess and comparisons with ulirgs , @xcite find that the @xmath6 and @xmath7 ratios for pg qsos are higher than those of ulirgs , implying that agns contribute significantly to the dust heating and hence to the fir emission . therefore , other sfr indicators besides the fir emission for qsos are important for further understanding the coeval growth of supermassive black holes and their host galaxies . recently , the quest ( quasar and ulirg evolution study ) group ( see * ? ? ? * ; * ? ? ? * ) reported the detection of polycyclic aromatic hydrocarbon ( pah ) emission features in pg qsos using the infrared spectrograph ( irs ) on _ spitzer _ @xcite . for 11 out of 26 pg qsos pahs have been clearly detected . furthermore , the average spectrum of the undetected 15 pg qsos also shows pah features . since the pah emissions are closely related to star formation , not to agns ( see , e.g. , * ? ? ? * ) , such detections strongly suggest that star formation occurs widely in qsos . their analysis shows that 30% or more fir emission in these pg qsos is from starbursts . furthermore , given that the low - excitation fine - structure emission line [ neii ] 12.81@xmath1 m is one of the dominant emission lines of hii regions and that the pah molecules are easily destroyed by high energy photons from agns ( e.g. , * ? ? ? * ) , [ neii ] emission may be an alternative , perhaps even better , tracer of star formation for qsos ( see 4 ) . in this paper we study the mir spectroscopic properties of ir qsos based on _ spitzer _ irs observations , and examine their connections and evolutionary relations to ulirgs and pg qsos . the sample selection , data acquisition and reduction are described in sect . 2 and 3 . the major results and discussions are given in sect . finally we summarise our results in sect . we adopt cosmological parameters _ km s@xmath9 mpc@xmath9 , @xmath10@xmath11=0.3 , and @xmath10@xmath12=0.7 throughout this paper . ir qsos are defined as type 1 agn with l@xmath3(@xmath13m)@xmath410@xmath14l@xmath15 @xcite . our basic ir qso samples are compiled from ulirg samples with spectroscopic information , plus the ir qsos obtained directly from the cross - correlation of the _ iras _ point - source catalog with the _ rosat _ all - sky survey catalog . the ulirg samples consist of 118 ulirgs from 1@xmath16jy ulirgs survey @xcite and 97 ulirgs from the qdot redshift survey @xcite . the total number of ir qsos is 31 , about one third of all the ir qsos found in a complete redshift survey with 15,411 _ iras _ galaxies and about 900 ulirgs ( pscz ; @xcite ) . thus it should be a representative sample of ir qsos ( see @xcite and @xcite for more detailed descriptions ) . we searched the _ spitzer _ archival data and found that 18 out of 31 ir qsos have been observed by irs and the data are available ( see table 1 ) . notice that 9 of the 10 ir qsos ( out of a total of 118 ulirgs ) from the 1@xmath16jy ulirg sample are included in our sample . the other 9 ir qsos are from qdot ( 4 ) and other qso samples . in addition we include the object iras f14026 + 4341 , which is classified as a hyper - luminous infrared galaxy ( with l@xmath17 @xmath4 10@xmath18l@xmath15 , @xcite ) and a broad absorption line quasar @xcite . our sample includes 90% ( 50% ) ir qsos out of 1@xmath16jy ( qdot ) ulirgs , and thus should be an overall representative sample . we checked that our results are essentially unchanged if we focus only on the 9 ir qsos from the 1@xmath16jy ulirg samples , and thus our compiled sample has no significant biases . 14 of the 19 objects have both low- and high - resolution irs observations , while four ( 3c 48 , iras f02054 + 0835 , pg 1543 + 489 , and iras f20036@xmath191547 ) have only low - resolution observations , and one ( iras f21219@xmath191757 ) has no long - low ( ll ) mode ( see @xmath203.1 ) observation ( see table 3 ) . the irs low - resolution spectra of a sample of palomar - green qsos ( pg qsos ) are retrieved from _ gto and go archival data ( programs 14 , 3187 , 3421 , and 20142 ; see table 4 ) . we remove objects whose mir spectra have a low s / n ratio or redshift larger than 0.27 to guarantee reliable measurements of rest - frame 30@xmath1 m fluxes . the number of pg qsos is 19 ( see table 1 ) , the same as the number of ir qsos . this pg qso sample is mainly used for studying the mir spectral slopes benefiting from the full wavelength coverage from @xmath05 - 30@xmath1 m of their low - resolution spectra . we also collected another pg qso sample with 22 objects studied by @xcite ( after excluding four objects that have been classified as ir qsos and grouped into ir qso sample ) . this second pg qso sample has deep sl mode observations ( 5.2 - 14.5@xmath1 m ) and is thus suitable for studying weak pah emission features in continuum - dominated qsos , and for analysing properties of the mir fine - structure lines ( especially [ neii]12.81@xmath1 m line in this work ) from high - resolution observations . the sample of ultra - luminous infrared galaxies ( ulirgs ) is selected based on the _ iras _ 1 jy sample of ulirgs @xcite , which have optical spectroscopic observations by @xcite and @xcite . their irs low - resolution spectra are retrieved from _ gto archival data ( program i d 105 ; see table 4 ) . the number in our ulirg sample is 35 ( see table 2 ) , including all spectral types except type 1 agns , namely seyfert 2 s , liners , and hiis as classified by @xcite and @xcite from diagnostic diagrams of optical lines . the mir spectra are acquired from the _ spitzer _ archival data using the leopard software ( see table 3 for the integrated exposure times and program ids for ir qsos ) . the data ( versions 13.2 to 15.3 of _ spitzer _ pipeline reduction ) include low - resolution ( short - low [ sl ] & long - low [ ll ] modes , r@xmath060 - 120 and wavelength range : 5.2 - 38.0@xmath1 m ) irs spectra for ir qsos , pg qsos , and ulirgs , and high - resolution ( short - high [ sh ] & long - high [ lh ] modes , r@xmath0600 and wavelength range : 9.9 - 37.2@xmath1 m ) irs spectra for most of the ir qsos . we use the smart software @xcite for data reduction , including the removal of rogue pixels , sky subtraction , and spectral extraction and analysis . the sky backgrounds for low - resolution ( sl & ll modes ) spectra are subtracted by differencing the adjacent sub - slit positions ( 1st & 2nd nods ) . no background subtraction is performed for the high - resolution ( sh & lh modes ) spectra , but this does not affect mir fine - structure line measurements ( see , e.g. , * ? ? ? the slit widths of [email protected] to [email protected] include most of the emission from the qso and its host galaxy , so no aperture corrections are performed . for the low - resolution spectra we use the 12@xmath1 m and 25@xmath1 m flux densities from _ iras _ ( or _ iso _ if the _ iras _ fluxes are upper limits or not available ) to scale the spectra by multiplying by a small factor , which is more suitable for the comparison between mir and fir properties in our statistics . the scaling factors for ir qsos and ulirgs are often close to unity , typically less than 1.1 . however , for pg qsos the scaling factors are larger , typically @xmath01.5 . this may be caused by variabilities of quasars in mir ( e.g. , * ? ? ? * ) and/or contamination of companions ( environments ) for _ iras _ ( or _ iso _ ) measurements . the fluxes of pah emission at 6.2@xmath1 m are measured by integrating the flux above a local continuum from 6.0 - 6.5@xmath1 m approximated by a second - order polynomial ( e.g. , * ? ? ? * ; * ? ? ? the uncertainties ( 1@xmath22 ) in the measurements are 20@xmath23 on average ( varying from @xmath05@xmath23 for pah strong objects to about 50@xmath23 for those with only marginally detectable pah features ) . the equivalent widths ( ew ) of 6.2@xmath1 m pah feature are obtained from dividing the integrated pah flux by the continuum flux density below the peak of the feature . upper limits ( 3@xmath22 ) are given by adopting typical widths of @xmath00.2@xmath1 m for the 6.2@xmath1 m pah feature @xcite , which is similar to the value , @xmath00.6@xmath1 m , used for the 7.7@xmath1 m feature by @xcite . note that we do not fit gaussian or lorentzian profiles to measure pah emissions ( e.g. , * ? ? ? * ; * ? ? ? * ) due to the relative weakness of pah features in qsos compared to their strong dust continuum . since the three ir qsos in the quest sample ( i zw 1 , mrk 1014 , pg 1613 + 518 ) have 7.7/6.2 flux ratios of 4.2 , 4.6 , 4.9 , respectively , similar to that of ngc 6240 ( 4.7 , @xcite ) , we estimate the 6.2@xmath1 m pah fluxes for pg qsos in the @xcite sample ( which has only 7.7@xmath1 m measurements ) by taking a 7.7/6.2 flux ratio of 4.7 . the fluxes of the ionised neon fine - structure lines in the mir ( [ neii ] , [ nev ] , [ neiii ] at 12.81 , 14.32 , and 15.56@xmath1 m ) for ir qsos are measured based on the high - resolution irs spectra , using the idea spectral analysis tool of the smart software . the fluxes are measured by fitting a single gaussian superposed on a local continuum approximated by a second - order polynomial . flux upper limits ( 3@xmath22 ) are derived adopting typical line widths of 600 kms@xmath9 @xcite . the fluxes of the [ neii]12.81@xmath1 m line for some ulirgs in our sample are from @xcite . fig . [ fig_spec ] shows the low - resolution _ spitzer _ irs mir spectra of 18 ir qsos in our sample ( except iras f21219@xmath191757 which has no ll mode observation ) . the dotted and dashed lines show the pah features and the mir fine - structure neon , sulphur , and oxygen lines . the shaded bar denotes the silicate emission / absorption feature centred at 9.7@xmath1 m . one can see from these spectra that the pah features at 6.2 , 7.7 , 8.6 , 11.2 & 12.7 @xmath1 m and the mir fine - structure emission lines , such as [ neii ] 12.81@xmath1 m , [ nev ] 14.32@xmath1 m , [ neiii ] 15.56@xmath1 m , [ nev ] 24.32@xmath1 m & [ oiv ] 25.89@xmath1 m are present in most ir qsos , although some emissions are weak for most of them . one can also see the silicate absorption feature at 9.7@xmath1 m for several ir qsos ( f00275 , f13218 , mrk 231 , f11119 and f15462 ) , which are rarely seen in pg qsos . for comparison , we show the average mir spectra of ir qsos , pg qsos and ulirgs in fig . [ fig_avgspec ] . the average spectra of ulirgs and pg qsos are from @xcite . it is clear from fig . [ fig_avgspec ] that the slope of mir continua from 15@xmath1 m to 30@xmath1 m of ir qsos is intermediate between that of ulirgs and qsos . we also show in the same figure the spectra of two representative ir qsos ( pg 1613 + 658 and mrk 231 ) . their mir spectra are intermediate between the average spectra of ulirgs and pg qsos . however , from the infrared , optical and x - ray observations , pg 1613 + 658 has the characteristics of classical qsos @xcite , while mrk 231 is an on - going merger with high sfr from its hundred - pc scale molecular disk @xcite . in order to clarify more quantitatively the differences in the spectral properties , we study the properties of 6.2@xmath1 m pah , [ neii ] 12.81@xmath1 m luminosities , and mir colour indices @xmath2(30 , 15 ) , for ir qsos , pg qsos , and ulirgs . the infrared colour index is defined as @xmath24 the central wavelengths ( 30 and 15@xmath1 m ) are selected in order to avoid the contamination by most strong spectral features ( e.g. , mir fine - structure lines , pah emissions , silicate emission / absorption features etc . , see also @xcite ) . the histograms for these quantities are shown in fig . [ fig_hist ] . the mean values of @xmath2(30 , 15 ) , ew ( pah 6.2@xmath1 m ) , pah 6.2@xmath1 m and [ neii ] 12.81@xmath1 m luminosities are labelled at the top of each panel of fig . [ fig_hist ] . from the left panels of fig . [ fig_hist ] , the mir spectral slopes for ir qsos are much flatter than those of ulirgs , but are significantly steeper than those of pg qsos . since classical qsos have much lower fir emissions than ir qsos at a given bolometric luminosity @xcite , and @xmath2(30 , 15 ) is correlated with @xmath2(60 , 25 ) ( see fig . [ fig_alpha ] and [ sec : statis ] ) , it is easy to understand why the @xmath2(30 , 15 ) slopes are steeper for ir qsos than those of classical qsos ( for details see [ sec : statis ] ) . on the other hand , the emissions from hot dust heated by central agn are mainly in the mir band . furthermore , _ spitzer _ observations reveal that star formation regions in star - forming galaxies also contribute significantly to the mir continuum emission ( e.g. , * ? ? ? * ; * ? ? ? * ) . therefore , for ir qsos , both starburst and central agn contribute to the mir continuum emission , leading to the mir continuum of ir qsos being stronger than those of ulirgs . thus the slopes of mir to fir continuum of ir qsos are flatter than those of ulirgs . the histograms of ew ( pah 6.2@xmath1 m ) show that the mean value of ew ( pah 6.2@xmath1 m ) of ir qsos ( [email protected]@xmath1 m ) is between those of ulirgs ( [email protected]@xmath1 m ) and pg qsos ( [email protected]@xmath1 m ) and close to that of warm ulirgs ( [email protected]@xmath1 m ; @xcite ) . considering the fraction of pah detections for ir qsos and pg qsos being 70% ( 13/19 ) and 40% ( 11/27 in quest qso sample ) respectively , while the ew ( pah 6.2@xmath1 m ) for pure agn is less than 0.005 - 0.02@xmath1 m @xcite , the mean value of ew ( pah 6.2@xmath1 m ) of ir qsos is significantly larger than that of classical qsos . from the middle and right panels of fig . [ fig_hist ] we see that the mean values of 6.2@xmath1 m pah and [ neii]12.81@xmath1 m luminosities of ir qsos are similar to those of ulirgs , but one order of magnitude higher than those of pg qsos . apparently the pah molecules in ir qsos can survive the fuv and x - ray photons emitted by central agn , suggesting that they are shielded by a large amount of gas and dust from radiation ( see * ? ? ? since the bolometric luminosities ( l@xmath26 ) of pg qsos observed by irs of _ spitzer _ are systematically smaller than those of ir qsos , where l@xmath26 is calculated from the optical continuum emission ( @xmath27 , @xcite ) , the differences in their ir continuum shapes may simply arise because more optically luminous qsos have more extended dust tori . to check this possibility , we investigated the 6.2@xmath1 m pah and [ neii]12.81@xmath1 m luminosities normalised by l@xmath26 . [ fig_hist2 ] shows the distributions of l@xmath28}$]/l@xmath26 and l@xmath29}$]/l@xmath26 for ir qsos and pg qsos , respectively . it is clear from fig . [ fig_hist2 ] that the mean values of l@xmath28}$]/l@xmath26 and l@xmath29}$]/l@xmath26 for ir qsos are also about one order of magnitude higher than those of pg qsos . therefore , the different properties of 6.2@xmath1 m pah and [ neii]12.81@xmath1 emissions between ir qsos and pg qsos are unlikely from different dust tori , instead the differences arise because of different star formation properties ( see section 4.2 ) . in summary , the mid - ir spectroscopic properties , including the continuum slope and emission line strengths , of ir qsos , pg qsos and ulirgs are consistent with that ir qsos are in a transitional phase from ulirgs to classical qsos , confirming the results from previous studies @xcite . in this subsection , we will use the mir spectroscopic features , including the mir continuum slope @xmath2(30 , 15 ) , 6.2@xmath1 m pah and fine - structure emission lines , to investigate the origin of mir emissions of ir qsos . we will also use these properties , combined with fir and optical properties , to disentangle the starburst and agn contributions in these objects . [ fig_alpha ] shows the relations of @xmath2(30 , 15 ) vs. @xmath2(60 , 25 ) ( left panel ) , @xmath2(30 , 15 ) vs. the fir excess l@xmath30/l@xmath26 ( middle panel ) , and @xmath2(30 , 15 ) vs. the ew ( pah 6.2@xmath1 m ) ( right panel ) for both ir qsos and pg qsos . it is clear from fig . [ fig_alpha ] that the colour indices of @xmath2(30 , 15 ) and @xmath2(60 , 25 ) are closely correlated , indicating that @xmath2(30 , 15 ) can express the relative strength of fir to mir emission for qsos . the middle and right panels of fig . [ fig_alpha ] show the correlations between @xmath2(30 , 15 ) with fir excess l@xmath30/l@xmath26 , and between @xmath2(30 , 15 ) with the ew ( pah 6.2@xmath1 m ) . the ew ( pah 6.2@xmath1 m ) is the ratio of pah 6.2@xmath1 m emission line to @xmath06@xmath1 m continuum . since the 6.2@xmath1 m pah emission is from star formation regions , and the 6@xmath1 m continuum traces the agn contribution ( e.g. , * ? ? ? * ) , thus the ew ( pah 6.2@xmath1 m ) expresses the relative contribution of star formation to agn @xcite . in fact , @xcite also found the strong correlation between infrared spectral slope and the ew ( pah 6.2@xmath1 m ) for ulirgs , especially for ulirgs with seyfert 1 and seyfert 2 optical spectra , while our results extend such relation to infrared luminous qsos and pah detected pg qsos . we conclude that @xmath2(30 , 15 ) , fir excess and ew ( pah 6.2@xmath1 m ) can serve as indicators of the relative contributions of starbursts to agns ( @xcite and see below ) . @xcite suggest that the ionised neon fine - structure lines [ neii]12.81@xmath1 m and [ neiii]15.56@xmath1 m can be used as a sfr indicator for star - forming galaxies . @xcite extend this relation to ulirgs . in addition , @xcite found a strong correlation between the far - infrared continuum ( l@xmath30 ) and low - ionisation [ neii ] line emission for both pg qsos and ulirgs , and argued that the [ neii ] line can also be used to estimate the sfr in qso host galaxies . one advantage to use [ neii ] 12.81@xmath1 m as a sfr estimator is that it suffers much less extinction than optical lines , such as @xmath31 and [ oii]3727 . however , there is still a debate about the origin of [ neii ] emission , because the narrow line region of qsos may also contribute substantially ( e.g. , * ? ? ? therefore , it is worth investigating the origin of [ neii]12.81@xmath1 m emission for ir qsos by comparing the multi - wavelength properties of pg qsos , ir qsos and ulirgs . [ fig_ne2pah ] shows the relation between pah 6.2@xmath1 m and [ neii]12.81@xmath1 m luminosities for ir qsos , ulirgs and pg qsos with firmly detected [ neii]12.81@xmath1 m and pah 6.2@xmath1 m emissions . because the pah emissions are purely from star formation regions @xcite , the tight correlation between 6.2@xmath1 m pah and [ neii]12.81@xmath1 m luminosities ( at a statistical level of @xmath4 99.99@xmath23 with the spearman rank - order test ) demonstrates that ( at least part of ) the [ neii ] 12.81@xmath1 m emission is also from star formation regions . note that for the pg qsos shown in fig . [ fig_ne2pah ] , their [ neii ] 12.81@xmath1 m luminosities normalised by the bolometric luminosities of agns ( l@xmath29}$]/l@xmath26 ratios , see below ) are about three times higher than that of pah undetected pg qsos ( see also @xcite ) . therefore , it is likely that the star formation contributes significantly to the [ neii ] emission not only for ir qsos , but also for pg qsos with detectable pah emissions @xcite . the mean values of [email protected]\mu m}$]/l@xmath26 ratios are @xmath32 , @xmath33 , @xmath34 and @xmath35 for ir qsos , pg qsos , pah - detected and pah - undetected pg qsos , respectively . thus for the same bolometric luminosity of a central agn , the mean [ neii]12.81@xmath1 m luminosity of ir qsos is about one order of magnitude higher than that of classical qsos . taken together with the tight correlation between [ neii ] and pah luminosities ( see above ) , we conclude that the [ neii]12.81@xmath1 m emission of ir qsos is mainly from star formation , while the contribution from the narrow line region of agns is not significant ( @xmath36 ) . by comparing the mir spectroscopic properties of ir qsos , ulirgs and pg qsos , we found that the indicators of relative contributions of starbursts to agns ( such as colour index @xmath2(30 , 15 ) and ew [ pah 6.2@xmath1 m ] ) for ir qsos are between those of ulirgs and pg qsos . these results are consistent with the findings of @xcite and @xcite that ( at least some ) infrared luminous qsos ( ir qsos ) are at a transitional stage from ulirgs to classical qsos . below we consider the star formation rates and agn / star formation feedback in more detail . as we argued , the agn contribution to [ neii ] emission for ulirgs and ir qsos is probably very small , and there is a tight correlation between [ neii]12.81@xmath1 m and pah 6.2@xmath1 m luminosities ( see [ sec : statis ] ) . we examine in more detail how they can be used as approximate sfr indicators for ulirgs and ir qsos . [ fig_sfr ] shows the [ neii]12.81@xmath1 m ( top panel ) and pah 6.2@xmath1 m ( bottom panel ) luminosities versus the 60@xmath1 m luminosity for ir qsos , pg qsos and ulirgs with firmly detected [ neii]12.81@xmath1 m and pah 6.2@xmath1 m emissions . a spearman rank - order analysis show that both correlations are significant at @xmath4 99.99@xmath23 level . the dashed lines in fig . [ fig_sfr ] represent the least - squares regression fits : @xmath37 @xmath38 the fitting formula ( 2 ) is consistent with that of @xcite for star - forming galaxies ( within the large errors ) . note that our sample objects have much higher @xmath39 and [ neii ] luminosities than their star - forming galaxies . thus both [ neii ] 12.81@xmath1 m and pah 6.2@xmath1 m luminosities can be used as approximate sfr indicators not only for normal star - forming galaxies , but also for galaxies with high infrared luminosities , such as ulirgs and ir qsos ( see @xcite and @xcite ) . however , the mean scatters ( about 0.7 to 0.8 dex ) in the relation of [ neii]12.81@xmath1 m , pah 6.2@xmath1 m with @xmath39 luminosities are larger than that ( about 0.6 dex ) of star - forming galaxies with lower infrared luminosity ( @xcite ) . comparing fig . [ fig_sfr ] with fig . [ fig_ne2pah ] , one can see that the scatter in the relation of [ neii]12.81@xmath1 m vs. pah 6.2@xmath1 m luminosities ( about 0.6 dex ) is smaller than that in the relations of [ neii]12.81@xmath1 m , pah 6.2@xmath1 m luminosities with l(60@xmath1 m ) . it is also clear that most large scatters are from ulirgs . this is perhaps not surprising since the range in the 9.7@xmath1 m silicate absorption depth among ulirgs is quite large ( see @xcite , @xcite ) . in short , the large scatters for ulirgs seen in the relations may be due to complicated , patchy extinctions among these galaxies in the mir band . a detailed discussion on extinction for ulirgs can be found in @xcite . one explanation for the observed correlation between spheroidal and black hole mass ( e.g. , * ? ? ? * ; * ? ? ? * ) is that star formation and the growth of central black holes may be self - regulated : agns / star formation can drive nuclear outflows which in turn suppress further cooling and star formation ( @xcite ) . while the detailed processes are still to be understood , it is now increasingly clear that feedback and outflows play an important role in galaxy formation and evolution . so far most observational evidence for agn feedback is from radio observations at the centre of clusters or groups of galaxies @xcite . on the galaxy scale , evidence is still limited . as discussed above , ir qsos have high sfrs and accretion rates @xcite , outflow properties in these objects may thus provide hints on the feedback processes on galaxy or group scale . it is well known that low - ionisation broad absorption line qsos ( lobal qsos ) comprise about 15% of bal qso population . they are defined as a subclass of bal qsos with an obvious blueshifted absorption in mg ii @xmath40@xmath402795,2802 and other low - ionisation species @xcite . such absorption troughs arise from resonance absorption by outflowing gas and dust @xcite . in addition , there is a rare class of lobal qsos , showing absorption features from excited iron ( termed as felobal qsos ) . the outflow velocities for most bal qsos span a large range , up to a few times @xmath41 kms@xmath9 , which may be formed on a scale of @xmath42 and directly associated with the wind from an accretion disk or molecular torus @xcite . however , recent spectral analyses based on keck observations for lobal qsos or felobal qsos reveal that the outflow velocities of some lobal qsos range from several hundred to several thousand kms@xmath9 and the wind is from regions of a few hundred pc ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , which is much larger than the central engine of agns , but similar to the size of the nuclear starburst region of ulirgs @xcite . on the other hand , @xcite studied four lobal qsos at [email protected] ( mrk 231 , iras 14026@xmath444341 , iras f07599 + 6508 , pg 1700 + 518 ; all four are in our sample ) and found that all are ulirgs with merging signatures . they argued that lobal qsos can not simply be explained by orientation effects , rather , they are directly related to young systems , still surrounded by gas and dust . it represents a short stage in the early life of a large fraction of qsos . moreover , @xcite reported the detections of mid / far - infrared emission from 9 felobal qsos by mips . they found that all of their objects are ir bright with infrared luminosities as high as ulirgs . thus all lobal qsos and felobal qsos with mir to fir information are infrared ultra - luminous , suggesting a link among lobal qsos , felobal qsos and ir qsos . for lobal qsos or felobal qsos , the velocities of outflows can be measured from blueshifted absorption lines . unfortunately it is still unclear whether most ir qsos are lobal qsos or felobal qsos , because not many ir qsos have been observed in the uv . however , there are measurements for the blueshifts in the permitted optical emission lines for our ir qso sample @xcite . the line profiles often contain two gaussian components , one broad and one narrow . the outflow velocities are determined by the blueshift of the broad gaussian component relative to the narrow component of permitted emission lines ( h@xmath45 ) for ir qsos ( see * ? ? ? we assume the blueshifted broad gaussian component is related to the outflow of clouds @xcite . [ fig_outflow ] shows the correlation between ew ( pah 6.2@xmath1 m ) and outflow velocities for ir qsos . as the outflow velocity increases , the ew ( pah 6.2@xmath1 m ) decreases . since the ew ( pah 6.2@xmath1 m ) expresses the relative contribution of star formation to agn in qsos ( see @xmath204.2 ) , the correlation shown in fig . [ fig_outflow ] implies that qsos with higher outflow velocities have a lower ratio of sfr to accretion rate . it may be because more energetic agns and central massive starbursts can drive faster outflows which then suppress or even quench star formation by heating up or expelling the cold gas and dust in the qso hosts , leading to a lower ratio of sfr to accretion rate . another possibility is that the higher velocity outflows are more efficient in ejecting dust cocoons of agns , leading to agns becoming the dominant source compared with starbursts . we should caution , however , that the sample shown in fig . [ fig_outflow ] is small ; a larger sample would allow us to better understand the relative importance of star formation and agn in feedback and driving outflows . in this work , we studied the mir spectral properties of low - redshift ir qsos based on spectroscopic observations with _ spitzer _ irs , and compared their properties with those of ulirgs and optically - selected pg qsos . the following are our main results : \1 . the average mir spectra , mir spectral slopes , 6.2@xmath1 m pah emission strengths and [ neii ] 12.81@xmath1 m luminosities of ir qsos are distinct from those of pg qsos . the mir properties of ir qsos are intermediate between ulirgs and optically - selected qsos , indicating that ir qsos are at a transitional stage from ulirgs to classical qsos . \2 . from the correlations between @xmath2(30 , 15 ) with @xmath2(60 , 25 ) , fir excess l@xmath30/l@xmath26 , and ew ( pah 6.2@xmath1 m ) for both ir qsos and pg qsos , we find that the mir colour index @xmath2(30 , 15 ) is a good indicator of the relative contribution of starbursts to agns for all qsos . both pah and [ neii]12.81@xmath1 m luminosities of ir qsos are in the same range as those of ulirgs , but are one order of magnitude higher than those of pg qsos ( see fig . [ fig_hist ] ) . from the tight correlation between pah 6.2@xmath1 m and [ neii ] 12.81@xmath1 m luminosities for ir qsos , ulirgs and pah detected pg qsos , and the fact that the l@xmath29}$]/l@xmath26 ratio for ir qsos is about one order of magnitude higher than that of classical qsos , we conclude that the [ neii ] 12.81@xmath1 m emission of ir qsos is dominated by star formation , thus their luminosity can be used as an approximate indicator of their sfrs . outflows in ir qsos play an important role in suppressing star formation by heating and/or expelling cold gas surrounding nuclei of qsos . thus ir qsos are an important observational sample to investigate agn and star formation feedback processes . we thank the anonymous referee for constructive comments and suggestions . we acknowledge drs . j. wang , d .- w . xu , y. shi , s. veilleux for advice and helpful discussions . we also thank z .- h . shang , y .- l . wu , k. zhang for helpful discussion on the _ spitzer _ irs data reductions , and sarah bryan for a careful reading of the manuscript . cc acknowledges the support of the tianjin astrophysics center ( tac ) at tianjin normal university , and the full living and travel supports of the european union marie curie program for him to participate in the agn / vlti summer school held in poland , 2007 . sm acknowledges the travel support of the nsfc , the chinese academy of sciences and the tianjin municipal government . 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 . the irs was a collaborative venture between cornell university and ball aerospace corporation funded by nasa through the jet propulsion laboratory and ames research center . smart was developed by the irs team at cornell university and is available through the spitzer science center at caltech . 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 project is supported by the nsf of china 10333060 , 10778622 and 973 program no.2007cb815405 , 2007cb815406 . 99 armus l. , et al . , 2006 , apj , 640 , 204 armus l. , et al . , 2007 , apj , 656 , 148 batcheldor d. , tadhunter c. , holt j. , morganti r. , odea c. p. , axon d. j. , & koekemoer a. , 2007 , apj , 661 , 70 boroson t. a. , & green r. f. , 1992 , apjs , 80 , 109 brandl b. r. , et al . , 2006 , apj , 653 , 1129 calzetti d. , et al . , 2007 , apj , 666 , 870 canalizo g. , & stockton a. , 2001 , apj , 555 , 719 canalizo g. , & stockton a. , 2002 , asp conf . ser . 255 : mass outflow in active galactic nuclei : new perspectives , 255 , 195 carilli c. l. , et al . , 2007 , apjs , 172 , 518 cui j. , xia x .- y . , deng z .- g . , mao s. , & zou z .- , 2001 , aj , 122 , 63 desai v. , et al . , 2007 , apj , 669 , 810 downes d. , & solomon p. m. , 1998 , apj , 507 , 615 de kool m. , becker r. h. , arav n. , gregg m. d. , & white r. l. , 2002 , apj , 570 , 514 evans a. s. , solomon p. m. , tacconi l. j. , vavilkin t. , & downes d. , 2006 , aj , 132 , 2398 farrah d. , lacy m. , priddey r. , borys c. , & afonso j. , 2007a , apj , 662 , l59 farrah d. , et al . , 2007b , apj , 667 , 149 ferrarese l. , & ford h. , 2005 , space science reviews , 116 , 523 gallagher s. c. , richards g. t. , lacy m. , hines d. c. , elitzur m. , & storrie - lombardi l. j. , 2007 , apj , 661 , 30 ganguly r. , & brotherton m. s. , 2008 , apj , 672 , 102 haas m. , et al . , 2003 , a&a , 402 , 87 hamann f. w. , netzer h. , & shields j. c. , 2000 , apj , 536 , 101 hao c. n. , xia x. y. , mao s. , wu h. , & deng z. g. , 2005a , apj , 625 , 78 hao c. n. , xia x. y. , mao s. , deng z. g. , & wu h. , 2008 , chinese journal of astronomy and astrophysics , 8 , 12 hao l. , et al . , 2005b , apj , 625 , l75 hao l. , weedman d. w. , spoon h. w. w. , marshall j. a. , levenson n. a. , elitzur m. , & houck j. r. , 2007 , apj , 655 , l77 heckman t. m. , kauffmann g. , brinchmann j. , charlot s. , tremonti c. , & white s. d. m. , 2004 , apj , 613 , 109 higdon s. j. u. , et al . , 2004 , pasp , 116 , 975 ho l. c. , 2005 , apj , 629 , 680 ho l. c. , & keto e. , 2007 , apj , 658 , 314 houck , j. r. , schneider , d. p. , danielson , g. e. , neugebauer , g. , soifer , b. t. , beichman , c. a. , & lonsdale , c. j. 1985 , apj , 290 , l5 houck j. r. , et al . , 2004 , apjs , 154 , 18 imanishi m. , dudley c. c. , maiolino r. , maloney p. r. , nakagawa t. , & risaliti g. , 2007 , apjs , 171 , 72 kim d .- c . , & sanders d. b. , 1998 , apjs , 119 , 41 lawrence , a. , et al . 1999 , mnras , 308 , 897 leighly k. m. , 2001 , probing the physics of active galactic nuclei , 224 , 293 lpari s. , terlevich r. , zheng w. , garcia - lorenzo b. , sanchez s. f. , & bergmann m. , 2005 , mnras , 360 , 416 lonsdale c. j. , farrah d. , & smith h. e. , 2006 , astrophysics update 2 , 285 low f. j. , cutri r. m. , kleinmann s. g. , & huchra j. p. , 1989 , apj , 340 , l1 magorrian j. , et al . , 1998 , aj , 115 , 2285 netzer h. , et al . , 2007 , apj , 666 , 806 neugebauer g. , green r. f. , matthews k. , schmidt m. , soifer b. t. , & bennett j. 1987 , apjs , 63 , 615 neugebauer g. , & matthews k. , 1999 , aj , 118 , 35 rieke g. h. , et al . , 2004 , apjs , 154 , 25 rowan - robinson m. , 2000 , mnras , 316 , 885 sanders d. b. , soifer b. t. , elias j. h. , madore b. f. , matthews k. , neugebauer g. , & scoville n. z. , 1988 , apj , 325 , 74 sanders d. b. , & mirabel i. f. , 1996 , ara&a , 34 , 749 saunders , w. , et al . 2000 , mnras , 317 , 55 schmidt m. , & green r. f. , 1983 , apj , 269 , 352 schweitzer m. , et al . , 2006 , apj , 649 , 79 shi y. , et al . , 2007 , apj , 669 , 841 silk j. , rees m. j. , 1998 , a&a , 331 , l1 smith j. d. t. , et al . , 2007 , apj , 656 , 770 spoon h. w. w. , keane j. v. , tielens a. g. g. m. , lutz d. , moorwood a. f. m. , & laurent o. , 2002 , a&a , 385 , 1022 spoon h. w. w. , marshall j. a. , houck j. r. , elitzur m. , hao l. , armus l. , brandl b. r. , & charmandaris , v. , 2007 , apj , 654 , l49 veilleux s. , kim d .- c . , & sanders d. b. , 1999 , apj , 522 , 113 veilleux s. , kim d .- c . , & sanders d. b. , 2002 , apjs , 143 , 315 voit g. m. , weymann r. j. , & korista , k . t. , 1993 , apj , 413 , 95 wang j. , wei j. y. , & he x. t. , 2006 , apj , 638 , 106 wang r. , et al . , 2007 , aj , 134 , 617 wang r. , et al . , 2008 , apj , in press ( arxiv:0806.3022 ) werner m. w. , et al . , 2004 , apjs , 154 , 1 weymann r. j. , turnshek d. a. , & christiansen w. a. , 1985 , astrophysics of active galaxies and quasi - stellar objects , 333 weymann r. j. , morris s. l. , foltz c. b. , & hewett p. c. , 1991 , apj , 373 , 23 wu h. , zou z. l. , xia x. y. , & deng z. g. , 1998 , a&as , 127 , 521 wu h. , cao c. , hao c .- n . , liu f .- s . , wang j .- l . , xia x .- y . , deng z .- g . , & young c. k .- s . , 2005 , apj , 632 , l79 wu h. , zhu y .- n . , cao c. , & qin b. , 2007 , apj , 668 , 87 zheng z. , wu h. , mao s. , xia x .- y . , deng z .- g . , & zou z .- l . , 1999 , a&a , 349 , 735 zheng x. z. , xia x. y. , mao s. , wu h. , & den , z. g. , 2002 , aj , 124 , 18 @lcccccc@ & r.a . & decl . & & & & + object & ( j2000.0 ) & ( j2000.0 ) & redshift & log@xmath46 & log@xmath47 & spectype + ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & ( 5 ) & ( 6 ) & ( 7 ) + + i zw 1 & 00 53 34.9 & @xmath4412 41 36 & 0.0611 & 11.310 & 11.050 & s1 + f00275@xmath192859 & 00 30 04.2 & @xmath1928 42 25 & 0.2781 & 12.342 & 11.568 & qso + 3c 48 & 01 37 41.3 & @xmath4433 09 35 & 0.3670 & 12.648 & 11.991 & qso + mrk 1014 & 01 59 50.2 & @xmath4400 23 41 & 0.1630 & 12.326 & 11.192 & qso + f02054 + 0835 & 02 08 06.8 & @xmath4408 50 02 & 0.3450 & 12.466 & 11.819 & qso + f07599 + 6508 & 08 04 33.1 & @xmath4464 59 49 & 0.1483 & 12.116 & 11.637 & qso + f11119 + 3257 & 11 14 38.9 & @xmath4432 41 33 & 0.1890 & 12.322 & 12.089 & qso + 3c 273 & 12 29 06.7 & @xmath4402 03 09 & 0.1583 & 12.263 & 12.427 & qso + mrk 231 & 12 56 14.2 & @xmath4456 52 25 & 0.0422 & 12.236 & 11.467 & qso + f13218 + 0552 & 13 24 19.9 & @xmath4405 37 05 & 0.2051 & 12.270 & 11.467 & qso + f13342 + 3932 & 13 36 24.1 & @xmath4439 17 31 & 0.1793 & 12.116 & 11.821 & qso + f14026 + 4341 & 14 04 38.8 & @xmath4443 27 07 & 0.3233 & 12.445 & 11.930 & qso + pg 1543 + 489 & 15 45 30.2 & @xmath4448 46 09 & 0.3996 & 12.344 & 11.843 & qso + f15462@xmath190450 & 15 48 56.8 & @xmath1904 59 34 & 0.0998 & 11.995 & 10.381 & s1 + pg 1613 + 658 & 16 13 57.2 & @xmath4465 43 10 & 0.1290 & 11.533 & 11.550 & qso + pg 1700 + 518 & 17 01 24.8 & @xmath4451 49 20 & 0.2920 & 12.090 & 12.115 & qso + f18216 + 6419 & 18 21 57.3 & @xmath4464 20 36 & 0.2970 & 12.659 & 12.607 & qso + f20036@xmath191547 & 20 06 31.7 & @xmath1915 39 08 & 0.1919 & 12.359 & 11.566 & qso + f21219@xmath191757 & 21 24 41.6 & @xmath1917 44 46 & 0.1120 & 11.661 & 10.952 & s1 + + pg 0804 + 761 & 08 10 58.6 & @xmath4476 02 42 & 0.1000 & 10.794 & 11.360 & qso + pg 0838 + 770 & 08 44 45.2 & @xmath4476 53 09 & 0.1310 & 10.994 & 11.020 & s1 + pg 0844 + 349 & 08 47 42.4 & @xmath4434 45 04 & 0.0640 & 10.323 & 10.854 & s1 + pg 1004 + 130 & 10 07 26.1 & @xmath4412 48 56 & 0.2408 & 11.639 & 11.964 & qso + pg 1119 + 120 & 11 21 47.1 & @xmath4411 44 18 & 0.0502 & 10.536 & 10.509 & s1 + pg 1149@xmath19110 & 11 52 03.5 & @xmath1911 22 24 & 0.0490 & 10.435 & 10.299 & s1 + pg 1211 + 143 & 12 14 17.7 & @xmath4414 03 13 & 0.0809 & 11.084 & 11.414 & qso + pg 1351 + 640 & 13 53 15.8 & @xmath4463 45 45 & 0.0882 & 11.248 & 11.203 & qso + pg 1411 + 442 & 14 13 48.3 & @xmath4444 00 14 & 0.0896 & 10.579 & 10.945 & s1 + pg 1426 + 015 & 14 29 06.6 & @xmath4401 17 06 & 0.0865 & 10.753 & 11.194 & qso + pg 1444 + 407 & 14 46 45.9 & @xmath4440 35 06 & 0.2673 & 11.556 & 11.636 & qso + pg 1501 + 106 & 15 04 01.2 & @xmath4410 26 16 & 0.0364 & 10.468 & 10.649 & s1 + pg 1519 + 226 & 15 21 14.2 & @xmath4422 27 43 & 0.1370 & 10.963@xmath48 & 11.001 & s1 + pg 1534 + 580 & 15 35 52.3 & @xmath4457 54 09 & 0.0296 & 9.663 & 10.069 & s1 + pg 1535 + 547 & 15 36 38.3 & @xmath4454 33 33 & 0.0389 & 9.742@xmath48 & 10.530 & s1 + pg 1552 + 085 & 15 54 44.6 & @xmath4408 22 22 & 0.1190 & 10.782@xmath48 & 10.974 & s1 + pg 1612 + 261 & 16 14 13.2 & @xmath4426 04 16 & 0.1309 & 11.053 & 11.019 & s1 + pg 1617 + 175 & 16 20 11.3 & @xmath4417 24 28 & 0.1124 & 10.637 & 11.140 & qso + pg 2130 + 099 & 21 32 27.8 & @xmath4410 08 19 & 0.0630 & 10.748 & 10.876 & s1 + + + notes basic properties of ir qsos and pg qsos in our sample . col.(1 ) : source name ; col.(2)-(3 ) : right ascension ( hours , minutes , seconds ) and declination ( degrees , arcminutes , arcseconds ) , from ned ; col.(4 ) : redshift , from ned ; col.(5 ) : monochromatic luminosity at 60@xmath1 m ; col.(6 ) : monochromatic luminosity at 5100 , @xmath40l@xmath49(5100 ) . the values are derived from @xcite except for f14026@xmath444341 it is measured based on the sdss spectrum , and for some pg qsos not included in @xcite their values were taken from @xcite and calculated in the same manner as in @xcite ; col.(7 ) : optical spectral classifications of ir qsos ( seyfert 1 or qso ) . the typical uncertainties of ir and optical luminosities are about 0.06 dex and 10 - 20% , respectively @xcite . + @xmath48the 60 @xmath1 m flux density is an upper limit . + + + notes mid - ir spectral properties of ir qsos . col.(1 ) : source name ; col.(2)-(3 ) : continuum flux densities around 15 and 30@xmath1 m rest wavelength , the uncertainties of the measurements are typically 10@xmath23 ; col.(4)-(5 ) : flux and equivalent width ( ew ) of pah emissions at 6.2@xmath1 m , upper limits ( 3@xmath22 ) are derived adopting widths of 0.2@xmath1 m ; col.(6 ) fluxes of [ neii ] 12.81@xmath1 m line . + @xmath48outside the rest wavelength range available , the value given here was linearly extrapolated based on the logarithm of the @xmath025@xmath1 m rest wavelength continuum level . + @lcccc@ & s@xmath50 & s@xmath51 & pah 6.2@xmath1 m & ew(pah ) + object & ( mjy ) & ( mjy ) & ( 10@xmath52w @xmath53 ) & ( @xmath1 m ) + ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & ( 5 ) + + pg 0804 + 761 & 125.5 & 150.8 & @xmath437.4 & @xmath430.012 + pg 0838 + 770 & 45.4 & 97.4 & @xmath432.0 & @xmath430.026 + pg 0844 + 349 & 105.5 & 155.1 & @xmath436.5 & @xmath430.018 + pg 1004 + 130 & 72.9 & 118.5 & @xmath432.0 & @xmath430.014 + pg 1119 + 120 & 176.9 & 303.9 & @xmath432.1 & @xmath430.008 + pg 1149@xmath19110 & 140.3 & 249.6 & @xmath432.1 & @xmath430.019 + pg 1211 + 143 & 239.6 & 268.8 & @xmath437.5 & @xmath430.014 + pg 1351 + 640 & 249.4 & 536.2 & 10.9 & 0.034 + pg 1411 + 442 & 125.3 & 143.2 & @xmath434.2 & @xmath430.010 + pg 1426 + 015 & 136.1 & 181.2 & @xmath435.2 & @xmath430.013 + pg 1444 + 407 & 69.8 & 99.1 & @xmath437.54 & @xmath430.036 + pg 1501 + 106 & 297.6 & 424.9 & @xmath433.4 & @xmath430.008 + pg 1519 + 226 & 54.2 & 70.8 & 2.46 & 0.016 + pg 1534 + 580 & 107.2 & 149.1 & @xmath432.08 & @xmath430.008 + pg 1535 + 547 & 98.4 & 111.5 & @xmath433.1 & @xmath430.011 + pg 1552 + 085 & 79.7 & 94.2 & @xmath431.3 & @xmath430.017 + pg 1612 + 261 & 83.4 & 136.4 & @xmath433.0 & @xmath430.020 + pg 1617 + 175 & 53.8 & 65.7 & @xmath433.2 & @xmath430.020 + pg 2130 + 099 & 196.4 & 293.0 & 3.56 & 0.006 + + f00188@xmath190856 & 122.8 & 1052.7 & 27.5 & 0.187 + f00397@xmath191312 & 90.1 & 523.5 & 9.14 & 0.018 + f01004@xmath192237 & 306.4 & 1176.4 & 11.2 & 0.048 + f01199@xmath192307 & 39.1 & 455.0 & 3.75 & 0.146 + f01298@xmath190744 & 65.4 & 718.5 & 7.53 & 0.092 + f01355@xmath191814 & 27.7 & 369.6 & @xmath432.7 & @xmath430.186 + z03521 + 0028 & 50.6 & 980.1 & 20.9 & 1.086 + f05189@xmath192524 & 1017.6 & 4994.1 & 54.8 & 0.034 + f08572 + 3915 & 657.6 & 3272.9 & @xmath437.7 & @xmath430.004 + f09463 + 8141 & 9.0 & 193.3 & 7.61 & 1.483 + f10091 + 4704 & 13.3 & 197.1 & 7.06 & 0.758 + f10378 + 1108 & 48.5 & 571.9 & 11.0 & 0.415 + f11095@xmath190238 & 111.8 & 1112.0 & @xmath433.7 & @xmath430.025 + f11223@xmath191244 & 19.0 & 214.3 & @xmath432.2 & @xmath430.065 + f11582 + 3020 & 32.5 & 352.8 & 1.25 & 0.022 + f12032 + 1707 & 45.5 & 428.8 & @xmath437.1 & @xmath430.161 + f12072@xmath190444 & 197.3 & 878.6 & 13.6 & 0.106 + f12112 + 0305 & 106.4 & 1683.5 & 56.8 & 0.431 + mrk 273 & 416.1 & 5068.8 & 74.8 & 0.129 + f13451 + 1232 & 286.5 & 996.7 & @xmath433.0 & @xmath430.013 + f14070 + 0525 & 21.6 & 304.0 & 5.59 & 0.187 + f14348@xmath191447 & 82.7 & 1309.2 & 33.5 & 0.365 + f15001 + 1433 & 59.6 & 475.1 & 16.1 & 0.184 + f15206 + 3342 & 101.6 & 628.9 & 38.2 & 0.342 + arp 220 & 1048.1 & 20907.7 & 161.0 & 0.272 + f16090@xmath190139 & 76.5 & 808.3 & 13.5 & 0.107 + f16300 + 1558 & 25.7 & 490.4 & 11.0 & 0.282 + f16333 + 4630 & 20.2 & 260.9 & 12.5 & 0.529 + ngc 6240 & 752.1 & 5763.0 & 313.0 & 0.365 + f17068 + 4027 & 52.3 & 410.9 & 10.8 & 0.198 + f17179 + 5444 & 72.6 & 341.8 & 6.43 & 0.099 + f20414@xmath191651 & 61.7 & 1397.4 & 37.5 & 0.587 + f22491@xmath191808 & 82.9 & 1473.3 & 23.1 & 0.465 + f23129 + 2548 & 40.2 & 434.9 & @xmath431.1 & @xmath430.021 + f23498 + 2423 & 46.4 & 239.7 & 6.11 & 0.053 +

we present the first multi - fluid analysis of a dense neutron star core with a deconfined colour - flavour - locked superconducting quark component . accounting only for the condensate and ( finite temperature ) phonons , we make progress by taking over results for superfluidhe . the resultant two - fluid model accounts for a number of additional viscosity coefficients ( compared to the navier - stokes equations ) and we show how they enter the dissipation analysis for an oscillating star . we provide simple estimates for the gravitational - wave driven r - mode instability , demonstrating that the various phonon processes that we consider are not effective damping agents . even though the results are likely of little direct astrophysical importance ( since we consider an overly simplistic stellar model ) our analysis represents significant technical progress , laying the foundation for more detailed numerical studies and preparing the ground for the inclusion of additional aspects ( in particular associated with kaons ) of the problem . 0.5 truein s

1005.1163

the state of matter at extreme densities continues to be an issue of vigorous investigation . the problem is complicated , not only from the theoretical point of view , but also by the fact that laboratory experiments are restricted . for example , while colliders like rhic at brookhaven , gsi in darmstadt and the lhc at cern probe hot quark - gluon plasmas they will never be able to explore the cold , extreme high density , region of the qcd phase diagram . in order to test our understanding of the relevant physics we need to turn to astrophysics , and the dynamics of compact stars . in fact , `` neutron stars '' represent unique laboratories of such extreme physics . with core densities reaching about one order of magnitude beyond nuclear saturation , they are likely to contain exotic states of matter like hyperon phases with net strangeness and/or deconfined quarks . it is well - established that these states of matter should exhibit superfluidity / superconductivity at the relevant temperatures ( neutron stars are born with temperatures @xmath1 k and rapidly cool below @xmath2 k ) . moreover , observed radio pulsar glitches provide strong evidence for the presence of a partially decoupled superfluid component in these systems . the modelling of the dynamics of these , potentially complex , objects presents a serious challenge . a key aspect of the problem concerns the fact that a superfluid system has additional dynamical degrees of freedom . this is well - known from experiments on laboratory systems like @xmath0he , which exhibit a second sound associated with thermal waves @xcite . analogous `` superfluid '' modes have been studied in detail for superfluid neutron - proton - electron mixtures relevant for the outer core of a neutron star @xcite . for simplicity , these studies have almost exclusively ignored thermal effects ( the work in @xcite is a notable exception ) . while this is a useful first approximation it is clear that the zero temperature assumption must be relaxed in a realistic model . basically , due to the density dependence of the various superfluid pairing gaps ( see @xcite for a guide to the literature ) , there will be regions in a neutron star where thermal effects are important ( in the vicinity of the critical density at which the phase transition occurs ) . understanding the nature of these transition regions , and their effect on various aspects of neutron star dynamics , is one of the main challenges for research in this area . in this paper , we describe a first attempt at modelling thermal dynamics in a superfluid neutron star . we focus on stars with a colour - flavour - locked ( cfl ) superconducting quark core @xcite at finite temperatures . this is an interesting problem for several reasons . first of all , the simplest possible model for this system considers a quark condensate coupled to a gas of phonons . this problem is analogous to @xmath0he at low temperatures , and hence we can bring our recent dissipative two - fluid model @xcite to bear on it ( more or less directly ) . the lessons we learn from this exercise should inform the development of finite temperature models for the superfluids in the outer core and the neutron star crust . secondly , even though there have been discussions of the dynamics of the different cfl phases , in particular in the context of the gravitational - wave driven r - mode instability ( see @xcite for references ) , the superfluid aspects have ( as far as we are aware ) not previously been accounted for . our discussion begins to address the relevance of the additional degrees of freedom in these systems , and provides some insight into the nature of the different `` fluids '' involved . as an application with immediate astrophysical relevance , we will work out the inertial r - modes and estimate the relevant viscous damping rates for the `` simplest '' model of cfl matter . we consider the `` cool '' regime where the temperature is significantly below all the quasiparticle energy gaps . in this regime , dissipation may mainly occur due to phonon interactions . one reason for considering this model is that there are results in the literature for both bulk- and shear viscosity @xcite as well as the mutual friction associated with superfluid vortices @xcite . of course , the simple `` condensate plus phonon '' model that we consider is not the whole story . it should apply at asymptotically high densities , but may not be the true ground state at lower densities . the discussion in @xcite adds extra dimensions to the problem by considering the bulk viscosity due to kaons ( allowing for flavour - changing processes ) , which will be present at higher temperatures . at first sight , this mechanism will only be relevant for very hot stars ( @xmath3 mev @xmath4 k ) since it assumes that there is a thermal population of kaons . below the critical energy where the kaons appear the contribution to the bulk viscosity is exponentially suppressed and may not be that important . however , in more recent work on the so - called cfl - k@xmath5 phase @xcite , it is argued that the main low temperature mechanism involves condensed kaons . this is important since the kaon condensate will remain present as @xmath6 . the upshot of this is that the problem requires a `` multi - fluid '' analysis at all temperatures . the simplest model would have three components ; the quark condensate , the kaon condensate and finite temperature excitations ( phonons and thermal kaons ) . this problem is more involved than the case that we focus on here . nevertheless , our results provide an essential starting point for investigations into the dynamical role of the kaons . most importantly , by providing the `` hydrodynamics '' view of the problem we illustrate the input needed to study various dynamical scenarios . this should stimulate further discussion between experts on different aspects of this multi - faceted problem , as required to make progress in the future . we consider a deconfined quark system system that contains a single cfl superconducting condensate and a phonon gas . formally , this problem is identical to that for @xmath0he at low temperatures . hence , we can take as our starting point the recent discussion of superfluid helium @xcite , which is based on the convective variational approach to multi - fluid dynamics @xcite and which we know is in one - to - one correspondence with the orthodox formulation developed by , in particular , khalatnikov @xcite . the variational model takes as its starting point an energy functional @xmath7 , representing the equation of state . this energy determines the relation between the various fluxes and the associated canonical momenta . in the present case , we distinguish between the massive particles in the system , with number density @xmath8 and flowing with @xmath9 from a massless entropy component with number density @xmath10 and flux @xmath11 . the latter represents the phonons ( which are treated in the fluid approximation , i.e. , we assume that there exists a suitable defined `` average '' transport velocity for each of the two components in the system ) the two momentum densities then follow from and @xmath12 , respectively . indices , which range from 1 to 3 , can be raised and lowered with the ( flat space ) metric @xmath13 , i.e. , @xmath14 . derivatives are expressed in terms of the covariant derivative @xmath15 which is consistent with the metric in the sense that @xmath16 . this formulation has great advantage when we want to discuss the geometric nature of the different dissipation coefficients . ] _ i^= mn v_i^- 2 w_i^ , and ^_i = 2 w_i^ , where @xmath17 and @xmath18 represents the entropy entrainment @xcite . it has been assumed that @xmath19 , as expected in an isotropic system , and we have defined = . e w^2 |_n , s , omitting the indices on @xmath20 for clarity . as discussed in @xcite , the associated momentum equations can be written f_i^= _ t _ i^+ _ j ( v_^j _ i^+ d^j _ i ) + n _ i ( _ - m v_^2 ) + _ j^_i v_^j , [ eulern ] and f_i^= _ t _ i^+ _ j ( v_^j _ i^+ d^j _ i ) + s _ i t + _ j^_i v_^j , [ eulers ] where @xmath21 is the matter ( quark ) chemical potential ; _ = ( e n ) _ s , w^2 , and we have used the fact that the temperature corresponds to the entropy chemical potential @xcite ; t= _ = ( e s ) _ n , w^2 . in these expressions , @xmath22 represent the viscous stresses while the `` forces '' @xmath23 allow for momentum transfer between the two components . in the following we will assume that the system is isolated , which means that @xmath24 . as we will discuss later , the force terms can also be used to account for `` external '' forces like gravity . in the present context , when we are dealing with a single matter quantity , we will have @xcite _ t n + _ j ( n v_^j ) = _ = 0 , as there is no particle creation / destruction . at the same time the entropy can increase , so we have _ t s + _ j ( s v_^j ) = _ 0 . we also have @xcite t _ = - f_i^w_^i - d^j _ i _ j v_^i - d^j _ i _ j w_^i , [ tgs ] where d_ij = d_ij^+ d_ij^ . for an isolated system we know that , if we impose the superfluid constraint of irrotationality the number of dissipation coefficients reduces significantly . as discussed in @xcite , we have - d_ij = g_ij ( ^ _ l j^l + _ ) + 2 ^_ij . [ dij ] where we have defined @xmath25 . we also have ; ( f_i^- _ l d^l _ i ) = _ i [ phi1 ] with = ^ _ l j^l + ^ _ [ phinal ] . at this point , only four dissipation coefficients remain in the problem . in the above expressions , we have used the standard decomposition ; _ i v^_j = ^_ij + g_ij _ + _ ijk w_^k in terms of , the expansion _ = _ j v_^j , [ exp]the trace - free shear ^_ij = ( _ i v^_j + _ j v^_i - g_ij _ ) , [ sh]and the `` vorticity '' w_^i = ^ijk ( _ j v^_k - _ k v^_j ) . we use analogous expressions for gradients of the relative velocity . the definition of the various quantities should be obvious from the constituent indices . however , since the system that we consider is not irrotational we need to consider relaxing the assumptions on the dissipation coefficients . this involves making some subtle decisions . in the case of an irrotational flow it is natural to assume that @xmath26 . when the superfluid rotates , and vortices are present , the force will not vanish . it is necessary to account for dissipation due to , for example , the scattering of phonons off of the vortex cores . the standard approach to the rotating problem is to add in this `` mutual friction '' force , keeping the other dissipative terms as in the irrotational case . this strategy ignores a number of dissipative terms that would , at least in principle , be allowed in the equations of motion @xcite . the role of these additional terms has not yet , as far as we are aware , been investigated . ignoring the potential relevance of most of the additional dissipation channels in the irrotational case , we simply account for the presence of vortices by i ) assuming that represents fluid elements with both a condensate and a smooth - averaged vorticity arising from the vortices , and ii ) accounting for the vortex mediated mutual friction by allowing a force @xcite f_i^ = _ n_v _ ijk ^j w_^k + _ n_v _ ijk ^klm ^j _ l w_m^ , [ mf ] to act on the particles ( with a balancing force affecting the excitations ) . here @xmath27 is the vortex area density and the vector @xmath28 ( the hat represents a unit vector ) is aligned with the rotation axis and has magnitude @xmath29 . since we will only consider the effect of phonon scattering off of vortices , we expect to be in the weak mutual friction regime where @xmath30 . this means that the first term in the force can be ignored , leaving only the second , dissipative , contribution . in order to set the stage for the discussion of r - modes , we need to provide a suitable rotating equilibrium configuration . to do this , we note that the two components flow together ( there is not heat flux ) when the system is in thermal equilibrium . this means that @xmath31 , which implies that @xmath32 . thus , it follows from the entropy momentum equation that we must have s _ i t = 0 t = . this is trivial ; if the system is isothermal then there will be no heat flux . rewriting the remaining momentum equation , making use of the continuity equation , we find ( after accounting for the gravitational force as the gradient of the gravitational potential @xmath33 ) _ t v_i^+ v_^j _ j v^_i + _ i ( _ + ) = 0 , where @xmath34 . using the lie - derivative along the flow , @xmath35 , we can rewrite this equation as _ v_n v_i^+ _ i ( _ + - 1 2 v_^2 ) = 0 . restricting ourselves to stationary models we have _ t v_i^= 0 . next , uniform rotation implies v_^i = _ _ ^i , and , for axisymmetric models , it then follows that _ v_i^= 0 . hence , the required equilibrium models are determined from the usual bernoulli - type equation _ + - 1 2 v_^2 = . this analysis shows that a rotating configuration can be obtained in the usual way . once we assume thermal equilibrium , we are dealing with a single - fluid problem . our main aim is to develop the tools required to make quantitative estimates for the r - mode instability in neutron stars with a cfl core . in doing this , we want to account for the fact that cfl matter requires a multi - fluid description . the analysis proceeds in three steps . first we need to formulate the linear perturbation problem for the system . this step is interesting because , as far as we are aware , this is the first time that perturbations of a matter plus massless entropy system have been considered in an astrophysical context . the results provide useful insights into finite temperature superfluid dynamics , and should be relevant in a broader context . the second step corresponds to determining the pulsation modes of the system , in this case the r - modes , and finally we need to estimate the damping timescales associated with the different dissipation channels . in the discussion below , we only work out the first of these steps in detail . in order to obtain useful estimates for the r - mode instability we simplify last two steps by considering a uniform density model . this has the advantage that the r - mode solution is simple , and we can evaluate the dissipation rates analytically . in a sense , we do not expect this approximation to be too bad because it is well known that the density profile for a canonical @xmath36/10 km strange star ( described by the mit bag model ) is almost flat ( see @xcite ) . of course , we really need to develop a consistent model for a realistic equation of state . this is essential if we want to study hybrid stars , where the cfl phase is only present in the core . however , the problem is complicated by the fact that the bulk viscosity damping of the r - modes requires the perturbations to be worked out to second order in the slow - rotation approximation @xcite . this necessitates a numerical solution , which makes some of the qualitative behaviour of the results less clear . another complicating factor is the well - known fact that it only makes sense to use a realistic equation of state in a fully general relativistic analysis ( see @xcite for discussion ) . the r - modes in single component relativistic stars have been studied @xcite , but there has not yet been any serious analysis of the corresponding multi - fluid problem ( although see @xcite ) . a focussed effort in this direction should be encouraged . in order to work out the r - mode solutions , it is advantageous to work in a rotating frame . focussing , for the moment , on the non - dissipative equations we find that , in a frame rotating uniformly with @xmath37 we have _ t _ i^+ _ j ( v_^j _ i^ ) + n _ i ( _ + m - m v_^2 ) + _ j^_i v_^j + 2_ijk^j v_^k=0 , [ rot_eulern ] and _ t _ i^+ _ j ( v_^j _ i^ ) + s _ i t + _ j^_i v_^j = 0 , [ rot_eulers ] it is notable that the coriolis force does not affect the entropy equation . this is , however , not surprising . it is well - known that inertial forces are proportional to the mass , and since our entropy component is taken to be massless it should not be affected by the coriolis force . from a technical point of view , it means that the problem we consider is subtly different from the two - fluid r - mode problem discussed in @xcite . let us now consider perturbations of the rotating equilibrium models discussed in the previous section . considering eulerian perturbations ( denoted by @xmath38 ) we have , in the rotating frame ( using the cowling approximation @xmath39 , and not explicity denoting the velocities as perturbations since they vanish in the background anyway ) ; _ t v_i^- 2_t w_i^ + _ i _ + 2 _ ijk ^j v_^k = 0 , [ comove]and 2_t w_i^ + s _ i t = 0 . [ counter]adding these we get an equation for the total perturbed momentum _ t v_i^+ _ i _ + s _ i t + 2_ijk ^j v_^k = 0 . noting that the pressure @xmath40 is defined by ( for a co - rotating equilibrium model ) _ i p = _ i _ + s _ i t , we see that _ i p = _ i _ + _ i _ + s _ i t + s _ i t + = _ i p + _ i _ + s _ i t , since @xmath41 in the background . hence , we have the usual euler equation _ t v_i^+ _ i p - _ i p + 2_ijk ^j v_^k = 0 . [ comovinge ] we also have the perturbed continuity equation for the particles _ t + _ i ( v_^i ) = 0 , and a conservation law for the entropy _ t s + _ i ( s v_^i ) = _ t s + _ i ( s v_^i ) - _ i ( s w_^i ) = 0 . the entropy is conserved since there is no heat flux in the background ( @xmath42 is quadratic in the heat flux @xcite ) . once we provide an equation of state for matter , we have all relations we need to study the linear dynamics of a non - dissipative finite temperature cfl quark core . our main aim is to work out the r - modes and establish the parameter range in which gravitational - wave emission triggers a secular instability . in order to assess the relevance of this instability we need to consider the various dissipative mechanisms that counteract the growth of an unstable mode . the damping due to shear- and bulk viscosity in a two - component system can be estimated using the strategy set out in @xcite . that is , we use energy integrals to estimate the various timescales . in essence , the damping timescale @xmath43 associated with any given process is obtained from 2 e ^-1 , where @xmath7 is the energy associated with the flow and @xmath44 is the rate of energy loss due to dissipation . this estimate should be accurate as long as the dissipation is weak enough that it does not affect the flow on a dynamical timescale . the energy associated with a given perturbation can be obtained from the equations of motion and . multiplying the first equation by @xmath45 and the second by @xmath46 ( where the bars denote complex conjugates ) and adding , we find that the kinetic energy is given by e= ( | v_|^2 - | w_|^2 ) dv . in the case of the r - modes , this is the leading order contribution to the energy . in fact , it turns out that the second term in the bracket is of higher order in the slow - rotation approximation ( for the same reasons as in @xcite ) . hence , the r - mode energy is well approximated by e | v_|^2 dv . in order to evaluate the energy loss due to viscous damping we add the dissipative part of the stress tensor @xmath47 to the equations of motion . the total energy dissipation then follows from the body integral of @xmath48 after some algebra , this leads to = - dv , [ diss1 ] where we have defined = _ recall the relations and and the definition @xmath49 . in order to combine this result with existing results for the viscosity coefficients we need to translate our variables into those of the `` orthodox '' two - fluid model due to , for example , khalatnikov @xcite . the required translation has been discussed at length in @xcite , and the key relations that we need are given in the appendix . carrying out the comparison , we find that the shear viscosity terms are exactly the same in the two descriptions . hence , we can use @xmath50 from @xcite without change . in the case of the bulk viscosity , we find that the two sets of coefficients have different mass scalings , and we identify = _ 2 , ^= m _ 1 = m _ 4 , ^ = m^2 _ 3 . this means that , when expressed in terms of the standard coefficients , the bulk viscosity dissipation rate becomes = - dv . [ diss2 ] to conclude , there are three bulk viscosity coefficients rather than the usual single one , and both dynamical degrees of freedom are needed if we want to evaluate the dissipation integrals . note that the dissipation is expressed in terms of @xmath51 and @xmath52 , rather than @xmath53 and @xmath54 , the variables that we will solve for when we determine the r - modes . it is , of course , straightforward to express in terms of these variables . unfortunately , the dissipation integrand is then rather messy . we find @xmath55 \\ + \zeta_2 \left| \theta_{\n } \right|^2 + \left [ \zeta_2 - 2 \rho \zeta_1 + \rho^2 \zeta_3 \right ] \left| \theta_{\n\s } \right|^2 - 2 ( \zeta_2 - \rho \zeta_1 ) \mathrm{re } \left ( \bar{\theta}_\n \theta_{\n\s } \right ) \\ + 2 \zeta_1 \mathrm{re } \left [ \bar{\theta}_\n \left ( w_{\n\s}^j \nabla_j \rho \right ) \right ] - 2 ( \zeta_1 - \rho \zeta_3 ) \left [ \bar{\theta}_{\n\s } \left ( w_{\n\s}^j \nabla_j \rho \right ) \right ] + \zeta_3 \left| w_{\n\s}^j \nabla_j \rho \right|^2 \big\ } dv \ . \label{complications}\end{gathered}\ ] ] where _ = _ j v_^j _ = _ j w_^j . however , as we will see later , this expression simplifies considerably in the r - mode problem . before moving on , it is worth making a general point . it might be tempting to suggest that the terms involving @xmath54 in should be less relevant than those involving only @xmath56 . however , without solving for an actual oscillation mode one can not make this argument precise . in general , the bulk viscosity contributions from terms involving @xmath54 can not be neglected . since a mode - oscillation typically involves both degrees of freedom ( unless the star is unstratified @xcite ) , we need to determine the nature of the oscillations before we make further simplifications . in order to obtain quantitative results for the r - mode instability we need to provide an equation of state . as our main focus is on the two - fluid aspects of the problem ( ignored in previous studies ) , the equation of state must account for the thermal excitations , i.e. the phonons . moreover , if we want to investigate the relative importance of the different bulk viscosity terms we need to allow for a relative flow at the perturbative level . this is also important if we want to consider the superfluid mutual friction @xcite . however , if we insist on the model being truly `` realistic '' then the problem becomes much more challenging . hence , we will make a number of simplifying assumptions . this is not only practical , it is also quite reasonable since this is our first attempt at a quantitative analysis . future work should aim to relax some of our assumptions . as discussed in section ii , the equation of state that we require takes the form of an energy functional @xmath19 . once this functional is provided , we can work out all the quantities that we need to construct a rotating model and calculate the r - modes . unfortunately , we do not have an equation of state of the required form . in fact , some of the parameters that we need are usually not considered , essentially since they are not required if one is only interested in equilibrium configurations . in order to make progress we have to build a suitably simple model equation of state . to do this , we assume that the thermal contribution due to the phonons adds to a zero - temperature equation of state for the condensate . this strategy has previously been developed for a relativistic superfluid @xcite , and builds on the classic expressions for a non - relativistic phonon gas , see @xcite . let us assume that , at zero temperature matter is described by a barotropic model such that @xmath57 . then the chemical potential follows from = d e_0 d n , the pressure is obtained from dp_0 = n d , and it is easy to work out the speed of ( first ) sound ; c_0 ^ 2 = dp_0 d , now consider thermal excitations represented by a phonon gas with a linear dispersion relation , with slope @xmath58 . working out the thermodynamics of such a gas one can show that its contribution to the pressure , @xmath59 , is given by @xcite = 4 ^5 45 ( kt)^4 ( 2 c_0)^3 ( 1 - w^2 c_0 ^ 2 ) ^-2 . [ psidef]here , and in the following , it is to be understood that @xmath60 corresponds to @xmath61 . from this we obtain the entropy density ( in the matter frame ) via s = t . this leads to the explicit result ; s = 4 kt = 16 ^5 45 ( kt 2 c_0 ) ^3 ( 1 - w^2 c_0 ^ 2 ) ^-2 , [ srel]which means that the heat capacity is given by c _ = t s t = 2 ^2 15 ( kt c_0 ) ^3 , in agreement with the phonon result in @xcite . finally , one can show that @xcite the `` normal fluid '' density , @xmath62 , required in the orthodox superfluid formalism ( see appendix ) is given by _ = 16 ^5 45 1 ( 2 ) ^3 ( kt c_0 ) ^4 ( 1 - w^2 c_0 ^ 2 ) ^-3 1 c_0 . this result is consistent with _ = 4 3 c_0 ^ 2 ( 1 - w^2 c_0 ^ 2 ) ^-3 . where the phonon energy is e _ = 4 ^5 15 ( kt 2 c_0 ) ^3 kt . in our formulation of the problem , we need the entrainment between particles and entropy . as discussed in @xcite the relevant entrainment parameter is related to @xmath62 according to = - _ 2 ( 1 - _ ) ^-1 - _ 2 , [ entrain]where the last approximation is accurate in the low temperature limit , when @xmath63 . this is the regime that we are considering here . it is worth noting that the above results lead to _ 29 ( t 10 ^ 9 ) ^4 ^3 . in order to be consistent we need @xmath64 ( the average density for a canonical neutron star ) , which translates into @xmath65 . in other words , the model should be valid for all astrophysical neutron stars . let us now use the phonon gas results to construct a `` complete '' model equation of state . that is , we want to deduce a consistent energy functional @xmath7 . this functional should be such that the temperature is obtained from t = . e s |_n , w^2 . now , combining with we find that . e s |_n , w^2 = 1 b s^1/3 , where b = ( 16 ^5 45 ) ^1/3 1 2 c_0 ( 1 - w^2 c_0 ^ 2 ) ^-2/3 . since @xmath66 ( by definition ) we can integrate to get e= e_1(n , w^2 ) + 3 4b s^4/3 = e_1 + 3= e_1 + e _ ( 1 - w^2 c_0 ^ 2 ) ^-2 . we learn that in the limit of a low relative velocity ( @xmath67 ) the equation of state is simply given by e = e_1 + e _ . this result is quite intuitive . using the fundamental relation of thermodynamics p + e = n+ st , it is also straightforward to show that the total pressure is given by p = p_0 + , as one might have expected . now that we have the required energy functional , we can determine the entrainment parameter from = . e w^2 |_n , s . working this out , we find that ( after some algebra ) = e_1 w^2 - _ 2 . comparing to we see that the model is consistent with = 0 . hence , it is natural to take @xmath68 , i.e. simply add the thermal phonon energy to the zero - temperature equation of state . the final result is then e= e_0 + 3= e_0 + e _ ( 1 - w^2 c_0 ^ 2 ) ^-2 . in order to proceed , we need to provide the zero - temperature equation of state . a `` realistic '' model should obviously be based on qcd . it is natural @xcite to use the mit bag model . in the case of cfl matter , this leads to ( using units where the speed of light is unity , @xmath69 ) p_0 - b + 3 ^ 2 4 ^ 2 ( 4 ^2 - m_s^2 ) , [ p0]and + b - 3 ^2 m_s^2 4 ^ 2 , where @xmath70 is the bag constant and @xmath71 is the pairing gap associated with the cfl condensate . for simplicity , we will use p_0 ( -4b ) , [ bag]in which case the speed of sound is constant ; c_0 ^ 2 = dp_0 d . as discussed in @xcite , this approximation may be quite good , essentially because the two contributions to the last term in almost exactly cancel each other . later , when we work out the various r - mode dissipation integrals , we will simplify the problem further by assuming that the density is uniform , with a constant speed of sound . in order to determine the damping timescale for the unstable r - modes , we need to supplement the non - dissipative model with viscous terms . from the discussion in section iiib we know that we need one shear viscosity coefficient and three bulk viscosity coefficients . in order to account for dissipation associated with the superfluid vortices , we should also consider the mutual friction . some of the required viscosity coefficients have been discussed in the literature . in particular , the dissipation due to various phonon interactions has been investigated @xcite . in fact , the availability of these results is one of the key reasons for us focussing on the condensate plus phonon model . having said that , the available results are incomplete . to what extent this is the case , and how we have dealt with this problem , will be discussed below . before we proceed it is worth iterating the point that we are focusing on the phonons because they represent the simplest in a hierarchy of relevant multi - fluid problems . the presence of kaons , either as thermal excitations or a condensate @xcite , would require us to extend the model to account for additional degrees of freedom . the issues involved are similar to those for a superfluid hyperon core @xcite , and we expect to consider them in the future . let us first consider the phonon shear viscosity . the relevant viscosity coefficient is determined in @xcite . we should use ( following the discussion above , we are using @xmath72 here ) = 2.46 10 ^ 26 ( _ q 300 ) ^8 ( 10 ^ 9 t ) ^5 , [ eta0]where @xmath73 is the quark chemical potential . we arrive at this result by taking both @xmath21 and the quark `` mass '' @xmath74 in the fluid model to be equal to the quark chemical potential @xmath73 . the temperature scaling of @xmath50 is the same as in the case of @xmath0he , which makes sense given that the involved phonon processes are the same . however , in the context of neutron stars , this model is severely limited . basically , the `` hydrodynamic '' treatment of the phonons is no longer valid if their mean - free path is larger than ( or comparable to ) the size of the system . as in kinetic theory , the mean - free path associated with the shear viscosity follows from ( up to a factor of order unity ) _ c_0 . for the phonon gas model this leads to 5 10 ^ 14 ( _ q 300 ) ^8 ( 10 ^ 9 t ) ^9 . in other words , for a 10 km star the model will not be valid below @xmath75 . this is obviously a problem , since mature neutron stars are expected to be significantly colder than this . in essence , we ought to model the phonons as ballistic in the low temperature regime . however , there is some evidence from laboratory experiments on @xmath0he that the fluid model remains relatively accurate also at lower temperatures . given this , we will use the two - fluid model also in the regime where it is no longer formally valid . still , in order to describe such systems we need a different model for the shear viscosity . in order to make progress we will simply assume that the mean free path is limited by the size of the system . that is , we will use an shear viscosity given by _ _ c_0 r , where @xmath76 is the radius of the star . this may seem like a rather drastic assumption , but it has recently been shown @xcite that in the case of helium it leads to results that agree quite well with the observed sound attenuation . in particular , we would have @xmath77 at low temperatures , which means that the viscosity weakens as the phonon density decreases . this makes ( at least qualitative ) sense . explicitly , we get _ 5 10 ^ 17 ( t 10 ^ 9 ) ^4 . [ etaeff]this model produces a maximum in the viscosity as a function of temperature , a feature that seems quite natural . having said that , more detailed work on the low - temperature phonon problem is obviously needed in order to improve on our results . this is , in fact , a very interesting problem ; when the mean - free path is large , the phonons interact with the `` surface '' more frequently than with each other . in principle , the hydrodynamics approach should not be valid in this regime . yet , there is some evidence from studies of heat conduction in nano - systems ( see @xcite for a recent discussion ) that a judicious choice of surface boundary condition for the phonons leads to a useful `` fluid '' model . we plan to explore this idea further in the future . we now turn to the bulk viscosity . only one of the three required coefficients has , so far , been calculated in detail . in the static limit , the analysis in @xcite leads to the result _ 2(=0 ) = 1.75 10 ^ 12 ( m_s 100 ) ^4 ( 10 ^ 9 t ) , [ zeta0 ] where @xmath78 is the strange quark mass . however , in order to use this result to study oscillation modes , we need consider the frequency dependence in more detail . this is essential since the bulk viscosity is a `` resonant '' mechanism that is particularly effective when the involved dynamics has a timescale similar to the relevant relaxation time . the static limit only provides partial information . following @xcite we will use _ i = _ i(t ) , i = 1 - 3 , [ zeds ] where the `` amplitudes '' @xmath79 , in general , depend on the different physical scales of the system , while @xmath43 is the relaxation time for the processes that give rise to bulk viscosity . according to @xcite , the relaxation time scales according to ~ . [ relax ] the scaling of all three viscosity coefficients , in the static limit , has been determined in @xcite . we should have _ 1&&_1 , + _ 2&&_2 , + _ 3&&_3 , where @xmath80 are constants . this , together with the relations in and allows us to determine the frequency dependent bulk viscosity coefficients , up to the unknown amplitudes @xmath80 and @xmath81 . these can not be determined with the approximations used in @xcite , although we can obviously infer @xmath82 from . from the analogous problem for @xmath0he , we know that the three bulk viscosity coefficients may be of the same order of magnitude . in absence of detailed results for the quark case , it thus makes sense to assume that @xmath83 and @xmath84 have a similar amplitude to @xmath85 . in particular , this would suggest that we parametrise @xmath84 , in the static limit , according to _ 3(=0 ) ( ) ^2 = | 10 ^ 12 ( 10 ^ 9 t ) ( 300 _ q ) , [ zeta30 ] where @xmath86 is an unspecified parameter . the frequency dependent result then takes the form _ 3=[z3 ] , with = 10 ^ 7 ( 10 ^ 9 t ) ^9 ( _ q 300 ) ^8 s[ts ] . here , @xmath81 is another undetermined parameter . in principle , one would expect both @xmath86 and @xmath81 to be of order unity , but given the lack of precise information we can vary them and assess how this affects the bulk viscosity damping of the r - modes . note that the coefficient in has a maximum at the resonance frequency @xmath87 . we know from previous work that it is essential to understand how this resonance frequency relates to the r - mode frequency . a precise statement to this effect is not possible , given the free parameters in our model , but the strong temperature dependence in means that even a change of several orders of magnitude in @xmath43 would produce a relatively small change in the temperature at which the resonance occurs for a given mode frequency . this suggests that the uncertainty associated with @xmath81 may not affect our analysis too severely ( assuming that the temperature scaling is correct , of course ) . if we , for example , consider a mode frequency @xmath88 s@xmath89 and an amplitude such that @xmath90 s , we would have a resonance around @xmath91 k. if , on the other hand , we take a drastically shorter relaxation timescale of @xmath92 s , then the resonance appears near @xmath93 k. that is , the resonance temperature would shift by less than an order of magnitude . this model is obviously phenomenological , and allows us to proceed , but an actual derivation of the frequency dependent viscosity coefficients is needed if we want more detailed results . it is obviously important to compare our results to other relevant estimates for the r - modes . the natural , and most immediate , comparison would be to unpaired quark matter . in that case , which corresponds to a single fluid problem , we have @xcite 1.4 10 ^ 17 ( 0.1 _ s ) ( 10 ^ 14 ^3)^5/3 ( t 10 ^ 9 ) ^-2 , [ unpeta]where @xmath94 is the strong interaction coupling constant , and = a(t ) ^2 + b(t ) , [ unpzeta]where a(t ) = 1.810 ^ 36 ( _ q 300 ) ^3 ( m_s 100 ) ^4 ( t 10 ^ 9 ) ^2 ^3 , and b(t ) = 2.6 10 ^ 7 ( _ q 300 ) ^6 ( 1 + m_s^2 4 _ q^2 ) ^2 ( t 10 ^ 9 ) ^4 ^-2 . we also want to quantify the relevance of the vortex mutual friction for the r - mode instability in cfl matter . in order to do this we need a representation of the counter - moving degree of freedom associated with an r - mode ( which is of order @xmath95 in the slow - rotation approximation ) @xcite . before we consider this problem , we need to translate the mutual friction results from @xcite to our formalism . in order to relate the parameter @xmath96 to the results in @xcite , let us consider equation where we now include a mutual friction force of the form . this leads to an equation for the evolution of the relative velocity of form + _ i t=- 2 . given this equation , and the results in the appendix , we can compare our formalism to the standard mutual friction description , see e.g. equations ( 2.2)-(2.3 ) in @xcite . this allows us to determine @xmath96 from the parameter @xmath97 that is calculated in @xcite . this leads to = 2 ^5 405 ( 1- 2 ) ( 1-c_0 ^ 2 ) ^2 c_0 ^ 6 ( t _ q)^5 , or 3.510 ^ -17(1- 2 ) ( 300 _ q ) ^5 ( t 10 ^ 9 ) ^5 . [ bcoff]we see that the mutual friction vanishes as the temperature decreases . this obviously makes sense since there will be no phonon - vortex scattering when the phonon gas becomes dilute . moreover , according to this result , we are safely in the extreme `` weak drag '' regime ( c.f . the discussion in @xcite ) where this implies that the phonon mutual friction will not damp the r - modes efficiently ( probably as expected ) @xcite . this result is confirmed by the detailed analysis in section [ mutualf ] . however , these results come with an important caveat . strictly speaking , our analysis is only valid as long as the phonons can be described as a fluid . as we have already explained , this is not the case for mature neutron stars . in fact , the results of @xcite are supposedly derived for ballistic phonons . this means that our analysis is somewhat inconsistent . at temperatures above ( say ) @xmath98 k the mutual friction parameter from @xcite may not apply , and below this temperature our two - fluid model may not be appropriate . however , the latter issue may not be that important . after all , one would expect the phonon mutual friction to weaken dramatically at lower temperature , c.f . , meaning the any quantitative errors in the analysis will be of no real importance . our ( simple ) model for a two - component cool cfl core is now `` complete '' , and we may conduct the r - mode analysis as in @xcite . in order to quantify the damping due to bulk viscosity , we need to account for terms of order @xmath95 in the analysis . this complicates the problem , since the centrifugal deformation of the star s shape enters at the same level . in view of this , and the fact that this is a first exploratory study , we will make a sequence of approximations that allow us to proceed analytically . this strategy also makes sense since we are using a simplified equation of state . the results of our analysis should help determine whether a full numerical study is worthwhile . first of all , we need to make a choice of primary variables . the linearised equations link the four scalar variables @xmath99 , @xmath100 , @xmath101 and @xmath102 . in principle , we can use the equation of state to express any two of these in terms of the other two . following the strategy set out in @xcite we will work with @xmath99 and @xmath100 . for our model equation of state we have ( since the background is co - rotating ) p = p_0 + ( t ) t = p_0 + s t , which means that we can express the equations in terms of @xmath103 and @xmath100 . we then have = ( p_0 ) p_0 + ( t ) t = 1 c_0 ^ 2 p_0 , since @xmath58 is contant , and s = ( s p_0 ) p_0 + ( s t ) t = c _ t t , [ ds]since @xmath104 . in order to avoid numerics we will make use of a `` trick '' used by lindblom , owen and morsink in one of the early r - mode instability papers @xcite . the basic idea is to neglect the rotational change in shape in different terms in the perturbation equations . once this is done one can `` estimate '' the bulk viscosity from the leading order r - mode solution . although this simplification is not not based on a rigorous argument , one can show that the bulk viscosity damping timescale is of the right order of magnitude ( within about a factor of 5 of the true second - order slow - rotation result @xcite ) . this is good enough for our present purposes . in our case , the calculation would proceed as follows : first consider the continuity equation i + _ i v_^i + v_^i _ i = 0 , where we have assumed that all perturbed quantities behave as @xmath105 . if we simply omit the last term ( that would vanish for a spherical star since an r - mode is purely toroidal to leading order ) , we get _ i v_^i - i = - i c_0 ^ 2 p_0 . this may not be a very accurate estimate , but it allows us to progress without having to determine the rotational corrections to the r - mode ( which depend on the term that we have neglected ) . now consider the second degree of freedom in a similar way . from the entropy equation we get i s + s ( _ i v_n^i - _ i w_^i ) + ( v_^i - w_^i ) _ i s = 0 . [ dels]in fact , for the model equation of state the last term vanishes identically since @xmath106 and @xmath107 for our background model . it follows that _ i w_^i -i . the approximation is completed by the @xmath108-components of the two euler equations . assuming that the r - mode velocity field takes the form v_^i = e^i(t + m ) , [ vector]and that @xmath54 is similar , although of higher slow - rotation order than @xmath56 @xcite , we get ir^2 ^2 v_+ i m p + 2 r^2 v_= 0 , and 2ir^2 ^2 w_+ ims t = 0 . [ dteq]the second relation shows immediately that @xmath100 is of higher slow - rotation order for the classic r - mode ( since @xmath109 is of higher order and @xmath110 for inertial modes ) . this means that we can justifiably neglect the temperature variation in the pressure term in the first equation . that is , we have ir^2 ^2 v_+ i m p_0 + 2 r^2 v_= 0 . [ dpeq]this is exactly the result used in @xcite . moreover , we now see that the first term in should also be of higher order , which means that we are left with _ i w_^i _ i v_^i . [ compress ] the above relations allow us to estimate the terms needed to evaluate the bulk viscosity integrals once we have the leading order r - mode solution . since the required contribution to the r - mode velocity field is purely toroidal , we have v_= - u_l r^2 _ y_l^m , v_= u_l r^2 _ y_l^m . moreover , the mode - solution is such that the only contribution comes from @xmath111 in which case @xmath112 ( noting that the normalisation is irrelevant at the linear perturbation level ) . finally , to the accuracy needed , the mode frequency is ( in the rotating frame ) = 2m+1 . [ rfreq ] using these results , and expanding the ( scalar ) pressure perturbation in spherical harmonics , i.e. using @xmath113 , we find that leads to = 1 2 m m+1 u_m , and thus we have _ i v_^i _ i w_^i - i 4 m ^2 ( m+1)^2 c_0 ^ 2 u_m y_m+1^m . [ dv]this completes the approximation scheme , which provides all the expressions we need to make a rough estimate of the bulk viscosity damping timescale . in order to estimate the mutual friction damping , we need to determine the detailed counter - moving degree of freedom associated with the r - modes . in principle , this calculation proceeds as in @xcite and involves the rotational corrections to the shape of the star . however , in order to be consistent we will instead build on the approximations introduced in the previous section . first of all , it is easy to show that the countermoving degree of freedom will be poloidal ( in contrast to the toroidal leading order r - mode solution ) . this means that we have w_r = 1r w_l y_l^m , w_= v_l r^2 _ y_l^m , w_= v_l r^2 _ y_l^m . if we also expand the temperature perturbation in such a way that @xmath114 , then it follows from that v_l = i s 2 t_l . similarly , we get from the radial component of ; w_l = i s 2 _ r t_l . combining these results with and , we find that s t_m+1 = 8m^3 ( m+1)^3 ( 2m+5 ) r^2 c_0 ^ 2 ( r r ) ^m+1 . [ tsol ] from this result we easily obtain the required counter - moving components of the r - mode . we arrive at by assuming that @xmath115 at the surface ( in accordance with the discussion in @xcite ) . this condition is unlikely to be correct in the present problem , and a detailed analysis of the appropriate condition to impose on the phonons needs to be carried out in the future ( c.f . the discussion in @xcite ) . at this point we simply note that our assumption that the whole star is composed of cfl matter is artificial anyway . moreover , the boundary condition does not affect the parameter scaling of the result . this is the dominant factor in determining the r - mode damping timescale . to estimate the growth / damping timescales for the r - mode instability we need to evaluate various energy integrals , see @xcite . first of all , the mode energy is given by e = 1 2 m(m+1 ) r^-(2m+2 ) _ 0^r r^2m+2 dr , while the growth time - scale due to ( current multipole ) gravitational - wave emission is obtained from = - 32 g ^ 2m+2 c^2m+3 ( m-1)^2m^2 ( m+2 m+1 ) ^2m+2 _ 0^r r^2m+2 dr . turning to the shear viscosity , we know that @xmath116 for the superfluid r - mode . this means that the shear viscosity integral simplifies to the usual , single fluid result . hence , the leading - order energy loss due to shear viscosity follows from the expression @xmath117 \right\ } dr \\ = - m(m^2 - 1)(2m+1 ) r^{-(2m+2 ) } \int_0^r \eta r^{2 m } dr \ , \label{1shear}\end{gathered}\ ] ] leading to = - ( m-1)(2m+1 ) _ 0^r r^2 m dr ^-1 . next , we need to estimate the bulk viscosity damping . from we know that the general bulk viscosity expression is complicated , especially since we can not rule out the possibility that the three different bulk viscosity contributions are of similar magnitude . however , the approximations that we have made simplify the problem considerably . neglecting the terms that arise due to the change in shape of the rotating star [ the last line in ] , and making use of we find that _ - _ 0^r _ | _ i v_^i |^2 dv - 16 m^2 ^4 ( m+1)^4 ( 2m+3 ) r^-(2m+2 ) _ 0^r _ c_0 ^ 4 r^2m+4 dr , which leads to = - 16 m ^4 ( m+1)^5 ( 2m+3 ) _ 0^r _ c_0 ^ 4 r^2m+4 dr ^-1 , where _ ^2 _ 3 . this result follows after some surprising cancellations , associated with . it is notable that only one of the three bulk viscosity coefficients plays a role in our simplified case . moreover , it is the one ( @xmath85 ) that remains in the navier - stokes limit . of course , one would not expect this drastic simplification for a more realistic stellar model . this fact provides strong motivation for a more detailed , numerical , r - mode analysis . we also learn that we can not , in general , ignore the additional bulk viscosities in the multi - fluid problem . finally , we have the dissipation integral for the mutual friction . from the results discussed in @xcite , we find that _ - 2 ( _ i^m - ^m _ i ) w_^i |w^_m dv . working out the angular integrals , this reduces to _ -2 \ { w_m+1 ^ 2 + v_m+1 ^ 2 } dr , [ mfint]where we have used q_l^2 = ( l+m)(l - m ) ( 2l-1)(2l+1 ) . the integral in is easily evaluated using the eigenfunctions from the previous section . in order to evaluate the relevant integrals , and obtain estimated timescales with explicit parameter scaling , we now assume that the density is uniform . taking all parameters constant , we easily find that _ = -22 ( 1.4 m_m ) ( 10 r ) ^4 ( p 1 ) ^6 , [ tgw]for the @xmath118 r - mode ( the sign indicates that the mode is unstable ) . we also get , for the shear viscosity , 1.1 10 ^ -26 ( 1 ) ( r10 ) ( 1.4 m_m ) ^-1 , and for the bulk viscosity we find 3.8 10 ^ -31 ( _ 1 ) ( r10 ) ^5 ( 1.4 m_m ) ( 1 p ) ^4 ^-1 . [ tbv]finally , in the case of the mutual friction we find 0.17 ( r10 ) ^4 ( p 1 ) ^-5 ^-1 . [ tmf ] this result differs significantly from the estimate in @xcite , in particular , in terms of the scaling with the rotation rate . however , our result is obtained from the actual counter - moving part of the r - mode solution and the final scaling accords with the standard result ( see , for example , @xcite ) . the main difference between our result and that in @xcite can be explained by noting that , if we take the r - mode energy to be proportional to @xmath119 rather than @xmath120 then we arrive at a mutual friction damping time scale very similar to that obtained in @xcite . the large discrepancy between the estimates thus follows from the fact that the velocites @xmath56 and @xmath54 associated with the r - mode solution are of different orders of @xmath121 in the slow - rotation approximation . r - mode unstable for systems located the different curves . the effect of shear- and bulk viscosity associated with phonon processes are shown as blue ( dashed ) and red ( solid / dashed ) curves , respectively . we show the result for both the canonical value @xmath122 ( solid red line ) and the rather extreme value @xmath123 ( dashed red line ) in , leading to relaxation times of @xmath124 s and @xmath125 s ( as indicated in the figure ) , respectively . we also show the result for the case when @xmath86 is increased by a factor of @xmath126 ( thin solid red line ) . our result for the mutual friction damping ( this dash - dot black line ) is compared to the result from @xcite ( thin dash - dot black line ) , demonstrating the dramatic effect of the different scaling with the rotation . for comparison , we also provide the instability curve for the combined shear- and bulk viscosity , obtained from and , in the case of unpaired quark matter ( black line at bottom of figure ) . finally , we show the observed spin period and inferred core temperature for a number of astrophysical systems ( filled squares ) , and we also indicate the region of parameter space where accreting neutron stars in low - mass x - ray binaries ( lmxb ) are thought to be located . the region below @xmath98 k , where the phonon - fluid model may not be appropriate , is shown with grey background to emphasize the need for improved modelling in the astrophysically relevant part of parameter space.,height=302 ] combining these estimates with , , and , i.e. balancing the four timescales , we arrive at the r - mode instability window in figure [ instab ] . the figure shows the critical rotational period for the instability as function of the core temperature for a canonical star with mass @xmath127 and radius @xmath128 km . because of the less efficient damping mechanisms , we show the period rather than the rotation rate . this makes sense because the long timescales involved beome much clearer than when the critical rotation rate is expressed as a small fraction of the rotational break - up limit . as a result , gravitational waves drive the @xmath118 r - mode unstable for systems located the different curves in the figure . our estimates show that the bulk viscosity is much weaker than the shear viscosity at all temperatures of interest . in order to show how the bulk viscosity result depends on the relaxation time , c.f . , we show the result for both the canonical value @xmath122 and the rather extreme value @xmath123 , leading to relaxation times of @xmath124 s and @xmath125 s , respectively . the fact that the associated instability curves do not differ much shows that the results are not very sensitive to this parameter . this is due to the strong scaling with temperature . in order to explore the importance of the overall strength of the bulk viscosity , we also show the result for the case when @xmath86 is increased by a factor of @xmath126 compared to our canonical value @xmath129 . this result shows that the bulk viscosity has to be vastly different from our assumed model in order to be relevant . the data illustrated in figure [ instab ] also shows that the instability curve obtained from balancing the gravitational - wave driving and our result for the phonon mutual friction damping is vastly different from the result in @xcite . the different dependence on rotation is obvious , and it is clear that the phonon mutual friction would be irrelevant for all astrophysical stars with cfl cores . for comparison , we show the instability curve for the combined shear- and bulk viscosity , obtained using and , in the case of unpaired quark matter . it is interesting to note that the phonon shear viscosity in the cfl case leads to the strongest r - mode damping at temperatures above @xmath98 k. of course , astrophysical neutron stars will cool to temperatures much lower than this soon after birth so this region may not be that relevant . the results in the figure show , quite clearly , that we need to improve our understanding of the low - temperature regime . the instability curve below the @xmath98 k is obtained using our phenomenological model . future work needs to model this regime in detail . for comparison , we also show the observed spin and inferred core temperature for a number of astrophysical systems in figure [ instab ] ( the data is taken from @xcite ) . this comparison is somewhat inconsistent because we have assumed that the core is shielded by a normal crust with a thickness of , at least , a few hundred meters ( so that the standard heat - blanket argument @xcite applies . the presence of a normal matter outer region should provide addition r - mode dissipation channels that we have not considered . in particular , one would expect a viscous boundary layer at the crust - core interface to be a more efficient damping agent ( see @xcite for the most recent discussion ) than the phonon mechanisms that we have considered . we also indicate the region of parameter space where accreting neutron stars in low - mass x - ray binaries would be located ( given the usual arguments @xcite ) . these systems would be at variance with the pure cfl plus phonon model in the sense that they would be located deep into the unstable region . if there were an active r - mode instability in these systems , one would expect it to have observational effects on , for example , the spin evolution . observations do not provide any evidence of this . we have presented the first true multi - fluid analysis of a dense neutron star core with a deconfined , colour - flavour - locked superconducting , quark core . by focussing on a cool system , and accounting only for the condensate and ( finite temperature ) phonons , we made progress by taking over much of the formalism from the analogous problem for superfluid @xmath0he , the archetypal two - fluid laboratory system . the additional fluid degree of freedom , in the present case represented by the phonon gas , leads to the system not being well represented by the navier - stokes equations . in particular , a complete model requires a number of additional viscosity coefficients . without an actual calculation it is not easy to establish whether the multi - fluid aspects are relevant or not . for example , in the case of the gravitational - wave driven instability of the f - modes it is known that the superfluid degree of freedom is very important , since the vortex mutual friction may completely suppress that instability below a critical temperature ( see @xcite ) . it is known that , because of the different nature of the associated velocity field , the mutual friction does not affect the r - mode instability in the same drastic fashion @xcite . these examples provide clear evidence that different problems need to be considered on a case - by - case basis . we have provided a detailed dissipative formulation for a system comprising a quark ( cfl ) condensate and phonons . the model builds on recent improvements in our understanding of the analogous problems of superfluid @xmath0he @xcite and causal heat conductivity @xcite . a key ingredient is the massless entropy component that represents the phonon excitations . we have discussed how the superfluid constraint of irrotationality reduces the number of required viscosity coefficients to four ( one shear and three bulk ) , and provided a translation between results in the literature @xcite and the coefficients in our formalism . we also emphasised that many more dissipative channels may come into play in a rotating system where superfluid vortices are present @xcite . in particular , we showed how the vortex mediated mutual friction is accounted for in the model , and translated the available results for the associated coefficients @xcite . in order to be able to make relevant estimates for the r - mode instability , we developed a simple two - component equation of state based on the mit bag model at zero temperatures with an additional phonon gas representing the thermal component . this example highlights the additional information that is needed in a multi - fluid analysis , in particular , regarding the entrainment coupling . future work needs to provide a consistent equation of state , including the key aspects . the model equation of state completed the formulation of the problem and we could , in principle , have carried out a numerical study of the r - modes . we opted not to do this , instead introducing a sequence of simplifying assumptions , because we felt that it would be useful to start by working out some less precise estimates for the relevant dissipation timescales . a numerical analysis of the problem should , of course , be encouraged . the problem has a number of interesting aspects , and may shed light on how one should deal with finite temperature superfluid neutron stars in general . this work was motivated by a desire to understand the different phases of cfl matter from a hydrodynamics point of view . it is , obviously , an interesting problem and the notion that observations of gravitational waves from relativistic stars may help shed light on the extreme qcd phase diagram is exciting . the r - mode instability has been discussed in this context for some time , see for example @xcite , but there has not been any previous discussion of the multi - fluid aspects of the problem . we hope that this work will stimulate a more detailed discussion between experts in the relevant areas . the fact that the various phonon processes that we have accounted for can be shown to have little effect on the r - mode instability should not discourage future efforts . in fact , the result could probably have been anticipated . the real challenge will be to account for additional degrees of freedom , especially associated with the kaons ( either thermal or in a condensate ) @xcite , and the modelling of `` hybrid '' stars with quark cores and various phase transitions . these problems have additional features that , while possibly understood in principle , have never been considered in practice . this makes the modelling more complex , but also intriguing since the richer dynamics may lead to surprises . the relationship between the flux conservative formulation and the orthodox framework for superfluid helium has been examined in detail elsewhere @xcite . still , it is useful to summarise the main results that are needed to relate the different dissipation coefficients . first of all , we identify the `` normal '' fluid in the standard description with the gas of excitations , which is directly associated with the entropy of the system . this leads to v^i_=v^i _ . secondly , the so - called `` superfluid '' velocity is given by v^i_==(1-)v_^i+v_^i , where @xmath130 is the entropy entrainment parameter . this means that the total mass flux takes the form v_^i=_v_n^i+_v^i_ , from which we learn that _ = _ = - , where @xmath131 and @xmath62 are the superfluid and normal fluid densities . we are grateful to mark alford and cristina manuel for a number of useful discussions . na and bh acknowledge support from stfc via grant number pp / e001025/1 . bh also acknowledges support from the european science foundation ( esf ) for the activity entitled `` the new physics of compact stars '' , under exchange grant 2449 , and thanks the dipartimento di fisica , universit degli studi di milano for kind hospitality during part of this work .

we report radial velocity studies of five candidate metal - poor field blue stragglers , all known to be deficient in lithium . four of the five stars are single - lined spectroscopic binaries , with periods ranging from 302 to 840 days , and low orbital eccentricities , in agreement with similar behavior found for other blue straggler candidates by preston & sneden ( 2000 ) and carney et al . ( 2001 ) . the limited data available for lithium abundances indicate that all blue straggler binaries have depleted lithium , but that constant velocity stars generally have normal lithium abundances . this suggests that the lithium gap " for hot metal - poor main sequence stars may not exist or lies at higher temperatures than found in the hyades . our results and those of preston & sneden ( 2000 ) show higher values of for the binary stars than those of comparable temperature constant velocity stars . the orbital periods are too long for tidal effects to be important , implying that spin - up during mass transfer when the orbital separations and periods were smaller is the cause of the enhanced rotation . the mass function distribution is steeper for the blue straggler binary stars than that of lower mass single - lined spectroscopic binaries , indicating a narrower range in secondary masses . we argue that if all secondaries are white dwarfs with the same mass , it is probably around 0.55 . the models of rappaport et al . ( 1995 ) , applied to white dwarf secondaries , suggest that the orbital elements of all metal - poor binary blue stragglers are consistent with stable mass transfer , with the possible exception of g202 - 65 .

astro-ph0412110

the blue stragglers " were first identified by sandage ( 1953 ) in the globular cluster m3 . since their discovery , blue stragglers have been found in many other globular clusters , and in open clusters as well . the frequency of blue stragglers is also correlated with position in clusters , being generally higher in the denser , more dynamically evolved portions , indicating that stellar interactions , perhaps stellar mergers or hardening of binaries leading to mass transfer , may be the cause or at least one of the causes of the phenomenon ( mateo et al . hut et al . ( 1992 ) , stryker ( 1993 ) , livio ( 1993 ) , trimble ( 1993 ) , bailyn ( 1995 ) , and leonard ( 1996 ) have provided extensive reviews of blue stragglers in globular clusters . some globular clusters show declining frequencies ( per unit luminosity ) of blue stragglers in their lower density outer regions , but others ( m3 : ferraro et al . 1993 , 1997 ; m55 : zaggia et al . 1997 ; 47 tuc : ferraro et al . 2004 and mapelli et al . 2004 ) show an enhanced frequency in the outermost , lowest density regions . this is explained qualitatively by invoking two different mechanisms for the creation of blue stragglers in clusters . in the central regions , dynamical effects involving collisions may enhance the probabilities for creating blue stragglers ( sills & bailyn 1999 ) . in the outer regions , normal stellar evolution may be responsible for their creation . a variety of stellar evolution explanations have been proposed to explain such unusual systems , one of which ( mass transfer : mccrea 1964 ) does not require any alterations to standard stellar evolution theory . but two explanations ( pulsation - driven mass loss : willson , bowen , & struck - marcell 1987 ; internal mixing : wheeler 1979a , b ; saio & wheeler 1980 ) do require significant modifications and make the understanding of the blue straggler phenomenon of high importance . the mass transfer model is perhaps favored by the increasing blue straggler frequencies seen in the outer regions of the above clusters because that is where binary systems are most likely to survive . mapelli et al . ( 2004 ) , for example , were able to reproduce the blue straggler binary frequency distribution in 47 tuc using a combination of collisional creation of blue stragglers and survival of mass transfer systems . the collisions , as expected , dominate the central regions , while the outer regions blue stragglers arise both from collisions and mass transfer in binary systems . preston & sneden ( 2000 ; hereafter ps2000 ) expanded on the earlier work of preston , beers , & shectman ( 1994 ; hereafter pbs ) and argued that _ field _ blue stragglers are created almost solely by the mass transfer process . not only does this make good qualitative sense , but they supported the argument by noting that field blue stragglers are much more common per unit luminosity than are blue stragglers in the lower - density regions of globular clusters . as ps2000 noted , this is consistent with the idea that even in those low density regions of clusters , binary systems may have been destroyed , thereby lowering the numbers of blue stragglers . in clusters , we must keep in mind that even the stars found in the low - density outer regions may be vulnerable to stellar encounters if their intra - cluster motions carry them into the central higher density regions . field blue stragglers therefore provide the cleanest " sample for the study of the mechanism that may create such stellar oddities via stellar evolution , independent of environment . the distributions of such observables as binary stars periods and orbital eccentricities are therefore much less likely to have been among field stars than in clusters . as we describe below , however , field star samples have introduced another puzzle , one at least as intriguing as the blue straggler phenomenon itself . we argued ( carney et al . 2001 ; hereafter cllgm ) and argue again here that binary star evolution and mass transfer is one path for the creation of blue stragglers , and apparently the most common among field stars . however , some metal - poor main sequence field stars that are hotter than globular cluster main sequence turn - offs are not binaries , but could be explained as the partial remnants of an accreted dwarf satellite galaxy whose star formation continued over a long period of time . following pbs , we distinguish between stars with delayed evolution and those that are simply young stars . we regard the blue straggler phenomenon as the process(es ) that delay(s ) normal stellar evolution . pbs identified a significant number of field metal - poor stars bluer than comparable metallicity globular cluster main sequence turn - off stars : candidate blue stragglers , in other words . preston & landolt ( 1998 ) found that the star cs @xmath1 is a very metal - poor ( [ fe / h ] @xmath2 ) sx phe pulsating variable , and is also a binary star with an orbital period of 431 days and a relatively low orbital eccentricity of 0.10 . ( sx phe variables lie in the instability strip s extension to the main sequence domain , and thus represent the hotter blue stragglers in globular clusters : see nemec , nemec , & lutz 1994 for a discussion and lists of such stars . ) the orbital properties are a vital clue to its origin because the normal orbital eccentricity of main sequence binary stars is much higher , with @xmath3e@xmath4 close to 0.37 ( duquennoy & mayor 1991 , latham et al . 2002 ) , independent of metallicity . an extensive study of blue metal - poor field stars by ps2000 found a surprisingly high fraction of binary stars , which led them to suggest that many of them were field blue stragglers . many of these stars showed periods of hundreds of days with low orbital eccentricities , as in the case of cs @xmath1 . cllgm identified ten metal - poor field blue straggler candidate stars from the large survey of carney et al . ( 1994 ) , based on colors that were slightly to significantly bluer than globular cluster main sequence turn - offs ( at similar metallicities ) . they found six of the stars to have periods in this same range , 167 to 844 days , and low orbital eccentricities ( @xmath5 ; @xmath6 ) . they argued that these periods and eccentricities are consistent with mass transfer . they noted that mass transfer systems such as ch , subgiant ch , and dwarf carbon stars likewise have very high binary fractions and that those binary systems have similarly long orbital periods and low orbital eccentricities . sneden , preston , & cowan ( 2003 ; hereafter spc2003 ) have strengthened the argument for mass transfer in their recent abundance analyses of six blue metal - poor stars from the ps2000 program , finding that the three constant velocity stars have normal element - to - iron abundance ratios for metal - poor stars , but that three binary stars show major enhancements of carbon and @xmath7-process elements , indicative of mass transfer from a highly - evolved agb star . in summary , then , mass transfer appears to be the primary cause of the blue straggler phenomenon among field stars . therefore , it may also be a major contributor to the formation of blue stragglers in clusters , although the situation there is more complicated by induced dynamical evolution of binary stars as well as outright capture and merger of stars . since the orbital characteristics are such an important component of the mass transfer model , more velocity data of known field blue stragglers and identification of additional candidates and their binary frequency and orbital characteristics are highly desirable . it is also important to employ consistent techniques for identifying candidate blue stragglers and for determining the orbital elements of the binary stars . in the latter case , we note that our criteria for identification of binaries and acceptance of final orbital solutions are more strict than those of some investigators ( see latham et al . 1988 , latham et al . 1992 , and latham et al . this has compelled us to re - evaluate the orbital solutions available in the literature , using the published velocity data , in order to re - analyze all the velocity data in a consistent manner . thus , while ps2000 found 42 binaries among their 62 program stars , we could unambiguously identify only 29 stars as being binaries using their published velocities and our software and experience . the other 13 stars may indeed be binaries , but our criteria and experience lead us to believe that additional radial velocities are needed to confirm such identifications . we also note that for some stars , ps2000 did not have enough time coverage to obtain a unique orbital period and offered alternative solutions when necessary . we have retained for use in this paper only the ten stars from ps2000 whose orbital elements satisfy our criteria . in all these cases , our orbital solutions are indistinguishable from those obtained by ps2000 . spc2003 have also addressed this point with additional radial velocity data for selected stars . we are confident that additional observations will resolve any remaining uncertainties . we add that g. preston ( priv . comm . ) has obtained additional velocities for cs @xmath8 and cs @xmath9 , and those results confirm the orbital solutions for those stars presented by spc2003 . for this paper , we have chosen to concentrate our efforts on seeking additional stars , rather than refinements of existing orbital solutions presented by ps2000 . the exciting results of spc2003 provides another motivation : a larger sample , and of brighter stars , could enable further progress in the search for chemical signatures of mass transfer , and , perhaps , an understanding of different mass transfer mechanisms ( e.g. , red giant or agb donor star ? ) . finally , we have not lost sight of perhaps the most intriguing discovery of pbs : they argued that many of their stars are merely masquerading as blue stragglers and that they may , in fact , be bona fide young , metal - poor stars , accreted perhaps from a small dwarf galaxy which had experienced a prolonged star formation history or perhaps a relatively recent burst of star formation . candidate field blue straggler stars may therefore provide clues to both stellar evolution and the evolution of our galaxy . we have two challenges in identifying metal - poor blue stragglers in the field population . the first is the rarity of such stars . large samples of field stars must be studied to find those few that are bluer than the main sequence turn - off of comparable metallicity globular clusters , and we must also be assured that they are main sequence stars and not the more luminous blue horizontal branch stars . thus accurate colors , reddenings , gravities , and metallicities are required for large numbers of stars . this is why the two major recent studies of metal - poor field blue stragglers have emerged from the bmp " ( blue metal poor ) studies like that of pbs and ps2000 , and the carney - latham - laird - aguilar " survey ( carney et al . 1994 ; cllgm ) . such methods pick up both true blue stragglers as well as any truly young metal - poor stars , which may have been acquired by the galaxy during a minor merger event ( pbs ) , as noted above . in their study of 62 blue metal - poor stars , many of which qualify as blue straggler candidates using our color criterion , ps2000 found an abnormally high fraction of binary stars , roughly four times higher than those of the disk and halo populations ( duquennoy & mayor 1991 ; latham et al . 1988 , 2002 ) . on the other hand , the frequency of double - lined systems was unusually low , suggesting a population of underluminous secondary stars , possibly white dwarfs . such evidence , plus a low fraction of binary systems with short orbital periods , and a very high fraction of binary systems with low orbital eccentricities , suggested to ps2000 that their large sample of blue metal - poor stars is a mixture of both relatively young stars , with , presumably , a normal frequency of binary stars and with orbital elements like those of galactic disk and halo stars , plus a significant number of blue stragglers whose origin is most readily explained by mass transfer . lithium is a very `` fragile '' element , vulnerable to proton capture at the relatively low temperature of order @xmath10 k. cool stars , with deep enough convection zones , may transport surface lithium to depths where lithium can be destroyed , thereby gradually depleting the lithium abundance in the photosphere . in metal - poor main sequence stars , lithium abundances are observed to decline with effective temperature for @xmath11 k ( see thorburn 1994 and references therein ) , presumably due to the deeper convection zones for such stars . the same effect is seen in metal - poor subgiants ( pilachowski , sneden , & booth 1993 ) . however , among the hotter main - sequence dwarfs and subgiants , lithium abundances in metal - poor stars show a near - constant value of log n(li ) = 2.2 , which is often referred to as the spite plateau " , after its discovery by spite & spite ( 1982 ) . there is now an extensive literature on lithium abundances in metal - poor stars ( see spite , molaro , & spite 1993 and ryan et al . 2001 for additional observations and references ) . mass transfer ( or mixing or pulsation - driven mass loss , for that matter ) should lead to depletion of surface lithium abundances . indeed , lithium is not found in many candidate blue stragglers studied to date ( hobbs & mathieu 1991 ; pritchet & glaspey 1991 ; glaspey , pritchet , & stetson 1994 ; cllgm ) . on the other hand , normal lithium abundances have been found in some stars that are blue straggler candidates , suggesting that they are either blue stragglers created by an as - yet unidentified mechanism , or truly younger metal - poor stars ( ryan et al . 2001 ; cllgm ) . we therefore include in table 1 the lithium abundances for our program stars , insofar as they have been determined by others . alas , the lithium abundance may not be entirely capable of distinguishing a field metal - poor blue straggler star ( low lithium ) from a young metal - poor dwarf ( normal lithium ) . in the hyades cluster , boesgaard & trippico ( 1986 ) and boesgaard & budge ( 1988 ) found that lithium abundances are very depleted in main sequence stars with effective temperatures lying roughly in the range of 6400 k to 6800 k. this lithium gap " , if it arises during the evolution of normal metal - poor main sequence stars , could lead us to confuse a young but otherwise normal metal - poor dwarf with a metal - poor blue straggler . we discuss this further in section 5.2 . spc2003 discovered strong enhancements of elements probably created by @xmath7-process nucleosynthesis in three stars that satisfy the color conditions described below . these abundance anomalies speak strongly to mass transfer , and also the evolutionary state of the donor star , and hence we have searched the literature to determine what is known about the abundances of two elements which are often associated with @xmath7-process nucleosynthesis , strontium and barium . these abundances are also summarized in table 1 . we rely once more on a large sample of field stars to select our program stars . stetson ( 1981 ) primarily used proper motion criteria to select possible high velocity stars from the sao catalog , and obtained a list of high - velocity early - type stars ( stetson 1991 ) . based on high - resolution spectra , glaspey , pritchet , & stetson ( 1994 ) found that these stars could be divided into three broad groups of stars , one of which is relevant to metal - poor field blue stragglers . their table 4 ( part _ iii _ ) includes six metal - poor stars that are also deficient in lithium . one of these stars , sao 80390 ( bd+25 1981 ) was studied by cllgm , and found to be a possible radial velocity variable . the other five stars constitute our new sample . in table 1 we provide the basic photometric and chemical information for these five stars , plus the ten blue straggler candidates from cllgm , and the ten metal - poor ( [ fe / h ] @xmath3 @xmath12 ) stars from ps2000 for which cllgm could obtain satisfactory orbital solutions using their software and generally more conservative criteria . the de - reddened @xmath13 values , effective temperatures , and metallicities for the ps2000 stars were taken from that paper , while for bd+23 74 , hd 8554 , hd 109443 , and hd 135449 , the values were taken from or determined using data taken from stetson ( 1991 ) and glaspey et al . values for the remaining stars were determined following the same precepts discussed by carney et al . ( 1994 ) . in figure 1 we show the color selection criterion for blue straggler candidates . as discussed by cllgm , the solid line represents a second - order polynomial fit to the de - reddened @xmath13 color indices of selected globular cluster main sequence turn - offs as a function of metallicity . because the tail of the field star metallicity distribution extends to lower metallicities than that of the globular clusters , we have also employed the y@xmath14 " isochrones ( kim et al . this use of the isochrones is far from rigorous since we ignore color shifts that may be required to bring the isochrones into agreement with observed cluster color - magnitude diagrams . further , we have made the convenient , perhaps naive , assumption that all globular clusters have the same age . we simply chose the isochrones with enhanced abundances of the @xmath15 " elements , [ @xmath15/fe ] = + 0.3 , and then chose the cluster ages to all be 11 gyrs , represented by the dashed line in figure 1 . this provides a reasonable match to the edge of the color distributions of the field stars over a wide range of metallicities , especially at the lowest metallicities , and a good match to the actual globular cluster turn - off distribution near the peak of the halo metallicity distribution ( [ m / h ] @xmath16 @xmath17 ) . stars blueward of _ both _ of these lines are considered blue straggler candidates . note that cs @xmath18 does not satisfy this criterion , and we exclude it from the following discussions . [ tab1 ] the normal stars , not known to be blue stragglers , are plotted as plus signs . however , compared to our earlier work ( cllgm ) , we have increased the number of normal stars for comparison . we have completed and are preparing for publication an additional sample of 470 metal - poor field stars selected from ryan ( 1989 ) and ryan & norris ( 1991 ) . three of those stars , g121 - 54 , bd@xmath19 3682 , and ltt 15049 ( = hd 154578 ) , plotted as plus signs but bluer than the two dividing lines , are now blue straggler candidates . we will report on their radial velocity behavior in a future paper . we have also continued our observations of the four constant velocity " stars in our earlier paper , bd+72 94 , bd+40 1166 , bd+25 1981 , and hd 84937 , and report the revised results as well . the observations were made using the same procedures as described by cllgm . we employed the center for astrophysics ( cfa ) digital speedometers ( latham 1985 , 1992 ) on the multiple mirror telescope and 1.5-m tillinghast reflector at the whipple observatory atop mt . hopkins , arizona , and on the 1.5-m wyeth reflector located in harvard , massachusetts . the echelle spectrographs were used with intensified photon - counting reticon detectors and recorded about 45 of spectrum in a single order centered near 5187 . the resultant resolution was about 8.5 kms@xmath20 . typical signal - to - noise values ranged from from 7 to 50 per resolution element . radial velocities were measured from the observed spectra using the one - dimensional correlation package * rvsao * ( kurtz & mink 1998 ) running inside the iraf environment . as before , we used a grid of synthetic spectra calculated using the model atmospheres computed using r. l. kurucz s code atlas9 . the details of the calculation of the models and the grid of synthetic spectra were presented by cllgm . for each of the five new blue straggler candidates we adopted a template from the grid of synthetic spectra with the effective temperature and metallicity nearest to the values for that star reported in table 1 . we used @xmath21 throughout . we computed correlations for all spectra of each star using all the available rotationally broadened synthetic spectra with the most appropriate temperature , gravity , and metallicity . to determine the radial velocity , we chose the rotationally - broadened template with the value of that gave the highest average value for the peak of the correlation function . this established the optimum set of parameters for the synthetic spectrum template for each star . the parameters adopted for the final correlations are coded in the final column of table 1 : effective temperature , log g , [ fe / h ] , and . table 2 is an illustrative summary of the individual radial velocity measurements , including the heliocentric julian day of mid - exposure , and the velocity and internal error estimate returned by the iraf task * rvsao * , both in kms@xmath20 . the full table with all the velocity information is available electronically . we summarize our radial velocity results in table 3 , including the total number of velocities measured for each star , the span of our velocity coverage ( in days ) , the derived rotational velocity , @xmath22 sin @xmath23 , the mean radial velocity , and the uncertainty of the mean velocity . note that for the binary stars , the mean radial velocity is not as appropriate as the systemic velocity that emerges from the orbital solution . for stars with orbital solutions , we therefore list here the systemic velocity and its uncertainty . we include in table 3 the measured rms external error , @xmath24 , and the mean internal error , @xmath25 , of the velocity measurements ( see kurtz & mink 1998 ) , and the ratio , @xmath26 . large values of @xmath26 ( @xmath27 and above ) are suggestive of radial - velocity variability . we also employ the probability , p(@xmath28 ) , that the @xmath28 value could be larger than observed due to gaussian errors for a star that actually has constant velocity . we employ the internal error estimate , @xmath29 , obtained from * rvsao * , for each of @xmath30 exposures when calculating @xmath31 : @xmath32 it is clear that at least four of our five new program stars display radial velocity variability . [ tab3 ] hd 142575 does not appear to be a radial velocity variable , and we show its velocity history in figure 2 . table 3 shows that the probability that the measured value of @xmath28 could arise by chance is large , which is a strong indicator of radial velocity stability . the four stars in our earlier work ( cllgm ) that appeared to have constant radial velocities have rather small p(@xmath28 ) values , in particular for bd+25 1981 . however , the ten new radial velocities have resulted in a higher p(@xmath28 ) value for this star than in our earlier paper , when it was only 0.000002 . the rest of figure 2 shows the velocity histories of these four stars . no obvious signs of orbital motion are seen , although bd+25 1981 remains an intriguing case . orbital solutions for four of the five new program stars are given in table 4 , as well as calculated uncertainties . we include the orbital period @xmath33 ( in days ) , the systemic ( mean ) radial velocity , @xmath34 , the orbital semi - amplitude , @xmath35 , the orbital eccentricity , @xmath36 , the longitude of the ascending node , @xmath37 , the time ( in hjd ) of the periastron passage , the projected semi - major axis of the primary star s orbit ( in giga - meters ) , the mass function , @xmath38(@xmath39 ) ( in units of a solar mass ) , the number of observations ( and the span in days of the observations ) , and , finally , the probable error of the match of the observed velocities to the orbital solution . please note that our radial velocity observations span several orbital cycles for all four stars . the orbital solutions and the phased velocity data are displayed in figure 3 . [ tab4 ] we have attempted to derive internally consistent values for the stellar rotational velocities , , using our synthetic spectra for all of our program stars . we have described above how we selected the optimum value of for the synthetic spectrum to derive the radial velocities . to provide the best estimate for , we interpolated between the average peak correlation values for all spectra using different values of in the synthetic spectra . we used quadratic fits of the correlation values vs. to select the final value . these are given in table 3 . we have tested the accuracy of our derived rotational velocities by employing two sets of template spectra for each star . the first choice of templates are those given in table 1 . the second choice was made by the computer , with the template selected independently on the basis of the height of the peak of the correlation function for the fifteen stars with templates listed in the table . in six cases this resulted in exactly the same adopted template . the other nine cases enable us to determine how a change in temperature or gravity alter our derived rotational velocity estimate . the greatest change was for hd 97916 , where the computer s preferred template was 750 k cooler and 0.5 dex lower in gravity , and the @xmath22 sin @xmath23 value changed by only 1.8 kms@xmath20 , down to 13.1 kms@xmath20 . for bd+40 1166 and hd 84937 , the computer chose templates 250 k cooler and hotter , respectively , than the values given in table 1 . even here , where the derived rotation velocities are comparable to or smaller than our instrumental resolution , the changes were small , @xmath40 and @xmath41 kms@xmath20 , respectively . for more rapidly rotating stars , the sensitivity was even smaller . for bd+23 74 , for example , a change of 1000 k changed the derived @xmath22 sin @xmath23 value by only 0.4 kms@xmath20 . the power of synthetic spectra and the @xmath28 fitting technique s sensitivity to many weak lines carrying the same information is demonstrated quite nicely by these tests . ps2000 drew attention to the relatively long periods and low orbital eccentricities of many of the binary stars they discovered . in combination with the high binary frequency and small fraction of double - lined systems , they argued that their results supported the identification of many of these binary systems as blue stragglers created by mass transfer , a model proposed originally by mccrea ( 1964 ) . cllgm similarly drew attention to the unusual orbital properties of the six field blue stragglers they studied and the ten systems discovered by ps2000 with [ fe / h ] @xmath42 @xmath12 and for which our own software resulted in what we considered compelling orbital solutions . neglecting the three ps2000 systems with periods of less than 20 days ( and whose orbital elements may have been altered by tidal interactions , independent of mass transfer ) , we found that 10 of the 13 binaries had @xmath43 and @xmath44 . for all 13 systems , @xmath45 and @xmath46 . the periods ranged from 167 days up to roughly 1600 days . one should recall that the median orbital eccentricity for normal main sequence , single - lined spectroscopic binaries is about 0.37 , independent of metallicity ( duquennoy & mayor 1991 ; latham et al . 2002 ) . mass transfer appears to be the best explanation for these orbital characteristics , as discussed originally by webbink ( 1986 ) . our earlier paper also discussed stellar systems where mass transfer is likely to have occurred , leaving an overabundance of elements synthesized during the agb evolutionary stage deposited in the outer layers of the receptor . if we confine our attention to the more metal - poor stars , there are three classes of stars to consider . the ch stars are not luminous enough to be agb stars themselves , yet show abundances expected of such stars , implying mass transfer has occurred and also that the companion star is likely to be a white dwarf . mcclure & woodsworth ( 1990 ) found that essentially all ch stars they studied are binaries , and those with orbital solutions showed periods of 328 to 2954 days , and low orbital eccentricities ( @xmath47 ; @xmath48 ) . the subgiant ch " stars discovered by bond ( 1974 ) may be descendents of some blue stragglers , as noted by luck & bond ( 1991 ) . mcclure ( 1997 ) found that 9 of the 10 such stars under long - term radial velocity monitoring are binary systems with long periods . he obtained orbital solutions for six of the sytems , finding orbital periods of 878 to 4140 days , and small orbital eccentricities ( @xmath49 ; @xmath50 , @xmath51 ) . finally , the one well - studied dwarf carbon star , g77 - 61 , has been found to have a nearly circular orbit and a moderately long period of 245 days ( dearborn et al . 1986 ) . one of the primary purposes of the observations reported in this paper was to test the idea that many blue stragglers are binary stars , and that their binary orbital properties confirm this kinship with the more obvious mass transfer systems described above . in figure 4 we update figure 4 from cllgm , which showed the orbital eccentricity vs. log p for the 6 new binary systems reported therein , and the 10 systems fom ps2000 subjected to the same orbital solution analyses . for comparison , we included the 156 orbital solutions for cooler single - lined spectroscopic binaries from latham et al . ( 2002 ) . in this new figure 4 , we highlight the locations of the four new orbital solutions from table 4 using filled circles . ( filled squares are our results from cllgm ; filled diamonds are stars from ps2000 with periods longer than 20 days , and open diamonds are stars from ps2000 with shorter periods . ) as before , the longer orbital period binary stars have unusually low values of the orbital eccentricity . in figure 5 , we take a different perspective , and consider the distribution of orbital eccentricities for stars in figure 4 with periods longer than 20 days . aside from the absence of high orbital eccentricities among the blue stragglers , we note that the orbital eccentricities cluster near zero but have a tail to higher values . this might indicate two different types of mass transfer and orbital evolution , despite the similarity of the orbital periods . alternately , it may reflect the original distribution of orbital eccentricities since binary systems with very high eccentricities evolve more slowly , and so are more likely to preserve a higher - than - average orbital eccentricity . in figure 6 we show the lithium gap " seen among some members of the hyades , using the results of cayrel et al . ( 1984 ) , boesgaard & trippico ( 1986 ) , and boesgaard & budge ( 1988 ) . open circles are measurements and v " symbols represent upper limits . the gap is easily seen between effective temperatures of about 6400 k and 6800 k. the normal lithium abundance for metal - poor dwarfs hotter than about 5400 k ( the spite plateau " ) . is shown as a dashed line . that value is lower than the peak of the hyades dwarfs and is in itself a long - standing puzzle , but beyond the point of this paper . we also show the stars from table 1 , with filled circles for stars that have measured lithium abundances , and which , it turns out , are all also stars with constant velocities . none of the binary stars in table 1 have detectable lithium lines and hence the abundances are only upper limits . these stars are plotted as downward - pointing arrows . note that for hd 142575 , glaspey et al . ( 1994 ) derived @xmath52 = 6550 k , and fulbright ( 2000 ) adopted @xmath52 = 6500 k , but we have adopted the photometric temperature of 6700 k to maintain consistency with the derivation of the temperature for all of our program stars . we have adjusted the lithium abundance derived for the star by fulbright ( 2000 ) as well to account for the change in temperature : log n(li ) rises from 1.45 to 1.58 . ( note as well , however , that glaspey et al . 1994 argued that log n(li ) @xmath3 0.7 . ) the first interesting point is that three of the metal - poor single stars have normal lithium abundances , at least for metal - poor stars . two of the stars , hd 84937 and cs @xmath18 , are very close to our adopted color limits used to identify blue stragglers . small changes in the measured @xmath13 color index or in the reddening estimate could easily shift either star into or out of the blue straggler candidate domain . however , bd+72 94 is considerably bluer and hotter , and in the middle of the hyades lithium gap , yet has a lithium abundance consistent with the spite plateau . conclusions from such a small sample may not be secure , but the results suggest the lithium gap may not exist for single metal - poor stars , or , perhaps , it is shifted to higher temperatures than the stars in our sample . a survey of the metal - poor single stars identified by ps2000 would answer this question . related to this point , hd 142575 , which appears to be single and slightly hotter than the other three single stars , has a reduced lithium abundance according to fulbright ( 2000 ) . is this star distinguishable in any way from the three constant velocity , normal - lithium stars ? if our photometric temperature estimate is correct , the star is slightly hotter than the other single stars , and so might be in the metal - poor equivalent to the hyades lithium gap . table 3 may provide another clue : hd 142575 has a relatively large value for . is rotation a factor in either the formation of the lithium gap or the depletion of lithium more generally ? we need more observations of lithium absorption line strengths in blue metal - poor stars , especially those studied by ps2000 . we conclude that our use of low lithium abundances as a selection criterion is supported by the ( admittedly limited ) data available . we note also that the relation between blue stragglers and lithium depletion has been discussed extensively by ryan et al . ( 2001 ) . ryan et al . ( 2002 ) have noted the modest but significant levels of rotation in three of the blue stragglers discussed in our previous paper ( cllgm ) . for bd+51 1817 , g66 - 30 , and g202 - 65 , they found @xmath22 sin @xmath23 values of @xmath53 , @xmath54,and @xmath55 kms@xmath20 , all of which are larger than values derived for cooler metal - poor dwarfs with normal lithium abundances . ( note that our method obtains somewhat higher values of 9.0 , 8.1 , and 11.6 kms@xmath20 , respectively . ) ryan et al . ( 2002 ) suggested that the likely origin of the rotation was the mass transfer process itself , and discussed the mechanisms by which such a process would lead to lithium depletion . these include mass transfer - induced deep mixing or transfer of lithium - depleted material from the donor star . regarding the latter point , pilachowski et al . ( 1993 ) found that surface lithium depletion in post - main sequence metal - poor stars begins before the stars reach the base of the red giant branch , so mass transfer from a more highly evolved red giant would deposit lithium - depleted material on the recipient main sequence star . we explore here the available rotational velocity information for the stars of table 3 . as described above , we are able to estimate using our grid of synthetic spectra . because different approaches appear to derive different values of , we proceed differentially by comparing values of constant velocity and binary stars only within a particular methodology . thus we make comparisons only between stars we have studied , or among the stars studied by ps2000 , but we do not intermingle those results . for the stars for which we have estimated , we must first explore the results for a control sample " . in figure 7 we show the histograms of the derived values using data taken from latham et al . we include only single stars judged to be main sequence dwarfs , and single - lined spectroscopic binary stars with periods exceeding 25 days . tidal circularization effects are clearly present for metal - poor dwarfs with orbital periods of 20 days or less , but should not be significant for the longer period systems we consider here . none of these three different groups were found to have different distributions of . however , metallicity does seem to have a small influence on the values we derive , as may be seen in figure 7 . we do not claim that figure 7 represents the true distribution of rotational velocities of metal - poor dwarfs because our instrumental resolution ( @xmath56 kms@xmath20 ) prevents accurate determinations for values much smaller than that . on the other hand , we are confident of our results for rotational velocities at or above that of our instrumental resolution . the fraction of stars with @xmath57 8.0 kms@xmath20 in the combined sample is only 4.0% ( 7.7% for the more metal - poor stars and 0.3% for the more metal - rich stars ) . note that we do not believe these results imply that metal - poor dwarfs rotate somewhat faster than metal - rich dwarfs . the differences may be artificial in the sense that the synthetic spectrum templates may introduce metallicity - dependent differences in the derived rotational velocities . in figure 8 we show the rotational velocities we have derived ( open and filled circles ) for the stars of table 3 , plus values from ps2000 ( open and filled squares ) . while the observations of ps2000 were obtained using a system with velocity resolution comparable to ours , they derived values using a different method than have we , which is why we distinguish the two sets of results graphically . the key question is whether mass transfer is a major cause of the blue straggler phenomenon and whether evidence for it exists in higher rotational velocities for the binary systems . because tidal interactions will spin up short - period binary systems , independent of mass transfer , we have excluded binary systems with orbital periods of less than 20 days from the figure or from further consideration here . we consider the results from ps2000 first . there are 14 stars in their study that appear to have constant velocity and have [ fe / h ] @xmath42 @xmath58 . recall that they have interpreted ( correctly , we believe ) these stars to be bona fide young metal - poor main sequence stars . note that all of them have @xmath42 12 kms@xmath20 , and @xmath3@xmath4 = 10.1 kms@xmath20 ( @xmath59 = 1.7 kms@xmath20 ) . cs @xmath18 , which is a binary system but which we do not consider to be a blue straggler candidate , has = 10 kms@xmath20 . [ we have assigned cs @xmath60 a temperature of 6925 k rather than 6900 k so it would not be overplotted on another star . ] if we restrict the sample to those stars from ps2000 for which we could obtain an independent orbital solution that agrees with their results ( cllgm , table 4 ) , we find that none of the seven binary stars with orbital periods longer than 20 days have values as small as the largest value found for the single stars , and that @xmath3@xmath4 = 17.3 kms@xmath20(@xmath59 = 3.5 kms@xmath20 ) . [ here we have altered the temperature for cs @xmath61 from 7050 k to 7000 k to inhibit it being overplotted on another star . ] if we further consider the 11 more binary stars for which ps2000 presented only one orbital solution , but which are not plotted in figure 8 , only two have values of 12 kms@xmath20 or less , while the remaining nine stars have values reaching as high as 160 kms@xmath20 . our own results , from our previous work ( cllgm ) and this paper , confirm this picture . we consider bd+25 1981 an uncertain case and have plotted it in figure 8 as an open triangle . the other four constant velocity stars ( bd+72 94 , bd+40 1166 , hd 84937 , hd 142575 ) have @xmath3@xmath4 = 7.8 kms@xmath20(@xmath59 = 3.8 kms@xmath20 ) , while the ten binary stars in table 3 have @xmath3@xmath4 = 18.7 kms@xmath20(@xmath59 = 10.0 kms@xmath20 ) . in summary , our own studies and the results of ps2000 show that blue straggler candidates which are binaries show significantly enhanced surface rotation compared to constant velocity stars . because the orbital periods imply separations far too great for tidal interactions to _ currently _ have any significant effects on the stars rotational periods , we believe the enhanced rotation indicates mass transfer has occurred and is the cause of the spin - up . the mass transfer model predicts that the originally more massive star transfers mass to the lower mass secondary , diminishing the orbital period and separation , until the original secondary star becomes the more massive component . at that point , the stars begin the separate , but mass transfer has transformed orbital angular momentum into rotational angular momentum , and tidal interactions or some mechanism has reduced the orbital eccentricities . as noted above , luck & bond ( 1991 ) suggested that some blue stragglers may be progenitors of the subgiant ch stars . until recently , there has been little information regarding abundances of carbon and @xmath7-process abundances in blue stragglers that would indicate mass transfer from an agb star , but such data are now beginning to appear , and so we have summarized in table 1 the abundances of the @xmath7-process elements strontium and barium . the available data for most candidate blue straggler stars are consistent with the normal abundances of these elements in metal - poor main sequence and red giant branch stars ( see mcwilliam 1997 ) . recently , spc2003 have found strong enhancements of these elements in three of the blue metal - poor stars discussed previously by ps2000 . thus for these stars the evidence favors mass transfer from an evolving agb star , supporting the hypothesis of luck & bond ( 1991 ) . if mass transfer from a highly - evolved agb star is responsible for the @xmath7-process enhancements , whereas the stars without such enhancements experienced mass transfer from a normal red giant , we might expect to see some differences in the orbital properties of the two samples . mass transfer from an agb star could imply , for example , that mass transfer did not occur during the preceding red giant evolutionary phase , and , therefore , the original orbital periods of binary systems destined to experience mass transfer from an agb star should be longer than those which undergo mass transfer from a first ascent red giant . one might expect , then , that blue stragglers showing signs of agb mass transfer ( @xmath7-process enhancements ) should have longer minimum orbital periods than those that ( presumably ) arose from mass transfer from a red giant ( no such enhancements ) . qualitatively , table 1 confirms this view . excluding cs @xmath62 , the eight stars with normal strontium and barium abundances have orbital periods as low as 167 days , and none exceed 500 days . the six subgiant ch stars with orbital solutions obtained by mcclure ( 1997 ) have periods only as short as 878 days , and extend up to 4140 days . the eight ( giant ) ch stars studied by mcclure & woodsworth ( 1990 ) have orbital periods ranging from 328 and 2954 days , roughly consistent with this simple picture . cs @xmath62 is consistent with this trend , with an orbital period of 1307 days ( spc2003 ) . its orbital eccentricity is also on the high side ( 0.21 ) for the ensemble of field halo blue stragglers , as figure 4 shows . however , the other two stars with @xmath7-process enhancements , cs @xmath8 and cs @xmath9 , show rather short periods , according to spc2003 , with 342 days ( and e @xmath63 ) and 194 days ( and e @xmath64 ) , respectively . we have not included these stars in table 1 because our orbital solutions using the data available to spc2003 indicated that alternative periods and eccentricities were possible , although less likely . preston ( private communication ) has continued to observe these stars and these orbital periods are increasingly robust . in particular , we have been able to confirm to our own satisfaction the orbital period for cs @xmath8 has cost the first author payment of a wager with dr . preston that longer periods would be preferred . ] . additional velocity data are needed , however , for cs @xmath9 . we are therefore confronted with the results that mass transfer from red giant and agb stars yield similar ranges in orbital periods and eccentricities . we can only speculate that diversity in the masses of the original primary stars and their original orbital periods contribute to this splendid if confusing variety . a point stressed by mcclure & woodsworth ( 1990 ) and mcclure ( 1997 ) was that the mass functions of the ch and subgiant ch stars , respectively , were small and indicative of a uniform secondary mass , roughly that of a white dwarf with 0.6@xmath0 . we undertook a similar analysis in our previous paper , which we revisit here . we begin by comparing the distribution of the mass functions of the blue stragglers and the somewhat lower masses of the primary stars in the single - lined spectroscopic binaries of latham et al . ( 2002 ) for stars with orbital periods longer than 20 days . the mass function for a spectroscopic binary is defined to be @xmath65 where the numerical value of @xmath66 is derived from the parameters of the orbital solution . figure 9 shows that the two mass function distributions are quite different , with the steeper one for the blue stragglers indicative of a narrower range of secondary masses , as in the cases of the ch and subgiant ch stars . so what is that secondary mass ? we must begin with an estimate of the masses of the primary stars , then exploit the mass functions to get some idea of the secondary masses of the ensemble . as we have noted , the lack of a detectable secondary spectrum in any of the blue straggler binary systems is suggestive of underluminous secondaries , and we explore here the idea that all the secondaries are stellar remnants with similar masses . in cllgm , we estimated the mass of each primary star using the derived effective temperatures and chemical compositions , plus the isochrones of girardi et al.(2000 ) . the isochrone ages were relatively young , generally 1.7 to 2.0 gyrs , to accomodate the high temperatures and apparent high masses of the primaries . the isochrones were computed using elemental abundances consistent with solar values and , to allow for enhanced abundances of the @xmath15 " elements ( [ @xmath15/fe ] @xmath67 ) , we employed the scaling relation of salaris , chieffi , & straniero ( 1993 ) to calculate @xmath68 for each primary . here we exploit the yonsei - yale " ( y@xmath14 ) isochrones of yi , kim , & demarque ( 2003 ) , which were computed using @xmath15-enhanced abundances and somewhat more appropriate ( and variable ) helium mass fractions . we used isochrones with an age of 2.0 gyrs in all cases . table 5 shows the results . we have also compared these new primary mass estimates with those computed using the girardi et al . ( 2000 ) isochrones . the maximum difference in derived primary mass was 0.04 @xmath0 , and the mean difference , in the sense of y@xmath14 minus girardi et al . ( 2000 ) , was @xmath69 @xmath0 , with @xmath70 @xmath0 . [ tab5 ] from the estimated value of @xmath71 , we can estimate the mass of the undetected companion ; i.e. for @xmath72 , using a revised version of equation [ eq : massfunction ] : @xmath73 these are also provided in table 5 . all of the minimum masses lie below the expected mass of a metal - poor white dwarf , @xmath74@xmath0 . they also lie far below those expected for a metal - poor main sequence turn - off star ( @xmath75@xmath0 ) or that of a neutron star ( @xmath76@xmath0 ) . the mean difference in secondary masses derived using the two sets of isochrones is only @xmath77 @xmath0 , with @xmath78 @xmath0 . if all the secondaries have roughly the same mass , is it closer to that of a white dwarf or a neutron star ? if we _ assume _ that all the secondaries have the same mass of @xmath79 , and calculate the orbital inclinations that this would imply , we obtain the @xmath80 values listed in table 5 . the average @xmath80 for the 17 stars in table 5 is @xmath81 ( @xmath82 ) , which is not far from the value of @xmath83 expected for randomly oriented orbits . ( use of the isochrones of girardi et al . 2000 leads to an identical result . ) increasing the assumed secondary mass to 0.65 @xmath0 , results in @xmath84 ( @xmath85 ) , which is inconsistent with the expectation by a significant amount , although we must recognize that our results suffer some observational bias in that very low values of sin @xmath23 result in smaller observed radial velocity variations and such systems are more likely to elude detection . if we further assume the secondaries are all neutron stars , with @xmath86@xmath0 , we find @xmath87 ( @xmath88 ) , and we rule out that class of secondaries , at least in the general case . we conclude that our orbital solutions are consistent with all the unseen companions in our low - eccentricity long - period orbits being white dwarfs viewed at random orbital orientations , which supports the mass - transfer model for the formation of field blue stragglers . if all the secondaries are white dwarfs with the same mass , then that mass can not be much smaller than @xmath89 , because the three largest minimum masses are @xmath90 ( cs @xmath62 ) , @xmath91 ( cs @xmath61 ) , and @xmath92 ( g202 - 65 ) , where the quoted errors include only the contribution from the mass functions and not from possible systematic errors in our estimates of the primary masses . on the other hand , the secondary masses can not be larger than about @xmath93 , because then the average value for the orbital inclinations would become seriously inconsistent with random orientations of the orbits . ryan et al . ( 2002 ) have discussed the mechanism for the mass transfer that might have increased the rotational velocities of the stars we see now as blue stragglers . one of the questions they raised was whether the mass transfer mechanism could have involved stable roche lobe overflow or whether the mass transfer process was unstable . sudden mass loss via a supernova , for example , would be such an unstable process , although we have essentially ruled out neutron star companions among this sample of field blue stragglers . rappaport et al . ( 1995 ) modelled the results of roche lobe overflow . they considered the evolution of binaries undergoing stable mass transfer , but for the case where the recipient is a neutron star . the same physics should also apply to systems where the recipient is a main - sequence star . the essence of the model is that short - period systems are so close that mass transfer begins early , when the expanding red giant core has not yet grown to a mature " size . therefore , the resultant white dwarf will appear with a small , premature " mass . in figure 10 , we plot the minimum white dwarf mass as a function of the orbital period and eccentricity , log [ @xmath94 . we also show the model predictions of rappaport et al . ( 1995 ) using their coefficient @xmath95 to be 3300 @xmath96 , consistent with the population ii case that they discussed . if all the secondary stars were white dwarfs and had evolved via stable mass transfer , their minimum masses , plotted in figure 10 , should all lie to the left of the theoretical curve , and we find that the data do so , more or less , except for g202 - 65 . thus almost all of the blue stragglers in table 5 , with orbital periods longer than 20 days , have orbital periods and eccentricities consistent with stable mass transfer . this includes cs @xmath62 , the star that shows signs of @xmath7-process enhancements and an implied agb donor star . we show in figure 10 the locations of the other two such stars from sp2003 , cs @xmath8 and cs @xmath9 . all three stars lie well within the regime of stable mass transfer . our new search for radial velocity variability of five metal - poor and lithium - deficient field stars has shown that four of them are single - lined spectroscopic binaries with orbital periods of a few hundred days and small orbital eccentricities . combining these results with those of ps2000 and cllgm , we have found that field metal - poor dwarfs hotter than comparable metallicity globular cluster turn - offs may represent stars with two very different origins . many of the binary stars studied by ps2000 and most , if not all , of the binary stars studied by cllgm and in this paper are distinguished from normal binary stars by their unusually low lithium abundances and unusually low binary orbital eccentricites . we have shown that the secondaries appear to have a unusually narrow range of masses , and are consistent with white dwarfs of 0.55 @xmath0 . despite the long periods of these binary systems , we find that their values are , on average , higher than those of single stars with comparable temperatures , suggesting spin - up as part of a mass transfer event . using the models of rappaport et al . ( 1995 ) , we argue that in all cases but one ( g202 - 65 ) , the orbital periods and eccentricities are consistent with stable mass transfer . in some cases , spc2003 found enhanced values of @xmath7-process abundances , signifying mass transfer from an evolved agb star , but not all blue stragglers show such enhancements . we have not discovered any correlation between the presence or lack of such enhancements and the binary system orbital properties . another interesting result is that among the constant velocity stars , lithium abundances are normal , including one star , bd+72 94 , that appears to be in the lithium gap " . hd 142575 , on the other hand , may lie in the metal - poor equivalent lithium gap , which may occur at somewhat higher temperatures than is seen in the hyades . more lithium abundance determinations for constant velocity stars in this temperature regime would be very interesting , both to strengthen the use of lithium as a distinguishing feature of blue stragglers created by mass transfer , and as a tool for exploring the cause of the lithium gap as well . these constant velocity stars may have originated within a dwarf satellite galaxy that underwent star formation over a long period of time and which only recently merged with the milky way , as suggested by pbs . additional radial velocities should be obtained of the constant velocity and binary stars in our samples ( cllgm and this paper ) as well as those stars studied by pbs , ps2000 , and spc2003 to refine the spectroscopic orbital solutions and more securely identify the true constant velocity stars . it is a pleasure to thank the national science foundation for financial support to the university of north carolina and to bowling green state university during the many years that we have been puzzling over the blue straggler phenomenon . as always , we are indebted to the many people who made observations for this project , especially joe caruso , perry berlind , bob davis , robert stefanik , jim peters , mike calkins , ed horine , joe zajack , skip schwartz , willie torres , ale milone , and dick mccroskey . lllllclcrcl bd+23 74 & 0.19 & @xmath97 & 7500 & 837 & no & @xmath98 & 1 & & & @xmath99 hd 8554 & 0.29 & @xmath100 & 6780 & 302 & yes & @xmath101 & 1 & & & @xmath102 hd 109443 & 0.36 & @xmath103 & 6650 & 724 & yes & @xmath104 & 1 & & & @xmath105 hd 135449 & 0.32 & @xmath106 & 6740 & 327 & yes & @xmath107 & 1,14 & & & @xmath108 hd 142575 & 0.33 & @xmath109 & 6550 & & yes & 1.45 & 1,3 & ; + 0.05 & 3 & @xmath110 & & & & & & & & & & bd+72 94 & 0.32 & @xmath111 & 6620 & & yes & 2.42 & 2 & & & @xmath112 bd+40 1166 & 0.42 & @xmath113 & 6200 & & no & & & & & @xmath114 bd+25 1981 & 0.29 & @xmath115 & 6860 & & no & @xmath116 & 1 & & & @xmath117 bd@xmath118 2669 & 0.28 & @xmath119 & 6710 & 381 & yes & & & & & @xmath120 + hd 84937 & 0.37 & @xmath121 & 6300 & & no & 2.23 & 10,11 & + 0.57/@xmath122 & 3,8,10 & @xmath123 hd 97916 & 0.415 & @xmath124 & 6250 & 663 & no & @xmath125 & 2 & ; + 0.05 & 3 & @xmath126hd 106516 & 0.45 & @xmath127 & 6120 & 844 & no & @xmath128 & 7,12,13 & ; @xmath129 & 4 & @xmath130 bd+51 1817 & 0.38 & @xmath131 & 6330 & 517 & no & @xmath132 & 2 & & & @xmath114 g66 - 30 & 0.37 & @xmath133 & 6280 & 688 & no & @xmath134 & 2 & @xmath135;+0.16 & 15 & @xmath136 g202 - 65 & 0.36 & @xmath137 & 6560 & 167 & yes & @xmath138 & 2 & & & @xmath139 & & & & & & & & & & @xmath140 & 0.16 & @xmath141 & 6560 & 486 & yes & & & @xmath142;+0.46 & 5 & @xmath143 & 0.17 & @xmath144 & 8050 & 1.0 & no & & & & & @xmath18 & 0.34 & @xmath145 & 6420 & 19 & yes & 2.15 & 2 & & & @xmath146 & 0.28 & @xmath147 & 6630 & 303 & yes & & & + 0.27;@xmath148 & 5 & @xmath149 & 0.20 & @xmath150 & 7700 & 2.0 & no & & & & & @xmath151 & 0.30 & @xmath152 & 6720 & 485 & yes & & & + 0.26;+0.15 & 5 & @xmath153 & 0.31 & @xmath154 & 6590 & 300 & yes & & & + 0.15;+0.16 & 5 & @xmath62 & 0.34 & @xmath155 & 6900 & 1307 & no & & & + 1.38;+0.37 & 6 & @xmath156 & 0.28 & @xmath157 & 6785 & 306 & yes & & & + 0.15;+0.19 & 5 & @xmath61 & 0.30 & @xmath158 & 6510 & 1576 & yes & & & & & w & 2450630.7792 & 33.98 & 2.89 + w & 2450655.7600 & 23.57 & 2.44 + w & 2450670.8359 & 31.28 & 1.85 + w & 2450691.8055 & 31.59 & 1.93 + w & 2450713.8439 & 28.01 & 1.61 + w & 2450728.7021 & 28.16 & 2.42 + w & 2450745.5618 & 37.96 & 2.48 + w & 2450771.5908 & 32.73 & 3.33 + w & 2450791.6284 & 26.96 & 3.94 + w & 2450811.5603 & 25.90 & 3.57 + & & & & & & & & & & + + & & & & & & & & & & + bd+23 74 & 00:32:43.3 @xmath15924:13:21 & 55 & 2558 & 29.6 & 32.52 & 0.60 & 4.47 & 2.27 & 1.97 & 0.000000 hd 8554 & 01:24:42.3 @xmath15907:00:05 & 32 & 1544 & 11.0 & 11.13 & 0.64 & 3.60 & 1.05 & 3.44 & 0.000000 + hd 109443 & 12:34:46.7 @xmath16023:28:32 & 43 & 2342 & 29.1 & 43.36 & 0.48 & 3.14 & 1.25 & 2.50 & 0.000000 + hd 135449 & 15:16:10.3 @xmath16032:53:33 & 33 & 2575 & 29.5 & @xmath161 & 0.69 & 3.94 & 1.37 & 2.88 & 0.000000 + hd 142575 & 15:55:02.8 @xmath162:04:12 & 16 & 4758 & 13.0 & @xmath163 & 0.17 & 0.54 & 0.68 & 0.80 & 0.784932 + & & & & & & & & & & + + & & & & & & & & & & + bd@xmath118 2669 & 08:46:39.6 @xmath164:21:25 & 135 & 2097 & 33.3 & 41.60 & 0.26 & 6.72 & 1.59 & 4.23 & 0.000000 + hd 97916 & 11:15:54.2 + 02:05:12 & 32 & 5083 & 14.9 & 61.04 & 0.16 & 4.12 & 0.72 & 5.74 & 0.000000 + hd 106516 & 12:15:10.5 @xmath165:18:44 & 39 & 5181 & 11.5 & 4.41 & 0.09 & 4.84 & 0.50 & 9.79 & 0.000000 + bd+51 1817 & 13:08:39.1 + 51:03:59 & 28 & 2194 & 9.0 & @xmath166 & 0.21 & 5.94 & 0.73 & 8.17 & 0.000000 + g66 - 30 & 14:50:07.8 + 00:50:27 & 27 & 4366 & 8.1 & @xmath167 & 0.14 & 4.50 & 0.81 & 5.55 & 0.000000 + g202 - 65 & 16:35:58.5 + 45:51:59 & 49 & 3871 & 11.6 & @xmath168 & 0.26 & 11.35 & 0.89 & 12.80 & 0.000000 + & & & & & & & & & & + + & & & & & & & & & & + bd+72 94 & 01:47:12.3 + 73:28:27 & 40 & 7441 & 5.5 & @xmath169 & 0.12 & 0.79 & 0.67 & 1.17 & 0.060699 + bd+40 1166 & 05:05:28.7 + 40:15:26 & 38 & 7997 & 8.6 & 105.31 & 0.12 & 0.75 & 0.66 & 1.13 & 0.067409 + bd+25 1981 & 08:44:24.6 + 24:47:47 & 81 & 6909 & 10.0 & 57.54 & 0.08 & 0.75 & 0.57 & 1.33 & 0.000139 + hd 84937 & 09:48:56.0 + 13:44:39 & 82 & 7355 & 4.8 & @xmath170 & 0.10 & 0.91 & 0.73 & 1.24 & 0.005681 + bd+23 74 & 840.4 & + 32.71 & 5.21 & 0.168 & 35 & 51974.6 & 59.4 & 0.0118 & 55 & 2.5 + & @xmath17114.8 & @xmath1710.36 & @xmath1710.50 & @xmath1710.099 & @xmath17132 & @xmath17172.5 & @xmath17130.0 & @xmath1710.0171 & 2558 & + hd 8554 & 302.5 & + 11.86 & 5.11 & 0.026 & 222 & 51111.1 & 21.2 & 0.0042 & 32 & 0.9 + & @xmath1711.6 & @xmath1710.17 & @xmath1710.25 & @xmath1710.047 & @xmath171103 & @xmath17185.9 & @xmath1711.8 & @xmath1710.0011 & 1544 & + hd 109443 & 684.5 & + 42.48 & 4.36 & 0.109 & 99 & 51743.3 & 40.8 & 0.0058 & 42 & 1.3 + & @xmath1717.8 & @xmath1710.26 & @xmath1710.34 & @xmath1710.079 & @xmath17139 & @xmath17173.1 & @xmath1714.0 & @xmath1710.0035 & 2342 & + hd 135449 & 326.1 & @xmath16042.50 & 5.49 & 0.053 & 283 & 51630.1 & 24.6 & 0.0056 & 33 & 1.3 + & @xmath1711.9 & @xmath1710.30 & @xmath1710.34 & @xmath1710.062 & @xmath17170 & @xmath17161.9 & @xmath1714.0 & @xmath1710.0027 & 2575 & + & & & & & & & & + + & & & & & & & & + bd @xmath118 2669 & 2.586 & 0.071 & @xmath119 & 0.28 & 6710 & 0.91 & 0.38 & 0.742 + hd 97916 & 2.822 & 0.042 & @xmath172 & 0.41 & 6235 & 0.93 & 0.39 & 0.771 + hd 106516 & 2.926 & 0.041 & @xmath127 & 0.45 & 6100 & 0.91 & 0.42 & 0.818 + bd + 51 1817 & 2.713 & 0.043 & @xmath131 & 0.38 & 6340 & 0.94 & 0.35 & 0.702 + g66 - 30 & 2.838 & 0.293 & @xmath133 & 0.38 & 6310 & 0.85 & 0.26 & 0.551 + g202 - 65 & 2.224 & 0.145 & @xmath137 & 0.36 & 6340 & 0.87 & 0.50 & 0.935 + cs 22166 - 041 & 2.687 & 0.024 & @xmath173 & 0.32 & 6560 & 0.96 & 0.48 & 0.901 + cs 22876 - 008 & 2.481 & 0.075 & @xmath154 & 0.28 & 6630 & 0.97 & 0.34 & 0.685 + cs 22892 - 027 & 2.686 & 0.259 & @xmath152 & 0.30 & 6720 & 1.05 & 0.21 & 0.508 + cs 22948 - 068 & 2.477 & 0.044 & @xmath147 & 0.31 & 6590 & 0.89 & 0.24 & 0.503 + cs 22956 - 028 & 3.116 & 0.207 & @xmath155 & 0.34 & 6335 & 0.83 & 0.52 & 0.955 + cs 22966 - 054 & 2.486 & 0.059 & @xmath174 & 0.28 & 6785 & 1.04 & 0.40 & 0.776 + cs 29518 - 039 & 3.198 & 0.075 & @xmath158 & 0.30 & 6510 & 0.86 & 0.55 & 1.000 + & & & & & & & & + + & & & & & & & & + bd+23 74 & 2.923 & 0.18 & @xmath97 & 0.19 & 7500 & 1.28 & 0.31 & 0.619 + hd 8554 & 2.481 & 0.017 & @xmath100 & 0.29 & 6780 & 0.98 & 0.18 & 0.390 + hd 109443 & 2.860 & 0.27 & @xmath103 & 0.36 & 6650 & 1.18 & 0.22 & 0.471 + hd 135449 & 2.514 & 0.043 & @xmath106 & 0.32 & 6740 & 1.08 & 0.21 & 0.447 + latham , d. w. , mazeh , t. , carney , b. w. , mccrosky , r. e. , stefanik , r. p. , & davis , r. j. 1988 , , 96 , 567 latham , d. w. , mazeh , t. , stefanik , r. p. , davis , r.j . , carney , b. w. , krymolowski , y. , laird , j. b. , torres , g. , & morse , j. a. 1992 , , 104 , 774

we present to resolution 3 mm continuum and line emission and near infrared k , , and [ feii ] images toward the massive star forming region w75 n. the co emission uncovers a complex morphology of multiple , overlapping outflows . a total flow mass of m extends 3 pc from end - to - end and is being driven by at least four late to early - b protostars . more than 10% of the molecular cloud has been accelerated to high velocities by the molecular flows ( relative to ) and the mechanical energy in the outflowing gas is roughly half the gravitational binding energy of the cloud . the w75 n cluster members represent a range of evolutionary stages , from stars with no apparent circumstellar material to deeply embedded protostars that are actively powering massive outflows . nine cores of millimeter - wavelength emission highlight the locations of embedded protostars in w75 n. the total mass of gas & dust associated with the millimeter cores ranges from 340 m to 11 m. the infrared reflection nebula and shocked emission have multiple peaks and extensions which , again , suggests the presence of several outflows . diffuse emission extends about 0.6 parsecs beyond the outer boundaries of the co emission while the [ feii ] emission is only detected close to the protostars . the infrared line emission morphology suggests that only slow , non - dissociative j - type shocks exist throughout the pc - scale outflows . fast , dissociative shocks , common in jet - driven low - mass outflows , are absent in w75 n. thus , the energetics of the outflows from the late to early b protostars in w75 n differ from their low - mass counterparts they do not appear to be simply scaled - up versions of low - mass outflows . # 1#2#3#1#2#3 # 1#2#3 = 500 = 500 =

astro-ph0210555

w75 n is a massive star forming region with an integrated iras luminosity of @xmath10 ( moore , mountain , & yamashita 1991 ; moore et al . 1988,1991 ) . the w75 n cloud is located at a distance of 2 kpc ( dickel , wendker , & bieritz 1969 ) , just @xmath11 north of the massive outflow system dr 21 powered by a cluster of ob stars ( e.g. garden et al . 1991 and references therein ) . both dr 21 and w75 n are part of the cygnus - x complex of dense molecular clouds . haschick et al . ( 1981 ) identified three regions of ionized gas in w75 n at a resolution of @xmath12 : w75 n ( a ) , w75 n ( b ) , and w75 n ( c ) . hunter et al . ( 1994 ) later resolved w75 n ( b ) with @xmath13 resolution into three regions : ba , bb , and bc . torelles et al . ( 1997 ) then imaged w75 n ( b ) at @xmath14 resolution , and detected ba and bb ( which they called vla 1 & vla 3 ) , along with another weaker , and more compact hii region , vla 2 . a parsec - scale molecular outflow originates near the cluster of ultracompact hii ( uc hii ) regions in w75 n ( b ) . the mass of the co outflow has been estimated to be 50 m to 500 m based on single - dish , co observations ( e.g fischer et al . 1985 ; hunter et al . 1994 ; davis et al . 1998a , b ; ridge & moore 2001 ) . the uc hii regions have a combined @xmath15 of @xmath16 l and most are in a protostellar phase based on the presence of oh , , & methanol masers , and compact millimeter continuum emission ( baart et al . 1986 ; hunter et al . 1994 ; torrelles et al . 1997 ; minier , conway , & booth 2000 , 2001 ; shepherd 2001 ; hutawarakorn , cohen , & brebner 2002 ; slysh et al . 2002 ; watson et al . 2002 ) . several studies have assumed the flow is dominated by a single massive star : the central source in the uc hii region vla 1 ( ba ) because the position angles of the ionized gas and the co emission are similar ( hunter et al . 1994 ; torrelles et al . 1997 ; davis et al . 1998a , b ) . shepherd ( 2001 ) suggested vla 3 ( bb ) and , perhaps , vla 2 may be the primary powering sources based on the presence of compact millimeter continuum emission . more recently , hutawarakorn et al . ( 2002 ) suggested vla 2 is the dominant source powering the outflow based on oh maser emission . given the sheer number of interpretations , it is clear that w75 n is a confused region . assuming a primary driving source for the co outflow , davis et al . ( 1998b ) suggested that the co red - shifted lobe and morphology supported a jet - driven , bow - shock entrainment scenario in which a steady , over - dense molecular jet , developed to explain highly - collimated outflows from low - mass protostars , was applied to w75 n ( lada & fich 1996 ; smith et al . 1997 ; & suttner et al . the proposed model implied a jet radius of 0.03 pc at 1.3 pc from the star with a jet opening angle of about 2.6 ( richer et al . if a powerful , well - collimated jet was being driven by an ob protostar in w75 n , it would provide strong constraints on outflow / accretion theories for luminous protostars ( see , e.g. , shang et al . 2002 ; cabrit , ferreira , & raga 1999 ; knigl 1999 ; shu et al . 2000 ; knigl & pudritz 2000 ) . to obtain a better understanding of the number of sources driving outflows and the energetics of the flow(s ) , we have made interferometric mosaics of the w75 n region in co(j=10 ) and millimeter continuum using the owens valley radio observatory and obtained images at near - infrared wavelengths using the telescopio nazionale galileo to compare the morphology of the shocked gas & infrared nebulosity with the co emission . observations in 2.7 mm continuum and line were made with the owens valley radio observatory ( ovro ) array of six 10.4 m telescopes between 1999 march 15 and 1999 december 4 . projected baselines ranging from 15 to 115 meters provided sensitivity to structures up to about @xmath17 . the final @xmath18 mosaic images of both line and continuum emission are made up of 17 fields with primary beam @xmath19 ( fwhm ) spaced @xmath20 apart . the total integration time on source was approximately 3.25 hours / pointing center . cryogenically cooled sis receivers operating at 4 k produced typical single sideband system temperatures of 200 to 600 k. the gain calibrator was the quasar bl lac and the bandpass calibrators were 3c 454.3 and 3c 345 . observations of uranus , neptune , or 3c 273 provided the flux density calibration scale with an estimated uncertainty of @xmath21% . calibration was carried out using the caltech mma data reduction package ( scoville et al . images were produced using the miriad software package ( sault et al . 1995 ) and deconvolved with a maximum - entropy - based algorithm designed for mosaic images ( cornwell & braun 1988 ) . the co @xmath22 data at 115.27 ghz were convolved with a @xmath23 taper resulting in a synthesized beam of @xmath24 ( fwhm ) at p.a . the spectral resolution was 2.6 and the rms noise was 0.13 . the spectral band pass was centered on the local standard of rest velocity ( @xmath5 ) of @xmath26 ( the assumed systemic velocity of the w75 n cloud ) , taken from the cs(j=76 ) emission peak ( hunter et al . 1994 ) . simultaneous 2.7 mm continuum observations were made in a 1 ghz bandwidth channel with central frequency 112.77 ghz . the @xmath22 data were convolved with a @xmath27 taper resulting in a synthesized beam @xmath28 ( fwhm ) at p.a . the rms noise was 3.6 mjy beam@xmath30 . an additional on - source integration time of 4.6 hours was obtained with ovro centered on the position of w75 n : mm 1 ( @xmath31 @xmath32 ) . observations were made on 1999 march 29 and 2001 march 18 . baselines between 35 and 240 meters provided sensitivity to structures up to @xmath27 . the co @xmath22 data ( spectral resolution 2.6 ) were convolved with a @xmath33 taper resulting in a synthesized beam of @xmath34 ( fwhm ) at p.a . the final rms noise was 50 mjy beam@xmath30 in each channel . observations were made in and with the kitt peak 12 m telescope on 2000 may 9 using the sis 3 mm receiver with 1 mhz filter banks centered on @xmath36 to give a velocity resolution of 2.6 and a total bandwidth of 650 . system temperatures ranged from 230 k for to 360 k for . the half - power beam width ( hpbw ) at 115 ghz is about @xmath37 . single dish spectra were obtained at three positions ( j2000 coordinates ) : centered on w75 n : mm 1 ( @xmath38 @xmath39 ) ; and in the south - east and north - west outflow lobes ( @xmath38 @xmath40 and @xmath38 @xmath40 , respectively ) . the data were reduced with the nrao unipops software package . the resulting spectra were used to estimate the optical depth in as a function of velocity and position in the w75 n region . near - infrared observations of the w75 n region were made on 2000 june 16 , at the 3.5 m telescopio nazionale galileo ( tng ) at the roque de los muchachos observatory on the spanish island of la palma . the arnica nir imager ( lisi et al . @xcite ; hunt et al . @xcite ) was used to obtain images in the h@xmath41 narrow band filter and in the k@xmath2 broad band filter . arnica is equipped with a hgcdte @xmath42 nicmos3 infrared array . the pixel scale , when coupled with the tng , is 0.355 arcsec / pixel , and the corresponding field of view is @xmath43 arcmin@xmath44 per frame . the seeing at the time of the observations was about 09 . to search for emission beyond the co emission , a 14 pointing mosaic pattern was employed covering an area of approximately 9@xmath452 arcmin@xmath44 that was roughly aligned with the co flow . the mosaic was repeated several times , dithering the telescope by a few pixels each time , until the desired integration time was achieved . the final integration times per sky position was 8 minutes in k@xmath2-band and 30 minutes in the h@xmath41 filter . data reduction and analysis were performed using the iraf software package . following standard flat - fielding and sky subtraction , the individual images were registered and the final mosaic was produced . the k@xmath2-band observations were calibrated using standard stars from the arnica list ( hunt et al . the h@xmath41 mosaic was calibrated assuming that a set of stars have the same flux density in the narrow- and broad - band filters . the broad - band mosaic was then used to subtract the continuum emission from the h@xmath41 mosaic . integrated line fluxes are then estimated assuming the width of the h@xmath41 filter as measured by vanzi et al . ( @xcite ) . the calibration accuracy is expected to be within 20% . accurate ( @xmath46 ) astrometry was derived for both mosaics using stellar positions from the 2mass second incremental data release . additional observations of two @xmath474@xmath48 fields centered north - east and south - west of w75 n were obtained in 2002 august 21 using the tng near infrared camera spectrograph ( nics , baffa et al . each of the two fields was observed through the h@xmath41 ( @xmath49=2.12 @xmath50 m ) and [ feii ] ( @xmath49=1.64 @xmath50 m ) narrow - band filters and in two narrow - band continuum filters , k@xmath51 and h@xmath51 ; a detailed characterization of all these filters can be found in ghinassi et al . ( @xcite ) . the observations were reduced and astrometrically calibrated following the procedure outlined above . the weather conditions were not photometric during the observations so the data could not be flux calibrated . line - only images were obtained by subtracting the narrow - band continuum images from the line@xmath52continuum images . the subtraction was not perfect on strong stellar sources or on stars with very red or very blue spectra . infrared reflection nebulosity is associated with two distinct regions of ionized gas ( fig . 1 ) : w75 n ( a ) at position @xmath31 @xmath53 and w75 n ( b ) at position @xmath31 @xmath53 ( e.g. haschick et al . 1981 , moore et al . 1988 ) . shock - excited emission ( figs . 2 & 3 ) is present in the north - east ( near @xmath31 @xmath53 ) and along the co emission boundaries ( see also davis et al . 1998a , b ) . our images also show that faint , patchy emission extends nearly an arcminute ( 0.6 pc at a distance of 2 kpc ) beyond the south - west co flow ( @xmath31 @xmath53 ) . the continuum subtracted h@xmath41 mosaic is shown in fig . 2 . following the nomenclature used by davis et al . ( 1998a , b ) for the south - west portion of the h@xmath41 flow , all previously known h@xmath41 knots and filaments are marked with solid lines . these are labeled sw a to sw h in the south - west flow , c a and c b in the central region , and ne a to ne f in the north - east flow . additionally , figs . 2 & 3 reveal faint diffuse h@xmath41 emission beyond the tip of the south - west flow and within the north - east flow . these new features are marked with dashed lines and labeled sw i to sw k and ne g to ne faint , diffuse h@xmath41 features as well as the filamentary structure of the emission in knots ne d , ne e , & ne f are confirmed by observations obtained in 2002 august . figure 3 presents an overlay of the continuum subtracted [ feii ] emission ( contours ) on the h@xmath41 ( greyscale ) images from the 2002 august observations . [ feii ] line emission is only detected close to the uc hii regions in w75 n b and near the exciting star of w75 n a. no [ feii ] emission is detected in the outer flow regions . the non - detection of [ feii ] far from the protostars does not appear to be due to higher extinction since we clearly detect 2.12@xmath50 m emission beyond the co outflow boundaries and we detect [ feii ] emission near the cloud core where the column density is higher . a chain of h@xmath41 knots , apparently unrelated to the main flow , are detected near @xmath31 @xmath54 ( fig . this jet is outside of the co and infrared mosaic fields . the knots may be associated with a jet from a young star north of the main complex discussed in this paper . a high - velocity co outflow , centered near w75 n ( b ) , measures 3 pc from end - to - end ( projected length ) and extends well beyond the infrared reflection nebula ( figs . 1 & 4 ) . red and blue - shifted co emission exists both in the north - east and south - west . the co mosaic did not include areas to the north - west and south - east so it is unclear if high - velocity co exists in these regions . the boundaries and flux density of the co outflow are well determined on the red - shifted side of the line , however , at velocities between 0 and 5.6 the dr 21 cloud ( @xmath55 ) confuses the identification of the outflow structure . nine millimeter continuum peaks showing the locations of warm dust emission are identified in fig . 5 & table 1 ( w75 n : mm 1 through w75 n : mm 9 ) . mm 1 mm 4 are near the origin of the outflow activity and lie @xmath56 from the w75 n ( b ) reflection nebulosity to the north and west . mm 5 is associated with the more extended hii region , w75 n ( a ) , while mm 6 mm 9 are not associated with any previously known sources . infrared counterparts do not exist for the millimeter sources ( except mm 5 ) suggesting that these sources are too deeply embedded to be detected at 2@xmath50 m . figure 5 also shows the @xmath0 resolution image from shepherd ( 2001 ) for comparison . the @xmath0 resolution resolved the individual millimeter cores mm 14 but resolved out the more extended emission associated with mm 5 and mm 6 . millimeter cores mm 79 were outside of the primary beam of the shepherd ( 2001 ) observations and thus , were not detected . figure 5 also compares the mm 5 millimeter source with narrow - band @xmath52continuum emission and k@xmath2 broad - band emission in w75 n ( a ) . the central star of w75 n ( a ) ( spectral type b0.5 ; haschick et al . 1981 ) is clearly visible in the infrared and is surrounded by a @xmath57 shell of thermal dust emission . diffuse reflection nebulosity is centered on the star and a wisp of emission is visible just east of the star ( @xmath31 @xmath53 ) . compact , high - velocity co emission appears to originate from the uc hii regions vla 1 ( ba ) & vla 3 ( bb ) and from mm 2 ( figs . 6 , 7 , & 8) . the @xmath58 resolution is not sufficient to determine if vla 2 , located only @xmath59 north of vla 3 ( bb ) , is also associated with high - velocity co gas . a detailed discussion of each of the proposed outflows is given below . * the outflow from vla 1 ( ba ) : * vla 1 ( ba ) is a thermal jet source associated with and oh masers ( baart et al . 1986 , torrelles et al . it is embedded in the mm 1 core detected in 1 & 3 mm continuum emission ( shepherd 2001 ) . the spectral type of the powering source is unknown since the observed centimeter continuum emission is likely due to the ionized jet rather than emission from an ionization - bounded uc hii region . red - shifted emission to the north - east of mm 1 can be traced to the jet - like ionized flow from vla 1 ( ba ) . figure 7 presents a @xmath27 image of the integrated co emission from the flow as well as a position - velocity diagram from a slice along the proposed flow axis ( p.a . a ridge of co emission extends to the north - east with projected velocity greater than 25 ( relative to @xmath5 ) almost an arcminute from the uc hii region . high - velocity co emission is also centered on the uc hii regions ( position offset @xmath60 in the pv diagram of fig . 7 ) . a @xmath0 resolution image ( fig . 6 , top left ) shows that this red - shifted emission near the base of the outflow appears to be produced by vla 1 ( ba ) as well as vla 3 ( bb ) and possibly vla 2 . the @xmath0 resolution of fig . 6 resolves out much of the extended emission in the flow leaving only compact clumps visible along the flow axis . the p.a . is similar to the elongation of the ionized emission in vla 1 ( ba ) ( p.a . @xmath61 ) and a line of masers detected along the jet ( torrelles et al . the molecular flow is also seen in fig . 1 as a well - collimated , red - shifted lobe extending to the north - east from the mm 1 core . the outflow is shown as one - sided in figs . 6 & 7 because only one side of the flow is detected . the counterflow may exist however it may be too extended to image at this high - resolution or it may be expanding into a less dense medium that would not create appreciable co emission . * the outflow from vla 3 ( bb ) : * vla 3 ( bb ) is a compact uc hii region in mm 1 with a central star of spectral type b0.5 to b0 . it is associated with a single maser and compact 1 & 3 mm continuum emission . a lower limit on the mass of warm gas and dust within 2000 au of the protostar is 5 m ( shepherd 2001 ) . figure 6 illustrates that compact red - shifted emission to the east and blue - shifted emission to the west of mm 1 can be traced to vla 3 ( bb ) with p.a . @xmath62 . there are no obvious features in the extended emission that correspond to an outflow with this orientation , however , the co mosaic did not extend to the north - west or south - east so a large - scale outflow could have been missed ( fig . 1 ) . the molecular gas morphology does not seem to be correlated with that of the ionized gas or masers near the source : the ionized gas is slightly elongated along p.a . 149and the maser is located near the southern boundary of the ionized gas . * the outflow from mm 2 : * mm 2 is a molecular core identified by compact , warm dust emission at 1 & 3 mm and maser emission ( torrelles et al 1997 , shepherd 2001 ) . the mass of the core is @xmath63 m. no ionized gas has been detected indicating that the spectral type is less than a b2 star or high accretion is preventing the formation of a uc hii region . high - velocity , blue - shifted emission ( 13 to 36 projected velocity relative to @xmath5 ) can be traced from the mm 2 core to the south - east ( p.a . figure 8 presents a @xmath27 image of the integrated emission from the flow as well as a position - velocity diagram from a slice along the proposed flow axis . the velocity of the jet relative to @xmath64 increases away from the position of mm 2 and remains well collimated . diffuse emission at velocities greater than @xmath65 is due to the w75 n & dr 21 clouds . figure 6 shows the more compact emission in the flow . three clumps of high - velocity gas are detected extending away from mm 2 along with faint emission at the location of the core . the compact clumps appear to trace a shell of dense gas surrounding the outflow axis . the molecular outflow is identified as one - sided because the red - shifted counterflow was not detected . the diffuse millimeter core mm 4 is located along the axis of the proposed mm 2 outflow . although no or oh maser emission has been detected toward mm 4 , nor has centimeter or 1 mm continuum emission been detected , there is diffuse warm dust emission traced by 3 mm continuum . thus , mm 4 may harbor an embedded protostar . assuming the mm 4 core is heated internally , the lack of maser activity , compact warm dust emission at 1 mm , and the absence of ionized gas emission at centimeter wavelengths suggest that mm 4 is a low - mass protostar ( shepherd 2001 ) . with the current resolution and sensitivity , we can not determine if the mm 4 protostar is contributing to the observed flow dynamics . the proposed position angles of the outflowing gas ( illustrated by arrows in figs . 6 , 7 , & 8) are 51 for vla 1 ( ba ) , 101 for vla 3 ( bb ) , and 124 for mm 2 . the position angle of the parsec - scale outflow detected in and co ( fig . 1 ) is 62.5 . although the orientation of the vla 1 ( ba ) outflow is similar to that of the parsec - scale flow , it does not appear likely that vla 1 ( ba ) is the powering source . assuming the vla 1 ( ba ) flow is symmetric , a blue - shifted counterflow is expected in the south - west , not a red - shifted flow . thus , our observations do not not identify the source responsible for the 3 pc outflow which dominates the large - scale morphology and kinematics of the region . the mass associated with co line emission is calculated following the method proposed by scoville et al . the co excitation temperature near the millimeter continuum emission varies from about 35 to 75 k with @xmath66 k being the median value near the mm 1 peak ( davis et al . rotational temperatures derived from ch@xmath67cn in the cloud core vary from 47 to 78 k , consistent with the davis et al . estimates ( kalenski et al . 2000 ) . we assume the gas is in lte , at a temperature of 50 k , with [ co]/ [ ] = @xmath68 , and [ co]/ [ ] = 71 at the galacto - centric distance of 8.5 kpc ( wilson & rood 1994 ) . the co optical depth as a function of velocity and position is calculated using single dish and spectra taken at three positions within the w75 n region ( fig . we assume is optically thin at all velocities which is probably valid in the line wings ; however , is likely to be optically thick near the line core . in channels where no emission is detected , we assume the co is optically thin . the co channel images ( fig . 4 ) show that the emission near @xmath5 is almost entirely resolved out by the interferometer . if high - velocity ( @xmath4 ) structures exist that are larger than the largest angular scale that can be imaged ( @xmath69 ) , then our mass estimate represents a lower limit . because multiple , overlapping flows are present , it is not possible to obtain a reasonable estimate of the inclination of each flow . thus , we assume an inclination angle of 45 which minimizes errors introduced by inclination effects . table 2 summarizes the physical properties of the molecular gas in the combined outflows originating within mm 1 and from the blue - shifted mm 2 outflow lobe . the total flow mass @xmath70 is given by @xmath71 where @xmath72 is the flow mass in velocity channel @xmath73 corrected for optical depth . the momentum @xmath74 is given by @xmath75 and the kinetic energy e by @xmath76 where @xmath77 is the central velocity of the channel relative to @xmath5 . the characteristic flow timescale @xmath78 is @xmath79 , where the intensity - weighted velocity @xmath80 is given by @xmath81 ( cabrit & bertout 1990 ) and @xmath82 is the flow radius . the mass outflow rate @xmath83 is @xmath84 and the force f is @xmath85 . assuming d=2 kpc , the total molecular mass in outflowing gas ( @xmath86 relative to @xmath5 ) is @xmath87 m. the values presented in table 2 are derived assuming all flows have the same systemic velocity . this is a reasonable assumption for the driving sources of the combined mm 1 outflows since the observed uc hii regions are embedded within the same molecular clump and have a projected separation of only 0.5 to @xmath88 ( 1000 - 2000 au at a distance of 2 kpc ) . the source driving the outflow from the mm 2 molecular core has a projected separation of about @xmath23 ( 10,000 au or 0.05 pc ) from mm 1 . it is possible that mm 2 could have a slightly different systemic velocity from mm 1 that we can not detect in our images or single dish spectra . if the systemic velocity of the mm 2 core is different by a factor of @xmath89 from the assumed velocity of 10 , then the error in the momentum estimates will be proportional to @xmath90 and mechanical energy to @xmath91 . the combined mm 1 outflows have a total mass of at least 165 m and energy , e @xmath92 ergs . the mm 2 outflow has a total mass @xmath93 m , e @xmath94 ergs . the co mosaic did not extend in the south - east direction of the mm 2 flow , thus , the full outflow was not imaged and age and @xmath95 should be considered a lower limit . despite the uncertainties , the flow masses and energies are consistent with those for outflows driven by young , early b stars . this is in agreement with the estimated spectral types of the stars powering the uc hii regions in mm 1 ( b2 to o9 ; hunter et al . 1994 ; torrelles et al . 1997 ; shepherd 2001 ; slysh et al . 2002 ) . the mass of gas and dust associated with warm dust being heated by the central protostars is estimated from the millimeter continuum emission using @xmath96 where d is the distance to the source , @xmath97 is the continuum flux density due to thermal dust emission at frequency @xmath98 , and @xmath99 is the planck function at temperature t@xmath100 ( hildebrand 1983 ) . assuming a gas - to - dust ratio of 100 , the dust opacity per gram of gas is taken to be @xmath101 cm@xmath44 g@xmath30 where @xmath103 is the opacity index ( see kramer et al . 1998 ; and the discussion in shepherd & watson 2002 ) . this value of @xmath104 agrees with those derived by hildebrand ( 1983 ) and kramer et al . ( 1998 ) to within a factor of 2 . the opacity index @xmath105 appears to be appropriate between wavelengths of 650 microns and 2.7 mm for sub - micron to millimeter - sized grains expected in warm molecular clouds and young disks ( pollack et al . we assume the emission is optically thin and the temperature of the dust can be characterized by a single value . using values of @xmath106 k and @xmath105 , we find the total mass of gas and dust associated with the 2.7 mm continuum emission is approximately 475 m ( table 1 ) . our results are consistent with those of shepherd ( 2001 ) and watson et al . ( 2002 ) to within the errors . the total molecular mass of outflowing gas from the mm 1 and mm 2 combined flows ( @xmath86 relative to @xmath5 ) is @xmath3 m. hunter et al . ( 1994 ) found @xmath107 with a rough scaling performed to take into account an optical depth correction . however , their image covered only the inner region of the flow so their estimate should be considered a lower limit . based on single dish observations of co(j=32 ) , davis et al ( 1998a ) estimated a total flow mass of @xmath108 , uncorrected for optical depth effects . this mass estimate is extremely high for an optically thin approximation . examination of their fig . 11 , t@xmath109 vs. ( @xmath110 ) , shows that the blue - shifted lobe has an order of magnitude increase in the integrated flux at the velocity of dr 21 ( 2.5 ) . it appears that their single - dish map may have been significantly contaminated by emission from the dr 21 cloud , which introduced uncertainties in the mass and kinematics estimates . ridge & moore ( 2001 ) estimated the outflow mass of the red - shifted lobe only to be 273 based on a co(j=21 ) single dish image corrected for optical depth . the mass of blue - shifted gas was not estimated by ridge & moore due to the contamination by the dr 21 cloud . this value is significantly higher than our estimate and may be due to missing extended emission in the interferometer image , especially at low velocities . despite this problem , interferometric imaging also provided benefits : it was easier to distinguish between outflow gas and the dr 21 cloud and to identify flows from multiple sources in the cluster . the total cloud core mass of w75 n has been estimated to be 18002500 based on observations at submillimeter wavelengths ( moore , mountain , & yamashita , 1991 ) and 1200 based on cs(j=76 ) emission ( hunter et al . 1994 ) . the gravitational binding energy of the cloud , @xmath111 , is 12@xmath112 ergs , where we take the radius @xmath113 pc ( moore et al . 1991 ) and @xmath114 is a constant which depends on the mass distribution ( @xmath115 for @xmath116 ) . more than 10% of the molecular cloud is participating in the outflow and the combined outflow energy is roughly half the gravitation binding energy of the cloud . the observed w75 n outflows are injecting a significant amount of mechanical energy into the cloud core and may help prevent further collapse of the cloud . our co(j=10 ) images suggest that high - velocity gas is associated with at least two uc hii regions : vla 1 ( ba ) and vla 3 ( bb ) and an embedded source in the millimeter core mm 2 . the position angles of the individuals outflows are not aligned , ranging from 51 to 124 . the morphology is diffuse and patchy both in the north - east and south - west . the irregular morphology of the infrared reflection nebula with fingers of nebulosity radiating out from the mm 1/mm 2 millimeter cores supports the conclusion that multiple energetic outflows are carving large cavities in the molecular cloud . low surface brightness emission extends well beyond the co outflow while [ feii ] emission is only detected close to the protostellar cluster . it is generally believed that [ feii ] line emission associated with low - mass outflows requires the presence of fast , dissociative shocks that disrupt dust grains and release heavy elements just behind a jump - shock ( j - shock ) boundary . emission , on the other hand , appears to be produced in slow , non - dissociative j - type shocks ( e.g. hollenbach & mckee 1989 ; smith 1994 ; gredel 1994 ; beck - winchatz et al . continuous - shocks ( c - shocks ) can not easily produce emission from ionized species such as [ feii ] nor can they produce the observed column densities typically seen in toward herbig haro objects from low - mass protostars ( gredel 1994 ) . in fact , nisini et al . ( 2002 ) find that there appears to be no correlation between and [ feii ] emission in outflows from low - mass ysos which supports the interpretation that physically different mechanisms are responsible for producing and [ feii ] emission . in a sample of herbig haro objects produced by jets from low - mass protostars , both and [ feii ] is found toward all sources and the morphology of and [ feii ] emission is similar on large scales although it differs in the detail ( gredel 1994 ; reipurth et al . these observations indicate that jets from low - mass protostars produce both fast , dissociative regions where [ feii ] emissions arises and slower , non - dissociative regions where emission arises . in contrast , [ feii ] emission toward w75 n is only detected close to the central sources and does not show a jet - like morphology as in outflows from low- and intermediate - mass young stellar objects ( e.g. lorenzetti et al . 2002 ; nisini et al . 2002 ; reipurth et al . the outflows from w75 n appear to exhibit only slow , non - dissociative j - type shocks which produce copious emission throughout the outflow region but the fast , dissociative shocks responsible for [ feii ] emission are absent in the outer regions of the flow . instead , the diffuse [ feii ] line emission in w75 n is coincident with the brightest k@xmath2-band reflection nebulosity . one possibility may be that the [ feii ] emission traces photo - dissociation regions ( pdrs ) along cloud surfaces illuminated by the massive protostars in the mm 1 core . this situation is also observed in the orion bar pdr ( e.g. walmsley et al . 2000 ) suggesting that the w75 n nebula may exhibit similar excitation conditions to those in orion . the and [ feii ] line emission in w75 n does not conform to what is expected for shock - excited emission resulting from the interaction between a well - collimated jet and diffuse molecular gas . in this respect , the physical characteristics of the w75 n flows differ from their low - mass counterparts which produce collimated jets observed in both and [ feii ] emission . many previous authors have assumed that only vla 1 ( ba ) was in an outflow stage based on the elongated morphology of the ionized gas , the presence of maser emission along the uc hii region axis , and because the position angles of the ionized gas and the co emission were similar . maser emission is also associated with vla 2 and vla 3 ( bb ) as well as mm 3 and mm 2 , however , outflowing material could not be traced to specific sources . assuming a single primary driving source for the co gas , davis et al . ( 1998b ) suggested that the south - west co red - shifted lobe and morphology supports a bow - shock entrainment scenario for a molecular outflow driven by a jet from a single massive star . our co and millimeter continuum observations do not support this theory that a single source drives the high - velocity co gas . further , our infrared observations suggest that the w75 n outflows are not likely to be scaled - up versions of jet - driven outflows from low - mass protostars . a question remains unanswered by this work : what source powers the 3 pc flow at p.a . the flow mass is @xmath117 m , the dynamical age is roughly @xmath118 years , and the mass loss rate @xmath119 to @xmath120 m@xmath121 yr@xmath30 . the flow parameters are consistent with those produced by an early b protostar . hutawarakorn et al . ( 2002 ) have modeled the oh maser position - velocity data and find evidence for a massive disk centered on vla 2 ( m@xmath122 m with p.a . = 155 , roughly perpendicular to the outflow axis ) . a high - velocity , time - variable oh maser cluster is coincident with vla 2 suggesting an outflow origin . further , recent observations with the very long baseline array ( vlba ) show that a clump of strong maser emission with high velocity dispersion is centered on vla 2 ( torrelles et al . 2003 ) . thus , the oh and maser activity suggests vla 2 is producing a powerful outflow . although our observations did not have adequate resolution to isolate high - velocity gas toward vla 2 , we have determined that vla 1 ( ba ) , vla 3 ( bb ) , and mm 2 are not likely to drive the 3 pc flow that dominates the region dynamics . thus , it is possible that vla 2 may drive the large - scale flow . follow - up observations at a resolution less than @xmath88 will be required to determine if , in fact , vla 2 drives the 3 pc flow . based on the size and velocity of the co outflows from the w75 n ( b ) uc hii regions , the region is @xmath123 years old . w75 n ( a ) is more evolved than the sources in mm 1 and the exciting star of w75 n ( a ) has no detectable high - velocity gas associated with it . the star , detected in the infrared , is centered within a shell of warm dust emission and an extended hii region ( hashick et al . 1981 ) . figure 10 shows a color - color diagram using data from the 2mass point source catalog for stars within @xmath124 of mm 1 that were detected at all three bands . the locus of main - sequence stars is represented by the thick , curved line ( bessell & brett 1988 , koornneef 1983 ) while the two diagonal lines show reddening vectors up to @xmath125 of dust ( adopting the @xmath126 extinction law from cardelli , clayton , & mathis 1989 ) . sources within the reddening vectors have colors consistent with main - sequence stars reddened by foreground dust . those to the right of the reddening vectors demonstrate excess emission at @xmath127 m , consistent with the presence of circumstellar material . the infrared colors of the w75 n ( a ) exciting star are consistent with those of a main - sequence star reddened by foreground dust . in comparison , the two bright stars to the south - east and south - west of mm 1 ( irs 2 and irs 3 ) have excess emission at @xmath127 m consistent with the presence of circumstellar material . the protostars within mm 1 and mm 2 are not detectable at infrared wavelengths . n represents a region of clustered star formation which appears to be forming mid to early - b stars which exist at a range of developmental stages . vla 1 ( ba ) appears to have a well - collimated , outflow based on the ionized gas morphology imaged by torrelles et al . ( 1997 ) and the presence of a relatively well - collimated , red - shifted co lobe which extends about 0.5 pc north - east of the source . however , the spectral type of the protostar is unknown since the ionized gas appears to be due to thermal jet emission . the well - collimated outflow which appears to be produced by the embedded source in mm 2 is not detected at centimeter wavelengths . either the powering source is not an early - b star ( e.g. it does not have sufficient ionizing radiation to produce a detectable uc hii region ) or accretion onto the protostar is sufficiently high that it prevents the formation of a uc hii region ( see e.g. churchwell 1999 and references therein ) . there is no evidence for well - collimated flows ( collimation ratios , length / width , @xmath128 ) from the remaining embedded sources ( early - b protostars ) or in the large - scale morphology of the co , , & [ feii ] emission . the lack of highly - collimated flows from the known , early - b protostars in w75 n suggests that it may be difficult for massive stars to collimate outflowing material . although a few mid to early - b protostars appear to be powering ionized jets , their molecular outflows tend to be complex and poorly collimated ( see , e.g. , the review by shepherd 2002 and references therein ) . to our knowledge , there is no well - collimated _ molecular _ outflow powered by a massive protostar ( spectral type early - b to o ) and most do not appear to have ionized outflow components that are well - collimated ( e.g. ridge & moore 2001 , shepherd , claussen , & kurtz 2001 ; churchwell 1999 ) . poorly collimated flows could be due to several factors : * confusion from multiple outflow sources in a cluster ( e.g. w75 n : this work ; or dr 21 : garden et al . 1991 ) ; * large flow precession angles ( e.g. pv ceph : reipurth , bally , & devine 1997 , gomez , kenyon , & whitney 1997 ; or iras 20126@xmath524104 : shepherd et al . 2000 ) ; * the presence of a strong wide - angle wind ( e.g. orion i : greenhill et al . 1998 ; or g192.163.82 : shepherd , claussen & kurtz 2001 ) ; and/or * the molecular flow represents only the truncated base of a much larger flow ( e.g. hh 8081 : yamashita et al . 1989 , rodrguez et al . 1994 ; or g192.163.82 : devine et al . 1999 ) . our observations of w75 n supports the interpretation that massive protostars may not be able to produce well - collimated molecular outflows . this conclusion does not rule out the possibility that an underlying , neutral jet may still exist as part of the outflow from the massive protostars in w75 n. high - resolution observations in shock tracers such as sio(j=10,v=0 ) or sio(j=21,v=0 ) may be able to determine whether collimated neutral jets are present in w75 n. w75 n represents an example of clustered , massive star formation . the cluster covers a wide range of evolutionary stages ; from stars with no apparent circumstellar material to deeply embedded protostars actively powering massive outflows . the co outflow measures more than 3 pc from end - to - end and is produced by at least four individual sources . emission extends well - beyond the co boundaries while [ feii ] emission is only located close to the protostellar cluster . the co , & [ feii ] morphology does not conform to what is expected for shock - excited emission resulting from the interaction between a well - collimated jet and diffuse molecular gas . the irregular morphology of the infrared reflection nebula with fingers of nebulosity radiating out from the millimeter cores supports the conclusion that multiple energetic outflows are carving large cavities in the molecular cloud . more than 10% of the molecular cloud is outflowing material and the combined outflow energy is roughly half the gravitational binding energy of the cloud . thus , the observed w75 n outflows are injecting a significant amount of mechanical energy into the cloud core and may help prevent further collapse of the cloud . research at the owens valley radio observatory is supported by the national science foundation through nsf grant number ast 99 - 81546 . star formation research at owens valley is also supported by nasa s origins of solar systems program , grant nagw-4030 , and by the norris planetary origins project . this paper is partly based on observations made with the italian telescopio nazionale galileo ( tng ) operated on the island of la palma by the centro galileo galilei of the inaf ( istituto nazionale di astrofisica ) at the spanish observatorio del roque de los muchachos of the instituto de astrofisica de canarias . the arnica and nics observations were performed in service mode by the tng staff , we especially acknowledge the help of francesca ghinassi , juan carlos guerra , and antonio magazz . 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 . baart , e.e . , cohen , r.j . , davies , r.d . , norris , r.p . , rowland , p.r . 1986 , mnras , 219 , 145 baffa , c. , comoretto , g. , gennari , s. , lisi , f. , oliva , e. , biliotti , v. , checcucci , a. , gavrioussev , v. , giani , e. , ghinassi , f. , hunt , l. k. , maiolino , r. , mannucci , f. , marcucci , g. , sozzi , m. , stefanini , p. , & testi , l. 2001 , a&a , 378 , 722 beck - winchatz , b. , bhm , k .- h . , & noriega - crespo , a. 1996 , aj , 111 , 346 bessell , m. s. , & brett , j. m. 1988 , pasp , 100 , 1134 cabrit , s. & bertout , c. 1990 , apj , 348 , 530 cabrit , s. , ferreira , j. , & raga , a. c. 1999 , a&a , 343 , l61 cardelli , j. a. , clayton , g. c. , & mathis , j. s. 1989 , apj , 345 , 245 churchwell , 1999 , in `` the origin of stars and planetary systems '' , ed . charles j. lada and nikolaos d. kylafis ( kluwer academic publishers ) , p. 515 cornwell , t. & braun , r. 1988 , in `` synthesis imaging in radio astronomy , '' eds . r. a. perley , f. r. schwab & a. h. bridle , pasp , p 167 davis , c. j. , moriarty - schieven , g. h. , eislffel , j. , hoare , m. g. , & ray , t. p. 1998a , aj , 115 , 1118 davis , c.j . , smith , m.d . , & moriarty - schieven , g.h . 1998b , mnras , 299 , 825 devine , d. , bally , j. , reipurth , b. , shepherd , d. , watson , a. 1999 , aj , 117 , 2919 dickel , h. r. , wendker , h. j. , & bieritz , j. h. 1969 , a&a , 1 , 270 fischer , j. , sanders , d. b. , simon , m. , & solomon , p. m. 1985 , apj , 293 , 508 garden , r. p. , hayashi , m. , gatley , i. , hasegawa , t. , & kaifu , n. 1991 , apj , 374 , 540 ghinassi , f. , licandro , j. , oliva , e. , baffa , c. , checcucci , a. , comoretto , g. , gennari , s. , & marcucci , g. 2002 , a&a , 386 , 1157 gomez , m. , kenyon , s. j. , & whitney , b. a. 1997 , aj , 114 , 265 gredel , r. 1994 , a&a , 292 , 580 greenhill , l. j. , gwinn , c. r. , schwartz , c. , moran , j. m. , & diamond , p. j. 1998 , nature , 396 , 650 haschick , a. d. , reid , m. j. , burke , b. f. , moran , j. m. , & miller , g. 1981 , apj , 244 , 76 hildebrand , r. h. 1983 , q.jl.r.astr.soc , 24 , 267 hollenbach , d. & mckee , c. f. 1989 , apj , 342 , 306 hunt , l.k . , lisi , f. , testi , l. , baffa , c. , borelli , s. , maiolino , r. , moriondo , g. , & stanga , r. m. 1996 , a&as , 115 , 181 hunt , l.k . , mannucci , f. , testi , l. , migliorini , s. , stanga , r.m . , baffa , c. , lisi , f. , & vanzi , l. 1998 , aj , 115 , 2594 hunter , t.r . , taylor , g.b . , felli , m. , & tofani , g. 1994 , a&a , 284 , 215 hutawarakorn , b. , cohen , r. j. , & brebner , g. c. 2002 , mnras , 330 , 349 kalenski , s. v. , promislov , v. g. , alakoz , a. v. , winnberg , a. , & johansson , l. e. b. 2000 , astronomy reports , 44 , 725 knigl , a. 1999 , new astronomy reviews , 43 , 67 knigl , a. & pudritz , r. e. 2000 , in `` protostars and planets iv '' , ed . v. mannings , a. boss & s. russell ( tucson : univ . of arizona press ) , in press koornneef , j. 1983 , a&a , 128 , 84 kramer , c. , alves , j. , lada , c. , lada , e. , sievers , a. , ungerechts , h. , & walmsley , m. 1998 , a&a , 329 , l33 lada , c. j. & fich , m. 1996 , apj , 459 , 638 lisi , f. , hunt , l.k . , baffa , c. , biliotti , v. , bonaccini , d. , del vecchio , c. , gennari , s. , hunt , l. k. , marcucci , g. , & stanga , r. 1996 , pasp , 108 , 364 lorenzetti , d. , giannini , t. , vitali , f. , massi , f. , & nisini , b. 2002 , apj , 564 , 839 minier , v. , conway , j. e. , & booth , r. s. 2000 , a&a , 362 , 1093 minier , v. , conway , j. e. , & booth , r. s. 2001 , a&a , 369 , 278 moore , t. j. t. , mountain , c. m. , yamashita , t. , & mclean , i. s. 1991 , mnras , 248 , 377 moore , t. j. t. , mountain , c. m. , & yamashita , t. 1991 , mnras , 248 , 79 moore , t. j. t. , mountain , c. m. , yamashita , t. , & selby , m. j. 1988 , mnras , 234 , 95 nisini , b , caratti o garatti , a. , giannini , t. & lonenzetti , d. 2002 , a&a , accepted pollack , j. b. , hollenbach , d. , beckwith , s. , simonelli , d. p. , roush , t. , & fong , w. 1994 , apj , 421 , 615 reipurth , b. , bally , j. , & devine , d. 1997 , aj , 114 , 2708 reipurth b. , yu , k. c. , heathcote , s. , bally , j. , & rodrguez , l. f. 2000 , aj , 120 , 1449 richer , j. s. , shepherd , d. s. , cabrit , s. , bachiller , r. , & churchwell , e. 2000 , in `` protostars and planets iv '' , ed . v. mannings , a. boss & s. russell ( tucson : univ . of arizona press ) , 867 ridge , n. a. & moore , t. j. t. 2001 , a&a , 378 , 495 rodrguez , l. f. , garay , g. , curiel , s. , ramirez , s. , torrelles , j. m. , gomez , y. , & velazquez , a. 1994 , apj , 430 , l65 sault , r. j. , teuben , p. j. , & wright , m. c. h. 1995 , in astronomical data analysis software and systems iv , ed . r. a. shaw , h. e. payne , & j. j. e. hayes , pasp conf series 77 , 433 scoville , n. z. , sargent , a. i. , sanders , d. b. , claussen , m. j. , masson , c. r. , lo , k. y. , & phillips , t. g. 1986 , apj , 303 , 416 scoville , n. z. , carlstrom , j. e. , chandler , c. j. , phillips , j. a. , scott , s. l. , tilanus , r. p. j. , & wang , z. 1993 , pasp , 105 , 1482 shang , h. , glassgold , a. e. , shu , f. h. , lizano , s. 2002 , apj , 564 , 853 shepherd , d. s. , yu , k. c. , bally , j. , & testi , l. 2000 , apj , 535 , 833 shepherd , d. s. , claussen , m. j. , & kurtz , s. e. 2001 , science , 292 , 1513 shepherd , d. s. 2001 , apj , 546 , 345 shepherd , d. s. & watson , a. m. 2002 , apj , 566 shepherd , d. s. 2002 , in `` galactic star formation across the stellar mass spectrum , '' asp conference series , in press shu , f. h. , najita , j. r. , shang , h. , & li , z .- y . 2000 , in protostars and planets iv , ed . v. mannings , a. boss & s. russell ( tucson : univ . of arizona press ) , 789 slysh , v. i. , migenes , v. , valtts , i. e. , lyubchenko , s. yu . , horiuchi , s. , altunin , v. i. , fomalont , e. b. , & inoue , m. 2002 , apj , 564 , 317 smith , m. d. 1994 , a&a , 289 , 256 smith , m. d. , suttner , g. , & yorke , h. w. 1997 , a&a , 323 suttner , g. , smith , m. d. , yorke , h. w. , & zinnecker , h. 1997 , a&a , 318 torrelles , j.m . , gomez , j.f . , rodrguez , l.f . , ho , p.t.p . , curiel , s. , & vazquez , r. 1997 , apj , 489 , 744 torrelles , j.m . , et al . 2003 , in preparation vanzi , l. , gennari , s. , ciofini , m. , & testi , l. 1998 , exp . astronomy , 8 , 177 walmsley , c. m. , natta , a. , oliva , e. , & testi , l. 2000 , a&a , 364 , 301 watson , c. , churchwell , e. , & pankoni , v. , 2002 , apj , in press wilson , t. l. , & rood , r. t. 1994 , araa , 32 , 191 yamashita , t. , suzuki , h. , kaifu , n. , tamura , m. , mountain , c. m. , moore , t. j. t. 1989 , apj , 347 , 894 @xmath129 mm 1 eastern outflow emission measured at velocities 2.2 to 4.8 and 15.2 to 36 . + mm 1 western outflow emission measured at velocities 8.2 to 4.8 and 15.2 to 36.0 . + mm 2 outflow emission measured between 26.4 and 2.2 . * figure 1 . * integrated co red - shifted ( red lines ) and blue - shifted ( blue lines ) emission contours from 36.0 to 17.8 and 0.4 to 26.4 , respectively . the images have an rms of 23.5 with a peak of 56.0 in the red - shifted emission image and 100.6 in the blue - shifted emission image . contours begin at 10% of the peak emission and continue at increments of 20% . the synthesized beam is @xmath24 at p.a . @xmath130 . line @xmath52 continuum emission is shown as grey scale displayed as the square root of the intensity with a peak of @xmath131 erg @xmath132s@xmath30arcsec@xmath133 . w75 n ( a ) is located at position @xmath31 @xmath53 and w75 n ( b ) at position @xmath31 @xmath53 . uc hii regions embedded in the core of mm 1 are shown as filled triangles while the millimeter cores mm 29 are shown as open circles . the large open circle ( mm 5 ) is coincident with the infrared emission associated with w75 n a. the solid black line delineates the boundaries of the co mosaic . * figure 2 . * continuum - subtracted line mosaic of w75 n. the image is displayed on a linear scale from 0.6 to a peak intensity of @xmath134 erg @xmath132 s@xmath30 . individual knots of emission are labeled ne a through i , c a & b , and sw a through k. features observed by davis et al . ( 1998a , b ) are identified by solid lines , previously undetected features which are more diffuse are shown as dashed lines . large dashed boxes outline the fields shown in fig . figure 3 . * continuum - subtracted emission shown in grey scale with [ feii ] emission shown as contours . the [ feii ] contours begin at @xmath135 and continue with a spacing of @xmath136 . the left panel shows the central and north - east outflow regions . the strong , diffuse [ feii ] emission is coincident with the w75 n ( a ) and ( b ) reflection nebulae . the right panel shows the south - west outflow . [ feii ] contours coincident with point sources are due to imperfect continuum subtraction . * figures 4a & 4b . * co channel images at 2.6 spectral resolution between 32.1 and @xmath137 . the central velocity is indicated in the upper right of each panel . the lsr velocity is 10 . the rms is 0.13 and the peak emission is 16.9 . in the first 12 and last 12 panels , contours are plotted from @xmath138 , 8 , 12 , 16 , 20 @xmath139 and continue with a spacing of @xmath140 . in the central 6 panels ( 8.7 to 4.3 ) , contours begin at @xmath141 and continue with a spacing of @xmath142 . panels at velocities 32.1 & 27.7 show the synthesized beam in the bottom right corner ( @xmath24 at p.a . @xmath130 ) and a scale size of 0.9 pc . the plus symbol in each panel represents the location of the peak emission in mm 1 . no other emission was detected outside of this velocity range . * figure 5 . * the bottom left image shows continuum emission at 2.7 mm . no other continuum sources were detected within the mosaic field . the image has an rms of 3.6 . contours begin at @xmath143 and continue with a spacing of @xmath144 . the greyscale is plotted on a linear scale from 7.2 to 265 . the synthesized beam in the lower right corner is @xmath28 at p.a . uc hii regions in the center of mm 1 are identified by filled triangles , mm 24 are shown as open circles . a scale size of 0.15 pc is represented by a bar in the lower left corner . the bottom right inset shows 3.3 mm continuum emission obtained with @xmath145 resolution ( figure 1 from shepherd 2001 ) . upper panels show the mm 5 millimeter source compared with narrow - band @xmath52continuum emission and wide - band 2.12@xmath50 m emission in w75 n ( a ) ( from fig . * figure 6 . * the relationship between compact , high velocity co emission , infrared emission , millimeter continuum peaks , and uc hii regions in w75 n ( b ) . red- and blue - shifted co emission ( upper panels ) is plotted from 20.4 to 36.0 and 5.6 to 23.8 , respectively . the rms in both images is 0.5 ; contours are plotted from 3 , 2 , 3 , 4 @xmath139 and continue at spacings of 1 @xmath139 . millimeter core positions for mm 24 are shown as filled circles , uc hii regions embedded in mm 1 are represented as filled triangles . proposed outflows are identified by arrows from uc hii regions vla 1 ( ba ) , vla 3 ( bb ) , and mm 2 . the synthesized beam ( @xmath146 at p.a . @xmath147 ) is shown in the top right image and a scale size of 0.1 pc is represented by a bar in the top left image . the bottom two images show the k@xmath2 and emission relative to the proposed outflows . both images are displayed as the square root of the intensity . * figure 7 . * * top : * integrated emission ( zeroth moment ) from 17.8 to 36.0 . the rms in the image is 1.2 , contours are plotted from 4 to @xmath142 with increments of @xmath148 and then from 20 to @xmath149 with increments of @xmath135 . greyscale is plotted from @xmath150 to a peak of 56.0 . millimeter cores are identified as filled circles , uc hii regions by filled triangles . the synthesized beam in the bottom left corner is @xmath151 at p.a . @xmath130 . * bottom * position - velocity plot along the length of the high - velocity , red - shifted outflow . contours are plotted at 2 , 5 , 10 , 15 , 20 , 30 , 40 , 50 , 70 , & 90% of the peak . * figure 8 . * * top : * integrated emission ( zeroth moment ) from 5.6 to 26.4 . the rms in the image is 0.96 , contours are plotted from 4 to @xmath142 with increments of @xmath148 , greyscale is plotted from @xmath150 to a peak of 20.0 . millimeter cores are identified as filled circles , uc hii regions by filled triangles . the synthesized beam in the bottom right corner is @xmath151 at p.a . @xmath130 . * bottom * position - velocity plot along the length of the high - velocity , blue - shifted outflow . contours are plotted at 3 , 5 , 7 , 9 , 20 , 40 , 60 , 80 , & 100% of the peak . * * co(j=10 ) optical depth as a function of velocity based on single - dish observations with the kitt peak 12 m telescope . optical depth is derived for three positions within the outflow : in the north - east lobe ( top ) ; centered on the mm 1 core ( center ) ; and in the south - west lobe ( bottom ) . co emission was measured in each channel image of the interferometer mosaic and an optical depth correction was made to the mass estimate based on the location and velocity of the emission . * figure 10 . * * left : * three color image using data from the two micron all sky survey ( 2mass ) . the image is @xmath152 on a side . the j - band image at 1.25@xmath50 m is shown as blue , h - band at 1.65@xmath50 m is green , & k@xmath2-band at 2.17@xmath50 m is red . the @xmath52 symbol represents the location of the mm 1 peak . * right : * a color - color diagram using data from the 2mass point source catalog for stars within @xmath124 of mm 1 that were detected at all three bands . the locus of main - sequence stars is represented by the thick , curved line while the two diagonal lines show reddening vectors up to @xmath125 of dust .

a new method for solving the time - dependent two - center dirac equation is developed . the time - dependent dirac wave function is represented as a sum of atomic - like dirac - sturm orbitals , localized at the ions . the atomic orbitals are obtained by solving numerically the finite - difference one - center dirac and dirac - sturm equations with the potential which is the sum of the exact reference - nucleus potential and a monopole - approximation potential from the other nucleus . an original procedure to calculate the two - center integrals with these orbitals is proposed . the approach is tested by calculations of the charge transfer and ionization cross sections for the h(1s)proton collisions at proton energies from 1 kev to 100 kev . the obtained results are compared with related experimental and other theoretical data . to investigate the role of the relativistic effects , the charge transfer cross sections for the ne()ne ( at energies from to mev / u ) and u(1s)u ( at energies from 6 to 10 mev / u ) collisions are calculated in both relativistic and nonrelativistic cases .

1004.5131

since the pioneering works @xcite , where the oscillatory behavior of the resonance charge - transfer probability for low - energy collisions was predicted , numerous publications have been devoted to the theoretical investigations of the charge - transfer , excitations and ionization in the h(1s)-h@xmath7 collisions ( see , e.g. , reviews @xcite ) . nonrelativistic two - center finite basis set calculations have been carried out in refs . nonrelativistic three - dimensional lattice methods in the position and momentum spaces have been applied for the time - dependent schrdinger equation in refs . @xcite . within the nonrelativistic approach , the probabilities and cross - sections for a homonuclear collision a@xmath8a@xmath9 for the nuclear charge @xmath10 can be easily obtained by scaling to the h(@xmath1)@xmath11h@xmath7 collision . in the straight - line trajectory approximation , the cross section @xmath12 scales exactly as @xmath13 @xcite , where @xmath14 is the projectile velocity . this scaling law is not valid , however , in the relativistic theory . collisions involving highly charged ions provide tests of relativistic and quantum electrodynamics effects in the scattering theory @xcite . the study of such processes can provide also a unique tool to probe the quantum electrodynamics ( qed ) in the supercritical coulomb field , if the total charge of the colliding ions @xmath15 is larger than the critical one @xmath16 ( see , e.g. , refs . @xcite and references therein ) . in the presence of such a field the energy of the one - electron @xmath17 state of the quasi - molecule can reach the negative - energy dirac continuum , when the distance @xmath18 between target ion @xmath19 and projectile ion @xmath20 becomes equal to the critical value @xmath21 . for the distance @xmath18 less than @xmath21 the ground state level dives into the negative - continuum spectrum . in the u@xmath5(1s)-u@xmath6 collision the critical radius for the point nucleus case was found to be @xmath22 fm @xcite . to date various approaches were developed to treat the heavy - ion collisions @xcite . in refs . @xcite , the two- and three - dimensional numerical lattice methods were employed to solve the time - dependent dirac equation at high energies . in refs . @xcite , high energy relativistic collisions of heavy ions were considered using the basis set approach , in which the time - dependent wave function was expanded in terms of the atomic eigenstates of the projectile and the target . for internuclear distances smaller than about 1000 fm some effects can also be evaluated within so - called monopole approximation , which accounts only for the spherically - symmetric part of the two - center potential @xcite . the atomic processes such as excitation , ionization and charge transfer in relativistic atomic collisions involving heavy and highly - charged projectile ions with energies ranging from 100 mev / u upward were studied in refs . @xcite and references therein . in the present work , we develop a new method for solving the two - center stationary and time - dependent dirac equations . the wave functions are expanded in terms of the dirac and dirac - sturm basis functions , which are central - field 4-component dirac bispinors centered at the ions . the radial parts of these orbitals are obtained by solving numerically the finite - difference radial one - center dirac and dirac - sturm equations . in the nonrelativistic calculations of atoms and molecules , so - called coulomb - sturmian basis set was introduced in ref . the hartree - fock calculations of atoms with this basis were considered by many authors ( see , e.g. , ref . the relativistic coulomb - sturmian basis was employed in the papers @xcite . in the present paper we use a non - coulomb relativistic sturm basis set , which is obtained by solving numerically the dirac - sturm equations with a special choice of the weight function , that was proposed in refs . this allows us to include any central - field potential in the radial equations for the large and small components of the basis functions . in particular , the coulomb potential of the other ion can be included in the radial equations within the monopole approximation . the basis set constructed in this way is described in detail in section [ subsec : basis ] . calculations of two - center integrals with the basis functions obtained require using special tools . in the nonrelativistic case , a special symmetrical procedure for such calculations , based on the lwdin reexpansion @xcite , was developed in refs . @xcite . in section [ subsec : integrals ] , we generalize this procedure to the relativistic case . to test the quality of the two - center expansion described above we perform relativistic calculations of the ground - state energy of molecular ion h@xmath23 and one - electron quasi - molecule th@xmath24 at the `` chemical '' distance @xmath25 a.u . and compare the results with high - precision calculations of refs . we also calculate the ground - state energy as a function of the internuclear distance @xmath18 and the critical radii @xmath21 for a number of one - electron quasi - molecules , including u@xmath26 . most calculations of the critical distances @xmath27 presented in the literature were performed either for the point - nucleus model @xcite or with a crude estimate of the nuclear - size effect @xcite . we calculate the critical distances for both point and extended nucleus models using the same basis set expansion . the obtained results and comparison with the calculations by other authors are presented in section [ subsec : stationary ] . the classical rutherford trajectories @xcite of the projectile and target ions are obtained by numerical solution of the newton s equations . the born - oppenheimer approximation is used to separate the motion of the electron and the nuclei . the magnetic interaction between the electron and the moving ions is neglected , because of low velocity of the projectile with respect to the target . the time - dependent dirac equation for the electron is solved using the two - center basis set expansion . the expansion coefficients can be defined employing , e.g. , the crank - nicholsen propagation scheme @xcite or the split - operator method @xcite . these methods conserve the norm of the time - dependent wave function at each time step , since the crank - nicholsen operator and the split - operator are unitary . however , in this work we use the direct evolution exponential operator method , which is more stable compared to the others . to obtain the matrix of the exponential operator in the finite basis set one has to diagonalize the generalized hamiltonian matrix at each step of time . since our basis set is not too large , the diagonalization procedure is not too time consuming . the amplitudes of the charge transfer to different bound states of the projectile ion are calculated nonperturbatively by projecting the time - dependent wave function onto the moving dirac orbitals of the projectile . in section [ subsec : charge - trnasfer ] we present the results of the relativistic calculations of the charge - transfer probabilities and cross sections for the h(1s)h@xmath7 , ne@xmath0(@xmath1)ne@xmath2 , xe@xmath28(@xmath1)xe@xmath29 , and u@xmath5(@xmath1)u@xmath6 low - energy collisions . all the calculations are performed in the laboratory frame @xmath30 , that is defined to be at rest with respect to the initial target position . the h(@xmath1)h@xmath7 collision is considered in section [ subsec : h2 ] . since the relativistic effects in this collision are negligible , the results of our calculations can be compared with nonrelativistic data obtained by other authors ( section [ subsec : h2 ] ) . the role of the relativistic effects is investigated in sections [ subsec : ne2 ] , [ subsec : xe2 ] , and [ subsec : u2 ] , where the relativistic and nonrelativistic calculations of the charge - transfer probabilities and cross sections are performed for higher-@xmath31 ions . within the born - oppenheimer approximation , the motion of the electron is considered as a motion in the field of the two nuclei being at given positions ( the stationary case ) or moving along the classical trajectories ( the non - stationary case ) . let @xmath32 and @xmath33 are the positions of the target ( @xmath19 ) and projectile ( @xmath20 ) nuclei , respectively . the time - dependent @xmath34 and stationary @xmath35 wave functions are the solutions of the time - dependent and stationary dirac equations , respectively . in the atomic units ( @xmath36 ) , these equations are given by @xmath37 here @xmath38 is the energy of the stationary state and the @xmath39 is the two - center dirac hamiltonian defined by @xmath40 where @xmath41 is the speed of light , @xmath42 , @xmath43 are the dirac matrices , and @xmath44 @xmath45 \displaystyle \int d^3 { { \mbox{\boldmath$r$}}}^{\prime } \ , \frac{\rho_{\rm nucl}({{\mbox{\boldmath$r$}}}^{\prime})}{|{{\mbox{\boldmath$r$}}}-{{\mbox{\boldmath$r$}}}^{\prime}| } & \quad \hbox{for the extended nucleus . } \end{array } \right .\ ] ] the nuclear charge density @xmath46 is defined by the nuclear model . in this paper we will use the fermi model for the nuclear charge distribution . here and in what follows we consider only the electric part of the classical electromagnetic interaction between the electron and the moving nuclei neglecting the magnetic interaction ( @xmath47 ) , which is small for low - energy collisions . the two - center expansion of the stationary wave function @xmath48 and the time - dependent wave function @xmath34 can be written in the form @xmath49 \psi({{\mbox{\boldmath$r$}}},t ) & = & \displaystyle \sum_{\alpha = a , b } \ , \sum_{a } c_{\alpha a}(t ) \ , \varphi_{\alpha , a } ( { { \mbox{\boldmath$r$}}}-{{\mbox{\boldmath$r$}}}_\alpha(t ) ) \ , , \end{array } \right . \label{expan1}\ ] ] where index @xmath50 enumerates the centers , index @xmath51 enumerates basis functions at the given center , and @xmath52 is the central - field bispinor , centered at point @xmath53 . the coefficients @xmath54 of the expansion ( [ expan1 ] ) for the stationary wave function @xmath48 can be obtained from the generalized eigenvalue equation @xmath55 where indexes @xmath56 and @xmath57 enumerate the basis functions of both centers , and the matrix elements of @xmath58 and @xmath30 are given by @xmath59 the expansion coefficients @xmath60 of the time - dependent wave function @xmath34 can be obtained by solving the linear system of first - order differential equations @xmath61 the matrix elements of @xmath62 are given by @xmath63 obviously the matrix @xmath62 is non - hermitian , if the overlapping matrix @xmath30 depends on time . the functions @xmath64 depend on time due to two reasons . first , the basis functions centered at the target and projectile nuclei move together with the nuclei . second , the basis functions depend parametrically on the distance between the nuclei , since their radial parts are obtained from the radial equations , where for each center the potential of the other nucleus is included in the so - called monopole approximation ( see section [ subsec : basis ] ) . therefore , the time derivative of the basis function can be divided into two parts @xmath65 where @xmath66 is the velocity of the ion @xmath53 . in the ion - ion collisions the internuclear distance vector @xmath67 , the length @xmath18 of vector @xmath68 , the target velocity ( @xmath69 ) , and the projectile velocity ( @xmath70 ) are time dependent . this dependence is defined by the trajectories of the nuclear motion . in low - energy collisions the nuclear trajectories can be obtained by solving classical non - relativistic newton s equations of motion . in the case of point charges this solution is well - known rutherford hyperbola ( see fig . [ rutherford ] ) , which can be given in the parametric representation by the equations @xcite is the minimal distance between target a and projectile b , v@xmath71 is the initial projectile velocity ) . the coordinate system @xmath72 is defined with respect to the moving target ion.,width=377 ] @xmath73 t & = & \displaystyle \frac{a}{v}_{0 } \ , ( \varepsilon \ , \sinh \ , \xi+\xi ) \ , , \end{array } \right . \label{rutherford1}\ ] ] where @xmath74 , @xmath75 @xmath76 is the initial velocity of the projectile , @xmath77 is the impact parameter , and @xmath78 is the reduced ion mass . in the coordinate system @xmath72 , which is shown in fig . [ rutherford ] , the @xmath79 and @xmath31 components of the internuclear distance vector @xmath68 are given by @xmath80 \displaystyle x & = & \displaystyle r \ , \sin \theta \end{array } \right . \ , , \qquad \hbox{where } \quad \theta = 2 \ , { \rm arctg } \left [ \frac { \sqrt{\varepsilon^2 - 1 } \ , ( { \rm th}(\xi/2)+1 ) } { ( \varepsilon+1)-(\varepsilon-1 ) \ , { \rm th}(\xi/2 ) } \right ] \,.\ ] ] the angle @xmath81 is related to the scattering angle @xmath82 by @xmath83 . in this work the two - center basis set @xmath84 is not orthonormal . let us consider the transformation of the basis set @xmath84 to the orthonormal basis @xmath85 by a matrix @xmath86 @xmath87 then the positive - defined matrix @xmath30 can be represented as the product of @xmath88 and @xmath89 : @xmath90 if the matrix @xmath89 is an upper - triangle matrix , then the decomposition ( [ cholesky1 ] ) is so - called cholesky factorization @xcite . the expansion of the time - dependent wave function over the orthonormal basis @xmath85 is given by @xmath91 where @xmath92 . the time - dependent dirac equation in the basis @xmath85 can be written in the form @xmath93 where @xmath94 , the hermitian hamiltonian matrix @xmath95 is @xmath96 and the matrix @xmath97 is defined by @xmath98 \ , l^{-1 } \right)_{ij } \,.\ ] ] it should be noted that matrix @xmath97 is hermitian , in contrast to the matrix @xmath62 defined by eq . ( [ matr2 ] ) . therefore , the matrix @xmath99 is also hermitian . the time - dependent matrix equation ( [ dirac_time ] ) can be considered as a linear system of the first - order differential equations at the range of time @xmath100 . we assume that at the initial moment of time ( @xmath101 ) the electron is localized on the target in the @xmath1 state and the projectile is the bare nucleus . then , the wave function @xmath34 at @xmath101 is given by @xmath102 if the dirac @xmath1-target wave function @xmath103 is included in the basis set , the initial conditions for the expansion coefficients can be written as @xmath104 equation ( [ dirac_time ] ) is solved numerically , using approximate evolution operator @xmath105 where hermitian matrix @xmath106 is chosen as @xmath107 since the approximate evolution operator @xmath108 is unitary , the time - dependent wave function conserves the norm at each time step @xmath109 the matrix @xmath110 is calculated at each time step using the eigen decomposition of matrix @xmath106 @xmath111 where @xmath112 is a diagonal matrix and columns of matrix @xmath113 are the eigenvectors of @xmath106 . then one obtains @xmath114 the time grid points @xmath115 are chosen as @xmath116 , where the parameter @xmath117 runs a uniform grid . the grid points @xmath118 can be obtained using equation ( [ rutherford1 ] ) . in our approach the basis set contains dirac and dirac - sturm orbitals . the dirac - sturm orbitals can be considered as pseudo - states , which should be included in the basis to take into account the contribution of the positive- and negative - energy dirac continuum . both types of basis functions @xmath119 are the central field dirac bispinors centered at the position @xmath120 ( @xmath50 ) @xmath121 \displaystyle i \ , \frac{q_{n \kappa}(r)}{r } \ , \chi_{-\kappa m}(\omega,\sigma ) \end{array } \right ) \,,\ ] ] where @xmath122 and @xmath123 are large and small radial components , respectively , and @xmath124 is the relativistic angular quantum number . the large and small radial orbital components are obtained by solving numerically the dirac or dirac - sturm equations in the central field potential @xmath125 . the dirac equation is given by @xmath126 \displaystyle c \left(~~\frac{d}{dr}+\frac{\kappa}{r } \right ) \ , p_{n \kappa}(r ) + \left ( v(r)-c^2 \right ) \ , q_{n\kappa}(r ) & = & \varepsilon_{n \kappa } \ , q_{n \kappa}(r)\ , . \label{dirac1}\ ] ] the radial components of the dirac - sturm orbitals which we denote by @xmath127 and @xmath128 are the solutions of the dirac - sturm generalized eigenvalue equation @xmath129 \displaystyle c \left(~~\frac{d}{dr}+\frac{\kappa}{r } \right ) \ , \overline p_{n \kappa}(r ) + \left ( v(r)-c^2 - \varepsilon_{n_0 \kappa } \right ) \ , \overline q_{n \kappa}(r ) & = & \lambda_{n \kappa } \ , w_{\kappa}(r ) \ , \overline q_{n \kappa}(r)\ , . \end{array } \right . \label{sturm1}\ ] ] here @xmath130 can be considered as the eigenvalue of the dirac - sturm operator and @xmath131 is a constant sign weight function . the energy @xmath132 is fixed in the dirac - sturm equation . if @xmath133 at @xmath134 , all sturmian functions have the same asymptotic at @xmath134 . it is clear that for @xmath135 the sturmian function coincides with the reference dirac orbital which has the radial parts @xmath136 and @xmath137 . the widely known choice of the weight function is @xmath138 , which leads to the well known charge quantization @xmath139 . the main advantage of this choice for the coulomb potential @xmath140 is that the coulomb - sturmian orbitals can be given in an analytical form . this is not the case , however , for the non - coulomb potential @xmath125 . in the relativistic case the choice @xmath138 is not very successful , because of the incorrect behavior of the coulomb - sturmian orbitals at @xmath141 . for this reason the standard form of the equation has to be modified @xcite . in our calculations we use the following weight function @xmath142 in contrast to @xmath143 , this weight function is regular at origin . it is well - known that the sturmian operator is hermitian and does not contain continuum spectra , in contrast to the dirac operator . therefore , the set of the sturmian eigenfunctions forms the discrete and complete basis set of one - electron wave functions . the central - field potential @xmath125 in equations ( [ dirac1 ] ) and ( [ sturm1 ] ) is arbitrary , and , therefore , it can be chosen to provide most appropriate dirac and dirac - sturm basis orbitals . at short internuclear distances the wave function of the electron experiences the strong coulomb field of both nuclei . to take into account this effect we have included the coulomb potential of the second ion in the total one center potential @xmath125 in so - called monopole approximation . for instance , the total central - field potential @xmath144 of the center @xmath19 is given by @xmath145 where @xmath146 is the coulomb potential of the nucleus @xmath19 and @xmath147 is the spherically - symmetric part of the reexpansion of the potential @xmath148 on the center @xmath19 @xmath149 for the point nucleus the potential @xmath147 is given by @xmath150 \displaystyle -\frac{z_b}{r } \quad & r < r \ , . \end{array } \right .\ ] ] the matrix elements of @xmath58 and @xmath30 ( eq . [ matr1 ] ) are easily reduced to radial integrals @xcite , which are calculated by numerical integration in the radial semi - logarithmic grid @xcite . two - center matrix elements are calculated using a symmetrical reexpansion procedure , proposed in refs . the reexpansion procedure is based on the technique developed by lwdin @xcite . we assume that in the local coordinate frame the @xmath151-axis is directed along the internuclear axis @xmath19@xmath20 ( see fig . [ int_region ] ) . the following geometrical relations take place @xmath152 \displaystyle \cos \theta_a= \frac{r_a^2+r^2-r^2_b}{2rr_a } \ , , \qquad \cos \theta_b= \frac{r_a^2-r^2-r^2_b}{2rr_b } \ , . \end{array}\ ] ] and @xmath153.,width=377 ] let the indexes @xmath51 and @xmath77 enumerate basis functions centered at the points @xmath19 and @xmath20 , respectively . the standard lwdin reexpansion of the nonrelativistic central - field function @xmath154 centered at the point @xmath20 in terms of the spherical harmonics @xmath155 centered at the point @xmath19 can be written in the form @xcite @xmath156 where @xmath157 is so - called lwdin @xmath53-function defined by @xmath158 here @xmath159 is the standard associated legendre polynomial and @xmath160 is the normalization constant @xmath161 similarly , the function @xmath162 centered at point @xmath19 can be expanded in spherical harmonics @xmath163 centered at the point @xmath20 . when the logarithmic or semi - logarithmic grid is used the radial grid step increases with increasing radius @xmath164 . therefore , the lwdin reexpansion procedure becomes unstable and poorly convergent for the values of radius @xmath164 in the region near @xmath165 , especially for the oscillating and strongly localized atomic - like wave functions . in addition , the lwdin procedure is not symmetric with respect to the centers @xmath19 and @xmath20 . to improve the convergence we modified the standard lwdin reexpansion procedure by dividing the range of the integration into two regions @xmath166 and @xmath153 as shown in fig . [ int_region ] . the region @xmath166 contains ion @xmath19 and the region @xmath153 contains ion @xmath20 . the dividing of the integration area into two parts can be done , for example , by a plane passing through the center of the segment ( @xmath167 ) . we apply the reexpansion procedure only to the `` tails '' of the wave functions occurring in a given region . to describe this procedure we introduce the step - wise functions @xmath168 and @xmath169 by @xmath170 and rewrite the product of the functions centered at the different points in the following way @xmath171 the reexpansion of the function tail @xmath172 centered at @xmath20 onto center @xmath19 has the form @xmath173 where @xmath174 and @xmath175 . in the relativistic case the spin - angular part @xmath176 of the large and small components of the central - field wave function is the pauli spinor @xcite @xmath177 where @xmath178 are the clebsch - gordan coefficients @xcite and @xmath179 is a spin function . the symmetric reexpansion of the relativistic wave function `` tails '' onto centers @xmath19 and @xmath20 can be written in the form @xmath180 \displaystyle i \,\frac{q_{b}(r_b)}{r_b } \ , \chi_{-\kappa_b \mu_b}({{\mbox{\boldmath$r$}}}_{b } ) \ , \end{array } \right ) \,\theta_a ( { { \mbox{\boldmath$r$ } } } ) = \sum_{\kappa } \left ( \begin{array}{l } \displaystyle \,\ , \frac{\overline p_{\kappa \mu_b}(b|r_a)}{r_a } \ , \chi_{\kappa \mu_b}({{\mbox{\boldmath$r$}}}_{a } ) \\[4 mm ] \displaystyle i \,\frac{\overline q_{-\kappa \mu_b}(b|r_a)}{r_a } \ , \chi_{-\kappa \mu_b}({{\mbox{\boldmath$r$}}}_{a } ) \ , \end{array } \right ) \ , \,\ ] ] [ rel_lowdin1 ] and @xmath181 \displaystyle i \,\frac{q_{a}(r_a)}{r_a } \ , \chi_{-\kappa_a \mu_a}({{\mbox{\boldmath$r$}}}_{a } ) \ , \end{array } \right ) \,\theta_b ( { { \mbox{\boldmath$r$ } } } ) = \sum_{\kappa } ( -1)^{l_a - l } \left ( \begin{array}{l } \displaystyle \,\ , \frac{\overline p_{\kappa \mu_a}(a|r_b)}{r_b } \ , \chi_{\kappa \mu_a}({{\mbox{\boldmath$r$}}}_{b } ) \\[4 mm ] \displaystyle i \,\frac{\overline q_{-\kappa \mu_a}(a|r_b)}{r_b } \ , \chi_{-\kappa \mu_a}({{\mbox{\boldmath$r$}}}_{b } ) \ , \end{array } \right ) \,.\ ] ] [ rel_lowdin2 ] the @xmath182- and @xmath183-functions , which are the relativistic analogs of the modified lwdin @xmath184-functions , are defined by @xmath185 \displaystyle \overline q _ { \kappa \mu_b}(b|r_{a } ) & = & \displaystyle \sum_{m_b , m_s } c^{j_b \mu_b}_{\overline l_b m_b,\frac{1}{2}m_s } \ , c^{j \mu_b}_{\overline l m_b,\frac{1}{2 } m_s } \ , \overline{\alpha}_{\overline l m_b}(q_b \ , \overline l_b |r_{a } ) \ , , \label{pq - functions } \end{array } \right .\ ] ] where @xmath186 . functions @xmath187 and @xmath188 are defined by equation ( [ alpha2 ] ) , where the function @xmath189 has to be replaced by the functions @xmath190 and @xmath191 , respectively . functions @xmath192 and @xmath193 are defined similarly to equation ( [ pq - functions ] ) , where indices @xmath19 and @xmath77 should be replaced by @xmath20 and @xmath51 , respectively . let us consider two - center overlap integrals @xmath194 . here and below symbol ( 0 ) means that the integral is considered in the local coordinate frame , where the @xmath151-axis is directed along internuclear axis @xmath19@xmath20 . the integral @xmath194 can be divided into two parts @xmath195 where the notations @xmath196 and @xmath197 mean the integration over the regions @xmath166 and @xmath153 , respectively ( see fig . [ int_region ] ) . using the reexpansions of the large and small components onto the center @xmath19 ( in the region @xmath166 ) and onto the center @xmath20 ( in the region @xmath153 ) we obtain @xmath198 \\[4 mm ] \displaystyle \langle a \mid b \rangle_{b } & = & \displaystyle ( -1)^{l_b - l_a } & \displaystyle \delta_{\mu_a,\mu_b } \ , \displaystyle \int \limits_{0}^{\infty } dr \ , \left [ p_b(r ) \cdot \overline p_{\kappa_b \mu_a}(a|r ) + q_b(r ) \cdot \overline q_{\kappa_b \mu_a}(a|r ) \right ] \ , . \end{array } \right .\ ] ] matrix elements of the nuclear attraction potentials , @xmath199 and @xmath200 , and of the mass operator @xmath201are calculated similarly to the overlap integral . as in case of the overlap integral , the region of integration for the gradient matrix element @xmath202 is divided into two parts , @xmath203 here index @xmath204 enumerates covariant spherical coordinates . using the gauss theorem @xcite for the integration over region @xmath19 , we obtain @xmath205 where @xmath206 is the surface integral over the region @xmath30 ( see fig . [ int_region ] ) . the volume integrals over regions @xmath19 and @xmath20 are given by @xmath207 \ , , \\[3 mm ] \displaystyle \langle a \mid { { \mbox{\boldmath$\nabla$}}}_{q } \mid b \rangle_{b } & = & \displaystyle { \mathop{{\sum}'}}_{\kappa } ( -1)^{l_a - l } g^{1 q}(j \mu_a , j_b \mu_b ) \int \limits_{0}^{\infty } dr \bigl [ \overline p_{\kappa \mu_a}(a|r ) \hat d_{\kappa,\kappa_b } p_b(r ) + \overline q_{\kappa \mu_a}(a|r ) \hat d_{-\kappa,-\kappa_b } q_b(r ) \bigr ] , \end{array } \label{nabla2}\ ] ] where the prime at the sum symbol indicates that the summation is restricted to odd values of @xmath208 and @xmath209 , and the operator @xmath210 is defined by @xmath211 the coefficients @xmath212 are the relativistic analogs of the gaunt coefficients @xcite @xmath213 the relativistic gaunt coefficient is non - zero only if @xmath214 is even . the surface integral is given by @xmath215 & \times \displaystyle \sum_{m , m_s } c^{j_a\mu_a}_{\overline l_a m , \frac{1}{2 } m_s } \ , c^{j_b\mu_a}_{\overline l_b m , \frac{1}{2 } m_s } \int \limits_{r/2}^{\infty } dr \ , \frac{1}{r } \ , q_a(r ) \ , q_b(r ) \ , u_{\overline l_a \overline l_b m } \bigl ( \frac{r}{2r}\bigr ) \bigr ] \ , , \end{array}\ ] ] where @xmath216 two - center ( @xmath217 ) matrix elements @xmath218 can be divided into three parts , similarly to the gradient matrix elements @xmath202 ( [ nabla1 ] ) , @xmath219 where the volume integrals @xmath220 and @xmath221 are given by @xmath222 \ , , \\[3 mm ] \displaystyle \langle a \mid { { \mbox{\boldmath$\alpha$ } } } \cdot{{\mbox{\boldmath$p$ } } } \mid b \rangle_{b } & = & \displaystyle ( -1)^{l_a - l_b } \delta_{\mu_a,\mu_b } \int \limits_{0}^{\infty } dr \bigl [ \bigl ( -\frac{dq_b}{d r } + \frac{\kappa_b q_b}{r } \bigr ) \overline p_{\kappa_b \mu_a}(a|r ) + \bigl ( \frac{dp_b}{d r } + \frac{\kappa_bp_b}{r } \bigr ) \overline q_{\kappa_b \mu_a}(a |r ) \bigr]\ , . \end{array}\ ] ] the last term in equation ( [ alp1 ] ) is the surface integral , which is given by @xmath223 & + & \displaystyle \delta_{\mu_a,\mu_b } \ , \sum_{m , m_s } m_s c^{j_a\mu_a}_{l_a m , \frac{1}{2 } m_s } \ , c^{j_b\mu_a}_{\overline l_b m , \frac{1}{2 } m_s } \int \limits_{r/2}^{\infty } dr \ , \frac{1}{r } \ , p_a(r ) \ , q_b(r ) \ , u_{l_a \overline l_b m } \bigl ( r/2r \bigr ) \ , . \end{array}\ ] ] as indicated above , the laboratory frame @xmath30 is defined to be at rest with respect to the initial target position . then , the two - center matrix elements calculated in the local coordinate frame @xmath224 ( see fig . [ rutherford ] ) have to be transformed to the laboratory frame @xmath225 . the corresponding two - center integrals can be obtained from @xmath194 , @xmath226 , and @xmath227 by rotating the coordinate system around the y - axis for angle @xmath228 ( fig . [ rutherford ] ) . for the overlap integrals @xmath229 and the two - center dirac - hamiltonian matrix elements @xmath230 in the laboratory frame @xmath30 we obtain @xmath231 \displaystyle h_{a , b } & = & h_{n_a \kappa_a \mu_a ; \ , n_b \kappa_b \mu_b } = \displaystyle \sum_{\mu } d^{j_a}_{\mu_a \mu}(\theta ) \,\ , d^{j_b^{\ast}}_{\mu_b \mu}(\theta ) \ , h^{(0)}_{n_a\kappa_a \mu ; \ , n_b \kappa_b \mu } \ , , \end{array}\ ] ] where @xmath232 are real wigner s d - functions @xcite . the transformation of the gradient matrix elements @xmath202 to the laboratory frame @xmath30 is given by @xmath233 transition amplitude for electron capture to an ion state @xmath234 is given by @xmath235 as previously mentioned , here index @xmath50 enumerates different centers ( target and projectile ions ) and @xmath236 are the wave functions of the free - moving ion @xmath53 . after the collision ( @xmath237 ) the wave functions @xmath236 of the free - moving ion @xmath53 are given by @xmath238 where @xmath239 are the stationary dirac wave functions of ion @xmath53 at the rest and @xmath240 is the translation factor . for the low - energy collisions the translation factor and the energy @xmath241 of the moving ion can be taken in the nonrelativistic approximation @xmath242 generally , the translation factor is introduced in the basis functions to improve the convergence of the time - dependent wave function expansion . we did not include the translation factor @xmath240 in the basis functions because of the computational complexity of the two - center integral calculations . however , at the limit @xmath237 we can reexpand the moving orbitals @xmath243 over the basis @xmath244 in the following way . at the limit @xmath237 the basis functions of the different centers do not overlap . the basis set @xmath84 is orthonormal and @xmath245 the expansion ( [ reexp1 ] ) is not exact , since the finite basis @xmath244 is incomplete and @xmath246 is non - unitary matrix . in particular , @xmath247 we can renormalize matrix @xmath246 and rewrite expansion ( [ reexp1 ] ) in the form @xmath248 here matrix @xmath249 plays a role of the normalization factor . the renormalized matrix @xmath250 is unitary . transition probabilities @xmath259 are defined by @xmath260 the probability @xmath259 , defined by equations ( [ prob1 ] ) and ( [ amplit2 ] ) , has an oscillatory behavior at @xmath261 , because the basis functions @xmath262 are not the solutions of the hydrogen - like dirac equation and the basis set is truncated . we can remove the oscillatory component of the probability @xmath259 for the large time ( @xmath237 ) in the same way , as it was done in refs . @xcite . at the large time ( @xmath237 ) the coefficients @xmath263 coincide with the coefficients @xmath264 . therefore , the coefficients @xmath264 are the solutions of equation ( [ dirac_time ] ) @xmath265 then for the coefficients @xmath266 we obtain the equation @xmath267 using the diagonalization procedure for the hermitian matrix @xmath268 , we can decompose @xmath269 where @xmath270 is a diagonal matrix with eigenvalues @xmath271 and @xmath113 is a unitary matrix . we introduce new coefficients @xmath272 by @xmath273 these coefficients have well defined limits at @xmath237 . the amplitudes @xmath255 , defined by equation ( [ amplit2 ] ) , in terms of the coefficients @xmath272 are given by @xmath274 then for the probabilities @xmath275 we obtain @xmath276 removing the oscillating term @xcite , we can introduce probabilities @xmath277 defined as @xmath278 since hydrogen - like dirac wave functions @xmath279 of each center @xmath53 ( including the positive and negative dirac continuum spectra ) form a complete basis set , we get @xmath280 the coefficients @xmath281 and the matrix elements have well - defined limits for @xmath261 . therefore , there exists the limit @xmath282 the direct ( @xmath283 ) , charge transfer ( @xmath284 ) , and ionization ( @xmath285 ) probabilities are given by @xmath286 where the prime at the sum symbol indicates that the summation runs over the discrete bound states of the ion @xmath53 . the cross sections for the charge transfer ( @xmath287 ) and ionization ( @xmath288 ) processes are then calculated as usual by integrating the probabilities over the impact parameter @xmath77 : @xmath289 it is well - known that in the nonrelativistic theory the scale transformation @xmath290 and @xmath291 allows one to transform the wave functions @xmath35 and the energies @xmath292 of a homonuclear one - electron quasi - molecule with a point nuclear charge @xmath10 to the wave functions @xmath293 and energies @xmath294 of the h@xmath23 molecule . the same scale transformation can be considered in the nonrelativistic homonuclear collisions . the time - dependent schrdinger equation describing a@xmath295-a@xmath9 collision is given by @xmath296 if in eq . ( [ schrod1 ] ) we set @xmath290 , @xmath297 , and @xmath298 , we obtain the time - dependent schrdinger equation for the h@xmath7-h collision @xcite @xmath299 it should be noted that the scaling @xmath300 is satisfied exactly for a straight - line trajectory . in this case the impact parameter @xmath77 , the velocity @xmath301 , and projectile energy @xmath302 are transformed by @xmath303 it follows that the probability @xmath304 and the cross section @xmath305 for a process in a symmetric system with nuclei of charge z can be obtained from the probability @xmath306 and the cross section @xmath307 for the same process in the h(@xmath1)h@xmath7 system by the relations @xmath308 it should be noted that this scaling law is not valid for the relativistic collisions . in our relativistic calculations , we used two dirac - sturm bases of different size . both bases include functions of the positive - energy dirac spectrum and sturm orbitals corresponding to the negative - energy dirac spectrum . it should be noted that the constructed bases satisfy the dual kinetic balance conditions @xcite and do not contain so - called `` spurious '' states @xcite . the positive - energy functions of the first basis ( basis 1 ) on each center in the standard nonrelativistic notations of atomic shells are given by : 1s-3s , 2p,3p , 3d , @xmath309-@xmath310 , @xmath311-@xmath312 , @xmath313-@xmath314 , @xmath315,@xmath316 . here the overline symbol ( @xmath317 ) is used to indicate the dirac - sturm ( pseudo - state ) basis functions . the total number of the positive - energy orbitals of both centers is @xmath318 and the total size of basis 1 , including the negative - energy orbitals , is @xmath319 . basis 1 is used for both stationary and time - dependent wave functions . the basis size can be increased in the calculations of the stationary states of quasi - molecules . positive - energy functions of the second basis ( basis 2 ) is constructed from @xmath320 atomic shells : 1s , @xmath321-@xmath322 , @xmath323-@xmath324 , @xmath325-@xmath326 @xmath327-@xmath328 @xmath329,@xmath330 . in basis 2 , the total number of orbitals of both ions , including the negative - energy spectrum , is equal to @xmath331 . this basis is used only in the calculations of the stationary states of quasi - molecules . in table [ mol_energies ] we present the results of our relativistic calculations of the @xmath332 state energy of the h@xmath23 , th@xmath24 , and u@xmath26 quasi - molecules for so - called chemical distance @xmath25 a.u .. since the quasi - molecule th@xmath24 was considered as a reference system for testing relativistic effects , it was calculated in a number of papers using high - precision large - scale methods ( see , e.g. , refs . @xcite ) . .relativistic energies ( a.u . ) of the 1@xmath333 state of quasi - molecules for the point - charge nuclei and @xmath25 a.u .. [ cols="^,^,^,^,^,^ " , ] [ u2_cross ] the values of @xmath334 were obtained for the rutherford trajectories of the target and projectile ions . one can see from the table , the relativistic effect amounts to about @xmath335 of the non - relativistic value of @xmath287 . in the 5-th column of table [ u2_cross ] , we also present our results obtained for the straight - line trajectory of the projectile ion ( in this case the target ion is at rest ) . as one can see from the table , the difference between the results obtained for the straight - line trajectory and the rutherford one is very small . the non - relativistic values of the charge - transfer cross section for the u@xmath5(@xmath1)+u@xmath6 collision , scaled to @xmath336 , are also compared with the cross section @xmath334 for the h(@xmath1)-h@xmath7 collision , presented in the 6-th column of table [ u2_cross ] . in this paper we presented a new method for the relativistic calculations of one - electron two - center quasi - molecular system in both stationary and time - dependent regimes . the method is suitable for a wide range of the internuclear distances including the critical regime , when the ground state of the quasi - molecule can dive into the negative - energy dirac continuum . using this method we calculated the energies of the h@xmath337 , th@xmath24 and u@xmath26 quasi - molecules , the critical distances for some homonuclear quasi - molecules a@xmath338 ( @xmath31=@xmath339 , @xmath340 , @xmath341 , @xmath342 , @xmath343 , @xmath344 ) , and the charge transfer probabilities , charge transfer and ionization cross sections for the h(@xmath1)h@xmath7 , ne@xmath0(@xmath1)ne@xmath2 , xe@xmath28(@xmath1)xe@xmath29 , and u@xmath5(@xmath1)u@xmath6 low - energy collisions . the results of our calculations of the charge transfer probabilities and cross sections for the h(1s)h@xmath7 collision are in a good agreement with experimental data and with theoretical results obtained by other authors . the influence of the relativistic effect on the charge transfer probabilities and cross sections for the ne@xmath0(@xmath1)ne@xmath2 , xe@xmath28(@xmath1)xe@xmath29 , and u@xmath5(@xmath1)u@xmath6 collisions is investigated . we demonstrated , that the relativistic and nonrelativistic charge - transfer probabilities as functions of the impact parameter have the same oscillatory behavior at low energies , but the relativistic curve shifted to lower energies compared to the nonrelativistic one . in the case of the u@xmath5(@xmath1)u@xmath6 collision the relativistic effect reduces the values of the cross section by about @xmath335 . in our further investigations we plan to study in more details the effect of diving the @xmath345 level of the u@xmath346 quasi - molecule into the negative - energy dirac spectrum and the influence of this effect on the values of the charge - transfer probability . with this goal , we are going to develop an approach which would allow us to compare the calculated probabilities with and without the diving of the ground state into the negative - energy continuum . we also plan to extend our method to collisions involving many - electron ions and neutral atoms . this will allow us to study the @xmath1-@xmath1 charge transfer in low - energy heavy ion - atom collisions . such experiments , that were successfully performed for low- and middle-@xmath31 collisions @xcite , are presently under preparation for high-@xmath31 collisions at gsi . we thank v. shevelko for providing us with his pwb results for the charge - transfer probabilities and for valuable discussions . this work was supported by dfg ( grants no . 436rus113/950/0 - 1 and vo 1707/1 - 1 ) , by gsi , by rfbr ( grants no . 08 - 02 - 91967 and 10 - 02 - 00450 ) , by the ministry of education and science of russian federation ( program for development of scientific potential of high school , grant no . 2.1.1/1136 ; program `` scientific and pedagogical specialists for innovative russia '' , grant no . p1334 ) , and by the extreme matter institute emmi in the framework of the helmholtz alliance ha216/emmi . y.s.k . acknowledges financial support by the dynasty foundation and daad . v.m.s . acknowledges financial support by the alexander von humboldt foundation . firsov , zh . eksp . teor * 21 * , 1001 ( 1951 ) . demkov , uchen . univ . * 146 * , 74 ( 1952 ) . bates , h.s.w . massey , and a.l . stewart , proc . soc ( london ) a*216 * , 437 ( 1953 ) . b.h . bransden and m.r.c . mcdowell , _ charge exchange and the theory of ion - atom collisions _ , ( oxford university press , ny 1992 ) . w. fritsch and c.d . lin , phys . rep . * 202 * , 1 ( 1991 ) . winter , adv . at . . phys . * 52 * , 391 ( 2005 ) . gallaher , l. wilets , phys . rev . * 169 * , 139 ( 1968 ) . r. shakeshaft , phys . a * 14 * , 1626 ( 1976 ) . reading , a.l . ford and r.l . becker , j. phys . b * 14 * 1995 ( 1981 ) . w. fritsch , c.d . lin , phys . a * 27 * , 3361 ( 1983 ) . a.m. ermolaev , j. phys . b * 23 * , l45 ( 1990 ) . n. toshima , phys . rev . * 59 * , 1981 ( 1999 ) . winter , phys . a * 80 * , 032701 ( 2009 ) . n. grn , a. mhlhans and w. scheid , j. phys . b * 15 * , 4043 ( 1982 ) . a. kolakowska , m.s . pindzola , f. robicheaux , d.r . schultz , c. wells , phys . a * 98 * , 2872 ( 1998 ) . a. kolakowska and m.s . pindzola , d.r . schultz , phys . a * 59 * , 3588 ( 1999 ) . schultz , m.r . strayer and j.c . wells , phys . lett . * 82 * , 3976 ( 1999 ) . x. tong , d. kato , t. watanabe , and s. ohtani phys . a * 62 * , 052701 ( 2000 ) . briggs , and j.h . macek , j. phys . b * 6 * , 982 ( 1973 ) ; corrigenda * 6 * , 2484 ( 1973 ) . j. eichler and w.e . meyerhof , _ relativistic atomic collisions _ , ( academic , new york , 1995 ) . shabaev , phys . rep . * 356 * , 119 ( 2002 ) . j. eichler and t. sthlker , phys . rep . * 439 * , 1 ( 2007 ) . y.b . zeldovich and v.s . popov , usp . nauk * 105 * , 403 ( 1971 ) [ sov . * 14 * , 673 ( 1972 ) ] . j. rafelski , l.p . fulcher , and w. greiner , phys . * 27 * , 958 ( 1971 ) . w. greiner , b. mller , j. rafelski , _ quantum electrodynamics of strong fields _ , ( springer - verlag , berlin , 1985 ) . u. mller - nehler and g. soff , phys . 246 * , 101 ( 1994 ) . j. rafelski , b. mller , phys . lett . * 65*b , 205 ( 1976 ) . u. becker , n. grn , w. scheid , and g. soff , phys . 56 * , 2016 ( 1986 ) . strayer , c. bottcher , v.e . oberacker , and a.s . umar , phys . a * 41 * , 1399 ( 1990 ) . j. thiel , a. bunker , k. momberger , n. grn , and w. scheid , phys . a * 46 * , 2607 , ( 1992 ) . k. momberger , a. belkacem , and a.h . sorensen , phys . a * 53 * , 1605 ( 1996 ) . wells , v.e . oberacker , m.r . strayer , and a.s . umar , phys . a * 53 * , 1498 ( 1996 ) . d.c . ionescu and a. belkacem , physica scripta * t80 * , 128 ( 1999 ) . m.s . pindzola , phys . a * 62 * , 032707 ( 2000 ) . o. busic , n. grn , and w. scheid , phys . a * 70 * , 062707 ( 2004 ) . k. rumrich , g. soff , w. greiner , phys . a * 47 * , 215 ( 1993 ) . k. momberger , n. grn , and w. scheid , j. phys . b * 26 * , 1851 ( 1993 ) . m. gail , n. grn and w. scheid , j. phys . b * 36 * , 1397 ( 2003 ) . g. soff , j. reinhardt , and w. betz , phys . * 17 * , 417 ( 1978 ) . t.h.j de reus , j. reinhardtt , b. mller , w. greiner , g. soff , and u. mller , j. phys . b * 17 * , 615 ( 1984 ) . e. ackad and m. horbatsch , phys . a * 78 * , 062711 ( 2008 ) . sthlker , d.c . ionescu , p. rymuza , f. bosch , h. geissel , c. kozhuharov , t. ludziejewski , p.h . mokler , c. scheidenberger , z. stachura , a. warczak and r. w. dunford , phys . a * 238 * , 43 ( 1998 ) . sthlker , d.c . ionescu , p. rymuza , z. stachura , a. warczak , and r. w. dunford , phys . a * 57 * , 845 ( 1998 ) . ionescu and th . sthlker , phys . * 67 * , 022705 ( 2003 ) . m. rotenberg , adv . at . and mol . * 6 * , 233 ( 1970 ) . gruzdev , g.s . soloveva , a.i . sherstyuk , opt . and spectrosc , * 63 * , 1394 ( 1987 ) , n.l . manakov , l.p . rapoport and s.a . zapryagaev , phys . a * 43 * , 139 ( 1973 ) . drake and s.p . goldman , adv . at . phys . * 25 * , 393 ( 1988 ) . j. avery and f. antonsen , j. math . * 24 * , 175 ( 1998 ) . i.p . grant and h.m . quiney , phys . a * 62 * , 022508 ( 2000 ) . tupitsyn , v.m . shabaev , j.r . crespo lpez - urrutia , i. draganic , r. soria orts , and j. ulrich , phys . a * 68 * , 022511 ( 2003 ) . tupitsyn , a.v . volotka , d.a . glazov , v.m . shabaev , g.plunien , j.r.crespo lopez - urrutia , a.lapierre , and j. ullrich , phys . a * 72 * , 062503 ( 2005 ) . lwdin , adv . in phys . * 5 * , 1 ( 1956 ) . s. kotochigova , i. tupitsyn , int . j. quantum chem . * 29 * , 307 ( 1995 ) . tupitsyn , d.a . savin , and v.g . kuznetsov , opt . and spectrosc . * 84 * , 344 ( 1998 ) . l. yang , d. heinemann , d. kolb , chem . . lett . * 178 * , 213 ( 1991 ) . o. kullie , d. kolb , eur . j. d * 17 * , 167 ( 2001 ) . lisin , m.s . marinov , v.s . popov , phys . * 69*b , 2 ( 1977 ) . matveev , d.u . matrasulov , and h.yu . rakhimov , phys . at . * 63 * , 318 . b. mller , and w. greiner , z. naturforsch . * 31*a , 1 ( 1976 ) . lisin , m.s . marinov , v.s . popov , phys.lett . * 91*b , 20 ( 1980 ) . v.s . popov , phys . at . nucl . * 64 * , 367 , ( 2001 ) . j. crank , p. nicholson , proc . . soc . * 43 * , 50 ( 1947 ) . feit , j.a . fleck , jr . , and a. steiger , j. comput . phys . * 47 * , 412 ( 1982 ) . j. wilkinson , c. reinsch , _ handbook for automatic computation _ , ( springer - verlag , berlin , 1971 ) , i.p . grant , b.l . gyorffy , _ the effects of relativity in atoms , molecules , and the solid state _ , ( ed . s wilson , new york : plenum , 1991 ) . r. szmytkowski , j. phys . b * 30 * , 825 ( 1997 ) . i.p . grant , adv . * 19 * , 747 ( 1970 ) . bratsev , g.b . deyneka , and i.i . tupitsyn , bull.acad.sci . ussr , phys . * 41 * , 173 ( 1977 ) . sharma , phys . a * 13 * , 517 ( 1977 ) . m.e . rose , _ relativistic electron theory _ , ( john - wiley & sons , ny - london , 1961 ) d.a . varshalovich , a.n . moskalev , v.k . khersonskii , _ quantum theory of angular momentum _ , ( world scientific , singapore , 1988 ) . marsden , a.j . tromba , _ vector calculus _ , ( 5th edition , w.h . freeman & company , new - york , 2003 ) . condon , g.h . shortley , _ theory of atomic spectra _ , ( cambridge university press , london , 1935 ) . shabaev , i.i . tupitsyn , v.a . yerokhin , g. plunien , and g. soff , phys . . lett . * 93 * , 130405 ( 2004 ) . tupitsyn , v.m . shabaev , opt . and spectrosc . * 105 * , 183 ( 2004 ) . parpia , a.k . mohanty , chem . . lett . * 238 * , 209 ( 1995 ) . l. lajohn , j.d . talman , theor . acta . * 99 * , 351 ( 1998 ) . a. rutkowski , chem . lett . * 307 * , 259 ( 1999 ) . b. mller , j. rafelski , and w. greiner , z. physik * 257 * , 183 ( 1972 ) . marinov , v.s . popov , and v.l . stolin , j. comp . phys . * 19 * , 241 ( 1975 ) . k - h wietschorke , b. muller , w. greiner and g. soff , j. phys . b * 12 * , l31 ( 1979 ) . parpia , a.k . mohanty , phys . a * 46 * , 3735 ( 1992 ) . i. angeli , at . data nucl . data tables * 87 * , 185 ( 2004 ) . kozhedub , v.m . shabaev , unpublished . kozhedub , o.v . andreev , v.m . shabaev , i.i . tupitsyn , c. brandau , c. kozhuharov , g. plunien , and t. sthlker , phys . a * 77 * , 032501 ( 2008 ) . johnson , g. soff , at . data nucl . data tables * 33 * , 405 ( 1985 ) . w. fritsch and c.d . lin , phys . a * 26 * , 762 ( 1982 ) . ldde and r.m . dreizler , j. phys . b * 14 * , 2191 ( 1981 ) . ldde and r.m . b * 15 * , 2703 ( 1982 ) . r. shakeshaft , phys . a * 18 * , 1930 ( 1978 ) . janev , j.j . smith , atomic and plasma material interaction data for fusion , nucl fusion suppl . special issue , * 4 * ( 1993 ) . m.b . shah j. phys . b * 20 * , 2481 ( 1987 ) . m.b . shah j. phys . b * 31 * , l757 ( 1998 ) . shevelko , private communication . shevelko , i.yu . tolstikhina , th . sthlker , nucl . methods b * 184 * , 295 ( 2001 ) . s. hagmann , c. l. cocke , j. r. macdonald , p. richard , h. schmidt - bcking , and r. schuch , phys . a * 25 * , 1918 ( 1982 ) . s. hagmann , s. kelbch , h. schmidt - bcking , c. l. cocke , p. richard , r. schuch , a. skutlartz , j. ullrich , b. johnson , m. meron , k. jones , d. trautmann and f. rsel , phys . a * 36 * , 2603 ( 1987 ) . r. schuch , h. ingwersen , e. justiniano , h. schmidtbocking , m. schulz , and f. ziegler , j. phys . b * 17 * , 2319 ( 1984 ) .

in this paper , we consider a large class of purely discontinuous rotationally symmetric lvy processes . we establish sharp two - sided estimates for the transition densities of such processes killed upon leaving an open set . when is a-fat open set , the sharp two - sided estimates are given in terms of surviving probabilities and the global transition density of the lvy process . when is a open set and the lvy exponent of the process is given by with being a complete bernstein function satisfying a mild growth condition at infinity , our two - sided estimates are explicit in terms of , the distance function to the boundary of and the jumping kernel of , which give an affirmative answer to the conjecture posted in . our results are the first sharp two - sided dirichlet heat kernel estimates for a large class of symmetric lvy processes with general lvy exponents . we also derive an explicit lower bound estimate for symmetric lvy processes on in terms of their lvy exponents . * ams 2000 mathematics subject classification * : primary 60j35 , 47g20 , 60j75 ; secondary 47d07 * keywords and phrases * : lvy processes , subordinate brownian motion , heat kernel , transition density , dirichlet transition density , green function , exit time , lvy system , boundary harnack inequality , parabolic harnack inequality

1303.6449

due to their importance in theory and applications , fine potential theoretical properties of lvy processes have been under intense study recently . the transition density @xmath8 of a lvy process is the heat kernel of the generator of the process . however , the transition density ( if it exists ) of a general lvy process rarely admits an explicit expression . thus obtaining sharp estimates on @xmath8 is a fundamental problem both in probability theory and in analysis . the generator of a discontinuous lvy process is an integro - differential operator and so it is a non - local operator . recently , quite a few people in pde are interested in problems related to non - local operators ; see , for example , @xcite and the references therein . when @xmath6 is a symmetric diffusion on @xmath7 whose infinitesimal generator is a uniformly elliptic and bounded divergence form operator , it is well - known that @xmath8 enjoys the celebrated aronson s gaussian type estimates . when @xmath6 is a pure jump symmetric process on @xmath7 , sharp estimates on @xmath8 have been studied in @xcite recently , which can be viewed as the counterpart of aronson s estimates for non - local operators . due to the complication near the boundary , two - sided estimates for the transition densities of discontinuous lvy processes killed upon leaving an open set @xmath0 ( equivalently , the dirichlet heat kernels ) have been established very recently for a few particular processes only . the first of such estimates is obtained in @xcite , where we succeeded in establishing sharp two - sided estimates for the heat kernel of the fractional laplacian @xmath9 with zero exterior condition on @xmath10 ( or equivalently , the transition density of the killed symmetric @xmath11-stable process ) in any @xmath2 open set @xmath0 . the approach developed in @xcite provides a road map for establishing sharp two - sided heat kernel estimates of other jump processes in open subsets of @xmath7 . the ideas of @xcite were adapted to establish sharp two - sided heat kernel estimates of relativistic stable processes and mixed stable processes in @xmath2 open subsets of @xmath7 in @xcite respectively . in all these cases , the characteristic exponents of these lvy processes admit explicit expressions , the boundary decay rates of the dirichlet heat kernels are suitable powers of the distance to the boundary . on the other hand , a varopoulos type two - sided dirichlet heat kernel estimate of symmetric stable processes in @xmath1-fat open sets was derived in @xcite ; this type of estimates is expressed in terms of surviving probabilities and the global transition density of the symmetric stable process . the objective of this paper is to establish sharp two - sided estimates on the transition density @xmath12 for a large class of purely discontinuous rotationally symmetric lvy processes . unlike the cases considered in @xcite , the characteristic exponents of the symmetric lvy processes considered in this paper are quite general , satisfying only certain mild growth condition at @xmath13 . moreover , the boundary decay rate of @xmath12 is no longer some power of the distance to the boundary . the analysis of the precise boundary behavior of @xmath12 is quite challenging and delicate . the main tools to obtain the precise boundary behavior of @xmath12 are two versions of the boundary harnack principle obtained in @xcite . in this paper we combine the approaches developed in @xcite with these boundary harnack principles to obtain sharp two - sided estimates for @xmath12 , which cover the main results in @xcite and much more . suppose that @xmath14 is a subordinator with laplace exponent @xmath4 , that is , @xmath15 is a nonnegative lvy process with @xmath16 and @xmath17=e^{-t\phi(\lambda)}$ ] for every @xmath18 . the function @xmath4 can be written in the form @xmath19 where @xmath20 and @xmath21 is a measure on @xmath22 satisfying @xmath23 . the constant @xmath24 is called the drift of the subordinator and @xmath21 the lvy measure of the subordinator ( or of @xmath4 ) . the function @xmath4 is a bernstein function , i.e. , it is @xmath25 , positive and @xmath26 for all @xmath27 . in particular , since @xmath28 and @xmath29 , the bernstein function @xmath4 has the property that @xmath30 the laplace exponent @xmath4 is said to be a complete bernstein function if the lvy measure @xmath21 of @xmath4 has a completely monotone density @xmath31 , i.e. , @xmath32 for every non - negative integer @xmath33 . for basic results on complete bernstein functions , we refer the reader to @xcite . throughout this paper , we assume that @xmath4 is a complete bernstein function satisfying the following growth condition at infinity ( see @xcite ) : * ( a ) : * there exist constants @xmath34 , @xmath35 , @xmath36 and @xmath37 such that @xmath38 note that it follows from the upper bound condition in * ( a ) * that @xmath4 has no drift . let @xmath39 be a brownian motion in @xmath7 independent of the subordinator @xmath15 . the subordinate brownian motion @xmath40 is defined by @xmath41 , which is a rotationally symmetric lvy process with lvy exponent @xmath42 . the infinitesimal generator of @xmath43 is @xmath44 . here and below for a function @xmath45 on @xmath46 , @xmath47 is defined as a pseudo differential operator in terms of fourier transform ; that is , @xmath48 , where @xmath49 is the fourier transform of a function @xmath50 on @xmath7 . it is known that the lvy measure of the process @xmath43 has a density given by @xmath51 where @xmath52 note that the function @xmath53 is continuous and decreasing on @xmath22 . we will assume that @xmath6 is a purely discontinuous rotationally symmetric lvy process with lvy exponent @xmath54 . because of rotational symmetry , the function @xmath5 depends on @xmath55 only , and by a slight abuse of notation we write @xmath56 . the infinitesimal generator of @xmath6 is @xmath57 . we further assume that the lvy measure of @xmath6 has a density with respect to the lebesgue measure on @xmath7 , which is denoted by @xmath58 . that is , @xmath59=e^{-t\psi(|\xi| ) } \quad \quad \mbox { for every } x\in \r^d \mbox { and } \xi\in \r^d,\ ] ] with @xmath60 we assume that @xmath61 is continuous on @xmath22 and that there is a constant @xmath62 such that @xmath63 this implies that @xmath64 and @xmath65 are comparable . clearly also implies that @xmath66 we remark that under the above assumptions , @xmath6 does not need to be a subordinate brownian motion because @xmath67 does not need to be monotone . for example , choose @xmath68 such that @xmath69 and a continuous function @xmath70 with @xmath71 , @xmath72 and @xmath73 for all @xmath74 . then @xmath75 is not monotone and its corresponding lvy process @xmath6 ( through lvy exponent ) is not a subordinate brownian motion . under the above setup , @xmath6 has a continuous transition density @xmath8 with respect to the lebesgue measure on @xmath7 ( see @xcite ) . clearly , @xmath8 is a function depending only on @xmath76 and @xmath77 , and so , by an abuse of notation , we also denote @xmath8 by @xmath78 . for every open subset @xmath79 , we denote by @xmath80 the subprocess of @xmath6 killed upon exiting @xmath0 . it is known ( see @xcite ) that @xmath80 has a transition density @xmath12 , with respect to the lebesgue measure , which is jointly locally hlder continuous . note that @xmath81 is the fundamental solution for @xmath82 in @xmath0 with zero exterior condition and so it can also be called the dirichlet heat kernel of @xmath83 in @xmath0 . the purpose of this paper is to establish sharp two - sided estimates on @xmath12 . the following two conditions will be needed for some of the results in this paper when @xmath0 is unbounded . * ( b ) * : there exist constants @xmath84 and @xmath85 $ ] such that @xmath86 \hbox { and } u\geq r>0.\ ] ] * ( c ) * : there exist constants @xmath87 and @xmath88 $ ] such that @xmath89 \hbox { and } r>0.\ ] ] throughout this paper we will use @xmath90 to denote the function [ e:1.9 ] ( r)=1(r^-2 ) , r>0 . note that in particular it follows from that [ e:1.8 ] ( 2r)= = 4(r ) r>0 . the inverse function of @xmath90 will be denoted by the usual notation @xmath91 . here and in the following , for @xmath92 , @xmath93 and @xmath94 . [ r:1.1 ] \(i ) the condition * ( b ) * is pretty mild . when @xmath6 is a rotationally symmetric lvy process such that @xmath95 is decreasing , condition * ( b ) * holds for all @xmath96 ( see ( * ? ? ? * proposition ) ) . in particular it holds for all subordinate brownian motions with @xmath97 . in this special case , we can also see this using the following elementary argument : when @xmath6 is a subordinate brownian motion , @xmath98 $ ] , where @xmath99 is the transition density of the brownian motion @xmath100 . it follows immediately that @xmath101 is decreasing in @xmath102 and so * ( b ) * holds with @xmath97 . \(ii ) under condition * ( a ) * , condition * ( b ) * is weaker than * ( c)*. under condition * ( a ) * , we will show in this paper that there exists @xmath103 such that @xmath104 \hbox { and } r>0\ ] ] ( see proposition [ p : new1 ] and theorem [ t : globalhke ] below ) . thus condition * ( c ) * amounts to say that there exist constants @xmath105 and @xmath106 $ ] such that for @xmath107 $ ] and @xmath74 , @xmath108 since @xmath109 is a decreasing function in @xmath102 , implies that condition * ( b ) * holds with @xmath110 and @xmath111 . \(iii ) assume that condition * ( a ) * holds . it follows from @xcite that , for every @xmath112 , there is a constant @xmath113 so that holds for @xmath114\times ( 0 , r]$ ] with @xmath115 . ( see proposition [ p : new1 ] below . ) so the assertions in conditions * ( b ) * and * ( c ) * are always satisfied for @xmath116 . \(iv ) by @xcite , @xmath117\times ( 0 , \infty),\ ] ] and consequently conditions * ( b ) * and * ( c ) * , hold for a large class of discontinuous processes including mixed stable - like processes ( with @xmath115 ) and relativistic stable - like processes ( with @xmath115 , see ( * ? ? ? * theorem 4.1 ) ) . before stating the main results of this paper , we need first to set up some notations . let @xmath118 . we denote the euclidean distance between @xmath119 and @xmath120 in @xmath7 by @xmath77 and denote by @xmath121 the open ball centered at @xmath122 with radius @xmath74 ; for any two positive functions @xmath50 and @xmath123 , @xmath124 means that there is a positive constant @xmath105 so that @xmath125 on their common domain of definition ; for any open @xmath126 and @xmath127 , @xmath128 stands for the diameter of @xmath0 and @xmath129 stands for the euclidean distance between @xmath119 and @xmath10 . [ def : ub ] let @xmath130 . we say that a open set @xmath0 is @xmath1-fat if there is @xmath131 such that for all @xmath132 and all @xmath133 $ ] , there is a ball @xmath134 . the pair @xmath135 is called the characteristics of the @xmath1-fat open set @xmath0 . the following factorization of the dirichlet heat kernel is the first main result of this paper . recall that @xmath136 and @xmath137 are the constants in condition * ( b)*. [ thm : oppz ] let @xmath6 be a purely discontinuous rotationally symmetric lvy process with lvy exponent @xmath5 and lvy density @xmath65 satisfying and respectively , where the complete bernstein function @xmath4 satisfies * ( a)*. suppose that @xmath0 is a @xmath1-fat open set with characteristics @xmath135 . \(i ) for every @xmath138 , there exists @xmath139 such that for @xmath140 , @xmath141 , @xmath142 \(ii ) if @xmath0 is unbounded , we assume in addition that condition * ( b ) * holds . for every @xmath138 , there exists @xmath143 such that for @xmath140 , @xmath141 , @xmath144 where @xmath145 . \(iii ) suppose in addition that @xmath0 is bounded . then there exists @xmath146 so that for all @xmath147 , @xmath148 where @xmath149 is the largest eigenvalue of the generator of @xmath80 . when @xmath6 is a rotationally symmetric @xmath11-stable process in @xmath7 , that is , when @xmath150 for some @xmath151 , parts ( i ) and ( ii ) of theorem [ thm : oppz ] are proved in @xcite . recall that @xmath152 and @xmath153 are the constants in condition * ( c)*. combining theorem [ thm : oppz](i)(ii ) and remark [ r:1.1](ii ) , we have the following corollary . [ c : main1 ] let @xmath6 be a purely discontinuous rotationally symmetric lvy process with lvy exponent @xmath5 and lvy density @xmath65 satisfying and respectively , where the complete bernstein function @xmath4 satisfies * ( a)*. suppose that @xmath0 is a @xmath1-fat open set with characteristics @xmath135 . if @xmath0 is unbounded , we assume in addition that condition * ( c ) * holds . for every @xmath138 , there exist @xmath154 and @xmath155 such that for @xmath140 , @xmath141 , @xmath156 where @xmath157 . the second main result of this paper is on explicit sharp dirichlet heat kernel estimates for subordinate brownian motions in @xmath158 open sets . so in the remainder of this section , we assume that @xmath159 , a subordinate brownian motion with lvy exponent @xmath160 . recall that an open set @xmath0 in @xmath161 ( when @xmath162 ) is said to be a ( uniform ) @xmath158 open set if there exist a localization radius @xmath163 and a constant @xmath164 such that for every @xmath165 , there exist a @xmath158-function @xmath166 satisfying @xmath167 , @xmath168 , @xmath169 , @xmath170 such that @xmath171 the pair @xmath172 is called the characteristics of the @xmath158 open set @xmath0 . note that a @xmath158 open set @xmath0 with characteristics @xmath172 can be unbounded and disconnected ; the distance between two distinct components of @xmath0 is at least @xmath173 . by a @xmath158 open set in @xmath174 we mean an open set which can be written as the union of disjoint intervals so that the minimum of the lengths of all these intervals is positive and the minimum of the distances between these intervals is positive . here is the second main result of this paper , which gives an affirmative answer to the conjecture posed in @xcite . in view of remark [ r:1.1 ] , it extends the main results of @xcite . recall that condition * ( b ) * holds with @xmath97 for any subordinate brownian motion . [ t : main ] suppose that @xmath6 is a subordinate brownian motion with lvy exponent @xmath160 with @xmath4 being a complete bernstein function satisfying condition * ( a)*. let @xmath0 be a @xmath158 open subset of @xmath161 with characteristics @xmath172 . \(i ) for every @xmath138 , there exists @xmath175 such that for all @xmath176\times d\times d$ ] , @xmath177 \(ii ) for every @xmath138 , there exists @xmath178 such that for all @xmath176\times d\times d$ ] , @xmath179 \(iii ) suppose in addition that @xmath0 is bounded . for every @xmath138 , there exists @xmath180 depending only on @xmath181 and @xmath182 so that for all @xmath183 , @xmath184 where @xmath149 is the largest eigenvalue of the generator of @xmath80 . recall that @xmath152 and @xmath153 are the constants in condition * ( c)*. combining theorem [ t : main](i)(ii ) and remark [ r:1.1](ii ) , we have the following corollary . [ c : main2 ] suppose that @xmath6 is a subordinate brownian motion with lvy exponent @xmath160 with @xmath4 being a complete bernstein function satisfying condition * ( a)*. let @xmath0 be a @xmath158 open subset of @xmath161 with characteristics @xmath172 . if @xmath0 is unbounded , we assume in addition that condition * ( c ) * holds . for every @xmath138 , there exist @xmath175 and @xmath185 such that for @xmath140 , @xmath141 , @xmath186 when @xmath6 is a rotationally symmetric @xmath11-stable process in @xmath7 , theorem [ t : main ] is first established in @xcite . sharp two - sided dirichlet heat kernel estimates in @xmath158 open sets are subsequently established in @xcite for censored stable processes , relativistic stable processes , mixed stable processes , and mixed brownian motion and stable processes , respectively . by integrating the two - sided heat kernel estimates in theorem [ t : main ] with respect to @xmath76 , we obtain the two - sided estimates on the green function @xmath187 ( see theorem [ t : gfe ] below ) , which extend ( * ? ? ? * theorem 1.1 ) . condition * ( a ) * is a very weak condition on the behavior of @xmath4 near infinity . using the tables at the end of @xcite , one can come up plenty of explicit examples of complete bernstein functions satisfying condition * ( a)*. here are a few of them . \(1 ) @xmath188 , @xmath189 $ ] ( symmetric @xmath11-stable process ) ; \(2 ) @xmath190 , @xmath191 and @xmath192 ( relativistic @xmath11-stable process ) ; \(3 ) @xmath193 , @xmath194 ( mixed symmetric @xmath11- and @xmath195-stable processes ) ; \(4 ) @xmath196 , @xmath191 , @xmath197 $ ] . now we give a way of constructing less explicit complete bernstein functions that have very general asymptotic behavior at infinity . suppose that @xmath191 and @xmath198 is a positive function on @xmath22 which is slowly varying at infinity . we further assume that @xmath199 is a right continuous increasing function with @xmath200 ( so @xmath201 ) . then the function @xmath202 is a stieltjes function , and so the function @xmath203 is a complete bernstein function ( see ( * ? ? ? * theorem 7.3 ) ) . it follows from ( * ? ? ? * lemma 6.2 ) that @xmath204 when @xmath205 . the rest of the paper is organized as follows . section [ s:2 ] recalls and collects some preliminary results that will be used in the sequel , including on - diagonal heat kernel estimates and the boundary harnack principle . section [ s:3 ] presents the interior lower bound heat kernel estimates , including an explicit lower bound estimate for symmetric lvy processes on @xmath7 . the proof of the short time factorization result for @xmath12 ( that is , theorem [ thm : oppz](i ) and ( ii ) ) is given in section [ s:4 ] , while the proof of theorem [ t : main](i ) and ( ii ) is given in section [ s:5 ] . the large time heat kernel estimates are proved in section [ s:6 ] . the green function estimates for subordinate brownian motions in bounded @xmath158 open sets are derived in section [ s:7 ] from the two - sided dirichlet heat kernel estimates in theorem [ t : main ] . the derivation , however , requires quite some effort . throughout this paper , @xmath206 and the constants @xmath207 , @xmath208 , @xmath173 , @xmath209 , @xmath1 , @xmath210 , @xmath211 , @xmath136 , @xmath137 , @xmath152 , @xmath153 , @xmath212 and @xmath213 will be fixed . we use @xmath214 to denote generic constants , whose exact values are not important and can change from one appearance to another . the labeling of the constants @xmath214 starts anew in the statement of each result . the dependence of the constant @xmath215 on the dimension @xmath216 will not be mentioned explicitly . we will use @xmath217 " to denote a definition , which is read as is defined to be " . we will use @xmath218 to denote a cemetery point and for every function @xmath50 , we extend its definition to @xmath218 by setting @xmath219 . we will use @xmath220 to denote the lebesgue measure in @xmath161 . for a borel set @xmath221 , we also use @xmath222 to denote its lebesgue measure . in the first part of this section , we assume that @xmath6 is a purely discontinuous rotationally symmetric lvy process with lvy exponent @xmath5 and lvy density @xmath65 satisfying and respectively , where the complete bernstein function @xmath4 satisfies * ( a)*. a function @xmath223 is said to be harmonic in an open set @xmath79 with respect to @xmath6 if for every open set @xmath224 whose closure is a compact subset of @xmath0 , @xmath225 \qquad \hbox{for every } x\in b.\ ] ] a function @xmath223 is said to be regular harmonic in an open set @xmath79 with respect to @xmath6 if @xmath226 \qquad \hbox{for every } x\in d.\ ] ] clearly , a regular harmonic function in @xmath0 is harmonic in @xmath0 . very recently the following form of the boundary harnack principle is established in @xcite . [ ubhp ] there exists @xmath227 such that for any @xmath228 , any open set @xmath229 , any @xmath230 and any nonnegative functions @xmath231 in @xmath7 which are regular harmonic in @xmath232 with respect to @xmath6 and vanish in @xmath233 , we have @xmath234 recall also from @xcite that @xmath235 enjoys the following properties : for every @xmath112 , @xmath236,\ ] ] and there exists @xmath237 such that @xmath238 it follows from and that both the function @xmath90 defined by and the function @xmath235 satisfy a doubling property ; that is , for every constant @xmath112 , there is a constant @xmath103 so that @xmath239.\ ] ] moreover , under condition * ( a ) * , for any given @xmath112 , @xmath109 satisfies all the conditions in @xcite for @xmath240 . thus we have the following two - sided estimates for @xmath8 from @xcite : [ p : new1 ] for any @xmath138 , there exists @xmath241 such that @xmath242\times \r^d \times \r^d.\ ] ] for any @xmath243 , there exists @xmath244 such that for all @xmath245\times \r^d\times \r^d$ ] with @xmath246 @xmath247 is given in the first display on page 1073 of @xcite . by ( * theorem 2.4 ) and , there exist @xmath248 and @xmath249 such that for all @xmath250\times \r^d\times \r^d$ ] with @xmath251 , holds . now we assume @xmath252 . we construct @xmath253 from @xmath6 by removing jumps of size larger than @xmath254 via meyer s construction ( see @xcite ) . let @xmath255 be the transition density of @xmath253 . by ( * ? ? * lemma 3.6 ) and ( * ? ? ? * lemma 3.1(c ) ) we have for every @xmath96 and @xmath256 , @xmath257 where @xmath258 applying ( * ? ? ? * theorem 1.4 ) and its proof to , we get @xmath259 now follows from . the function @xmath260 gives rise to a lvy system for @xmath6 , which describes the jumps of the process @xmath6 : for any non - negative measurable function @xmath50 on @xmath261 with @xmath262 for all @xmath263 and stopping time @xmath182 ( with respect to the filtration of @xmath6 ) , @xmath264 = \e_x \left [ \int_0^t \left ( \int_{\br^d } f(s , x_s , y ) j_x(x_s , y ) dy \right ) ds \right].\ ] ] ( see , for example , ( * ? ? ? * proof of lemma 4.7 ) and ( * ? ? ? * appendix a ) . ) when @xmath265 and @xmath4 is a complete bernstein function satisfying condition * ( a ) * , the following boundary harnack principle on a @xmath158 open subset with explicit decay rate is established in @xcite ( see also @xcite ) . [ t : bhp ] suppose that @xmath6 is a rotationally symmetric lvy process with lvy exponent @xmath160 with @xmath4 being a complete bernstein function satisfying condition * ( a)*. assume that @xmath0 is a ( possibly unbounded ) @xmath2 open set in @xmath161 with characteristics @xmath172 . then there exists @xmath266 such that for @xmath267 $ ] , @xmath268 and any nonnegative function @xmath269 in @xmath7 that is harmonic in @xmath270 with respect to @xmath6 and vanishes continuously on @xmath271 , we have @xmath272 in this section , we assume that @xmath6 is a purely discontinuous rotationally symmetric lvy process with lvy exponent @xmath5 and lvy density @xmath65 satisfying and respectively , where the complete bernstein function @xmath4 satisfies * ( a)*. we will give some preliminary lower bounds on @xmath12 and @xmath8 . we first recall the following from @xcite . recall that @xmath273 was defined in . [ l : exit ] for any positive constant @xmath274 , there exist @xmath275 and @xmath276 such that for all @xmath277 $ ] and @xmath278 , @xmath279 and @xmath280\ , \leq\ , c \,\phi ( r).\ ] ] follows directly from ( * ? ? ? * lemma 2.5 ) . we then have @xmath281 \geq \eps \phi ( r ) \p_z ( \tau_{b(z , r ) } > \eps \ , \phi ( r ) ) > 2^{-1 } e^{-\|\sj_1\|_{\infty}}\eps \phi ( r ) .\ ] ] on the other hand , by the lvy system for @xmath6 in , , and , we have @xmath282\\ & \geq & c_1\e_z \left [ \int_0^{\tau_{b(z , r ) } } \int_{b(z , 3r)\setminus b(z , 2r ) } j(x_s , y ) dy ds\right ] \geq c_2 \frac{|b(z , 3r)\setminus b(z , 2r)|}{r^d \phi ( r ) } \ , \e_z \left [ \tau_{b(z , r ) } \right ] , \end{aligned}\ ] ] which yields @xmath283\leq c \phi ( r)$ ] . [ l : cks2lemma3.1 ] for any positive constants @xmath284 and @xmath254 , there exists @xmath285 such that for all @xmath286 and @xmath287 $ ] , @xmath288 by and , there exists @xmath289 such that for all @xmath290 $ ] , @xmath291 thus it suffices to prove the lemma for @xmath292 . applying the parabolic harnack inequality ( * ? ? ? * theorem 5.2 ) at most @xmath293 $ ] times , we conclude that there exists @xmath294 such that for every @xmath295 , @xmath296 thus @xmath297 this proves the lemma . for the next four results , @xmath0 is an arbitrary nonempty open set and we use the convention that @xmath298 when @xmath299 . [ p : cks2prop3.2 ] let @xmath138 and @xmath300 be constants . there exists @xmath301 such that [ e : lb1 ] p_d(t , x , y ) c ( ^-1(t))^-d for every @xmath302\times d\times d$ ] with @xmath303 . we fix @xmath302\times d\times d$ ] satisfying @xmath304 . note that @xmath305 and that @xmath306 so by the parabolic harnack inequality ( * ? ? ? * theorem 5.2 ) , there exists @xmath307 such that @xmath308 this together with lemma [ l : cks2lemma3.1 ] and yields that @xmath309 where @xmath310 for @xmath311 . [ l : cks2lemma3.3 ] let @xmath138 and @xmath300 be constants . there exists @xmath312 so that @xmath313 for every @xmath302\times d\times d$ ] with @xmath314 and @xmath315 . it follows from lemma [ l : cks2lemma3.1 ] that , starting at @xmath316 , with probability at least @xmath317 the process @xmath6 does not move more than @xmath318 by time @xmath319 . thus , it suffices to show that there exists @xmath320 such that for @xmath302\times d\times d$ ] with @xmath321 and @xmath315 , @xmath322 let @xmath323 , @xmath324 and @xmath325 . it follows from lemma [ l : cks2lemma3.1 ] and that there exists @xmath326 such that @xmath327 \,\ge\,t \ , \p_x \left(\tau^t_x \ge t \right ) \,\ge\ , c_3\,t \qquad \hbox { for all } t\in ( 0 , t].\ ] ] since @xmath328 , by the lvy system of @xmath6 and , @xmath329\nn\\ & \ge & c_4\e_x \left[\int_0^{t\wedge \tau^t_x } \int_{b^t_y } j ( x_s , u ) duds \right ] . \label{e : dew}\end{aligned}\ ] ] we consider two cases separately . \(i ) suppose @xmath330 . since @xmath331 , we have for @xmath332 and @xmath333 , @xmath334 thus by and , @xmath335 \ , | b^t_y| \ , j(2|x - y| ) \,\ge\ , c_5\ , ( \phi^{-1}(t))^d \ , t j(2|x - y|)\end{aligned}\ ] ] for some positive constant @xmath336 . therefore , in view of , the assertion of the lemma holds when @xmath330 . \(ii ) suppose @xmath337 . in this case , for @xmath332 and @xmath333 , @xmath338 thus from and then , @xmath339 \int_{b^t_y } j\left(|x - y| + a\phi^{-1}(t)/4 \right ) \ , du\\ & \ge & c_7 \ , t\ , | b^t_y| \ , j\left(|x - y| + a\phi^{-1}(t)/4 \right ) \\ & \ge & c_8 \ , t(\phi^{-1}(t))^d j\left(|x - y| + a\phi^{-1}(t)/4 \right)\end{aligned}\ ] ] for some constants @xmath340 , @xmath341 . thus we conclude from and that the assertion of the lemma holds for @xmath337 as well . [ p : cks2prop3.4 ] let @xmath182 and @xmath284 be positive constants . there exists @xmath342 such that @xmath343 for every @xmath344\times d\times d$ ] with @xmath345 and @xmath315 . by the semigroup property , proposition [ p : cks2prop3.2 ] , lemma [ l : cks2lemma3.3 ] and , there exist positive constants @xmath346 and @xmath347 such that @xmath348 combining propositions [ p : cks2prop3.2 ] and [ p : cks2prop3.4 ] , we have the following preliminary lower bound for @xmath349 . [ step31 ] let @xmath182 and @xmath284 be positive constants . there exists @xmath342 such that @xmath350 for every @xmath344\times d\times d$ ] with @xmath351 . in particular , proposition [ step31 ] with @xmath352 gives [ t : globalhke ] for any constant @xmath138 , there exists @xmath353 such that for all @xmath245\times \r^d\times \r^d$ ] , @xmath354 in this section , we continue assuming that @xmath6 is a purely discontinuous rotationally symmetric lvy process with lvy exponent @xmath5 and lvy density @xmath65 satisfying and respectively , where the complete bernstein function @xmath4 satisfies the growth condition * ( a)*. throughout this section , @xmath138 is a fixed constant and @xmath0 is a fixed @xmath1-fat open set with characteristics @xmath135 . recall that @xmath355 is defined in definition [ def : ub ] . for @xmath356\times \overline d$ ] , set @xmath357 and define @xmath358 let @xmath359 note that @xmath360 and @xmath361 . see figure [ pic : ts ] . [ lem : etd1r ] for every @xmath138 and @xmath362 , we have that , for @xmath363\times d$ ] , @xmath364 , \end{aligned}\ ] ] where @xmath365 and @xmath366 are the sets defined in and the ( implicit ) comparison constants in depend only on @xmath367 and @xmath4 . without loss of generality we assume @xmath368 . fix @xmath363\times d$ ] , and set @xmath369 . recall that @xmath365 and @xmath366 are the sets defined in . observe that @xmath370 note that by , and ( [ e : levy ] ) , we have @xmath371 \asymp \ , \phi ( r)^{-1 } \ , \e _ { x}\big [ \tau_{u(x , t)}\big ] \ , \asymp \ , t^{-1}\ , \e _ { x}[\tau_{u(x , t ) } ] . \label{eq : cesp } \end{aligned}\ ] ] if @xmath372 , then @xmath373 and so @xmath374 which is greater than or equal to a positive constant depending only on @xmath375 by lemma [ l : cks2lemma3.1 ] . thus we have , in view of , @xmath376 moreover , when @xmath372 , we have @xmath377 and so by lemma [ l : exit ] and , @xmath378 \leq \e_x[\tau_{u(x , t ) } ] \leq \e_x[\tau_{b(x , r ) } ] \le c_3\phi(r ) \leq c_4 t.\ ] ] combining the last two displays with and the fact that @xmath379 we arrive at the assertion of the lemma when @xmath380 . now we assume that @xmath381 . we note that @xmath382 for @xmath383 $ ] , by theorem [ ubhp ] , we have @xmath384 note that when @xmath385 , @xmath386 and so @xmath387 . thus , by , and lemma [ l : exit ] , @xmath388 \geq c_7 \phi ( r)^{-1}\ , \e_0\big[\tau_{b(0,\kappa r/3)}\big ] \ge c_8 \phi ( r)^{-1}\,\phi(\kappa r/3 ) \ge c_9,\end{aligned}\ ] ] where the last inequality is due to . therefore , @xmath389 since @xmath390,$ ] we have by ( [ eq : rsp ] ) , and , @xmath391 .\label{eq : cesp00 } \end{aligned}\ ] ] on the other hand , by the strong markov property of @xmath6 , @xmath392\\ & & + \e_x\left [ \p_{x_{\tau_{u(x , t)}}}\left(\tau_{b(x_{\tau_{u(x , t)}},\kappa r/6)}\le mt \right ) : x_{\tau_{u(x , t ) } } \in b(a'_r(x ) , \kappa r/6)\right]\\ & \le & \e_x\left [ \p_{x_{\tau_{u(x , t)}}}\left(\tau_{v(x , t ) } > mt\right ) ; x_{\tau_{u(x , t ) } } \in b(a'_r(x ) , \kappa r/6)\right]\\ & & + \p_{0}\left(\tau_{b(0,\kappa r/6 ) } \le mt\right ) \p_x\left ( x_{\tau_{u(x , t ) } } \in b(a'_r(x ) , \kappa r/6)\right)\\ & = & \p_x\left ( \tau_{v(x , t ) } > mt , x_{\tau_{u(x , t ) } } \in b(a'_r(x ) , \kappa r/6)\right)\\ & & + \left(1-\p_{0}\left(\tau_{b(0,\kappa r/6)}>mt\right)\right ) \p_x\left ( x_{\tau_{u(x , t ) } } \in b(a'_r(x ) , \kappa r/6)\right)\\ & \le & \p_x\left ( \tau_{v(x , t ) } > mt \right)+\left(1-\p_{0}\left(\tau_{b(0,\kappa r/6)}>mt\right)\right ) \p_x\left ( x_{\tau_{u(x , t ) } } \in b(a'_r(x ) , \kappa r/6)\right ) . \end{aligned}\ ] ] thus @xmath393 since @xmath394 by and lemma [ l : cks2lemma3.1 ] , combining with , we get @xmath395 \leq c_{13}\p_x(x_{\tau_{u(x , t)}}\in b(a'_r(x ) , \kappa r/6 ) ) \leq c_{14 } \p_x(\tau_{v(x , t)}>mt).\end{aligned}\ ] ] this together with and completes the proof of this lemma . [ lemppu100 ] suppose that @xmath396 are open subsets of @xmath161 with @xmath397 and @xmath398 . let @xmath399 . if @xmath400 and @xmath401 , then for all @xmath402 , @xmath403 \right)\left(\sup_{u\in u_1,\ , z\in u_3 } j_x(u , z)\right)\ ] ] and @xmath404 for , see the proof of ( * ? ? ? * lemma 3.4 ) . for , see the proof of ( * ? ? ? * lemma 3.3 ) . * proof of theorem [ thm : oppz](ii ) . * it follows from lemma [ lem : etd1r ] and the semigroup property that it suffices to establish the assertion for @xmath405 . so in the remainder of this proof , we assume that @xmath405 . fix @xmath406 $ ] and set @xmath407 . by , , and theorem [ t : globalhke ] , @xmath408 by the semigroup property , , and lemma [ lem : etd1r ] , when @xmath409 , @xmath410 thus by and , @xmath411 now we assume @xmath412 . let @xmath413 be the set defined in , @xmath414 and @xmath415 then by condition * ( b ) * we have @xmath416 extend the definition of @xmath101 by setting @xmath417 for @xmath418 and @xmath419 . for each fixed @xmath256 and @xmath96 with @xmath420 , one can easily check , in view of , that @xmath421 is a parabolic function in @xmath422\times b(x , 2r)$ ] . so by the parabolic harnack inequality from ( * ? ? ? * theorem 5.2 ) , there is a constant @xmath423 so that for every @xmath406 $ ] , @xmath424 hence we have @xmath425 where in the last inequality , we applied condition * ( b ) * again . if @xmath426 , then @xmath427 . thus if @xmath428 , then @xmath429 thus by , , and theorem [ t : globalhke ] , @xmath430 let @xmath431 . it follows then from , lemmas [ lem : etd1r ] and [ lemppu100 ] that for @xmath412 , @xmath432\right)\nn\\ & \leq & c_6\p_x(\tau_{d}>t)\,p(t/2 , c_5x , c_5y ) . \label{eq : ppun3 } \end{aligned}\ ] ] hence by condition * ( b ) * , , , symmetry , the semigroup property and lemma [ lem : etd1r ] , we conclude that for all @xmath302\times d\times d$ ] , @xmath433 * proof of theorem [ thm : oppz](i ) . * fix @xmath302\times d \times d$ ] and set @xmath434 . let @xmath365 be the set defined in , @xmath355 and @xmath435 be the points defined in definition [ def : ub ] and respectively . note that @xmath436 for @xmath437 , @xmath438 where in the first inequality we used proposition [ step31 ] , in the second inequality we used and respectively in the cases @xmath439 and @xmath440 . on the other hand , for @xmath441 , by lemma [ lemppu100 ] with @xmath442 and @xmath443 , we have @xmath444 since @xmath445 we have by lemma [ l : cks2lemma3.1 ] , and , @xmath446 it follows from condition * ( a ) * that there exists @xmath447 such that @xmath448 for every @xmath449 and @xmath450 $ ] . thus with @xmath451 we have that @xmath452 for every @xmath449 , and hence @xmath453 . thus lemma [ lem : etd1r ] implies that @xmath454 consequently , by , @xmath455 similarly , using the symmetry we also have @xmath456 . hence we have by and , @xmath457 in this section we assume that @xmath0 is a @xmath158 open set in @xmath161 with characteristics @xmath172 and that @xmath6 is a subordinate brownian motion with lvy exponent @xmath160 , where @xmath4 is a complete bernstein function satisfying condition * ( a)*. * proof of theorem [ t : main](ii ) . * without loss of generality we assume @xmath458 . in view of theorem [ thm : oppz](ii ) , it suffices to show that @xmath459\times d.\ ] ] fix @xmath356\times d$ ] and set @xmath460 . by theorem [ t : globalhke ] , we only need to show the theorem for @xmath461 . take @xmath462 such that @xmath463 . let @xmath464 and @xmath465 be the unit inward normal of @xmath466 at the point @xmath467 . put @xmath468 note that @xmath469 . applying the boundary harnack principle ( theorem [ t : bhp ] ) and we get @xmath470 take @xmath471 so that @xmath472 then , by , and , we have @xmath473\\ & \ge & c_3 \e_x [ \tau_{u_1}]\\ & \ge & c_4 \phi(\phi^{-1}(t))^{-1 } \e_x [ \tau_{u_1 } ] = c_4 t^{-1 } \e_x [ \tau_{u_1 } ] . \end{aligned}\ ] ] now by the same argument as that in , we get , @xmath474 \le c_4^{-1 } t \p_x ( x_{\tau_{u_1 } } \in b(x_2 , r ) ) \le c_5 t \p_{x_1}(x_{\tau_{u_1 } } \in b(x_2 , r ) ) \sqrt{\frac { \phi(\delta_d(x ) ) } { t } } \le c_5\sqrt{t\ , \phi(\delta_d(x ) ) } . \label{e : nnw2 } \end{aligned}\ ] ] thus , by and , we have @xmath475 + \p_x \left ( x_{\tau_{u_1 } } \in d\setminus u_1\right ) \le c_6 \sqrt{\frac { \phi(\delta_d(x))}{t}}.\end{aligned}\ ] ] since @xmath476 , follows immediately . let @xmath477 be the euclidean distance between @xmath119 and @xmath478 . it is well - known that any @xmath2 open set @xmath0 with @xmath158-characteristics @xmath172 satisfies both the _ uniform interior ball condition _ and the _ uniform exterior ball condition _ : there exists @xmath479 such that for every @xmath127 with @xmath480 and @xmath481 with @xmath482 , there are @xmath483 so that @xmath484 , @xmath485 and that @xmath486 and @xmath487 for @xmath488 and @xmath489 . in the remainder of this section , we fix such an @xmath490 and set @xmath491 . for any @xmath492 with @xmath493 , let @xmath494 be a point on @xmath478 such that @xmath495 and @xmath496 . [ l : cks2lemma4.6 ] let @xmath497 and @xmath300 . there exists a constant @xmath498 such that for every @xmath499\times d$ ] with @xmath500 and @xmath501 , @xmath502 where @xmath503 . let @xmath504 and assume first that @xmath505 . as in the proof of lemma [ l : cks2lemma3.3 ] , we get that , in this case , using the fact that @xmath506 $ ] , there exist constants @xmath507 , @xmath508 such that for all @xmath509 , we have @xmath510 by taking @xmath511 , this shows that holds for all @xmath300 in the case when @xmath512 . so it suffices to consider the case that @xmath513 . we now show that there is some @xmath514 so that holds for every @xmath515 and @xmath516 . for simplicity , we assume without loss of generality that @xmath517 and let @xmath518 . let @xmath519 . by the strong markov property of @xmath520 at the first exit time @xmath521 from @xmath522 and lemma [ l : cks2lemma3.1 ] , there exists @xmath523 such that @xmath524\right ) \nonumber \\ & \geq & c_3 \ , \p_{x}\left ( \tau_u < at \hbox { and } x_{\tau_u}\in b(0 , 2^{-1 } \kappa_0 \phi^{-1}(t ) ) \right ) \label{e:4.8}.\end{aligned}\ ] ] let @xmath525 and @xmath526 . it follows from the boundary harnack principle ( theorem [ t : bhp ] ) and that there exist @xmath527 , @xmath528 , such that for all @xmath529 $ ] , @xmath530 by , lemma [ l : exit ] and - , we have @xmath531 \\ & \geq & c_6 j(\kappa_0 \phi^{-1}(t ) ) \ , |b(0 , \kappa_0 \phi^{-1}(t ) /2)| \ , \e_x \left [ \tau_{b_1 } \right ] \\ & \geq & \frac{c_7 } { ( \kappa_0 \phi^{-1}(t))^d \phi ( \kappa_0 \phi^{-1}(t ) ) } \ , ( \kappa_0 \phi^{-1}(t))^d \ , \phi ( \kappa_0 \phi^{-1}(t ) ) = c_7.\end{aligned}\ ] ] thus @xmath532 it follows from that there exists @xmath533 such that @xmath534 \le \ , a^{-1 } c_9\ , \sqrt{\frac{\phi(\delta_d(x))}{t}}.\end{aligned}\ ] ] define @xmath535 . we have by and the display above that for @xmath515 , @xmath536 and show that holds for every @xmath515 and for every @xmath127 with @xmath537 . now we deal with the case @xmath538 and @xmath513 . if @xmath539 , we have from for the case of @xmath540 that there exist @xmath541 and @xmath542 such that @xmath543 if @xmath544 ( in this case @xmath545 ) , we get from by taking @xmath546 . the proof of the lemma is now complete . * proof of theorem [ t : main](i)*. by theorem [ thm : oppz](i ) , it suffices to show that @xmath547\times d.\ ] ] assume @xmath356\times d$ ] . since @xmath0 satisfies the uniform interior ball condition with radius @xmath490 and @xmath548 , we can choose a point @xmath549 as follows : if @xmath550 , let @xmath551 so that @xmath552 and @xmath553 for every @xmath554 ) . if @xmath555 , choose @xmath556 so that @xmath557 . note that in this case , @xmath558 and @xmath559 for every @xmath560 . we also define @xmath561 the same way . if @xmath562 , by lemma [ lem : etd1r ] ( with @xmath563 when @xmath564 ) and lemma [ l : cks2lemma4.6 ] ( with @xmath565 ) , @xmath566 if @xmath567 , by lemma [ lem : etd1r ] , proposition [ step31 ] and , @xmath568 in this section , we first give the proofs of theorems [ thm : oppz](iii ) and [ t : main](iii ) . * proof of theorem [ thm : oppz](iii ) . * since @xmath0 is bounded , in view of , the transition semigroup @xmath569 of @xmath80 consists of hilbert - schmidt operators , and hence compact operators , in @xmath570 . so @xmath571 has discrete spectrum @xmath572 , arranged in decreasing order and repeated according to their multiplicity . let @xmath573 be the corresponding eigenfunctions with unit @xmath574-norm ( @xmath575 ) which forms an orthonormal basis for @xmath570 . clearly , for every @xmath576 @xmath577 by using the eigenfunction expansion of @xmath578 we get @xmath579 noting that @xmath580 is increasing and @xmath581 , we have for all @xmath96 , @xmath582 on the other hand , by theorem [ thm : oppz](ii ) , remark [ r:1.1](iii ) and we have that there is a constant @xmath583 so that for every @xmath127 , @xmath584 it now follows from that for every that for every @xmath96 , @xmath585 for @xmath586 and @xmath587 , we have that @xmath588 by theorem [ thm : oppz](ii ) , remark [ r:1.1](iii ) , and we have that there are constants @xmath589 , @xmath311 , so that for every @xmath586 and @xmath587 , @xmath590 by theorem [ thm : oppz](i ) , theorem [ t : globalhke ] , the boundedness of @xmath0 and we have that there are constants @xmath589 , @xmath591 , so that for every @xmath586 and @xmath587 , @xmath592 this combined with establishes theorem [ thm : oppz](iii ) . * proof of theorem [ t : main](iii ) . * by theorem [ t : main](i ) , it suffices to prove the theorem for @xmath593 . by , and the boundedness of @xmath0 , @xmath594 . this and theorem [ thm : oppz](iii ) imply theorem [ t : main](iii ) . in this section , we use theorem [ t : main ] to get sharp two - sided estimates on the green functions of subordinate brownian motions in bounded @xmath2 open sets . we first establish the following two lemmas . the second inequality holds trivially . without loss of generality , we assume @xmath597 . if both @xmath598 and @xmath599 are less than 1 or if both are larger than one , @xmath600 so we only need to consider the case when @xmath601 . note that @xmath602 . if @xmath603 , then by , @xmath604 and so @xmath605 when @xmath606 , by again , @xmath607 and so @xmath608 where the assumption @xmath609 is used in the last inequality . this establishes the first inequality of . by condition * ( a ) * , we have that for every @xmath138 , there exist @xmath610 such that @xmath611 moreover , for every @xmath612 , we have [ e:6.20 ] r(r ) ( r ) r(0 , m ] ( see the paragraph after ( * ? ? ? * lemma 1.3 ) ) . on the other hand , for @xmath618 with @xmath623 , using the change of variable @xmath624 and then applying integration by parts for the first integral below , we have @xmath625 since @xmath626 , by and the fact that @xmath627 is increasing , @xmath628 for some @xmath629 . using and in the second integral in , we get that for @xmath618 with @xmath623 , @xmath630 [ t : gfe ] suppose that @xmath6 is a subordinate brownian motion with lvy exponent @xmath160 with @xmath4 being a complete bernstein function satisfying condition * ( a)*. let @xmath0 be a bounded @xmath158 open subset of @xmath161 with characteristics @xmath172 and @xmath638 , @xmath639 . using the boundedness of @xmath0 , remark [ r:1.1](iii ) and , the results of theorem [ t : main](i)(ii ) can be rewritten as follows : there exists @xmath655 such that for @xmath302\times d\times d$ ] , @xmath656 [ c : gfe ] suppose that @xmath6 is a one - dimensional subordinate brownian motion with lvy exponent @xmath160 with @xmath4 being a complete bernstein function satisfying condition * ( a)*. let @xmath0 be a bounded @xmath158 open subset of @xmath174 with characteristics @xmath172 and @xmath638 , @xmath668 . [ r:6.4 ] let @xmath678 . note that @xmath679 let @xmath680 then @xmath681 the upper and lower matuszewska indices can be computed as @xmath682 thus we have @xmath683 ( see @xcite ) . let @xmath684 , @xmath685 . with the change of variable @xmath686 , we see that is equivalent to [ e:6.9n ] ( r)=_t^-2^r1(t ) t^1/2 dt = 2_r^-1/2^t ds c_1 ( r^-1/2 ) r^1/2 = ( r ) r t^-2 . let @xmath687 . is equivalent to [ e:6.10n ] ( r)= ^_r 1(t ) t^1/2 dt = 2_0^r^-1/2 ds c_2 ( r^-1/2 ) r^1/2=(r ) r t^-2 . in fact , by ( * ? ? ? * corollaries 2.6.2 and 2.6.4 ) , @xmath688 for every @xmath689 if and only if @xmath690 , and @xmath691 for every @xmath689 if and only if @xmath692 . by following the same proof in ( * section 6 ) , one can construct @xmath4 whose upper and lower matuszewska indices are @xmath693 and @xmath694 so that @xmath695 and @xmath696 . for such @xmath4 , neither nor hold . * proof of corollary [ c : gfe ] . * put @xmath697 , where @xmath128 is the diameter of @xmath0 and @xmath616 is the constant in . for notational simplicity , we let @xmath698 and @xmath699 . recall that @xmath700 is defined by . in view of theorem [ t : gfe](iii ) , we only need to consider @xmath701 , which we will assume from now on in this proof . \(ii ) since @xmath704 , by we have that for @xmath701 , @xmath705 thus by theorem [ t : gfe](iii ) , @xmath706 since @xmath707 is increasing in @xmath102 , the above together with theorem [ t : gfe](iii ) implies that @xmath708 we next give an example of a one - dimensional subordinate brownian motion with lvy exponent @xmath709 that satisfies condition * ( a ) * but its associated function @xmath710 satisfies neither condition nor . we can get its explicit green function estimates on bounded @xmath158 open sets by using theorem [ t : gfe](iii ) but not by corollary [ c : gfe ] . [ e : g1 ] by the discussion at the end of the introduction , we know that for any @xmath711 , the function @xmath712 is a complete bernstein function . moreover , @xmath713 when @xmath714 . when @xmath715 $ ] , @xmath716 is an explicit example of a complete bernstein function such that @xmath713 when @xmath714 . let @xmath6 be a one - dimensional subordinate brownian motion with lvy exponent @xmath709 . note that [ e:7.16 ] ( r)=1/ ( 1/r^2 ) r ( ( 1/r ) ) ^-p 0<r1/2 . let @xmath717 be the lambert - w function , that is , @xmath717 is the unique solution to @xmath718 . suppose @xmath719 . then @xmath720 so @xmath721 , which implies that @xmath722 . so in view of , there is a constant @xmath723 so that @xmath724.\ ] ] let @xmath734 . we have by theorem [ t : gfe](iii ) , and - that for @xmath728 , @xmath735 while for @xmath730 , @xmath736 where @xmath737 . it is elementary to check that for @xmath738 and @xmath739 , @xmath740 when @xmath728 and @xmath741 thus we have from the four displays above together with that for @xmath728 , @xmath742 while for @xmath730 , @xmath743 * acknowledgement . * the main results of this paper , in particular theorems [ thm : oppz ] and corollary [ c : main2 ] , were reported at the _ the sixth international conference on stochastic analysis and its applications _ held at bedlewo , poland , from september 9 to 14 , 2012 . at the same meeting , k. bogdan , t. grzywny and m. ryznar announced that they had also obtained the factorization form estimates in terms of surviving probabilities and global heat kernel @xmath744 as in our theorem [ thm : oppz](i)(ii ) for the dirichlet heat kernels of a similar class of purely discontinuous subordinate brownian motions considered in this paper but only for bounded @xmath158 open sets . chen , p. kim and t. kumagai . on heat kernel estimates and parabolic harnack inequality for jump processes on metric measure spaces . _ acta mathematica sinica , english series * 25 * _ ( 2009 ) , 10671086 .

we use high resolution images obtained with the advanced camera for surveys on board the hubble space telescope to determine morphology , nuclear luminosity and structural parameters of the spheroidal component for a sample of 20 seyfert galaxies at . we combine these measurements with spectroscopic information from the keck telescope ( paper i ) to determine the black hole mass - spheroid luminosity relation ( m-l ) , the fundamental plane ( fp ) of the host galaxies and the black hole mass - spheroid velocity dispersion relation ( m- ) . the fp is consistent with that of inactive spheroids at comparable redshifts . assuming pure luminosity evolution , we find that the host spheroids had smaller luminosity and stellar velocity dispersion than today for a fixed m . the offsets correspond to ( ) and ( ) , respectively for the m-l and m- relations ( the double error bars indicate random and systematic uncertainties ) . a detailed analysis of known systematic errors and selection effects shows that they can not account for the observed offset . we conclude that the data are inconsistent with pure luminosity evolution and the existence of universal and tight scaling relations . in order to obey the three local scaling relations by assuming no significant black hole growth the distant spheroids have to grow their stellar mass by approximately 60% ( ) in the next 4 billion years , while preserving their size and holding their stellar mass to light ratio approximately constant . the measured evolution can be expressed as , consistent with black holes of a few m completing their growth before their host galaxies . based on the disturbed morphologies of a fraction of the sample ( 6/20 ) we suggest collisional mergers with disk - dominated systems as the physical mechanism driving the evolution .

0706.0519

in the local universe , most galactic nuclei harbor a supermassive black hole ( e.g. , * ? ? ? * ) . the mass of the black hole correlates with global properties of the host , such as the velocity dispersion and luminosity of the spheroidal ( or bulge ) component , indicating a connection between nuclear activity and galaxy formation and evolution @xcite . understanding the origin of this relation is a major challenge for cosmological models and is believed to hold the key to solving several astrophysical problems such as the role of feedback from nuclear activity in suppressing star formation in massive galaxies ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . in the standard cosmological scenario , spheroids grow by mergers of smaller galaxies while black holes grow by accreting surrounding matter . depending on the relative timing of the two processes , the scaling relations between black hole mass and spheroid luminosity and velocity dispersion ( hereafter m@xmath1-l or m@xmath1-l@xmath2 to specify the blue band and m@xmath1-@xmath3 relations , respectively ) could also evolve with cosmic time . for example if the spheroid evolved passively due to aging of stellar evolutions , changing l but not @xmath3 , while the black hole grows by accretion we would expect evolution in both the m - l and the m-@xmath3 relations , with the latter evolving more slowly than the former . by contrast , if black holes completed their growth first and the dominant mode of growth now is the transformation of stellar disks into spheroids , the two relations would evolve in the opposite sense @xcite . the tightness of the local relationships has been interpreted as evidence for feedback that synchronizes the relative growth . in this context , the tight relationships would be naturally reproduced if galaxies and black holes moved up the m-@xmath3 relation during merging events , so that the correlation would appear not to evolve with redshift @xcite . detailed theoretical predictions are extremely difficult due to the daunting range of scales ranging from the mpc scale halo to the pc scale dynamical sphere of influence of the black hole , to the @xmath13pc scale of the accretion disk and physical processes involved radiative transfer , heating and cooling , accretion , just to name a few . in spite of the challenge , numerous groups have been able to develop models that are increasingly more successful at reproducing a variety of observations ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? however , to this date , the evolution of scaling laws remains very uncertain and sensitive to the schemes and approximations adopted to deal with the complex physics ( compare for example the recent works by @xcite and @xcite ) . accurate empirical measurements are needed to discriminate between scenarios , and provide input on the relative importance of various physical phenomena as well as on the accuracy of approximations . with this goal in mind , a number of groups have started observational programs to trace the evolution of scaling laws over cosmic time @xcite . however , observers face two fundamental limitations . on the one hand , since the sphere of influence of supermassive black holes is too small to be resolved at cosmological distances with present technology black hole mass estimates can only be obtained for active galaxies . typically , this involves broad line agn and the dynamics of the broad line region , with consequent loss of information about the host galaxy which is swamped by nuclear light . on the other hand at a deeper level the evolution of the scaling laws , depends on the interplay of at least four physical processes : i ) black hole accretion ; ii ) evolution of the stellar populations ; iii ) dynamical evolution of the spheroid , e.g. through mergers ; iv ) black hole feedback on star formation . for this reason , even when a scaling law can be measured as a function of redshift , the interpretation is often times ambiguous . for example , as discussed by @xcite , how much of the evolution of the m - l relation is due to evolution in the spheroid mass and how much is due to evolution of the stellar populations ? in this paper we address the two fundamental limitations by adopting the following strategy . first , as in our pilot study @xcite and in the first paper of this series ( * ? ? ? * hereafter paper i ) , we focus on relatively moderate redshift ( @xmath14 ) and luminosity ( monochromatic luminosity at 5100 @xmath15 erg s@xmath16 ) agn . although the lookback time is considerably smaller than that of the most distant quasars studied by other groups , this choice allows us to determine the host galaxy properties with considerably smaller uncertainties . second , we concentrate on a relatively small sample and measure several independent properties of the host galaxies . paper i reported stellar velocity dispersion measurements based on keck spectroscopy . this paper presents host spheroid luminosity and size measurements based on hst - acs imaging . we use this combined dataset to study at the same time and for the same sample the m@xmath1-l and m-@xmath3 relations and the fundamental plane of host spheroids . the combination of these diagnostics which can be thought as projections of a more fundamental manifold ( e.g. , * ? ? ? * ) allows us to disentangle stellar mass growth , stellar population evolution and black hole growth , and helps to identify the processes at work ( see also * ? ? ? * ; * ? ? ? * ; * ? ? ? the paper is organized as follows . in [ sec : data ] , we summarize the properties of the sample , describe the hst - acs observations , and present our surface photometry . we also derive black hole masses using nuclear luminosities as measured from hst images and the new calibration of the broad line region size - nuclear luminosity relation @xcite , together with h@xmath17 line widths from paper i. to construct a suitable local comparison sample , we use sloan digital sky survey images to derive new measurements of the spheroid luminosity of a sample of local seyferts with m@xmath1 available from reverberation mapping @xcite . [ sec : results ] describes our main results , i.e. the fundamental plane , m@xmath1-l@xmath18 , and m@xmath1-@xmath3 relation of the distant seyferts . detailed estimates of systematic errors and selection effects are given in [ sec : sys ] . [ sec : mani ] analyzes the three scaling relations under the assumption that the distant seyferts will evolve to match the local relations and derives constraints on the evolution of stellar mass , size and stellar populations of the host spheroids as a function of black hole growth . the results are discussed and compared with the literature in [ sec : disc ] , and summarized in [ sec : sum ] . throughout this paper magnitudes are given in the ab scale . we assume a concordance cosmology with matter and dark energy density @xmath19 , @xmath20 , and hubble constant h@xmath21=70 kms@xmath16mpc@xmath16 . the sample discussed in this paper coincides with that discussed in paper i of this series , with minor exceptions . two of the objects ( s16 and s31 ) studied in this paper do not have keck spectroscopy due to the unfavorable weather conditions at keck when follow - up observations were planned . one of the objects with deep keck spectroscopy ( s28 ) has no acs images because it was observed after the schedule of the hubble program had been completed . as in our previous paper , except for s99 that was selected before sdss data release 1 ( dr1 ) @xcite , the objects were selected from the sdss archive , based on the following criteria : i ) spatially resolved in the sdss images ; ii ) redshift between 0.35 and 0.37 ; iii ) h@xmath17 equivalent width and gaussian width greater than 5 in the rest frame . most of the objects were selected from dr-1 , and the remaining ones ( s21,s23 , s26 , s27 ) were selected from dr-2 . after the initial selection based on these criteria , the sdss spectra of the objects observable from keck were visually inspected by two of us ( tt and mam ) to check line identification . objects showing strong fe ii nuclear emission ( the main obstacle to velocity dispersion measurement ) were eliminated from the sample . for example , for the run of september 2003 , out of 33 observable objects , 8 were rejected on the basis of visual inspection ( the mean and rms @xmath22 magnitude of the selected and rejected samples are consistent , 18.76 and 0.42 vs. 18.98 and 0.63 , respectively ) . of the remaining 25 objects , 12 were observed during the run , based on observability . the fraction of rejected is similar for following runs , although the total number of objects increases significantly in subsequent data releases . no color selection was imposed , although a post - facto analysis shows that the sdss colors of the sample are intermediate between those of a quasar and those of an old stellar population , consistent with the comparable fraction of nuclear and stellar light inferred from the hst images in the rest of this paper . coordinates , redshifts and other basic properties of the sample are given in table [ tab : sample ] . the sample was observed using the wide field camera of the acs on board hst between august and december of 2004 as part of general observer program 10216 ( pi : treu ) . each object was observed for one orbit , split in four exposures dithered by semi - integer pixel shifts to ensure cosmic ray and defect removal as well as to improve sampling of the point spread function ( psf ) . the total exposure times range between 2148s and 2360s . filter f775w ( @xmath23 ) was chosen so as to sample the region corresponding to the rest frame 5100 in order to estimate the size of the broad line region , and to avoid contamination from the broad emission lines h@xmath17 and h@xmath24 . this filter choice also provides spheroid luminosity redward of the 4000 break , and it is thus appropriate to infer spheroid structural parameters and rest frame b and v luminosity with minimal uncertainty due to filter transformations . postage stamp images of the targets are shown in figure [ fig : montage ] . the following observables are needed to investigate our science questions : i ) spheroid luminosity ; ii ) nuclear luminosity , to estimate the size of broad line region , and hence the black hole mass ; iii ) effective radius and effective surface brightness of the spheroid , to construct the fundamental plane . this section presents the measurements with an extensive discussion of systematic and random uncertainties . the images were reduced using multidrizzle to remove cosmic rays and defects , correct for distortion and improve sampling of the psf . based on monte carlo simulations we adopted a final pixel size of @xmath25 , which provides the best compromise between sampling of the psf , signal to noise ratio of the individual pixels , and noise correlation . surface photometry was derived using the galfit software @xcite to fit two - dimensional models to the data . optimizing a large number of non - linear parameters is a notoriously difficult problem , so we proceeded in steps , adding one component at a time . first , we fitted a point source to determine the center of the system , which is assumed common to all components . we then added a spheroid modeled as a @xcite @xmath26 profile , and/or an exponential disk , if required by the @xmath27 statistic . if needed , an additional component described by a @xcite profile with index n=0.5 was added to model the bar , identified as an elongated feature with a change in position angle with respect to the overall surface brightness distribution ( see , e.g. , s01 in figure 1 ) . bars were found to be present in 7/20 cases and contribute between 4% and 16% of the total light . extensive tests were conducted to ensure that the solution corresponded to the true global minimum of the @xmath27 over the parameter space . bright unsaturated stars in the field of the images were use to create a library of 43 point spread functions . each galaxy was fitted with all the psfs in the library to find the best fitting one as well as the best fitting parameters . in three cases ( s11 , s31 , s99 ) , a prominent dust lane prevented us from determining accurate surface photometry . in two cases ( s03 and s10 ) no stable solution with a sizable spheroid could be found , as the spheroidal component tended to become vanishingly small in size . therefore we fixed the spheroid half light radius to 2.5 pixels ( 0.1 arcsec ) , the minimum size that could be resolved given our psf and we considered the measured luminosity as an upper limit . similarly , for three objects ( s12 , s21 , s23 ) the measured spheroid half light radius is very close to our resolution limit ( i.e. @xmath283pixels ) , and therefore we also consider their luminosity as an upper limits . extensive testing shows that the upper limit to the spheroid luminosity is robust with respect to the choice of the fixed half light radius , and of the psf . twelve out of twenty cases provided stable best fitting models and hence robust measurements . for illustration purposes one - dimensional surface brightness profiles ( obtained with iraf task ellipse ) are shown in figures [ fig : profiles1 ] and [ fig : profiles2 ] . the formal statistical uncertainties on the spheroid luminosity are typically 0.05 mags . the rms scatter of the parameters for all the statistically acceptable psfs was adopted as the systematic uncertainty due to psf modeling . we estimated the total uncertainty including systematic errors by varying systematically all the fitting parameters and psf , finding that the results are typically stable within 0.2 mags . spheroid luminosity was then transformed into rest frame b - band luminosity by correcting for galactic extinction and applying k - color corrections , calculated as described in @xcite . errors on extinction and k - color corrections are a few hundredths of a magnitude . conservatively , we adopt 0.5 mags ( i.e. 0.2 dex ) as the total uncertainty on spheroid luminosity . this uncertainty is smaller than the estimated uncertainty on black hole mass from single epoch measurements @xmath290.4 dex and therefore sufficient to meet our goal of constructing the m - l@xmath2 relationship . observed and intrinsic spheroid luminosities are given in table [ tab : meas ] . another source of systematic uncertainty is the choice of the surface brightness profile used to describe the spheroid . although @xmath26 profiles are the traditional and widely used choice in the analysis of agn host galaxies ( e.g. * ? ? ? * ) , detailed studies of nearby bulges show that sersic profiles with lower sersic index ( typically 2 - 3 ) can provide a better fit . to estimate systematic errors on the spheroid luminosity associated with our choice of profile , we repeated our analysis using a sersic profile with index 2 and 3 ( figure [ fig : sersic ] ) . as expected , we find that the best fit spheroid luminosity decreases with the sersic index . quantitatively , the spheroid luminosity decreases on average by 0.15 ( 0.33 ) mags when changing the sersic index from 4 to 3 ( 2 ) . similarly , the point source luminosity increases on average by 0.02 ( 0.05 ) magnitudes and the effective radius of the spheroids decreases by 9% ( 11% ) when changing the sersic index from 4 to 3 ( 2 ) . we thus conclude that adopting a @xmath26 profile provides a conservative measurement of the maximum luminosity of the spheroid and of the minimum nuclear luminosity ( and hence m@xmath1 ) . we will come back to this point in the discussion of our results in section [ ssec : sysl ] . other measurements of interest are total magnitude , nuclear luminosity at rest frame 5100 ( l@xmath30 ) , the combination of effective radius ( r@xmath31 ) and effective surface brightness ( sb@xmath31 ) that enters the fundamental plane fp@xmath32 - 0.32 sb@xmath31 ( see [ ssec : fp ] ) and the fraction of nuclear light ( f@xmath33 ) . the total uncertainties , conservatively estimated as for the spheroid luminosity , are 20% on l@xmath30 and on f@xmath33 , and 0.1 on fp@xmath34 . the relevant quantities are also listed in table [ tab : meas ] . as in paper i of this series , black hole masses were derived from the width of h@xmath17 and the observed nuclear luminosity at 5100 , using the so - called virial method or empirically calibrated photo - ionization method @xcite . briefly summarized , the method assumes that the kinematics of the broad line region trace the gravitational field of the central black hole . the width of the line provides the velocity scale , while the nuclear continuum luminosity provides the size via the empirical correlation observed for the local reverberation mapped sample @xcite . the virial coefficient is obtained by requiring that local agn hosts and quiescent galaxies obey the same m-@xmath3 relation @xcite . this method has been shown to provide an estimate of the black hole mass within a factor of 2 - 3 . for this paper we will assume as in the previous paper of the series a nominal uncertainty of 0.4 dex on black hole mass obtained with this method ( see peterson 2007 , for a recent discussion of the method ) . however , since we allow for unknown intrinsic scatter in the scaling relations when fitting for the intercept ( [ sec : results ] ) , adopting a larger random uncertainty on each individual black hole mass estimate ( e.g. 0.6 dex , * ? ? ? * ) has a negligible impact in terms of overall uncertainty ( [ ssec : random ] ) . as in paper i , we adopt the second moment of the broad component of h@xmath17 as our velocity scale since this is more robustly measured than the alternative fwhm ( paper i ; see also peterson et al . 2004 ; collin et al . the velocity scale is measured from keck spectra when available and from sdss spectra in the few cases when keck spectra are not available ( see paper i for details ) . as shown in paper i , our methodology gives unbiased black hole masses when applied to the local sample of calibrators . a major improvement with respect to paper i is that we can use high resolution hubble images to measure the nuclear luminosity and infer broad line region size . in practice we use the following equation to estimate black hole masses , obtained combining the most recent calibrations of the size luminosity relation and virial coefficients @xcite . @xmath35 where @xmath36 is the second moment of the broad h@xmath17 line profile . we note that for the sample in common with paper i , the use of nuclear luminosities and this new calibration implies black hole masses smaller by 0.09 dex on average , consistent with our estimate of the magnitude of this systematic error in paper i. black hole masses are listed in table [ tab : meas ] . to study evolutionary effects we consider two comparison samples to define the local m - l@xmath2 relation . first we consider the sample of quiescent galaxies collected by @xcite . the sample consists of galaxies with black hole mass determined from spatially resolved kinematics and with spheroid luminosity determined from two - dimensional surface photometry . the second sample is that of local seyfert galaxies with black hole mass measured via reverberation mapping discussed in @xcite and @xcite . to ensure self - consistent determination of spheroid luminosity for the local and distant seyfert sample we measured the parameters of the local sample in exactly the same way as we do for distant galaxies . we searched the sdss archive for g-band images of local seyferts , which provide a very good match to our distant galaxies in terms of resolution ( the seeing is typically 10 times the hst psf , but the angular size distance is typically 10 times smaller ) and rest frame wavelength . sdss images for nine seyferts were found in the archive . the resulting spheroid luminosities for the local seyfert samples are listed in table [ tab : local ] , together with redshifts , black hole masses and velocity dispersions from the literature . the two samples are complementary in terms of vices and virtues . the first sample is larger in size and black hole masses , and spheroid luminosities are most robustly measured : i ) m@xmath1 is obtained from spatially resolved kinematics ; ii ) the determination of spheroid luminosity does not suffer from poor resolution or the presence of a prominent point source in the center . however , the second sample is the most appropriate for a direct comparison for a variety of reasons : i ) the virial coefficient is unknown , but assuming that it is not evolving with redshift , the comparison between seyfert samples is independent of its numerical value ; ii ) if the local m - l relation is not universal , seyferts may define a different m - l relation than quiescent early - type galaxies perhaps because they are at a different evolutionary stage . thus , we conclude that at this stage the more conservative approach is to consider both samples for the following analysis and consider the uncertainty on the local relation as an additional source of systematic errors . more comprehensive studies of the local m - l relation are needed to eliminate this source of uncertainty . this is left for future work . this section presents the main results of this paper . first in [ ssec : fp ] we discuss the fundamental plane of the host galaxies in comparison with that of normal quiescent galaxies , for the subsample of objects that have both @xmath3 and structural parameters . in [ ssec : ml ] we present the m@xmath1-l@xmath2 relation for distant seyfert galaxies . having established that the fundamental plane of the host galaxies is indistinguishable from that of quiescent galaxies , we adopt the luminosity evolution inferred from fp analysis @xcite to compare with the local m@xmath1-l@xmath2 relation . in [ ssec : ms ] we revisit the m@xmath1-@xmath3 relation derived in paper i , using the new improved black hole mass estimates based on nuclear luminosities determined from hst imaging . in the local universe early - type galaxies obey the fundamental plane , i.e. an empirical correlation between effective radius , central velocity dispersion and effective surface brightness of the form : @xmath37 with @xmath38 , @xmath39 and @xmath40 for the coma cluster in the b(ab ) band . the evolution of the fp out to redshift @xmath41 is well established for quiescent early - type galaxies , for agn hosts @xcite , and for the spheroidal component of spiral galaxies with bulge - to - total luminosity ratio greater than 0.2 ( macarthur et al . 2007 , in preparation ) . in figure [ fig : fpb ] we plot the fp parameters of the subset of distant seyferts for which structural parameters and stellar velocity dispersion is available , together with a comparable sample of quiescent galaxies taken from @xcite . the good agreement between the fp of quiescent and active galaxies gives us confidence that our surface photometry is not systematically biased by the presence of a nuclear point source . the offset with respect to the local relation ( solid line ) is normally interpreted as evolution of the stellar populations . under the assumption of pure luminosity evolution , luminosity evolves with redshift such that the expected value at @xmath8 is given by @xmath42 @xcite . applying the same assumption to the sample of distant seyferts at @xmath0 implies that : @xmath43 , where random and systematic errors have been added in quadrature for simplicity . in order to compare with the local relations we need to account for the fact that the luminosity of a stellar population decreases as it ages . as our first working hypothesis , in this section we will present our results assuming that stellar populations evolved as inferred from the fp studies under a pure evolution scenario , and use the variable l@xmath44 obtained using equation [ eq : lbev ] . a more general discussion will be given in [ sec : mani ] . the relation between black hole mass and host spheroid luminosity for our points as well as for local comparison samples is shown in figure [ fig : bhl ] . the main result is that the seyfert galaxies at @xmath0 and those at @xmath8 cover approximately the same range in spheroid luminosity , but the average black hole mass is higher for the distant seyferts . the mismatch is exacerbated if one considers that five of the distant seyferts measurements are upper limits ( while only one of the local seyferts is an upper limit ) . conservatively we will generally consider the measured offset as the best estimate of the offset although it should be kept in mind that it is most likely a lower limit . quantitatively , the offset with respect to the fiducial local relationship ( solid line ) corresponds to @xmath45 ( @xmath46 ) , considering the intrinsic scatter of the relation to be a free parameter and then marginalizing over it . listed systematic errors are derived as discussed in section [ sec : sys ] . if measured with respect to the local seyferts ( magenta line ) , the best estimate of the offset changes to @xmath47 this is visualized in figure [ fig : bhlhisto ] where we plot the distribution of residuals with respect to the local fiducial relation for the distant and local samples . we conclude that pure luminosity evolution is inconsistent with the observations , if we required that at @xmath8 all galaxies obey the m@xmath1-l@xmath2 relation . we can not solve this inconsistency by invoking different luminosity evolution . the stellar populations would be required to become brighter with time in order to reconcile the data with the local relationship . this can be ruled out on physical grounds . therefore , taking this result at face value , we have to conclude that either not all spheroids obey the m - l relations , or that a significant amount of new stars have been added to the spheroids since @xmath0 . the interpretation of this result will be discussed in sections [ sec : mani ] and [ sec : disc ] , after we conclude presenting the evidence and discussing systematic errors and selection effects in the remainder of this section . figure [ fig : ms ] shows the m@xmath1-@xmath3 relation for the distant seyferts as well as the local comparison samples . the samples and symbols are the same as in figure [ fig : bhl ] with few minor exceptions : i ) only the 14 distant seyferts with available stellar velocity dispersion ( from paper i ) are plotted , including s28 and s99 for which hst photometry is not available . ii ) local relations are from @xcite and @xcite ; iii ) local seyferts obey the same relation as quiescent galaxies by construction , as this is the constraint used to derive the virial coefficient @xcite , so there is no need to show a separate local relation for seyferts . the only substantial difference with respect to figure 8 in paper i is that black hole masses have been recalculated based on hst photometry . the m@xmath1-@xmath3 relation shows the same basic result as the m@xmath1-l@xmath2 relation . at fixed host galaxy properties , distant seyferts appear to host larger black hole mass than local ones . quantitatively , the offset corresponds to @xmath6 ( i.e. @xmath48 ) . note that offset is slightly reduced ( 0.08 in @xmath49 ) with respect to paper i and the systematic error is smaller due to the improved estimate of black hole masses that avoid stellar contamination . as shown in figure [ fig : ms ] and discussed in paper i , adopting the local relation from @xcite or @xcite changes the offset by much less than the error bars because the two local relations are very well matched in the range of black hole mass and stellar velocity dispersion covered by our sample . this section is devoted to understanding and estimating systematic errors . first , in [ ssec : sysmbh ] and [ ssec : sysl ] , we list potential sources of systematic error and estimate as accurately as possible the associated uncertainty . for brevity we will not repeat the analysis of systematics already discussed in paper i , unless when substantial changes / improvements are introduced . [ ssec : random ] derives the uncertainty on our estimate of the random errors due to our assumed accuracy of single - epoch black hole mass determinations . then , in [ ssec : sel ] , we introduce a monte carlo scheme to simulate unknown selection effects and estimate potential biases . [ ssec : syssum ] gives a very short summary of this section for the impatient reader . understanding the m@xmath1 estimate is clearly important . while velocity dispersion and spheroid photometry are measured independently and present mutually consistent results ( as supported by the fp relation , which does not involve m@xmath1 ) , only one measurement of m@xmath1 is available and that determines the evolution of both relationships . our measurement of m@xmath1 depends on two observables ( velocity scale and nuclear luminosity ) , on the empirically calibrated relationship between nuclear luminosity and broad line region size and on the choice of the virial coefficient . we now go through each source of error and estimate the systematic error that they may introduce for a sample of our size . in paper i we demonstrated that our measurement of the velocity scale from the second moment of h@xmath17 in single epoch spectra is unbiased . this means that when applied to the sample of local calibrators our estimator yields the same m@xmath1 as obtained by the original studies based on the second moment of h@xmath17 measured from the variable part of the spectra . recently , @xcite suggested that single epoch spectra may overestimate the line width by @xmath2920% , corresponding to an error of @xmath290.15 dex on black hole mass , if the same virial coefficient is assumed . this effect is likely to depend on the setup of the experiment and on the variability pattern . however , to be conservative , we will consider this as a global uncertainty to the zero point of our measured black hole mass . @xcite studied the time variability of the width of h@xmath17 for a subset of our objects . they found that the r.m.s . variations of the second moment of h@xmath17 are less than 14% ( i.e. @xmath50 dex in m@xmath1 ) . for our sample of 14 - 17 objects this has therefore negligible effect on the average with respect to the aforementioned source of error . our best estimate of overall systematic uncertainty due to line width measurement is 0.15 dex . to check for systematic effects due to systematic change of the line shape with redshift or sample properties such as eddington ratio ( e.g. , * ? ? ? * ) we also computed black hole masses using the full width half maximum instead of the second moment of the line as velocity scale . after removing the narrow component of h@xmath17 using [ ] as a template as described in @xcite and @xcite , we measured the fwhm ( mcgill et al . 2007 , in preparation ) and derived black hole mass estimates using eq . 1 in @xcite . this alternative estimate of black hole mass ( m@xmath51 ) agrees very well with that adopted in this paper , with an average offset of @xmath52 dex , i.e. consistent with no offset . the new calibration of the size - luminosity relation @xcite , and our hst based nuclear luminosities are a substantial improvement with respect to paper i , allowing us to avoid issues related to host galaxy contamination . the dominant source of error is now the intrinsic scatter of the size luminosity relation ( 30% ) corresponding to @xmath50 dex in m@xmath1 per individual object . for our sample size , this translates into a negligible source of error in the mean . as far as the overall calibration of the relation is concerned , for the luminosity range of our sample , the 68% range on the slope and intercept of the relationship translate into an overall uncertainty in size of order 0.02 dex . we thus conclude that the uncertainty due to the ( known ) scatter of the size luminosity relation is negligible . the final source of error is related to the virial coefficient needed to convert size and velocity scales into mass and thus to the zero point of the local relation . as discussed in detail in paper i , we adopt the virial coefficient determined by @xcite requiring that the m-@xmath3 relation be the same for local quiescent and active galaxies . the uncertainty on the average of the virial coefficient is 34% , thus corresponding to an uncertainty of 0.13 dex on the calibration of the black hole mass . however as discussed in paper i assuming that the virial coefficient does not evolve with redshift , this factor cancels out between the local and distant samples and therefore is irrelevant . incidentally , we note that our results can also be interpreted as cosmic evolution of the virial coefficient , if one is willing to drop this assumption . we consider this an unlikely explanation , as it would require the geometry or kinematics of agn to evolve with cosmic time , but unfortunately it will not be possible to discard it until direct measurements of the virial coefficient can be obtained in some other way . the situation is slightly more complex for the m@xmath1-l relation . once the virial coefficient is fixed there is no more freedom . therefore the difference between the intercept of the relation for local active and quiescent ( 0.23 dex in m@xmath1 ) is an additional source of uncertainty . larger samples of galaxies with well determined black hole mass and spheroid luminosity are needed to overcome this limitation . for the purpose of this paper we consider the difference as an additional source of uncertainty . to produce a single measurement , we weight the two samples equally ( for reasons discussed in 3 ) and take the average and semi - difference of the two intercepts as best estimate of the local average and systematic uncertainty on the zero point ( @xmath53 dex ) . the main sources of uncertainty that could affect the spheroid luminosity of the sample as a whole are systematic errors in k - corrections , in the adopted luminosity evolution , and in the choice of the surface brightness profile for the bulge . the uncertainty on the k - correction is negligible , at most 0.02 dex , as estimated using a range of stellar population models to compute the transformation from observed f775w to rest frame b. the luminosity evolution as measured from the fundamental plane @xcite carries an overall uncertainty of 0.03 dex . even assuming that this translates to a shift of the whole sample , this is still a negligible source of error . as discussed in section [ sssec : surf ] adopting a sersic profile instead of a @xmath26 profile tends to systematically reduce the bulge luminosity while increasing the point source luminosity , and therefore black hole mass , thus moving points further away from the local m - l relation of quiescent galaxies . in contrast , since we used the same identical technique for the local seyferts , adopting a sersic profile would not alter the observed offset . we conclude that the measured offset is not overestimated by an appreciable amount due to known systematic errors on spheroid luminosity determination . finally , the uncertainty of black hole mass estimates from single epoch data is assumed to be 0.4 dex @xcite . this represents all sources of random error that contribute to the scatter in eq . [ eq : mbh ] . after quantifying sources of systematic errors on the mean of eq [ eq : mbh ] , we now consider whether the random error associated with our measurement of the offset of the scaling relations could be underestimated . to test this , we repeat the analysis assuming that the random error on each individual m@xmath1 estimate is 0.6 dex @xcite . the inferred random error on @xmath54 increases by 0.03 dex , e.g. from 0.14 dex to 0.17 dex for the m@xmath1-l relation . note that our analysis includes a nuisance parameter to account for the unknown intrinsic scatter of scaling relations and therefore the estimated errors are slightly different than what would be naively derived considering only errors on the y axis ( 0.4 / @xmath55 = 0.4/4 = 0.1 dex vs 0.6/@xmath55=0.15 ) . this is a negligible error on the error , considering that the systematic term is dominant . the total error , defined as the quadratic sum of the random and systematic uncertainties , would only change by 0.01 dex . in this section we estimate possible bias due to selection effects . our galaxies are selected based on their nuclear properties ( having a broad line agn ) . therefore , intrinsic scatter and observational errors in the scaling relations could conceivably lead us to favor large black hole masses and therefore overestimate evolution . our present sample covers approximately a decade in black hole mass ( @xmath56 ) . the upper limit is naturally expected because of the steep drop of the black hole mass function . is the lower limit in mass a result of some unknown selection effect ? our nominal selection limits on line width and flux are small enough that we would have been able to select objects with black hole masses as low as 10@xmath57 m@xmath12 , as verified by running our measurement tools on the entire sdss - dr5 spectroscopic database at this redshift . however the objects with m@xmath1 well below 10@xmath58 are a small fraction of the total . this maybe be due to an intrinsic drop in the black hole mass function , to a selection effect in the sdss spectroscopic sample or to a combination of the two . the decline in the number of objects at @xmath0 with m@xmath1 well below 10@xmath59 is also seen in figure 1 of @xcite , where most of the points lie above the group identified as having m@xmath1 between 10@xmath60 and 10@xmath61 m@xmath12 . estimating the black hole mass function is beyond the scope of this work ( e.g. , * ; * ? ? ? however , to understand the implications of this selection effect on our measurement , we can take a conservative approach and model this drop as a sharp selection in m@xmath1 . in the following discussion we assume for reference that our implicit selection function @xmath62 ( the results are unchanged if the limit is chosen to be 7.8 or 8.0 ) . we then use monte carlo simulations to determine the amount of bias on the offset of the intercept of the scaling relations introduced by the implicit selection process . the scaling relations are populated according to the velocity dispersion function and spheroid luminosity function determined by @xcite and @xcite respectively , and errors on both axis are taken into account . we assume that the intrinsic scatter of the relation is smaller than the measurement errors ( 0.4 - 0.5 dex ) which is consistent with local estimates . note that in this scheme all the selection procedure is modeled a single hard threshold in black hole mass and we only work in the scaling relation variables . this simple scheme allows us to bypass all the assumptions that would be needed to simulate from first principles the observational selection effects which are a complex combination of total flux , broad line flux , broad line width ( at both ends ) , bulge luminosity and stellar velocity dispersion . the results of the simulations are shown in figure [ fig : bias ] . the curves show the recovered ( `` measured '' ) offset as a function of the input offset @xmath63 m@xmath1 . the upper panel shows the results for the m@xmath1-@xmath3 relation , the lower panel shows the results for the m@xmath1-l relation . the bias is almost negligible for the m@xmath1-@xmath3 relation , while it is at most 0.1 dex for the m@xmath1-l relation . the difference between the two relations is due to : i ) the smaller errors on velocity dispersion than those on spheroid luminosity ; ii ) the different behavior of the two distributions at the faint ( low velocity ) end . the velocity dispersion function peaks in the range covered by our sample , while the luminosity function of bulges extends to much smaller luminosities . as a softer cutoff would imply smaller bias , we conclude that selection effects are responsible for at most 0.1 dex of the observed offset of the m@xmath1-l relation , and that this source of bias is negligible for the m@xmath1-@xmath3 relation . in conclusion , the total systematic error on the evolution of m@xmath1-@xmath3 relation is 0.21 dex in m@xmath1 ( i.e. slightly reduced with respect to paper i due to the improved nuclear luminosity estimates ) , and 0.19 on the m@xmath1-l relation , dominated by the uncertainty on the black hole mass and on the uncertainty on the local relation . although we caution the reader to keep in mind all the caveats discussed above , we now condense all the information discussed in this section in seven numbers . our best estimates of the offset of the relations without accounting for selection effects are : * m@xmath1-l : @xmath64 * m@xmath1-@xmath3 : @xmath7 unknown selection effects could remove as much as 0.1 dex to the offset of the m@xmath1-l relation . in the previous sections we concluded that pure luminosity evolution is inconsistent with the observations and the requirement that the distant seyferts lie on the local relations . in this section , we explore the evolutionary constraints that can be obtained by examining the evolution of black holes and spheroids in a four dimensional parameter space with axes given by luminosity , velocity dispersion , size and black hole mass instead that on lower dimension spaces . in other words , we know that in the local universe , black holes and their host spheroids lie on the fp , the m@xmath1-@xmath3 and m@xmath1-l relations 3 relationships involving 4 parameters . thus , by requiring that our distant galaxies land on the local relationships at @xmath8 , we can derive constraints on the evolution in size , luminosity and mass of the host galaxies , as a function of black hole growth in the same time span . the results of this exercise are shown in figure [ fig : manifold ] . for simplicity , systematic and random errors are combined in quadrature in order to compute confidence bands , and neglecting covariance . the main result is that in order to satisfy all scaling relations in the local universe the host galaxies have to grow in luminosity and velocity dispersion , while leaving the size substantially unchanged . consider the no black hole growth scenario ( consistent with the low eddington ratios , see paper i ) . assuming that m@xmath65 , the spheroid mass has to increase by @xmath66 dex ( @xmath67% ) , by today . in other words , @xmath68 similarly , the spheroid luminosity has to increase by @xmath69 dex , implying that its average stellar mass to light ratio has to stay approximately constant ( with an uncertainty of approximately 50% ) over the next 4 gyrs . this requires an injection of younger stars to counteract the ageing of the resident population . the effective radius does not change significantly @xmath70 . those trends are amplified if significant black hole growth is assumed . although at this stage the measurement uncertainties do not allow firm conclusions , it is clear that this empirical methodology holds great promise as a way to disentangle evolution of the stellar populations , black hole growth , and dynamical evolution of the host . in the future , with larger samples covering a wider range of masses and redshifts , it will be possible to study in the detail the co - evolution of spheroids and black holes , perhaps in the framework or more fundamental underlying correlations such as that proposed by @xcite . in previous sections , we have presented evidence that just four billion years ago , there was a population of supermassive black holes living in smaller and less luminous spheroids than today . based on a detailed analysis of uncertainties and selection effects , this evolutionary signature appears significant at the 95% cl . in this section we will briefly discuss our results in the broader context of galaxy formation and evolution and identify a possible physical mechanism . however , before discussing the interpretation of this result , it is worth mentioning two caveats ( see also the discussion in paper i ) . i ) our measurement relies on m@xmath1 estimates based on an empirically calibrated method . the sample of local calibrators is very small and does not necessarily match exactly the properties of our distant sample . more work on the local and distant samples remains to be done before unknown systematics can be firmly ruled out . ii ) the local scaling relations are based on very limited samples . the quiescent sample consists of approximately 40 objects . in addition to the intrinsic difficulty of resolving the sphere of influence , the local sample is composed of earlier type galaxies than the sample considered here . as discussed in paper i , the two samples may have an evolutionary sequence built in at the selection . in this respect , the local agn sample is very important , as the selection process and properties are very similar to those of our own sample . however the local agn sample is comparable or smaller in size to our distant sample , and therefore it is hard to identify subtle differences which could point to some unknown selection effect . keeping these two caveats in the back of our minds , we move on to discuss and interpret the main result of this paper , that black hole mass growth appears to be completed before bulge growth . an important clue is given by the morphology of the host galaxies , provided by the hst images . the majority of galaxies in our sample are not spheroid - dominated elliptical or lenticulars , but rather intermediate or late type spirals . most remarkably , as shown in figure , 6/20 galaxies are observed to show signs of a major ongoing merging ( 3 ) or to be morphologically disturbed ( 3 ) . considering only the three most extreme cases and assuming that major mergers are visually identifiable for @xmath290.5 - 1 gyr @xcite , this finding implies that most or possibly all the galaxies in our sample will undergo a major merger in between the time of observation and today . gas - rich mergers are believed to be the main mechanism transforming stellar and gaseous disks into stellar spheroids , leading to substantial increase in spheroid luminosity and velocity dispersion . by converting rotation supported stars into pressure supported stars , a merger with a disk dominated system ( and hence negligible supermassive black hole ) could grow the bulge more efficiently then the corresponding growth of the black hole by accretion of cold gas @xcite . as calculated in [ sec : mani ] , in order to satisfy the local scaling relations , assuming that black hole growth is negligible , the stellar mass of the spheroid will have to grow by approximately 60% , while the spheroid mass to light ratio wold have to remain approximately constant , qualitatively consistent with a single dissipative merger @xcite , and associated bursts of star formation . this scenario is similar to that discussed in paper i , to which the reader is referred for further discussion and references , with two important additions : i ) the high resolution hst images provide direct evidence for gas rich mergers ; ii ) our joint analysis of several scaling relations allows us to quantify the expected growth of the spheroid mass and constrain the evolution of the stellar populations of the spheroid via the evolution of the mass - to - light ratio . our imaging results may also suggest an apparent evolution of seyfert 1 host galaxy morphologies from @xmath0 to the current epoch . this is because most local seyfert 1 galaxies have substantial bulges , with normal stellar populations , and no signs of interactions @xcite . although currently most seyfert 1s at z=0 do not show signs of ongoing mergers , they do have hints that these occurred @xmath29 1 gyr earlier @xcite , but that their host galaxy morphologies have settled back down to normal by today @xcite . so our result appears to be consistent with the idea that seyfert host galaxies , as quiescent galaxies ( e.g. * ? ? ? * ; * ? ? ? * ) , were more disturbed by interactions 4 gyrs ago than they are today . unfortunately our sample is too small to make a proper comparison , and to derive statistically significant results . however , we checked the consistency of this statement using the goods database . the merger rate for our sample is 3/20 ( @xmath71 ) considering only close pairs / mergers , and it increases to 6/20 considering all disturbed systems ( @xmath72 ) . as a control sample , we selected from goods all galaxies with luminosity @xmath73 ( i.e. the same range in stellar luminosity that our sample , excluding the point source ) , in the redshift range @xmath74 ( morphologies , photometric and spectroscopic redshifts from bundy et al . 2006 ) and performed the same visual classification . we found 8/42 close pairs / mergers ( @xmath75 ) , increasing to 12/42 ( @xmath76 ) if we consider all disturbed systems . this is in good agreement with the fraction observed for the distant seyferts , and somewhat larger than observed in the local universe ( e.g. * ? ? ? the small size of our sample does not warrant a more detailed attempt to measure the merger fraction , so this is left for future work and larger samples . independent evidence appears to support the scenario discussed above . from the galaxy evolution point of view , although the most massive spheroids appear to have completed most of their growth by @xmath77 , it is clear that spheroids of a few @xmath78 m@xmath12 are undergoing significant evolution at @xmath41 . this view is supported by the signatures of recent star formation detected by fundamental plane studies ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * and references therein ) , by the evolution of the mass function ( e.g. bundy , treu & ellis 2007 and references therein ) , and by the evolution of the quenching or transition mass ( e.g. , * ? ? ? * and references therein ) . from the point of view of black hole demographics , it is hard to pinpoint with sufficient precision the growth of black holes in this mass range , although general arguments based on the global star formation and accretion history suggest that black hole growth may predate bulge growth ( e.g. , * ? ? ? * ) . at larger masses , the existence of impressively luminous quasars at @xmath79 @xcite with the high black hole to host galaxy mass ratio determined by radio observations @xcite , appears to be consistent with our scenario , although it is hard to make a direct comparison , considering that evolution may very well be mass - dependent . recent studies of the scaling relations between black hole mass and host galaxy properties tend to focus on higher masses and redshifts than are in our sample . for example , @xcite study the host galaxies of lensed quasars out to @xmath80 , ruling out pure luminosity evolution and finding that the ratio between m@xmath1 and @xmath81 was @xmath82 times larger at @xmath83 than today . similar results are obtained by other studies of quasar host galaxies at comparable redshifts @xcite . these results are consistent with our own , although the comparison requires caution considering that the higher redshift studies typically rely on uv broad emission lines and fluxes for estimating m@xmath1 , instead of h@xmath17 , and that the contrast between agn and host light is more unfavorable than in our case . this prevents accurate decomposition of the host spheroid light and direct determination of the stellar velocity dispersion , which needs to be inferred from proxies such as co and [ ] line widths @xcite . an alternate approach followed by @xcite leads to the opposite conclusion . they use the correlation length of agn hosts at @xmath83 to estimate the virial mass of the halo , and the line width and uv flux at 1350 to estimate m@xmath1 . they compare the inferred relation between halo mass and black hole mass with the local relation @xcite , finding no evidence for evolution . given their error bars , they rule out evolution of over one order of magnitude in the ratio i.e. evolution of the form @xmath84 , with @xmath85 at 90% cl . however , they can not rule out evolution by a factor of 6 that would be predicted extrapolating our best estimate . in conclusion , it seems that several lines of evidence are beginning to point in the same direction : black holes appear to complete their growth before their host galaxies . the uncertainties are still large and much work remains to be done before evolution can be considered conclusively detected . however , with the advent of new technologies such as laser guide star adaptive optics and robotic telescopes it will hopefully be possible to improve dramatically our ability to measure black hole masses in the local and distant universe , and thus reduce the main source of error . this paper is devoted to the study of the cosmic evolution of black holes and their host galaxies . to this aim we have performed a detailed analysis of a sample of 22 seyfert 1 galaxies at @xmath14 . the choice of this particular redshift and moderate luminosity agns allows us to derive precision measurements of the host galaxy properties , as well as to obtain an estimate of the black hole mass from the dynamics of the broad line region . using high resolution images taken with acs we derived luminosity , effective radius and effective surface brightness of the host spheroid as well as nuclear luminosity . we combined this information with emission line widths and stellar velocity dispersion based on high signal to noise keck spectroscopy ( paper i ) to construct the m@xmath1-l , m@xmath1-@xmath3 and fundamental plane relations of distant seyferts . we compared the @xmath14 scaling relations with those followed by local samples of quiescent and active galaxies to determine evolutionary trends . the main results can be summarized as follows : 1 . the m@xmath1-l@xmath2 relation at @xmath14 is inconsistent with the local relation and the assumption of pure luminosity evolution of the host galaxy . adopting pure luminosity evolution consistent with fundamental plane studies , the offset from the local relation corresponds to an offset in present day b - band luminosity of @xmath4 , i.e. @xmath86 , in the sense that black holes lived in smaller bulges at @xmath14 than today . the m@xmath1-@xmath3 relation at @xmath14 is inconsistent with the local relation and the assumption of pure luminosity evolution . the offset with respect to the local relation corresponds to @xmath87 , i.e. @xmath88 , in the sense that black holes lived in smaller bulges at @xmath14 than today . monte carlo simulations show that the offset is not the result of selection effects , which are negligible for the m@xmath1-@xmath3 relation , and can account for at most 0.1 dex of the observed offset of the m@xmath1-l relation . 4 . in order to satisfy the local m@xmath1-@xmath3 , m@xmath1-l and fp relations by @xmath8 assuming no black hole growth our distant spheroids have to grow their stellar mass by approximately 60% ( @xmath89 ) in the next 4 billion years , while preserving their size and holding their stellar mass to light ratio approximately constant . this corresponds to an evolution of the black hole to spheroid mass ratio of the form @xmath90 . assuming that our results are not due to unknown systematic errors or unknown selection effects , the observed evolution can be qualitatively explained if our seyferts undergo a single collisional merger with a disk - dominated system between z=0.36 and today this is consistent with the observed merging / interacting fraction and a timescale for merging visibility of @xmath291 gyr . a single merger could increase the spheroid mass by transporting stellar mass from the progenitors disks , without the corresponding growth of the central black holes due to the lack of black hole in the disk dominated system . at the same time , this process would add younger stars to the spheroid ( either from the merging disks or from newly formed stars ) thus counteracting the fading of the old stellar populations and producing an approximately constant stellar mass to light ratio in the spheroid . numerical simulations including realistic prescriptions for star formation , agn activity and mass loss will be needed to see if these mergers do , indeed , preserve r@xmath31 and m@xmath91/l@xmath2 . if these indications are supported by future studies , then they will confirm that black holes completed their growth before their host galaxies and are perhaps to be seen less as a by - product of galaxy formation than as an orchestrator @xcite . we thank david hogg and the sdss project for providing calibrated images of the local comparison sample . we thank aaron barth , brad peterson , gregory shields , and risa wechsler for discussions . we thank chien peng for useful discussions and advice on using galfit and the anonymous referee for useful suggestions and constructive criticism . we are grateful to kevin bundy for providing the catalog of galaxies in the goods fields with photometric and spectroscopic redshift used to find the control sample described in [ sec : disc ] . this work is based on data obtained with the hubble space telescope and the 10 m w.m . keck telescope . we acknowledge financial support from nasa through hst grant go-10216 . tt acknowledges support from the nsf through career award nsf-0642621 , and from the sloan foundation through a sloan research fellowship . , m. c. , denney , k. d. , cackett , e. m. , dietrich , m. , fogel , j. k. j. , ghosh , h. , horne , k. , kuehn , c. , minezaki , t. , onken , c. a. , peterson , b. m. , pogge , r. w. , pronik , v. i. , richstone , d. o. , sergeev , s. g. , vestergaard , m. , walker , m. g. , & yoshii , y. 2006 , , 651 , 775 , x. , strauss , m. a. , richards , g. t. , hennawi , j. f. , becker , r. h. , white , r. l. , diamond - stanic , a. m. , donley , j. l. , jiang , l. , kim , j. s. , vestergaard , m. , young , j. e. , gunn , j. e. , lupton , r. h. , knapp , g. r. , schneider , d. p. , brandt , w. n. , bahcall , n. a. , barentine , j. c. , brinkmann , j. , brewington , h. j. , fukugita , m. , harvanek , m. , kleinman , s. j. , krzesinski , j. , long , d. , neilsen , jr . , e. h. , nitta , a. , snedden , s. a. , & voges , w. 2006 , , 131 , 1203 , k. , bender , r. , bower , g. , dressler , a. , faber , s. m. , filippenko , a. v. , green , r. , grillmair , c. , ho , l. c. , kormendy , j. , lauer , t. r. , magorrian , j. , pinkney , j. , richstone , d. , & tremaine , s. 2000 , , 539 , l13 , j. m. , davis , m. , faber , s. m. , guhathakurta , p. , gwyn , s. , huang , j. , koo , d. c. , le floch , e. , lin , l. , newman , j. , noeske , k. , papovich , c. , willmer , c. n. a. , coil , a. , conselice , c. j. , cooper , m. , hopkins , a. m. , metevier , a. , primack , j. , rieke , g. , & weiner , b. j. 2006 , arxiv astrophysics e - prints , r. k. , bernardi , m. , schechter , p. l. , burles , s. , eisenstein , d. j. , finkbeiner , d. p. , frieman , j. , lupton , r. h. , schlegel , d. j. , subbarao , m. , shimasaku , k. , bahcall , n. a. , brinkmann , j. , & ivezi , . 2003 , , 594 , 225 , s. , gebhardt , k. , bender , r. , bower , g. , dressler , a. , faber , s. m. , filippenko , a. v. , green , r. , grillmair , c. , ho , l. c. , kormendy , j. , lauer , t. r. , magorrian , j. , pinkney , j. , & richstone , d. 2002 , , 574 , 740 lrrcrrr s01 & 15 39 16.23 & + 03 23 22.06 & 0.3596 & 18.74 & @xmath92 + s02 & 16 11 11.67 & + 51 31 31.12 & 0.3544 & 18.94 & - + s03 & 17 32 03.11 & + 61 17 51.96 & 0.3583 & 18.20 & - + s04 & 21 02 11.51 & -06 46 45.03 & 0.3580 & 18.41 & 186@xmath93 8 + s05 & 21 04 51.85 & -07 12 09.45 & 0.3531 & 18.35 & 132@xmath93 5 + s06 & 21 20 34.19 & -06 41 22.24 & 0.3689 & 18.41 & 169@xmath9314 + s07 & 23 09 46.14 & + 00 00 48.91 & 0.3520 & 18.11 & 145@xmath9313 + s08 & 23 59 53.44 & -09 36 55.53 & 0.3591 & 18.42 & 187@xmath9311 + s09 & 00 59 16.11 & + 15 38 16.08 & 0.3548 & 18.16 & 187@xmath9315 + s10 & 01 01 12.07 & -09 45 00.76 & 0.3506 & 17.92 & - + s11 & 01 07 15.97 & -08 34 29.40 & 0.3562 & 18.34 & 127@xmath93 9 + s12 & 02 13 40.60 & + 13 47 56.06 & 0.3575 & 18.12 & 173@xmath9322 + s16 & 11 19 37.58 & + 00 56 20.42 & 0.3702 & 19.22 & - + s21 & 11 05 56.18 & + 03 12 43.26 & 0.3534 & 17.21 & - + s23 & 14 00 16.66 & -01 08 22.19 & 0.3515 & 18.08 & 172@xmath93 8 + s24 & 14 00 34.71 & + 00 47 33.48 & 0.3621 & 18.21 & 214@xmath9310 + s26 & 15 29 22.26 & + 59 28 54.56 & 0.3691 & 18.88 & 128@xmath93 8 + s27 & 15 36 51.28 & + 54 14 42.71 & 0.3667 & 18.80 & - s28 & 16 11 56.30 & + 45 16 11.04 & 0.3682 & 18.59 & 210@xmath9310 s29 & 21 58 41.93 & -01 15 00.33 & 0.3575 & 18.77 & - s31 & 10 15 27.26 & + 62 59 11.51 & 0.3504 & 18.14 & - s99 & 16 00 02.80 & + 41 30 27.00 & 0.3690 & 18.33 & 224@xmath9312 [ tab : sample ] lrrcrrrrc s01 & 18.50 & 19.92 & 10.28 & 21.85 & 5.29 & 0.74 & 0.29 & 8.21 + s02 & 19.03 & 19.85 & 10.31 & 20.27 & 2.63 & 0.36 & 0.22 & 7.99 + s03 & 17.94 & 20.23 & 10.16 & 17.04 & 0.50 & 1.69 & 0.39 & 8.29 + s04 & 18.06 & 20.12 & 10.20 & 18.36 & 0.96 & 1.42 & 0.36 & 8.45 + s05 & 17.93 & 20.45 & 10.07 & 18.84 & 1.03 & 2.04 & 0.47 & 8.77 + s06 & 18.35 & 20.48 & 10.06 & 18.81 & 1.01 & 0.54 & 0.18 & 8.17 + s07 & 17.79 & 20.35 & 10.11 & 18.69 & 1.01 & 2.26 & 0.45 & 8.55 + s08 & 18.31 & 21.75 & 9.55 & 20.50 & 1.23 & 1.25 & 0.40 & 8.10 + s09 & 18.17 & 19.00 & 10.65 & 19.87 & 3.24 & 0.78 & 0.22 & 8.15 + s10 & 18.01 & 19.30 & 10.53 & 16.08 & 0.49 & 1.11 & 0.27 & 8.27 + s12 & 18.12 & 21.16 & 9.78 & 18.14 & 0.54 & 1.05 & 0.28 & 8.69 + s16 & 19.14 & 22.26 & 9.34 & 19.97 & 0.76 & 0.70 & 0.48 & 8.26 + s21 & 17.45 & 18.95 & 10.67 & 15.82 & 0.51 & 2.30 & 0.34 & 8.81 s23 & 17.99 & 20.85 & 9.91 & 17.93 & 0.57 & 1.20 & 0.29 & 8.72 s24 & 18.06 & 18.59 & 10.81 & 22.41 & 12.6 & 0.44 & 0.11 & 8.33 s26 & 18.87 & 20.06 & 10.23 & 17.73 & 0.75 & 0.50 & 0.27 & 8.01 s27 & 18.51 & 19.46 & 10.46 & 21.17 & 4.78 & 0.92 & 0.36 & 8.10 [ tab : meas ] lrcc ark120 & 0.032 & 10.82 & 8.18 + mrk79 & 0.022 & 9.79 & 7.72 + mrk110 & 0.035 & 9.85 & 7.40 + mrk590 & 0.026 & 10.40 & 7.68 + mrk817 & 0.031 & 10.49 & 7.69 + ngc3227 & 0.004 & 8.85 & 7.62 + ngc4051 & 0.002 & 8.43 & 6.28 + ngc4151 & 0.003 & 9.49 & 7.66 + ngc5548 & 0.017 & 10.53 & 7.83 + [ tab : local ]

the particle production process is reviewed , through which cosmic inflation can produce a scale invariant superhorizon spectrum of perturbations of suitable fields starting from their quantum fluctuations . afterwards , in the context of the inflationary paradigm , a number of mechanisms ( e.g. curvaton , inhomogeneous reheating etc . ) through which such perturbations can source the curvature perturbation in the universe and explain the formation of structures such as galaxies are briefly described . finally , the possibility that cosmic vector fields also contribute to the curvature perturbation ( e.g. through the vector curvaton mechanism ) is considered and its distinct observational signatures are discussed , such as correlated statistical anisotropy in the spectrum and bispectrum of the curvature perturbation .

1009.5466

the standard model of cosmology at present is comprised by hot big bang cosmology and cosmic inflation . the cosmology of the hot big bang accounts for the hubble expansion of the universe , the observed cosmic microwave background ( cmb ) radiation , the primordial abundance of light elements ( formed in the early universe through the process of big bang nucleosynthesis ) and the age of the universe , which agrees well with the astrophysical estimates of the ages of the oldest globular clusters . in turn inflation overcomes or at least ameliorates certain fine - tuning problems regarding the initial conditions of the hot big bang , namely the so - called horizon and flatness problems @xcite @xcite . what is cosmic inflation ? in a nutshell , inflation is a brief period of superluminal expansion of space in the early universe . what inflation does is that it makes the observable universe large and uniform . however , if the universe were perfectly uniform there would be no structures like galaxies or galactic clusters , no stars with planets orbiting around them , no ... us . it is imperative , therefore , that there is a deviation from perfect uniformity , which can give rise to these structures . indeed , we need a primordial density perturbation ( pdp ) for structure formation to occur . it so happens that evidence for such a pdp exists as the latter reflects itself on the cmb through the so - called sachs - wolfe effect @xcite which states that cmb light is redshifted when crossing growing overdensities . this effect directly connects the fractional amplitude of the pdp with the fractional perturbation of the temperature of the cmb : @xmath0 even though the pdp appears to be very small , numerical simulations of structure formation have shown that it is enough to account for the observed structure in the universe . what is the origin of this pdp ? well , it turns out that this too can be accounted for by cosmic inflation . to have an idea of how cosmic inflation is achieved consider the so - called friedman equation , which is the temporal component of the einstein equations for a homogeneous and isotropic spacetime . in flat space this equation reeds , where is the reduced planck mass is newton s gravitational constant and we consider natural units where . ] and is the hubble parameter corresponding to the rate of the expansion of the universe , with @xmath1 being the scale factor of the universe , parameterising the universe expansion , and the dot denotes derivative with respect to the cosmic time @xmath2 . suppose now that , at some period in its early history , the universe was dominated by an effective cosmological constant @xmath3 . then , , which means that the hubble parameter is constant and , therefore , , i.e. space expands exponentially in time . thus , inflation occurs when the universe is dominated by an effective vacuum density . when inflation ends this density has to be transferred in the density of the thermal bath of the hot big bang . this , in effect , amounts to a change of vacuum , to the true vacuum of the present , with @xmath4 . ] . therefore , vacuum states during inflation are not necessarily vacuum states after inflation , but instead they can become populated . this process is called particle production and it is similar to the production of particles ( in the form of hawking radiation ) on the event horizon of black holes @xcite . indeed , the cosmological horizon during inflation is an event horizon ( of size @xmath5 ) and can be viewed as an `` inverted '' ( i.e. inside - out ) black hole in the sense that nothing can escape being `` sucked out '' by the superluminal expansion . virtual particle pairs , corresponding to quantum fluctuations , are broken up by the expansion and are pulled away to superhorizon distances , where they can no more find eachother and annihilate , becoming thereby real particles , giving rise to classical perturbations of the corresponding fields @xcite . the amplitude of these perturbations is determined by the hawking temperature @xmath6 , which for de sitter space is @xcite . let us review now the particle production process in a more rigorous manner . the standard paradigm considers a real , minimally coupled , scalar field @xmath7 of mass @xmath8 with lagrangian density @xmath9 using the above one obtains the equation of motion of the scalar field . the field is expected to become homogenised by the inflationary expansion as any inhomogeneities are inflated away . in this case the equation of motion becomes @xmath10 where . deviations from homogeneity are introduced originally only from its vacuum fluctuations . to follow their evolution we perturb the field from its homogeneous value as . then we obtain the equation of motion of the fourier components of the field perturbations . this equation reads @xmath11 \delta\phi_k=0\ , , \label{eom}\ ] ] where . the next step is to promote the perturbations of the field to quantum operators defined as @xmath12,\ ] ] where @xmath13 and @xmath14 are creation and annihilation operators respectively and we consider canonical quantisation with . the mode functions @xmath15 satisfy the same equation of motion as the fourier components of the field perturbations @xmath16 because this equation is linear . we can solve this equation considering that in the subhorizon limit ( ) the solution matches the so - called bunch - davies vacuum @xcite which corresponds to the quantum fluctuations of a free scalar field in minkowski spacetime . thus , the boundary condition reads @xmath17 where is the conformal time , which factors out the expansion of the universe . using the above , the solution of eq . ( [ eom ] ) for the mode functions is @xmath18 , \label{phisolu}\ ] ] where @xmath19 denotes bessel functions of the first kind and . to investigate particle production we evaluate the above solution in the superhorizon limit ( ) . we find @xmath20 using the above we can calculate the power spectrum in the superhorizon limit . we find @xmath21 thus , we see that the scale dependence of the field perturbations is only included in the last term of the above . considering a light field , i.e. , we see that the scale dependent term is eliminated and the spectrum becomes scale invariant . indeed , for a light field the above spectrum can be written as @xmath22 where and we considered . thus , a light minimally coupled scalar field obtains a scale invariant spectrum of perturbations when they exit the horizon . the typical value of the field perturbations is , i.e. it is determined by the hawking temperature as expected . what is the fate of these perturbations after horizon exit ? the perturbations can now be treated classically and their evolution is determined by the equivalent of eq . ( [ eomhom ] ) @xmath23 considering that the solution of the above for a light field is @xmath24\approx\frac{h}{2\pi}.\ ] ] thus , we see that the perturbations of a light scalar field freeze - out after horizon exit . this guarantees that scale invariance is maintained since , regardless of the time of horizon exit ( i.e. regardless of how large the size of the perturbations becomes by inflation ) , the perturbations have the same amplitude because until the end of inflation ( see fig . [ superhor ] ) . schematic log - log plot of the evolution of the perturbations of a scalar field ( enveloped by red solid lines ) which follow the universe expansion compared with the cosmological horizon ( solid green lines ) . during inflation the cosmological horizon is an event horizon of constant physical radius @xmath5 . the perturbations of the scalar field start off as quantum fluctuations at subhorizon size , which are stretched to superhorizon scales by the ( quasi)exponential expansion . after inflation , the size of the superhorizon perturbations continues to grow following the expansion of the universe ( i.e. proportional to @xmath25 ) but the cosmological horizon ( which is now a particle horizon ) grows faster , with the speed of light . as a result , the perturbations reenter the horizon some time after the end of inflation . in the graph two different perturbations are depicted . they exit the horizon at different times , corresponding to the horizontal dashed lines . as a result they are inflated to different sizes so one is much larger that the other when they reenter the horizon . the size of the perturbations at horizon reentry is depicted by the vertical dashed lines . however , both the perturbations have approximately the same amplitude , determined by the hawking temperature during inflation , which remains approximately constant . the fact that the scalar field perturbations retain the same amplitude even though they attain different sizes during their superhorizon evolution is the reason behind the scale - invariance of the perturbation spectrum.,width=384 ] how are such field perturbations related with the pdp ? the pdp arises because of the generation of a corresponding curvature perturbation @xmath26 , as is discussed below . if the curvature perturbation is due to the perturbations of a light scalar field then their power spectra are proportional , i.e. . this means that they have the same scale dependence . the latter can be parametrised in the form of a power - law : , so that @xmath27 assuming that @xmath26 is due to a light scalar field one obtains ( cf . ( [ pdphi ] ) ) @xmath28 where quantifies the deviation from pure de sitter expansion which we have ignored so far . since , during inflation for a light scalar field , we find that , if the curvature perturbation is due to this field then and the scale dependence of @xmath29 is eliminated . thus , the pdp in this case would be approximately scale - invariant . indeed , the latest wmap observations suggest @xcite @xmath30 which agrees with the predictions of inflation with the quantum fluctuations of a light scalar field as the source of the pdp . note that the observations deviate from exact scale invariance ( ) at 1-@xmath31 . this reveals some dynamics during inflation and agrees with the expectations of realistic inflation models . in general relativity the curvature of spacetime and the energy density of its content are interchangeable quantities , through the einstein equations , depending on the choice of foliation of spacetime . therefore , for the curvature perturbation it is useful to define a quantity @xmath26 which is independent of such foliation ( gauge invariant ) . this can be written as @xcite @xmath32 where the first term on the right - hand - side is the curvature perturbation in uniform density slices of spacetime and the second term is the density perturbation in flat slices of spacetime . we call @xmath26 the ( gauge invariant ) curvature perturbation from now on . one can also define the power spectrum of the curvature perturbation which is given by the two - point correlator as @xmath33 the latest observations of wmap give @xcite @xmath34 where is the pivot scale . the corresponding density perturbation ( at horizon re - entry ) is given by @xmath35 which , as mentioned , is the observed measurement of the pdp . another useful quantity is the so - called bispectrum @xmath36 of the curvature perturbation , defined by the three - point correlator of @xmath26 as follows @xmath37 the bispectrum can be related with the power spectrum is the following manner @xmath38,\ ] ] where . the bispectrum is useful because it is exactly zero if the curvature perturbation is gaussian , i.e. if it obeys gaussian statistics . one expects that the perturbations of the fields which are generated during inflation from their quantum fluctuations are indeed gaussian since the quantum fluctuations are random . any non - gaussian signal therefore would arise by the process which translates the perturbations of such fields , e.g. @xmath39 , into the curvature perturbation @xmath26 . this process may be highly non - linear in which case @xmath26 would feature non - gaussian statistics . the level of such non - gaussianity is quantified in the bispectrum of @xmath26 by the so - called non - linearity parameter @xmath40 . the latter also depends on the configuration of the three @xmath41-vectors used to determine the bispectrum . the two most popular configurations used to determine @xmath40 are 1 ) the equilateral configuration , where and 2 ) the squeezed configuration , where . the wmap findings for the values of @xmath40 in these configurations are @xcite @xmath42 note that there is a hint ( at 1-@xmath31 ) of non - gaussianity in the squeezed configuration . it is likely that non - gaussianity may be detected in the near future by the observations of the planck satellite , which are expected to improve the precision of the measurement of @xmath40 by about an order of magnitude . it is important to stress here that the pdp is highly gaussian . indeed , remembering that and also we see that , even if the observational upper bounds on @xmath40 are saturated , non - gaussianity in the pdp is tiny . this also agrees with the conjecture that the pdp is due to perturbations of suitable fields ( e.g. light scalar fields ) arising from their quantum fluctuations during a period of inflation . before discussing the mechanisms through which the quantum fluctuations of suitable fields can generate the curvature perturbation from inflation , it is necessary to briefly present the so - called inflationary paradigm . this is the typical manner in which inflation is modelled in particle cosmology . according to the inflationary paradigm the universe undergoes inflation when dominated by the potential density of a scalar field , which is called the inflaton field . we return to the lagrangian of eq . ( [ lphim ] ) but this time instead of the mass term we consider a generic function @xmath43 , which corresponds to the potential density of the scalar field . in this case , eq . ( [ eomhom ] ) has the form @xmath44 where the prime denotes derivative with respect to the scalar field @xmath7 . the above equation is similar to the equation of motion of a body sliding down the potential @xmath45 and subject to a friction term determined by the rate of the universe expansion @xmath46 . based on this analogy we will consider the field as rolling down the potential in field space . now , according to the inflationary paradigm , for inflation to take place the universe must be dominated by the potential density of the inflation field , i.e. the kinetic density @xmath47 of the field needs to be subdominant : @xmath48 this means that the field hardly moves and , therefore , its potential density remains roughly constant , providing thereby the effective cosmological constant needed for inflationary expansion . when the condition is valid it is said that the field undergoes slow - roll and its variation ( motion in field space ) is overdamped by the friction term in eq . ( [ eomhomv ] ) . in this case one can ignore the acceleration term in eq . ( [ eomhomv ] ) which is then recast as @xmath49 because the roll of the field is overdamped , the value of @xmath50 is very small , which means that the slope of the potential needs to be very small as well , according to the slow - roll equation above . hence , the inflaton field corresponds to a flat direction in field space . schematic representation of the inflaton potential @xmath43 . during inflation the inflaton field @xmath7 ( represented by a circle ) rolls down its flat potential in field space until it reaches a critical value @xmath51 ( corresponding to ) when slow - roll is terminated and inflation ends . afterwards the field undergoes rapid oscillations around its vev , corresponding to the minimum of the potential . the field decays while oscillating reheating thereby the universe.,width=384 ] inflation ends when the slow - roll condition is violated , i.e. at some critical value @xmath51 of the inflaton field , the potential becomes steep and curved such that . afterwards , the inflaton condensate oscillates around its vacuum expectation value ( vev ) . such coherent oscillations correspond to massive particles ( inflatons ) , which eventually decay into the standard model particles that comprise the thermal bath of the hot big bang ( see fig . [ infpot ] ) . this process is called reheating and , if it occurs in a perturbative manner , it is usually modelled by adding a phenomenological decay term in the field equation @xmath52 where @xmath53 stands for the decay rate of the inflaton field . reheating occurs when the density of inflation is transferred to the newly formed thermal bath , whose temperature is , called the reheating temperature . we are ready now to discuss some ways that the perturbations of a light scalar field can give rise to the curvature perturbation in the universe , which corresponds to the pdp that is responsible for structure formation . the traditional mechanism through which the curvature perturbation is generated employs the inflaton field itself for the job . this , so - called inflaton hypothesis , simply assumes that the field responsible for the formation of the curvature perturbation is the same field which determines the dynamics of inflation , i.e. the inflaton field . as mentioned in the previous section , to undergo slow - roll the inflaton must correspond to a flat direction in field space and therefore is characterised by , i.e. it is a light scalar field . this means that , through the particle production process , it obtains a scale - invariant superhorizon spectrum of perturbations . as a result , the perturbed inflaton field will reach the critical value @xmath51 which terminates inflation at slightly different times at different points in space . thus , inflation will continue a little bit more in some locations than in others ( see fig . [ infend ] ) . this is the cause of the generation of the curvature perturbation @xmath26 . indeed , it can be shown that the latter is determined by the difference in logarithmic expansion between the uniform density and the spatially flat slices , i.e. @xcite . schematic representation of the way that the curvature perturbation is generated in the inflaton hypothesis . the inflaton field perturbations vary its expectation value typically by . this means that , while the inflaton field ( represented by a circle ) rolls down its potential in field space , its value is different throughout coordinate space , represented by the @xmath54-labelled axis . thus , in some locations the rolling inflaton is ahead compared to others , which means that , in these locations , it will reach the critical value @xmath51 that terminates inflation , somewhat earlier . at a given time the inflaton value is represented by the wiggly line which crosses the line in several places . therefore , the universe inflates more in some places than in others , because , at the given time that corresponds to the wiggly line shown , in some locations inflation has ended while in others it still continues.,width=384 ] the pdp in this case is given by @xmath55 where the asterisk denotes the epoch of horizon exit of the inflaton perturbations . since in the inflaton hypothesis a single degree of freedom ( the inflaton ) determines both the dynamics of inflation and the pdp , the latter can be written in terms only of the characteristics of the inflaton potential . indeed , eq . ( [ drh ] ) can be recast as @xcite @xmath56 which again is to be evaluated at horizon exit . observational constraints on the amplitude of the pdp ( cf . ( [ drobs ] ) ) suggest that , if @xmath57 is not extremely small , the energy scale of inflation is comparable to that of grand unification , i.e. . this seems a natural scale to introduce new physics but it turns out that , for single field models , it requires fine - tuning of model parameters . the spectral index in the inflaton hypothesis is @xcite @xmath58 where corresponds to the curvature of the potential along the direction of the inflaton field . if as in eq . ( [ lphim ] ) , then as discussed after eq . ( [ pdphi ] ) . under the inflaton hypothesis the generated non - gaussianity is expected to be negligible with , i.e. below the limit of observability . this means that if non - gaussianity is indeed observed in the pdp then all the single field inflation models are going to be falsified . the beauty and the curse of the inflaton hypothesis is that it relies on a single degree of freedom , namely the inflaton field , to account for all the problems that inflation aims to address , i.e. provide the period of accelerated expansion that deals with the horizon and flatness problems and also generate the appropriate spectrum of curvature perturbations , which agrees with observational constrains on its amplitude and spectral index . as a result , models of single field inflation are typically overconstrained and suffer from substantial fine - tuning . this is why alternative hypotheses have been put forward for the generation of the pdp from inflation . this hypothesis assumes that the field responsible for the formation of the curvature perturbation has nothing to do with the dynamics of inflation , i.e. it is _ other _ than the inflaton field . this scalar field is called curvaton @xmath31 @xcite ( see also refs . @xcite@xcite@xcite@xcite ) . in order for the curvaton to play this role it needs to obtain a superhorizon spectrum of perturbations during inflation . thus , it needs to be a light field during inflation so that it can undergo particle production . it is evident that by introducing another degree of freedom , the fine - tuning problems of inflation model - building are substantially improved @xcite@xcite . however , it must be stressed that the curvaton is not necessarily a new ad hoc addition to the theory , which is introduced by hand . indeed , the fact that the curvaton is not connected to the inflaton sector allows it to correspond to physics at energy scales much smaller than that of inflation ( e.g. the tev scale which is accessible by lhc ) . consequently , there exist a number of candidates for the curvaton which correspond to realistic fields already present in simple extensions of the standard model . prominent examples are a right - handed neutrino @xcite , an mssm flat direction @xcite@xcite or the so - called orthogonal axion in supersymmetric realisations of the peccei - quinn symmetry @xcite@xcite . under the curvaton hypothesis the curvature perturbation is given by @xcite @xmath59 where , where is the density parameter of the curvaton field at the time of its decay after inflation is also denoted as @xmath60 in much of the literature . ] . in the above @xmath61 is the curvature perturbation attributed to the curvaton field , which is determined by the fractional perturbation of the field itself : @xmath62 where we used that and we assumed that near its decay the curvaton density is , where is the mass of the curvaton field . the spectral index of the pdp in this case is @xcite @xmath63 where @xmath64 corresponds to the curvature of the potential along the curvaton direction . log - log plot of the evolution of the inflaton energy density which decays into radiation @xmath65 ( purple line ) at the end of inflation ( denoted by ` end ' ) and the curvaton energy density @xmath66 ( green line ) ( prompt reheating is assumed ) . during inflation , the curvaton density is negligible . after inflation . in contrast , remains constant ( the curvaton is frozen at some value @xmath67 ) until , when the curvaton unfreezes and begins oscillating , after which time ( denoted ` osc ' ) . at some moment ( denoted ` dom ' ) the curvaton density dominates the universe ( ) until , some time later ( denoted ` dec ' ) when it decays into the thermal bath of the hot big bang . the dashed slanted line depicts the possibility that the curvaton decays before domination ( ) , when substantial non - gaussianity can be generated.,width=384 ] since by definition the curvaton should not affect the dynamics of inflation , during inflation we expect . thus the density parameter of the curvaton is extremely small and its contribution to the overall curvature perturbation ( cf . ( [ zcurv ] ) ) is also small . for the curvaton to significantly contribute to @xmath26 we need to consider the period after inflation , when its contribution to the density can increase . in the simplest case when the equation of motion for the curvaton is of the same form as eq . ( [ eomhom ] ) . since during inflation the field is light and , the curvaton is overdamped and remains frozen in some value @xmath67 . after inflation , however , the hubble parameter decreases in time . as a result , there will be a moment when , at which time the curvaton condensate will unfreeze and will begin coherent oscillations around its vev . these oscillations correspond to massive particles ( curvatons ) whose density is diluted as , which is less drastic that the density of the radiation background . thus , the oscillating curvaton has a chance to dominate ( or nearly dominate ) the universe before its decay ( see fig . [ curvevol ] ) . thus , at curvaton decay can be as large as unity and the curvaton contribution to @xmath26 can be substantial . consequently , the curvaton imposes its curvature perturbation onto the universe at ( or near ) its domination . because of its spectrum of perturbations @xmath68 the amplitude of the curvaton oscillations is also perturbed , i.e. it is larger in some places than in others . this means that the density of the oscillating field is perturbed too and the field will dominate the universe at different times at different locations . this is what generates the curvature perturbation ( cf . ( [ zs ] ) ) . in contrast to the inflaton hypothesis , under the curvaton hypothesis non - gaussianity can be substantial . indeed , in this case @xcite @xmath69 which can be large if the decay of the curvaton happens before domination when . in fact , the wmap observations in eq . ( [ fnlobs ] ) set the lower bound . there are numerous other ways to generate the pdp from a superhorizon spectrum of scalar field perturbations . this section briefly reviews two of them , namely the inhomogeneous reheating and the end of inflation mechanisms . as with the curvaton mechanism , these mechanisms assume that the contribution to the curvature perturbation from the inflaton field itself is negligible . this mechanism generates the pdp by assuming that the inflaton decay rate @xmath53 is modulated by another scalar field @xmath31 @xcite@xcite . this scalar field is the one which undergoes particle production during inflation and which obtains thereby a superhorizon spectrum of perturbations . in this scenario we can ignore the perturbations of the inflaton field . thus , we can consider that inflation is terminated at the same time throughout space . the inflaton then begins its coherent oscillations around its vev until it decays when . however , since in this case and @xmath31 is perturbed in space , the decay rate is different at different locations so that the hot big bang begins at different times . this is what generates the curvature perturbation which corresponds to the pdp . roughly speaking we have @xmath70 in this scenario the spectral index is again given by eq . ( [ nscurv ] ) . non - gaussianity in this model is quantified as @xcite @xmath71 where now the prime denotes derivative with respect to @xmath31 . if @xmath53 has a power - law dependence on @xmath31 then , which is marginally observable . this mechanism applies to a particular type of inflation model , the so - called hybrid inflation . therefore , before discussing the mechanism a brief summary of hybrid inflation is necessary . hybrid inflation couples the inflaton field @xmath7 to another scalar field @xmath72 , called the waterfall field , in such a way that inflation is terminated by a phase transition which sends the waterfall to its vev @xcite . since inflation is likely to be at the scale of grand unification , in many cases the waterfall field is assumed to be the higgs field of a grand unified theory ( gut ) . thus , inflation is terminated by the breaking of grand unification , i.e. the gut phase transition . the scalar potential for hybrid inflation has the following form @xmath73 where @xmath74 is the gut energy scale , @xmath75 is the self - coupling of the waterfall field and @xmath76 is the interaction coupling between the waterfall and the inflaton . the potential @xmath77 is responsible for the slow - roll of the inflaton and its precise form is not relevant to this discussion . from the above potential it is evident that the effective mass - squared of the waterfall field is @xmath78 the above implies that there is a critical value of the inflaton @xmath79 such that if then and . in this case , eq . ( [ vhybrid ] ) becomes . the constant contribution to the scalar potential provides the effective cosmological constant for ( quasi)de sitter inflation : . if however , then and a phase transition occurs which results in , which gives the inflaton a large mass ( through the interaction term ) and so . assuming , after the phase transition and inflation ends . the end of inflation mechanism for the production of the pdp introduces an extra coupling between the waterfall field and another scalar field @xmath31 @xcite@xcite . the scalar potential now becomes @xmath80 where @xmath81 parametrises the strength of the interaction between @xmath31 and the waterfall field @xmath72 . with this addition the effective mass - squared of the waterfall field becomes @xmath82 and the critical value which triggers the phase transition that ends inflation is now @xmath83 i.e. it is modulated by the value of @xmath31 . therefore , if @xmath31 is light during inflation , then it undergoes particle production and obtains a superhorizon spectrum of perturbations with typical magnitude . this means that the value of @xmath84 is also perturbed . hence , the phase transition which terminates inflation occurs earlier in some parts of the universe than in others depending on when the inflaton reaches the critical value @xmath84 . consequently , the universe inflates more in some locations than in others and this generates the curvature perturbation . as in the curvaton scenario , the spectral index in this case is given by eq . ( [ nscurv ] ) @xcite . tantalising evidence exists of a preferred direction in the cmb temperature perturbations . in particular , the low multipoles in the cmb appear to be aligned @xcite@xcite at a level which is statistically rather improbable @xcite@xcite . a preferred direction in the cmb can not be accounted for if inflation and the generation of the pdp is due to scalar fields only . moreover , despite their abundance in theories beyond the standard model , scalar fields have not been observed as yet . if the higgs field is not found in the lhc , the credibility of the ubiquitous usage of scalar fields in cosmology will be shaken . until recently only scalar fields have been considered both as responsible for the dynamics of inflation and also for the generation of the observed pdp . however , in the pioneering work of ref . @xcite the possibility that a vector field contributes in the generation of the pdp was first considered . since then , a number of attempts have been made to investigate the role and the implications of cosmic vector fields in the generation of the pdp ( see ref . @xcite and references therein ) . why vector fields were not considered originally for the generation of the pdp ? after all , these are fields which are similar to the massive gauge bosons , which have indeed been observed in lep . the main difficulty had to do with the inherent anisotropic nature of vector fields . inflation would homogenise a vector field condensate , and a homogeneous vector field picks up a preferred direction in space . thus , if this vector field condensate were to dominate the density of the universe ( so that it can affect the expansion and generate the pdp ) it was felt that the anisotropic stress generated would lead to strongly anisotropic expansion which would be impossible to reconcile with the predominant isotropy of the cmb . however , as is discussed below , this difficulty can be circumvented in the vector field remains subdominant during inflation . still , if ( in analogy to scalar fields ) light vector fields were needed for a scale invariant spectrum then a more subtle problem arises . massless abelian vector fields are conformally invariant and they can not undergo particle production during inflation . thus , for light vector fields particle production is expected to be suppressed . nevertheless , there are numerous mechanisms which break the conformality of vector fields so that particle production can be efficient . these mechanisms are model dependent which means that they can have distinct observational signatures as is discussed below . suppose that there is a suitable theory which breaks the vector field conformality during inflation . how are we to study particle production of the vector field ? we follow the same recipe as in sec . [ pp ] assuming that the inflationary expansion remains isotropic . firstly we perturb the vector field around its homogenised value . then we fourier transform the perturbations and we obtain the equations of motion for the fourier components in the theory which we are considering . then we promote the perturbations of the vector field into quantum operators @xmath85 , \label{vecexpan}\ ] ] where @xmath86 are the mode functions and we consider again canonical quantisation . in the above we have introduced the polarisation vectors @xmath87 where @xmath88 denote the left and right transverse polarisations and @xmath89 denotes the longitudinal polarisation in eq , ( [ vecexpan ] ) runs over these values . ] . as with the scalar field case , the mode functions of the above expansion are expected to satisfy the equations of motion for the fourier components of the vector field perturbations . the form of these equations depends on the theory which breaks the conformality of the vector field . to solve them we again consider vacuum boundary conditions in the subhorizon limit ( ) . they are @xcite : @xmath90 we see that the boundary condition is again the bunch - davies vacuum but the longitudinal component it is multiplied by the lorentz boost factor , which takes us from the frame with where all components are equivalent , to the frame of momentum @xmath41 . after solving the equations of motion we evaluate the solutions in the superhorizon limit ( ) . then we can obtain the power spectrum for each polarisation as , where @xmath91 is evaluated in the superhorizon limit . because we have three degrees of freedom there are three possibilities : case a : @xmath92 case b : @xmath93 case c : @xmath94 even though case a appears to be the most generic , in practice it is case b that is the most common because usually the theories which break the vector field conformality are parity conserving . we have isotropic particle production only in the special case c. what would happen if the spectrum of the produced perturbations of the vector field affected the curvature perturbation ? since particle production is in general anisotropic we expect an anisotropic contribution to @xmath26 . this contribution can be parametrised as follows . for the power spectrum we can write @xcite @xmath95 , \label{stanisp}\ ] ] where and with and the ellipsis denotes higher order contributions . the anisotropic part of the spectrum corresponds to a new observable , namely _ statistical anisotropy _ , which amounts to direction dependent patterns in the cmb @xcite ( see fig . [ cmbanis ] ) . statistical anisotropy is quantified by the so - called anisotropy parameter @xmath76 in eq . ( [ stanisp ] ) . the wmap observations set a surprisingly weak upper bound on the anisotropy parameter : @xcite but the authors acknowledge that , because the direction is suspiciously near the ecliptic plane , this is probably due to an unknown systematic error . hence we treat this number as an upper bound . ] . thus we see that , at present , observations allow up to 30% statistical anisotropy in the spectrum of the pdp . the forthcoming data of the planck satellite will improve precision by an order of magnitude and are likely to detect statistical anisotropy . if this will be so then a cosmic vector field will have to be involved in the generation of the pdp during inflation . in a similar manner one can parametrise the statistically anisotropic contribution from a vector field in the bispectrum . indeed , we can write @xcite@xcite @xmath96 where @xmath97 denotes the isotropic part and @xmath98 is the projection of @xmath99 onto the plane of the three @xmath41-vectors which are used to define the bispectrum . note that , as non - gaussianity has not been observed yet , there are no bounds on @xmath100 . if this means that non - gaussianity is predominantly anisotropic ( even though @xmath29 is not ) . if non - gaussianity is indeed observed ( e.g. by the planck satellite ) and no angular modulation of @xmath40 is found on the microwave sky then all models which predict will be falsified . it is important also to stress that the directions of statistical anisotropy in the spectrum and the bispectrum are correlated ( determined by the direction of @xmath99 ) , which is a smoking gun for the contribution of a vector field to the pdp @xcite . in the cases a and b the spectra of the superhorizon perturbations of the vector field are predominantly anisotropic . since @xmath29 is isotropic at least at the level of 70% the contribution of the vector field to the pdp in these two cases has to be subdominant and its significance is only the possibly observable statistical anisotropy in the spectrum and bispectrum . thus , in the cases a and b we still require some other , isotropic source ( such as a scalar field ) to provide the dominant contribution to the pdp . on the other hand , in case c the perturbation spectra of the vector field are isotropic . in this case , the vector field alone can generate @xmath26 and no input from any scalar field is necessary . note that , in this case , the vector field may also produce statistical anisotropy if the @xmath101 are not exactly the same but differ slightly , albeit by no more that 30% . in this section we discuss a couple of proposals for the generation of a flat superhorizon spectrum of vector field perturbations . consider the following theory @xmath102 where is the field strength of the abelian vector field @xmath103 , @xmath8 is its bare mass , @xmath104 is constant and @xmath105 is the ricci scalar . from the above we see that the vector field has effective mass - squared . following the procedure outlined in the previous section , we obtain the following solution for the mode functions of the transverse perturbations of the physical vector field @xcite . ] @xmath106 . \label{vectranssolu}\ ] ] the above is identical to eq . ( [ phisolu ] ) with the crucial difference that . using this solution we obtain the following expression for the power spectrum of the transverse perturbations in the superhorizon limit @xmath107 which again is identical to eq . ( [ pphi ] ) but for the different value of @xmath108 . from the above it is evident that a scale invariant spectrum is attained for . this translates into a requirement for both @xmath104 and @xmath8 as follows . in a spatially flat , homogeneous and isotropic spacetime , the scalar curvature is , where @xmath109 is the barotropic parameter of the content of spacetime . during inflation , so that . this means that we can attain a scale invariant transverse spectrum if 1 ) and 2 ) , i.e. the vector field is a light field with a negative effective mass - squared @xcite . ] . under these conditions the transverse spectra are indeed scale - invariant and are given by , exactly as a massless scalar field . employing the same conditions for @xmath104 and @xmath8 one obtains the following solution for the longitudinal component of the physical vector field @xcite : @xmath110 \frac{e^{ik / ah}}{\sqrt{2k}}\,.\ ] ] clearly the above is totally different from eq . ( [ vectranssolu ] ) . in the superhorizon limit the power spectrum of the longitudinal component is . this means that we are in case b since and because the theory is parity invariant . thus , particle production in this theory is anisotropic at a level of 100% . therefore , the vector field can not alone be responsible for the pdp . its contribution to @xmath26 has to be subdominant and it can only be the source of statistical anisotropy . before concluding , we should point out that this model has been criticised for suffering from instabilities ( ghosts ) @xcite@xcite ( see however ref . @xcite ) . now , consider the theory @xmath111 where , is the kinetic function and we also consider during inflation . the maxwell - type kinetic term in combination with the positive mass - squared guarantees the stability of the model @xcite and therefore is motivated even if the vector field is not a gauge boson . note also that a massive abelian vector field is renormalisable even is it is not a gauge field @xcite . the solutions for the mode functions of the field perturbations are too complicated to reproduce here . it suffices to say that scale invariance in the transverse components requires the kinetic function to scale with the expansion as and also the physical vector field to be light at horizon exit , where . scale invariance for the longitudinal component additionally requires @xcite@xcite . if the vector field is a gauge boson then @xmath112 is the gauge kinetic function , which is related to the gauge coupling as . this means that only the case when is acceptable because only then the gauge field remains weakly coupled during inflation . note that the model is naturally realisable in supergravity theories where @xmath112 is a holomorphic function of the scalar fields of the theory @xcite . in general , khler corrections to the scalar potential result in masses of order @xmath46 for the scalar fields @xcite - problem of inflation , since slow - roll requires for the inflaton . ] , which are therefore expected to be fast - rolling down the potential slopes during inflation , causing significant variation to @xmath112 indeed , for a power - law dependence of the gauge kinetic function to the scalar fields , it is easy to show that @xcite , i.e. @xmath112 has a power - law dependence on @xmath113 as assumed in this model . now , if @xmath112 is modulated by the inflaton field itself , then it can be shown that , under fairly general conditions , the backreaction to the inflaton s roll renders the scaling an attractor solution @xcite . similarly , for a gauge boson , @xmath8 can be modulated by a fast - rolling higgs field with tachyonic mass @xcite . given the conditions and the power spectra for the transverse and longitudinal components depend on whether the vector field remains light until the end of inflation or not ( note that ) . in particular we find @xcite@xcite @xmath114 from the above we see that , if the vector field remains light until the end of inflation , , i.e. we are in case b. therefore , particle production is strongly anisotropic and the vector field contribution to the pdp has to be subdominant but it can still give rise to substantial statistical anisotropy . in contrast , if the field becomes heavy by the end of inflation then particle production is isotropic and the vector field can be solely responsible for the generation of the pdp . note that , because we need the field to be light when the cosmological scales exit the horizon , for it to become heavy we need during inflation , which is possible only in the case when . if then and we have throughout inflation . but if then we have so that we may end up having at the end of inflation even though we started of with a light field at horizon exit . in the latter case the vector field begins coherent oscillations a few exponential expansions ( e - folds ) before the end of inflation . assuming that at the end of inflation and the field becomes canonically normalised , we find that there is ample parameter space for case c to be realised , namely : @xcite@xcite . through the examples in the previous section it is evident that a scale invariant spectrum of perturbations ( isotropic or not ) of a vector field can indeed be generated if one assumes some theory which appropriately breaks the conformality of the vector field . however , in order for these perturbations to affect or even generate the pdp the vector field needs to affect the universe expansion , i.e. its density should become dominant ( or nearly dominant ) at some point . but , even if particle production is isotropic the homogeneous zero - mode condensate is not . how can we avoid excessive anisotropic stress when the vector field dominates the expansion ? it turns out that this is possible if the vector field plays the role of the curvaton . consider a minimally coupled massive abelian vector field for which @xmath115 where and . it is clear that both models in the previous section eventually approach the above theory , when and also after the end of inflation when and . the vector field condensate is homogenised by inflation so that . in this case it can be shown that the temporal component is zero and the spatial components satisfy the following equation @xcite @xmath116 which is similar to eq . ( [ eomhom ] ) for the scalar field . the energy - momentum tensor of this theory can be written in the form @xcite @xmath117 where @xmath118 eq . ( [ tmn ] ) is reminiscent of a perfect fluid with the crucial difference that the pressure in the longitudinal direction is of opposite sign compared to the transverse pressure . thus , it seems that if the homogeneous vector field were to dominate the universe it would indeed generate excessive anisotropic stress . therefore , we assume that the vector field remains subdominant during inflation so that ( quasi)de sitter expansion is not spoilt . after inflation the hubble parameter decreases until . when this happens one can ignore the friction term in eq . ( [ veceom ] ) , which therefore suggests that the vector field condensate starts rapid ( quasi)harmonic oscillations . it is easy to show that , during a hubble time , on average , which means that , over many oscillations , the average pressure is zero : @xcite . thus the oscillating vector field condensate behaves as pressureless , _ isotropic _ matter . its density , therefore , scales as which is not as drastic as the radiation background , which scales as . therefore , the oscillating vector field can gradually increase its density parameter and come to dominate ( or nearly dominate ) before its decay ( see fig . [ vecurvevol ] ) . because it is isotropic , it dominates without causing any excessive anisotropic stress , so that the universe expansion remains isotropic @xcite . log - log plot of the evolution of the inflaton energy density which decays into radiation @xmath65 ( purple line ) at the end of inflation ( denoted by ` end ' ) and the vector curvaton energy density @xmath119 ( green line ) ( prompt reheating is assumed ) . during inflation , the vector curvaton density is negligible ( its evolution depends on the model which breaks its conformality ) . after inflation . similarly , decreases also as radiation after inflation when the vector field is light ( in contrast to the scalar curvaton case where @xmath66 remains constant when the curvaton is light , cf . [ curvevol ] ) . however , when , the vector curvaton becomes heavy and begins oscillating , after which time ( denoted ` osc ' ) . at some moment ( denoted ` dom ' ) the vector curvaton density dominates the universe until , some time later ( denoted ` dec ' ) when it decays into the thermal bath of the hot big bang . as the vector curvaton is rapidly oscillating it does not give rise to anisotropic stress at domination . the dashed slanted line depicts the possibility that the vector curvaton decays before domination ( ) , when substantial non - gaussianity can be generated.,width=384 ] from the above we see that a massive abelian vector field can follow the curvaton scenario and play the role of the curvaton without problem . indeed , the perturbations of the vector field imply a perturbation in its local density @xmath120 , which means that in some locations it dominates the universe earlier than in others . this is how it can generate the curvature perturbation according to the curvaton mechanism . it is important to note that , since @xmath119 is a scalar quantity , the curvature perturbation generated is scalar and not vector in nature . also , note that the perturbations of the vector field @xmath121 follow a similar equation to eq . ( [ veceom ] ) , which means that they too are rapidly oscillating and do not introduce anisotropic stress . thus , at domination the perturbations do not cause anisotropic expansion either , not even in a small scale . if particle production of the vector field is anisotropic then there are direction dependent patterns in the amplitude of the oscillating zero mode that lead to statistical anisotropy in the pdp @xcite . in ref . @xcite statistical anisotropy in the bispectrum in the vector curvaton model was investigated . it was found that it manifests itself only quadratically , i.e. the expansion in eq . ( [ fnlanis ] ) is truncated to . @xmath100 was evaluated in the two models discussed in the previous section . in the non - minimally coupled to gravity model in the equilateral configuration , whereas in the squeezed configuration . this means that the angular modulation of @xmath40 is prominent and should be detected if non - gaussianity is found . in the varying kinetic function and mass model when , i.e. if the field remains light until the end of inflation . here we see that non - gaussianity is predominantly anisotropic . should no angular modulation of @xmath40 be observed this possibility will be excluded . however , if , i.e. if the field becomes massive by the end of inflation , particle production is isotropic ( case c ) and , which is identical to the scalar curvaton case ( cf . ( [ fnlcurv ] ) ) . is defined in the same way as @xmath122 . ] the vector curvaton is an elegant mechanism for vector fields to contribute of even generate the pdp since it does not need to couple the fields to the inflaton sector . however , it is by no means the only way that a vector field can affect the curvature perturbation . for example , in ref . @xcite , the end of inflation mechanism is employed ( see sec . [ einf ] ) , where , instead of coupling the waterfall field @xmath72 of hybrid inflation to a scalar field @xmath31 , the authors considered introducing a coupling of the form . this model too produces distinct observational signatures . for example , in this model , where the maximum value for @xmath123 in the squeezed configuration is independent from the model parameters : . cosmic structure originates from the growth of quantum fluctuations during a period of cosmic inflation in the early universe . the particle production process generates an almost scale invariant spectrum of superhorizon perturbations of suitable fields , e.g. light scalar fields . these perturbations give rise to the primordial density / curvature perturbation via a multitude of mechanisms ( e.g. inflaton , curvaton , inhomogeneous reheating etc . ) . observables such as the spectral index @xmath124 or the non - linearity parameter @xmath40 will soon exclude whole classes of models . indeed , the planck satellite observations are expected to increase precision up to . recently , the possibility that cosmic vector fields contribute or even generate the curvature perturbation @xmath26 ( e.g. through the vector curvaton mechanism ) is being explored . if it is so then vector fields can produce distinct signatures such as correlated statistical anisotropy in the spectrum and bispectrum of @xmath26 . wmap observations allow up to 30% statistical anisotropy in the spectrum but the planck satellite mission is expected to reduce this bound down to 2% @xcite , if statistical anisotropy is not observed . anisotropy in the non - gaussianity can be dominant , which means that @xmath40 may feature an angular modulation on the microwave sky . i would like to thank my collaborators mindaugas kariauskas , david h. lyth , yeinzon rodriguez - garcia and jacques m. wagstaff . this work was supported by the european union through the marie curie research and training network universenet " ( mrtn - ct-2006 - 035863 ) . my participation to the neb-14 conference was supported by a royal society conference grant . guth a h 1981 the inflationary universe : a possible solution to the horizon and flatness problems _ phys . rev . _ d * 23 * 347 starobinsky a a 1980 a new type of isotropic cosmological models without singularity , _ phys . lett . _ b * 91 * 99 . sachs r k and wolfe a m 1967 perturbations of a cosmological model and angular variations of the microwave background _ astrophys . j. _ * 147 * 73 ; [ 2007 _ gen . rel . grav . _ * 39 * 1929 ] hawking s w 1975 particle creation by black holes _ commun . phys . _ * 43 * 199 [ 1976 erratum - ibid . * 46 * 206 ] dimopoulos k 2009 where galaxies really come from _ contemp . phys . _ * 50 * 633 ( _ preprint _ 0906.0903 [ hep - ph ] ) gibbons g w and hawking s w 1977 cosmological event horizons , thermodynamics , and particle creation , _ phys . rev . _ d * 15 * 2738 bunch t s and davies p c w 1978 quantum field theory in de sitter space : renormalization by point splitting _ proc . lond . _ a * 360 * 117 komatsu e _ et al . _ 2010 seven - year wilkinson microwave anisotropy probe ( wmap ) observations : cosmological interpretation _ preprint _ arxiv:1001.4538 [ astro-ph.co ] lyth d h , malik k a and sasaki m 2005 a general proof of the conservation of the curvature perturbation _ journal of cosmol . phys . _ jcap05(2005)004 ( _ preprint _ astro - ph/0411220 ) sasaki m and stewart e d 1996 a general analytic formula for the spectral index of the density perturbations produced during inflation _ prog . phys . _ * 95 * ( 1996 ) 71 ( _ preprint _ astro - ph/9507001 ) liddle a r and lyth d h 2000 _ cosmological inflation and large scale structure _ ( cambridge university press ) . lyth d h and wands d 2002 generating the curvature perturbation without an inflaton _ phys . lett . _ b * 524 * 5 ( _ preprint _ hep - ph/0110002 ) enqvist k and sloth m s 2002 adiabatic cmb perturbations in pre big bang string cosmology _ nucl . phys . _ b * 626 * 395 ( _ preprint _ hep - ph/0109214 ) moroi t and takahashi t 2001 effects of cosmological moduli fields on cosmic microwave background _ phys . lett . _ b * 522 * 215 [ 2002 erratum - ibid . b * 539 * 303 ] ( _ preprint _ hep - ph/0110096 ) mollerach s 1990 isocurvature baryon perturbations and inflation _ phys . rev . _ d * 42 * 313 linde a d and mukhanov v f 1997 nongaussian isocurvature perturbations from inflation _ phys . rev . _ d * 56 * 535 ( _ preprint _ astro - ph/9610219 ) dimopoulos k and lyth d h 2004 models of inflation liberated by the curvaton hypothesis _ phys . rev . _ d * 69 * 123509 ( _ preprint _ hep - ph/0209180 ) moroi t , takahashi t and toyoda y 2005 relaxing constraints on inflation models with curvaton _ phys . rev . _ d * 72 * 023502 ( _ preprint _ hep - ph/0501007 ) mcdonald j 2003 right - handed sneutrinos as curvatons _ phys . _ d * 68 * 043505 ( _ preprint _ hep - ph/0302222 ) enqvist k , kasuya s and mazumdar a 2003 adiabatic density perturbations and matter generation from the mssm _ phys . lett . _ * 90 * 091302 ( _ preprint _ hep - ph/0211147 ) enqvist k , jokinen a , kasuya s and mazumdar a 2003 mssm flat direction as a curvaton _ phys . rev . _ d * 68 * 103507 ( _ preprint _ hep - ph/0303165 ) chun e j , dimopoulos k and lyth d h 2004 curvaton and qcd axion in supersymmetric theories _ phys . rev . _ d * 70 * 103510 ( _ preprint _ hep - ph/0402059 ) dimopoulos k and lazarides g 2006 modular inflation and the orthogonal axion as curvaton _ phys . _ d * 73 * 023525 ( _ preprint _ hep - ph/0511310 ) lyth d h , ungarelli c and wands d 2003 the primordial density perturbation in the curvaton scenario _ phys . _ d * 67 * 023503 ( _ preprint _ astro - ph/0208055 ) dvali g , gruzinov a and zaldarriaga m 2004 a new mechanism for generating density perturbations from inflation _ phys . _ d * 69 * 023505 ( _ preprint _ astro - ph/0303591 ) dvali g , gruzinov a and zaldarriaga m 2004 cosmological perturbations from inhomogeneous reheating , freezeout , and mass domination _ d * 69 * 083505 ( _ preprint _ astro - ph/0305548 ) zaldarriaga m 2004 non - gaussianities in models with a varying inflaton decay rate _ phys . _ d * 69 * 043508 ( _ preprint _ astro - ph/0306006 ) linde a d hybrid inflation 1994 _ phys . rev . _ d * 49 * 748 ( _ preprint _ astro - ph/9307002 ) dvali g r , shafi q and schaefer r k 1994 large scale structure and supersymmetric inflation without fine tuning _ phys . rev . lett . _ * 73 * 1886 ( _ preprint _ hep - ph/9406319 ) lazarides g , schaefer r k and shafi q 1997 supersymmetric inflation at the grand unification scale _ phys . rev . _ d * 56 * 1324 ( _ preprint _ hep - ph/9608256 ) lyth d h 2005 generating the curvature perturbation at the end of inflation _ journal of cosmol . astropart . phys . _ jcap11(2005)006 ( _ preprint _ astro - ph/0510443 ) salem m p 2005 on the generation of density perturbations at the end of inflation , _ phys . rev . _ d * 72 * 123516 ( _ preprint _ astro - ph/0511146 ) alabidi l and lyth d h 2006 curvature perturbation from symmetry breaking the end of inflation _ journal of cosmol . _ jcap08(2006)006 ( _ preprint _ astro - ph/0604569 ) land k and magueijo j 2005 the axis of evil _ phys . _ * 95 * 071301 ( _ preprint _ astro - ph/0502237 ) land k and magueijo j 2007 the axis of evil revisited _ mon . not . * 378 * 153 ( _ preprint _ astro - ph/0611518 ) eriksen h k , hansen f k , banday a j , gorski k m and lilje p b 2004 asymmetries in the cmb anisotropy field _ astrophys . j. _ * 605 * 14 [ 2004 erratum - ibid . * 609 * 1198 ] ( _ preprint _ astro - ph/0307507 ) hansen f k , banday a j and gorski k m 2004 testing the cosmological principle of isotropy : local power spectrum estimates of the wmap data , _ mon . not . * 354 * 641 ( _ preprint _ astro - ph/0404206 ) dimopoulos k 2006 can a vector field be responsible for the curvature perturbation in the universe ? _ d * 74 * 083502 ( _ preprint _ hep - ph/0607229 ) dimopoulos k , kariauskas m and wagstaff j m 2010 vector curvaton with varying kinetic function _ phys . _ d * 81 * 023522 ( _ preprint _ 0907.1838 [ hep - ph ] ) ackerman l , carroll s m and wise m b 2007 imprints of a primordial preferred direction on the microwave background _ phys . _ d * 75 * 083502 [ 2009 erratum - ibid . d * 80 * 069901 ] ( _ preprint _ astro - ph/0701357 ) dimopoulos k , kariauskas m , lyth d h and rodriguez y 2009 statistical anisotropy of the curvature perturbation from vector field perturbations _ journal of cosmol . astropart . phys . _ jcap05(2009)013 ( _ preprint _ 0809.1055 [ astro - ph ] ) groeneboom n e and eriksen h k 2009 bayesian analysis of sparse anisotropic universe models and application to the 5-yr wmap data _ astrophys . j. _ * 690 * 1807 ( _ preprint _ 0807.2242 [ astro - ph ] ) groeneboom n e , ackerman l , wehus i k and eriksen h k 2009 bayesian analysis of an anisotropic universe model : systematics and polarization _ preprint _ 0911.0150 [ astro-ph.co ] kariauskas m 2010 quantum fluctuations of vector fields and the primordial curvature perturbation in the universe _ preprint _ 1009.1779 [ astro-ph.co ] . kariauskas m , dimopoulos k and lyth d h 2009 anisotropic non - gaussianity from vector field perturbations _ phys . _ d * 80 * 023509 ( _ preprint _ 0812.0264 [ astro - ph ] ) valenzuela - toledo c a , rodriguez y and lyth d h 2009 non - gaussianity at tree- and one - loop levels from vector field perturbations _ phys . rev . _ d * 80 * 103519 ( _ preprint _ 0909.4064 [ astro-ph.co ] ) dimopoulos k and kariauskas m 2008 non - minimally coupled vector curvaton _ journal of high energy phys . _ jhep 07(2008)119 ( _ preprint _ 0803.3041 [ hep - th ] ) himmetoglu b , contaldi c r and peloso m 2009 instability of anisotropic cosmological solutions supported by vector fields _ phys . * 102 * 111301 ( _ preprint _ 0809.2779 [ astro - ph ] ) himmetoglu b , contaldi c r and peloso m 2009 instability of the acw model , and problems with massive vectors during inflation _ phys . _ d * 79 * 063517 ( _ preprint _ 0812.1231 [ astro - ph ] ) kariauskas m and lyth d h 2010 on the health of a vector field with ( r a@xmath125)/6 coupling to gravity _ preprint _ 1007.1426 [ astro-ph.co ] carroll s m , dulaney t r , gresham m i and tam h 2009 instabilities in the aether _ phys . _ d * 79 * 065011 ( _ preprint _ 0812.1049 [ hep - th ] ) cornwall j m , levin d n and tiktopoulos g 1974 derivation of gauge invariance from high - energy unitarity bounds on the s matrix _ phys . rev . _ d * 10 * 1145 [ 1975 erratum - ibid . d * 11 * 972 ] dimopoulos k , kariauskas k and wagstaff j m 2010 vector curvaton without instabilities _ phys . lett . _ b * 683 * 298 ( _ preprint _ 0909.0475 [ hep - ph ] ) dimopoulos k 2007 supergravity inspired vector curvaton _ phys . rev . _ d * 76 * 063506 ( _ preprint _ 0705.3334 [ hep - ph ] ) dine m , randall l and thomas s 1996 baryogenesis from flat directions of the supersymmetric standard model _ nucl . phys . _ b * 458 * 291 ; 1995 supersymmetry breaking in the early universe _ phys . rev . _ * 75 * 398 dimopoulos k and wagstaff j m 2010 in preparation yokoyama s and soda j 2008 primordial statistical anisotropy generated at the end of inflation _ journal of cosmol . astropart . _ jcap08(2008)005 ( _ preprint _ 0805.4265 [ astro - ph ] ) pullen a r and kamionkowski m 2007 cosmic microwave background statistics for a direction - dependent primordial power spectrum _ phys . rev . _ d * 76 * 103529 ( _ preprint _ 0709.1144 [ astro - ph ] )

m giants recovered from the two micron all - sky survey ( 2mass ) have recently been used to map the position and velocity distributions of tidal debris from the sagittarius ( sgr ) dwarf spheroidal galaxy entirely around the galaxy . we compare this data set to both test particle orbits and n - body simulations of satellite destruction run within a variety of rigid milky way potentials and find that the mass of the milky way within 50 kpc of its center should be in order for any sgr orbit to simultaneously fit the velocity gradient in the sgr trailing debris and the apocenter of the sgr leading debris . orbital pole precession of young debris and leading debris velocities in regions corresponding to older debris provide contradictory evidence in favor of oblate / prolate galactic halo potentials respectively , leading us to conclude that the orbit of sgr has evolved over the past few gyr . in light of this discrepancy , we consider constraints from the younger portions of the debris alone within three models of the flattening of the galactic potential ( 0.90/1.0/1.25 , i.e. oblate / spherical / prolate ) in our further n - body simulations . based upon the velocity dispersion and width along the trailing tidal stream we estimate the current bound mass of sgr to be independant of the form of the galactic potential ; this corresponds to a range of mass to light ratios - 36 for the sgr core . models with masses in this range best fit the apocenter of leading sgr tidal debris when they orbit with a radial period of roughly 0.85 gyr and have perigalactica and apogalactica of about 15 kpc and 60 kpc respectively . these distances will scale with the assumed distance to the sgr dwarf and the assumed depth of the galactic potential . the density distribution of debris along the orbit in these models is consistent with the m giant observations , and debris at all orbital phases where m giants are obviously present is younger ( i.e. was lost more recently from the satellite ) than the typical age of a sgr m giant star .

astro-ph0407566

the sagittarius dwarf spheroidal galaxy ( sgr ) , discovered only a decade ago @xcite , is the most compelling example of a satellite currently being cannibalized by the milky way . there have been numerous studies reporting the discovery of stars and star clusters plausibly associated with debris from this satellite , either trailing or leading it along its orbit ( see majewski et al . 2003 hereafter `` paper i '' for a comprehensive summary ) . using a study of faint , high - latitude carbon stars for which a significant overdensity was found to be aligned in angular position with the projection of sgr s orbit , totten & irwin ( 1998 ) were the first to present data that suggested that the tidal tails of the disrupting sgr system extend a full @xmath5 across the sky . the conclusions of the carbon star study were recently dramatically verified using m giants selected from the 2mass database ( paper i ) . because sgr is relatively metal - rich , m giant stars are prevalent in its debris stream , are far more common than carbon stars , and can be easily identified to distances of more than 50 kpc within the 2mass database . moreover , the large sample of m giants in the core of sgr itself permits a much more reliable distance scale to be derived for these stars than is possible for the carbon stars . as a result , for the first time , primary leading and trailing tidal arms can clearly be traced using the 2mass m giants , with the trailing tail spanning at least 150@xmath6 across the southern galactic hemisphere and the leading tail arcing up to create a rosette orbital loop in the northern galactic hemisphere . follow - up spectroscopy of sgr - candidate stars has determined line - of - sight ( i.e. `` radial '' ) velocities for sgr m giant stars throughout the trailing tail ( * ? ? ? * hereafter paper ii ) , and work on stars in the leading trail is in progress ( * ? ? ? * hereafter paper v ) . a number of groups have sought to model the sgr milky way interaction ( e.g. johnston , spergel , & hernquist 1995 , velazquez & white 1995 , ibata et al . 1997 , edelsohn & elmegreen 1997 , johnston et al . 1999 , helmi & white 2001 , g ' omez - flechoso , fux , & martinet 1999 , mart ' inez - delgado et al . the interaction of the sgr dwarf spheroidal with the milky way offers a sensitive probe of the shape and strength of the galactic potential , and also provides a nearby laboratory for exploring the internal dynamics of satellite galaxies under the strong tidal influence of a parent system . ibata & lewis ( 1998 ) made an extensive series of simulations to match data available at the time . among their models , model k6-a provides the closest match to the general morphology of the sgr tidal tails as mapped by m giant stars selected from 2mass . however , the 2mass m giant work represents such a substantial increase in our knowledge of the phase - space distribution of sgr debris that a new study of the system fully constrained by these data is warranted . recently , a controversy has begun to develop over the oblate / prolate nature of the galactic halo as measured using sgr tidal debris . helmi ( 2004 ) has presented evidence in favor of a prolate ( @xmath1 1.25 ) halo using sgr leading debris velocity trends , while in a companion paper ( johnston , law , & majewski 2004 , hereafter `` paper iii '' ) we have demonstrated that such prolate halos fail to reproduce the observed orbital pole precession of leading vs. trailing debris , for which oblate ( @xmath1 0.90 ) halos best reproduce observational data . in earlier studies , @xcite and @xcite determined that values of @xmath7 1.0 and @xmath1 0.85 respectively best fit the available data . in this paper , we explore whether it is possible to resolve this conflict using a single - component ( i.e. mass - follows - light ) model for sgr , traveling along a single orbit in a non - evolving potential . we present the results of numerical simulations to find the best fit to the measured positions and velocities of the m giants presented in papers i , ii , and v ( preliminary results have been presented in law et al . 2004 ) while allowing orbital , potential and sgr internal parameters to vary . our aim is to constrain the current mass and orbit of sgr as tightly as possible as a precursor to further studies in which higher order effects ( such as multi - component models for sgr travelling along evolving orbits ) are also accounted for . in 2 we describe our simulation technique , and outline the properties of the observed tails that will be used to constrain the simulations . in 3.1 we use simple test particle simulations to examine what range of galactic and orbital parameters could be consistent with sgr debris . in 3.2 we use the results from full n - body simulations of satellite destruction along viable orbits in the chosen galactic potentials to more tightly constrain the mass and orbit of sgr . in 4 we compare our results to previous observational and numerical work and assess possible evolution of the sgr orbit , and in 5 we summarize our conclusions . our simulation technique closely follows that outlined in @xcite . the milky way is represented by a smooth , rigid potential , and sgr by a collection of @xmath8 self - gravitating particles whose mutual interactions are calculated using a self - consistent field code ( hernquist & ostriker 1992 ) . a three - component model is used for the galactic potential and consists of a miyamoto - nagai ( 1975 ) disk , hernquist spheroid , and a logarithmic halo : @xmath9 @xmath10 @xmath11 following johnston et al . ( 1999 ) , we take @xmath12 @xmath13 , @xmath14 @xmath13 , @xmath15 kpc , @xmath16 kpc and @xmath17 kpc . in 3.1 we investigate how different choices @xmath18 , @xmath19 , @xmath20 kpc and @xmath21 km s@xmath22 ( the circular speed at the solar circle @xmath23 in eqn . [ haloeqn ] was chosen to match @xmath24 for a given bulge and disk contribution and adopted @xmath25 ) affect our fit to the debris data . initially , the particles in our model of sgr are distributed to generate a plummer ( 1911 ) model @xmath26 where @xmath27 is the initial mass of sgr and @xmath28 is its scale length . these particles represent both the dark and light matter components of the satellite . we do not attempt to generate a more specific two - component model that matches sgr s internal density and velocity distribution since both will evolve during the simulation . rather , we explore to what extent sgr s debris can constrain the present global characteristics of the satellite . these global characteristics can then be used in a more careful consideration of the core structure in future work when better data on the core are available . in this paper we use the spherical , sun - centered , sgr - coordinate systemsrm4n / sgr/ ] defined in paper i , since this is the coordinate system in which satellite debris is observed and therefore can be compared to simulations most clearly . the zero - plane of the latitude coordinate @xmath29 coincides with the best - fit great circle defined by sgr debris , as seen from the sun . the longitudinal coordinate @xmath30 is zero in the direction of the sgr core and increases along the sgr trailing debris stream , i.e. away from the galactic plane . figure [ xydiag ] shows a representative n - body simulation of the sgr dwarf in cartesian @xmath31 , @xmath32 coordinates ( see paper i for the definition of the sgr , gc and sgr,@xmath33 coordinate systems ) , and illustrates the orientation of the spherical coordinate system with respect to the galactic plane . the colors of the simulated data used in this and other figures in this paper represent different debris `` eras '' , i.e. orbits ( denoted as one apogalacticon to the next apogalacticon ) on which the debris was stripped from the satellite . yellow points represent debris stripped from the satellite since apogalacticon about 0.5 gyr ago , while magenta , cyan , and green points represent debris stripped from the dwarf 2 , 3 , and 4 orbits ago respectively . note that while each color represents debris unbound from the satellite between two successive apogalactic passages , the majority of debris of each color is released during the corresponding perigalactic passage . this color scheme allows us to discriminate readily between different wraps of tidal debris , and is also useful for determining the expected age of debris at any given point along the tidal stream . the observed position and radial velocity data for sgr m giants ( papers i , ii , and v ) provide strong constraints on the orbit of the sgr dwarf . most other sgr detections around the sky fall within the m giant - traced tails ( see fig . 17 of paper i ) ; therefore we compare our models to the m giants alone because they offer the most consistent , wide - ranging map of sgr debris , and at the same time encompass the previous detections . we compare our results to the recently announced sdss detections @xcite in 4.1 . we define eleven observed properties that we adopt as constraints on our simulated milky way sgr system : 1 . the model sgr dwarf should be located at @xmath34 ( paper i ) . the line - of - sight velocity of the model dwarf should be @xmath35 171 km s@xmath22 ( ibata et al . 1997 ) . sgr debris should be aligned with the plane passing through the sun having a pole @xmath36 ( paper i ) . 4 . the average heliocentric distance for sgr leading debris at apogalacticon ( @xmath37 ) in the northern galactic hemisphere should be 42 @xmath38/24 kpc ) kpc , where @xmath39 is the assumed distance to sgr ( which sets the distance scale of the m giants and is taken to be 24 kpc in paper i ) . radial velocities along the trailing stream from @xmath40 - @xmath41 should match data presented in paper ii . radial velocities along the leading stream from @xmath42 - @xmath43 should match data presented in paper v. 7 . the leading and trailing debris tails should define two distinct planes with poles offset from each other by @xmath44 10 degrees ( paper iii ) . the physical width of the trailing debris stream perpendicular to the orbital plane should be consistent with m giant observations ( i.e. have a projected spatial dispersion @xmath45 kpc ) . the average radial velocity dispersion along the trailing stream from @xmath40 - @xmath46 should match the dispersion found for m giants in paper ii ( @xmath47 = 10.0 km s@xmath22 ) . the model debris to which the m giant data are matched should be younger ( i.e. have left sgr more recently ) than a typical sgr m giant age ( 2 - 3 gyr ; see paper i ) . 11 . there should be a break in the surface density of trailing debris at @xmath48 , which has previously been interpreted ( mateo et al . 1998 , paper i ) to correspond to the transition between debris lost on the current pericentric passage and that lost on the previous passage . table [ param.table ] outlines the galactic and satellite parameters varied to produce a model that fits the constraints detailed above . rather than run lengthy n - body simulations to randomly search for a global minimum in this degenerate , multi - dimensional parameter space ( 8 of which are allowed to vary ) , a more efficient , multi - step approach was taken to converge to the best fit to the observational data , relying on physical insight gained both from analytical descriptions of debris dispersal ( tremaine 1993 , johnston 1998 , helmi & white 1999 , johnston , sackett & bullock 2001 ) and from previous modeling of sgr by the authors ( johnston , spergel , & hernquist 1995 , johnston et al . 1999 ) and other groups ( velazquez & white 1995 , ibata et al . 1997 , edelsohn & elmegreen 1997 , g ' omez - flechoso , fux , & martinet 1999 , mart ' inez - delgado et al . 2004 ) . these studies have found that while there is a systematic distance offset for leading / trailing debris inside / outside the orbit of the sgr dwarf ( a reflection of the debris moving to more / less tightly bound orbits see fig . [ xydiag ] ) the line - of - sight velocity remains approximately aligned with that of the satellite s orbit at all orbital phases . hence in 3.1 below we are able to eliminate a wide range of orbits in a variety of galactic potentials through test - particle integrations alone : we use constraint 4 as an upper limit on a possible orbit s apocentric distance and examine how well the line - of - sight velocities along the orbit match the data in constraints 5 and 6 . this technique allows us to find reasonable values for all of the free parameters listed in table [ param.table ] except for sgr s current mass . in 3.2 we describe full - scale simulations of the destruction run for satellites of various masses along the orbits and in the potentials selected in 3.1 . we assume sgr s current angular position , line - of - sight velocity and direction of proper motion to be fixed by constraints 1 , 2 and 3 respectively , and adopt an amplitude for the motion of sgr perpendicular to our line - of - sight ( @xmath49 ) somewhere within @xmath50 times the error bars on the @xcite measurement of @xmath51 km s@xmath22 . the sgr velocity and position relative to the sun are then tranformed to galactocentric coordinates to provide initial conditions for the test particle orbits , assuming some values for the solar distance from the galactic center ( @xmath52 ) and from sgr ( @xmath39 ) . ( note that changing @xmath39 from the assumed value of 24 kpc scales the distances to all of the sgr m giants by the same fractional amount , since these distances are estimated from a color - apparent magnitude relation derived from m giants in sgr s core paper i ) . these orbits are then integrated backwards and forwards in time in the chosen galactic potential ( see 3.1.2 ) and the quality of fit of the orbital path to the m giant position and velocity data quantified ( as described in 3.1.3 ) . we anticipate that sgr s debris will tell us something about the contours of the gravitational potential in the region that its orbit explores ( @xmath44 10 - 50 kpc ) . hence we do not vary all parameters in equations ( 1 ) - ( [ haloeqn ] ) , but instead hold the bulge component fixed and explore the effect of changing the contribution of the disk to the rotation curve through the parameter @xmath53 , as well as the radial length scale , flattening , and overall depth of the halo potential respectively through the parameters @xmath25 , @xmath54 and @xmath24 . as a final check on the generality of our results , we repeat our experiments with the halo component replaced by models of the form proposed by @xcite hereafter referred to as nfw models . in this case , the flattening is introduced in the density @xmath55 , rather than potential contours , and the approximate form of the potential is adopted from @xcite . to explore a similar effective range in @xmath54 and radial gradient as the logarithmic models , @xmath55 and @xmath25 ( the length scale of the nfw potential ) are chosen from a wider range than for their logarithmic counterparts in particular , the range @xmath56 kpc was explored because this encompasses the range of scale lengths ( of order tens of kpc ) found for dark matter halos of similar mass - scale to the milky way in cosmological models of structure formation at the current epoch ( e.g. * ? ? ? the mass scale of the nfw potential is then constrained to match the adopted @xmath24 . a guideline for assessing the goodness of fit of an orbit to the positional data is that the maximum heliocentric distance observed for the leading debris ( @xmath57 , constraint 4 ) must be systematically less than that of the orbit of the sgr core , @xmath58 i.e. @xmath59 . we can also find an upper limit to this ratio since we expect the size of this offset to scale as @xmath60 , where @xmath61 is the mass of the milky way enclosed within the pericenter of the orbit @xcite . for example , if we take this limit as @xmath62 then we might expect to cover all models with @xmath63 i.e. sgr masses up to 10% of the mass of the milky way . since the internal dispersion measured for sgr ( 11 km s@xmath22 , * ? ? ? * ) suggests a mass far less than this we take @xmath64 as a generous range for considering an orbit apogalacticon distance acceptable . orbits with apogalactica outside this range are immediately rejected . we next quantify the fit of orbits that are not already rejected to the trailing and leading velocity data ( constraints 5 and 6 ) through the parameters @xmath65 and @xmath66 : @xmath67_i^2\over{\sigma^2_{\rm a } } } \label{khia}\ ] ] where `` a '' represents the observed data set being considered ( i.e. `` lead '' or `` trail '' ) , @xmath68 is the number of m giants in the data set , @xmath69 is the velocity of an m giant at @xmath70 and @xmath71 is the velocity of the orbit at this @xmath70 . the data compared to in the leading portion of the debris are selected by fitting a 3rd order polynomial to the full data set of velocities as a function of @xmath30 in the range @xmath72 . outliers from the main trend are thrown out using a 2.5-@xmath73 iterative rejection technique until convergence is reached and the weight @xmath74 calculated as the dispersion of the velocities of this final set of @xmath75 stars about the best - fit polynomials . the process is then repeated for stars in the region @xmath76 most sensitive to the trailing debris . the selected stars in both regions are plotted as black squares in figure [ orb.fig ] . clearly , these data sets are not intended to represent a complete sample of sgr stars , but rather as a guide to the general trends of velocities and dispersion in these regions . we also express these quantities as a single parameter to measure the combined goodness - of - fit : @xmath77 note that since test particle orbits only serve as an indication of where the debris should lie , we do not simply search for the parameters corresponding to the minima of these quantities : for example , we do not consider a difference of order @xmath78 ( corresponding to average systematic offsets @xmath79 km s@xmath22 very much less than the dispersion in the data ) between the fit to two different orbits to be very significant . rather , we use more extreme differences to rule out or favor broad regions of parameter space . although the velocity trends in the leading debris ( constraint 6 ) appear to strongly favor galactic models with prolate ( @xmath80 ) halo components @xcite , we have shown in paper iii that the direction of the precession of debris orbits ( as measured by the offset in the poles of best - fit planes to leading and trailing debris constraint 7 ) strongly favors models with oblate halos since prolate models induce precession in the _ opposite _ sense to that observed . because no other adjustment to the potential can change the fundamental sense of precession in prolate _ vs _ oblate potentials , we restrict ourselves to asking whether we can resolve this contradiction between the implications of constraint 6 and constraint 7 by revisiting the fit to the velocity and distance data alone over a much wider range of parameter space than has been considered previously . the aim is to examine whether there are any circumstances in which an orbit in an oblate potential can be found that can fit all the constraints at once . figure [ khi.fig ] plots the minimum values of @xmath65 ( solid lines ) , @xmath66 ( dotted lines ) and @xmath81 ( dashed lines ) obtained as a function of @xmath54 ( left hand panel , logarithmic halo model ) or @xmath82 ( right hand panel , nfw halo model ) when all other parameters are allowed to vary freely within the ranges outlined in table [ param.table ] . the solid lines show that the trailing velocity data have a slight preference for models with oblate halos , although the difference @xmath83 between the minima for models with @xmath84 and @xmath80 is not sufficiently large that we can confidently rule out prolate models with test particle orbits alone , since it corresponds to a velocity offset much less than the dispersion in the data . in contrast , the dotted lines show that leading velocity data strongly prefer prolate halo models , to such an extent that this preference dominates the combined @xmath81 ( dashed lines ) . these results are the same for the logarithmic and nfw models . overall , we conclude that we can not find a single orbit in a static potential model that simultaneously fits the velocity data in the trailing data together with the sense of precession suggested by the offset of the planes of the leading _ vs _ trailing data . the exciting implication of the conclusion of the previous section that no _ single _ orbit and/or potential can fit all the data is that some evolution of sgr s orbit has occurred over the time since debris in the leading portion of the streamer , furthest in @xmath85 from sgr , was released . we discuss some possible culprits for this orbital evolution in 4.3 , but defer a detailed investigation of these effects for future work . for the remainder of this study , we narrow our present analysis to concentrate on the younger portions of the debris , lost within the last 1 - 2 orbits , where ( 1 ) the effect of orbit evolution is negligible , ( 2 ) the modelling can be acheived with the fewest free parameters , and ( 3 ) the interpretation of the data is less ambiguous . the goal is to ask what the younger debris alone can tell us about the galactic potential and sgr s current mass and orbit . these results can subsequently be used as starting points for studies that use the older debris to examine higher order effects such as orbital evolution , evolution of the potential , and/or multi - component models for sgr . we expect debris in the trailing streamer in the range explored by the velocity data to be roughly the same age as that in the early parts of the leading streamer to about the first apocenter ( as demonstrated by * ? ? ? * and see also 3.2 below ) . in these regions , the velocity data can be similarly fit by both oblate and prolate potentials ( as demonstrated by the solid lines in figure [ khi.fig ] ) , and there is no significant offset in the orbital poles between the leading and trailing components . hence , we now drop constraints 6 and 7 on our models since these were derived from regions where orbit evolution could be significant . we continue our discussion of test - particle constraints on the galactic potential and our position relative to the galactic center and sgr using the condition @xmath86 and examining @xmath65 alone . in order to sort through our large parameter space , we first look at parameters that do not appear to be strongly constrained by the data and fix reasonable values for those ( see discussion in a. and b. below ) before going on to look at preferred ranges for the remaining parameters ( in c. ) . in the upper left hand panels of figures [ ktrail_log.fig ] ( for logarithmic halo experiments ) and [ ktrail_nfw.fig ] ( for nfw halo experiments ) we project results in our 7-dimensional parameter space onto the 2-dimensions of @xmath39 and @xmath87 by plotting the minimum value of @xmath65 at each point in this plane when all other parameters are allowed to vary freely . the plots reveals a preference for larger values of the ratio @xmath88 , with the absolute scale ( as set by @xmath39 ) being arbitrary . for consistency with the distance scales adopted earlier in paper i we choose to take @xmath89 kpc ( which also lies within the 2-@xmath73 error bars of the recent measurement by * ? ? ? * ) and set @xmath90 kpc . so long as @xmath91 we expect all subsequent results involving distances ( e.g. scale - length of the halo @xmath25 , or predicted distances to debris ) can be scaled by whatever value @xmath39 is assumed in a given study . with @xmath89 kpc and @xmath90 kpc fixed , the upper right hand panels of figures [ ktrail_log.fig ] and [ ktrail_nfw.fig ] project the remaining five - dimensions of parameter space onto the @xmath24 - @xmath53 plane . for high enough @xmath24 there is no preference for a particular @xmath53 , but models with lower @xmath24 are inconsistent with heavier galactic disks ( i.e. higher @xmath53 ) . figure [ rot.fig ] offers some clue as to why this is the case by plotting rotation curves for only those potentials in which orbits with @xmath92 could be found . these are very flat out to large radii for all models , with circular velocities at 50 kpc in the range 180 - 220 km s@xmath22 ( which corresponds to enclosed masses for the milky way at these radii of @xmath93 in effect , sgr debris velocities are now providing additional evidence for the existence of a dark matter halo to the milky way ) . if enough of the contribution @xmath94 is provided by the disk , then the remaining halo component is simply not massive enough to support such an extended flat rotation curve . ( larger mass halos could be built by allowing @xmath25 an even wider range , but these models would [ i ] have rising rotation curves at the solar circle ; and [ ii ] be inconsistent with scale lengths measured for milky way - sized dark matter halos formed in cosmological models of structure formation see eke , navarro & steinmetz , 2001 ) . since no values of @xmath53 and @xmath24 are at this point clearly preferred , we adopt @xmath95 and @xmath96 km s@xmath22 . the colored lines in the lower panels of figures [ ktrail_log.fig ] and [ ktrail_nfw.fig ] demonstrate that , with @xmath90 kpc , @xmath89 kpc , @xmath97 and @xmath96 km s@xmath22 fixed , particular values for @xmath25 ( which determines the radial gradient of the potential and hence the shape of the rotation curve ) and @xmath49 ( which determines the scale of the orbit within this potential ) are quite strongly preferred , with only a mild dependence on @xmath54 . hence we perform full n - body simulations in potentials with logarithmic halos in which @xmath98 , @xmath99 kpc ( from the minima in the lower left hand panels ) and @xmath49 in the range @xmath100 km s@xmath22 around @xmath101 km s@xmath22 . all three values , @xmath98 , are considered since all represent equally viable fits to the younger debris . clearly , our choices are not unique . the black curves in the lower panels of figures [ ktrail_log.fig ] and [ ktrail_nfw.fig ] outline where the colored lines would fall if all other parameter choices were the same but @xmath102 km s@xmath22 ( dashed lines ) or @xmath103 ( dotted lines ) . in both cases , the scale - length changes significantly in order to maintain the necessary flatness of the rotation curve , and @xmath49 is similarly affected . in addition , our decision to use logarithmic halos rather than nfw halos is arbitrary , since figures [ ktrail_log.fig ] and [ ktrail_nfw.fig ] reveal no preference for either form of the potential , but rather more generally indicate that any model that generates a flat rotation curve out to 50 kpc will suffice . we anticipate that data exploring even larger distances from the galactic center will be able to address whether an nfw ( with a falling rotation curve in this region ) or logarithmic potential is more appropriate . despite these multiple minima in parameter space , we are able to reach some general conclusions at this point : ( i ) sgr debris data prefers models with large values of @xmath88 and flat rotation curves out to 50 kpc , and ( ii ) with all other parameters fixed , sgr orbits in prolate halos will have systematically lower @xmath49 than in spherical or oblate halos . these conclusions offer a tantalizing glimpse of how sgr debris might be used to map out the galactic potential on large scales once parameters such as @xmath52 and @xmath49 are known with more certainty . using the galactic parameters determined in 3.1 above , we now perform fully self - consistent n - body simulations to refine the estimates obtained in 3.1 of sgr s orbital velocity and to determine the mass of the dwarf . these simulations follow the evolution of satellites with a range of initial masses and physical scales ( varied through the parameters @xmath104 and @xmath105 in equation [ [ plummereqn ] ] ) along a small range of plausible orbits within the three models of the galactic potential ( @xmath1 0.9/1.0/1.25 ) discussed in 3.1.5c . in 3.2.1 we find the mass of sgr ( independent of @xmath105 ) that best fits constraints 8 and 9 in each of these three models of the galactic potential , and demonstrate that this best - fit mass is common to all three cases . fixing the satellite mass to this best - fit value , we refine our estimate for sgr s tangential velocity using constraints 4 and 5 in 3.2.2 and summarize the properties of our best - fit models in 3.2.3 . while we do not attempt to model the sgr core in detail , we are nonetheless able to constrain its current total mass under the assumptions that the dwarf is roughly spherical and non - rotating . motivated by previous work ( e.g. , johnston , hernquist , & bolte 1996 , johnston 1998 ) we expect that debris width ( constraint 8) and velocity dispersion ( constraint 9 ) at a given orbital phase primarily reflect the mass within the tidal radius of the satellite on the orbit immediately prior to that debris becoming unbound , and that they do not depend strongly on the internal structure of the satellite ( in our case parameterized by the scale length of the initial plummer model ) . for the same reasons , we do not expect that our results are strongly sensitive to the particle distribution we have adopted . we do expect the internal orbital distribution will independently affect debris morphology , but do not address that issue in this paper . to compare the simulations to the data constraints , we calculate the average radial velocity dispersion @xmath106 and the average dispersion of distances perpendicular to the sgr plane @xmath107 in the trailing tail for m giant data and our numerical simulations . we do not consider leading debris in obtaining our mass estimates since only our prolate halo model successfully matches the _ trend of leading debris , while all three halo models reproduce the trailing debris trend . @xmath106 is calculated in the range @xmath40 - @xmath46 for consistency with the velocity dispersion analysis presented in paper ii , while @xmath107 is calculated in the range @xmath108-@xmath109 since this range of debris longitudes is one for which all sgr stars in the sample , @xmath110 , @xmath111 , @xmath112 , and 13 kpc @xmath113 40 kpc . ] are at a similar distance @xmath114 from the sun ( this minimizes artificial width inflation on the sky due to differential distance errors ) and is also in a region of the galaxy where sample contamination by milky way disk stars is negligible . figure [ mplot ] plots the calculated velocity dispersion ( left - hand panels ) and width ( right - hand panels ) as functions of simulated bound satellite mass for choices of @xmath1 0.9 ( lower panels ) , 1.0 ( middle panels ) , and 1.25 ( upper panels ) . in all panels the m giant dispersion / width is plotted as a solid line with 1-@xmath73 error bars indicated by the hatched regions , while the points in all panels indicate n - body simulation results ( incorporating a 17% artificial distance scatter to simulate the photometric distance errors given in paper i ) for model satellites evolved along the orbits found earlier in 3.1.5c for a variety of choices of initial satellite mass ( @xmath115 @xmath13 - @xmath116 @xmath13 ) and physical scale ( @xmath117 0.2 kpc - 1.5 kpc ) . clearly , similar values of @xmath118 are preferred for models in oblate , spherical , and prolate galactic potentials alike . to quantify more precisely the range of acceptable masses indicated by figure [ mplot ] we fit the data points in each panel with a third - order polynomial with 2.5-@xmath73 rejection criteria iterated to convergence and extrapolate from the resulting power - series coefficients the mass range whose @xmath106 and @xmath107 lie within the 1-@xmath73 uncertainty range around the m - giant measurements . these results , presented in tabular form in table [ mass.table ] , indicate that in all models of the galactic potential considered the present bound mass of the sgr dwarf should not be very different from @xmath119 2 - 5 @xmath120 if the model dwarf is to successfully reproduce the m giant observations . we now fix the initial mass and scale of the model dwarf such that the present - day dwarf has a bound mass in the range found above in 3.2.1 , and endeavor to refine our orbits using the single remaining free parameter @xmath49 . we explore a range of values @xmath100 km s@xmath22 around the values @xmath121 280/270/254 km s@xmath22 chosen from test - particle orbits previously in 3.1.5c . note that it is not possible to fix the final bound mass of the satellite in these simulations , since the change in the orbital path produced by varying @xmath49 will naturally affect the mass - loss history of the model dwarf . however , as demonstrated by figure [ mplot ] ( filled triangles ) these small variations in @xmath49 have only a minor effect on the final mass of the model dwarf . returning to constraint 4 on the average apogalacticon distance of leading debris , the average distance of observed leading sgr debris ( @xmath37 ) is calculated from the 2mass database by averaging over the distances of all stars in the range @xmath122 - @xmath123 with heliocentric distances @xmath124 kpc @xmath113 60 kpc and subject to the restrictions @xmath125 , @xmath110 , @xmath111 kpc , @xmath126 kpc ( this combination of restrictions was chosen to separate sgr leading arm stars most clearly from the underlying disk population ) . figure [ davgplot ] plots the average apogalacticon distance of the m giants as a solid line with 1-@xmath73 error bars indicated by the hatched region , along with the values calculated from the simulated data ( again incorporating a 17% distance uncertainty ) for the simulations with fixed initial mass and physical scale but varying @xmath49 ( filled triangles ) . simulations with a range of initial masses and physical scales whose present bound mass falls within the acceptable range found in the previous section are also plotted ( filled squares and crosses ) : these points are difficult to distinguish since @xmath118 and @xmath28 are not the primary factors governing the behavior of @xmath37 , demonstrating the minor variation in @xmath37 permitted by the remaining uncertainty in satellite mass . while figure [ davgplot ] shows a strong correlation between leading debris distance and orbital velocity however , the relatively large uncertainty in the m giant debris distance allows us only to place constraints on the dwarf velocity to within about @xmath100 km s@xmath22 . a more compelling velocity constraint can be obtained by again using constraint 5 , that the trailing arm velocities match those observed for m giants . we calculate the average offset of the centroid of simulated trailing debris velocities - @xmath46 . ] from the m giant centroid and plot these offsets as a function of the tangential velocity of the dwarf in figure [ voffsetplot ] . well defined minima corresponding to the best fits to the velocity data are obtained for specific velocities in each choice of the galactic potential , and are fairly insensitive to the remaining uncertainties in satellite mass ( filled squares and crosses ) . we therefore conclude that the best choices of tangential velocity for the model dwarf are @xmath121 275 - 280/265 - 270/250 - 260 km s@xmath22 ( note that , in this case , the best - fit test particle orbits obtained in 3.1.5c actually picked out the best orbits for the n - body simulations ) . although each of these estimates are reasonably consistent with the observed value @xmath127 280 @xmath128 20 km s@xmath22 measured by ibata et al . ( 2001 ) , it is interesting to note that the ibata et al . ( 2001 ) measurement appears to slightly favor oblate models of the galactic halo over prolate models at the 1-@xmath73 level for our current choice of @xmath129 220 km s@xmath22 . note , however , that a higher value of @xmath24 will systematically shift these estimates of @xmath49 to higher velocities ( see figs . 4 & 5 , dashed line in lower right - hand panels ) , resulting in better agreement of estimates of @xmath49 in prolate halos with the ibata et al . ( 2001 ) measurement . based upon figures [ mplot ] , [ davgplot ] , and [ voffsetplot ] , simulations with @xmath119 2.6 - 5.0/2.5 - 5.3/2.5 - 5.5 @xmath120 and @xmath121 275 - 280/265 - 270/255 - 260 km s@xmath22 best fit our constraints , and these models are hereafter referred to as our `` best - fit models''srm4n / sgr/ to aid future comparisons of these models with new observations and new disruption models . ] . although the uncertainty in the galactic potential gives rise to uncertainties in @xmath49 considerably greater than the ranges given here , within a given potential @xmath49 can be constrained to within about @xmath128 5 km s@xmath22 . our best - fit models have a maximum extent of bound material @xmath130 along the semi - major axis , within which we calculate a luminosity for sgr of @xmath131 using data presented in paper i. the mass - to - light ratio of sgr in these models should therefore be @xmath132 19 - 36/18 - 38/18 - 39 @xmath133 . while the 500 maximum extent for bound material is somewhat dependent on the adopted internal structure of the satellite , it is on the order of the true tidal radius previously pointed out ( 4.3.3 of paper i ) as required to avoid sgr having a quite extraordinary ( and unlikely ) bound mass , and is also of order the observed _ minor _ axis dimension ( i.e. , 0.35 times that of the 1801 major axis radius ) of the limiting radius of the fitted king profile to the central satellite . these orbits have periods of 0.85/0.88/0.87 gyr with perigalactica and apogalactica of 10 - 16/14/14 - 19 kpc and 56 - 58/59/56 - 59 kpc respectively , and a present space velocity @xmath134 @xmath135 ( 238 , -42 , 222)/(235 , -40 , 213)/(231 , -37 , 198 ) km s@xmath22 , corresponding to @xmath136(230 , 75 , 222)/(227 , 73 , 213)/(224 , 69 , 197 ) km s@xmath22 and @xmath137 ( 171 , 272 , -65)/(171 , 263 , -63)/(171 , 247 , -59 ) km s@xmath22 with respect to the galactic standard of rest . these velocities will scale roughly with the assumed value of @xmath24 , although will also depend systematically upon @xmath54 , @xmath25 , and @xmath53 . figure [ simplot1 ] plots simulated sgr debris for our best - fit models along with the m giant distance and velocity data from papers i , ii and v , and demonstrates visually that our models generally fit the m giant observations well . the m giant data is clearly traced by debris released during the last two pericentric passages of the model dwarf ( yellow and magenta points ) and possibly by debris released three pericentric passage ago ( cyan points ) , although there appear to be far fewer m giants corresponding to cyan points than magenta or yellow . this corresponds to m giants becoming unbound from the sgr dwarf over the last 1.5 - 2.5 gyr consistent with constraint 10 , that the debris age be younger than the typical age of an m giant star . note , however , that as predicted by the orbits in 3.1 models in oblate and spherical halo potentials fail to fit the leading velocity trend ( particularly for cyan points ) , while the model orbiting in a prolate potential both reproduces this velocity trend and provides a more convincing fit to the apparent trend of m giant distances at @xmath138 - @xmath139 . note also the presence of cyan and green debris within a few kpc of the sun over a wide range of @xmath30 for oblate and spherical halo models - this is a consequence of the leading streamer diving almost directly through the solar neighborhood in these two models . conclusive proof of the presence or absence of sgr debris around the sun would provide a significant additional constraint on the models . the density of stars in the trailing stream for the best - fit models is plotted as a function of @xmath30 in figure [ denplot ] , and is similar in structure to the density of the m giant stream ( constraint 11 , plotted in fig . 13 of paper i ) , with a break in the slope of the density profile around @xmath140 degrees and a relatively constant density thereafter ( we only consider this first break in the observed density profile since we expect this to depend primarily upon satellite mass ) . the details of the run of density along the trailing streamer will depend on the internal light distribution of the satellite . however , since we consider only single - component models in this paper , we omit further consideration of the density profile and internal structure of the dwarf at this time . as noted in 2.2 , most other sgr detections around the sky fall within the m giant - traced tails ( see fig . 17 of paper i ) , so that our best - fit models also provide a good match to these other data . in this section , we compare our predictions for older sgr debris ( green points ) not traced by the m giants with observations of older tracers . in figure [ simplot2 ] , carbon star data and @xmath141 . ] ( open boxes ) are plotted for comparison with our best - fit sgr models ( colored points ) . while some of the carbon stars appear consistent with both m giant and simulated debris , many others have distances and velocities that differ substantially from the m giant and model distributions , and attempts to fit simulation models to these carbon stars will likely produce results that differ noticeably from our own best - fit models and the m giant data . although some of this discrepancy could be due to the uncertain distance scale for the carbon stars ( see 8.3 of paper i ) , it is also possible that these stars could trace debris older than the @xmath44 2.5 gyr old m giant stream , since carbon stars can have larger ages ( 5 - 6 gyr ) than m giants . the open triangles near @xmath142 in figure [ simplot2 ] represent data for a set of metal - poor , k - giant stars first pointed out by @xcite . using semi - analytical modeling , @xcite proposed that these stars represent debris stripped from sgr three pericentric passages ago ( corresponding to cyan - colored points in our model ) . indeed , our model suggests that these points may plausibly be fit by cyan or green debris ( i.e. debris from 3 - 4 pericentric passages ago ) in the @xmath1 0.90 leading streamer that is currently raining down from the north galactic pole onto the solar neighborhood , although the interpretation of these data is uncertain in models where @xmath1 1.0 or 1.25 . we also note an interesting comparison with possible sgr red clump stars detected in a pencil - beam survey by majewski et al . ( 1999 ) at @xmath143 , and for which the radial velocity data are plotted in figure [ simplot2 ] ( top panel , solid triangles ) . these stars at @xmath144 exhibit a range of line - of - sight velocities from 0 to 150 km s@xmath22 , which closely matches the predicted range of velocities of simulated leading tidal debris wrapped almost @xmath5 in orbital longitude from the sgr dwarf ( cyan and green points ) for simulations where @xmath1 1.0 or 1.25 . the degree of this agreement is highly sensitive to the mass of the model satellite : simulations with present mass @xmath145 predict a larger dispersion in velocities than observed by majewski et al . , while simulations with mass @xmath146 predict a smaller dispersion than observed . it is tempting therefore to point to these data as further evidence in favor of the satellite mass estimates determined earlier in 3.2.1 however , the distance to these stars is measured to be roughly 20 kpc ( majewski et al . 1999 ) about half that of the cyan - green leading debris whose velocities they reproduce so well and therefore , while they are interesting to compare to model data , their true origin and interpretation remains unclear . recently , the discovery of an overdensity of a - colored stars in the sloan digital sky survey with apparent magnitude @xmath147 at @xmath148 degrees and within 15 kpc of sgr s nominal orbital plane was announced ( newberg et al . these authors estimate an average heliocentric distance of 83 kpc to these stars , but note that other detections in directions which overlap the m giant stream suggest that their adopted distance scale is 12.5% larger than that used to calibrate the m giants in paper i. the open circle in figure [ simplot2 ] ( left - hand panels ) plot the average of their data , with the distance rescaled to 73 kpc so that the m giant and sdss distance scales match . figure [ simplot2 ] suggests that it is plausible to identify the sdss detection with debris of age @xmath44 1.5 - 2.5 gyr ( i.e. cyan - colored points ) in the trailing sgr stream , although future radial velocity measurements could help determine whether this identification is correct or if the newberg et al . feature is instead a part of some older , more distant section of the stream or even halo substructure unrelated to sgr . newberg et al . ( 2003 ) also note a hint of precession in the sgr stream by comparing their detections of leading and trailing debris closer to sgr s core , in agreement with our own results presented in paper iii . unfortunately , the angular extent of the 83 kpc debris has not yet been mapped accurately enough to pinpoint the angular position of the centroid of the debris ; such a measurement could in the future provide a strong constraint on the flattening of the galactic potential . previous attempts to model the orbit and disruption history of the sgr dwarf ( e.g. velazquez & white 1995 , johnston , hernquist & bolte 1996 , ibata et al . 1997 , edelsohn & elmegreen 1997 , ibata & lewis 1998 , g ' omez - flechoso , fux & martinet 1999 , johnston et al . 1999 , helmi & white 2001 , ibata et al . 2001 , mart ' inez - delgado et al . 2004 ) have made considerable progress in constraining models of the dwarf using only the previously available pencil - beam detections of satellite debris . in this paper we have presented the first model based upon a complete all - sky view of the satellite s tidal streams , and in this section we review and compare some of the predictions of these earlier models to those of our own best - fit models . we first consider those results for which the majority of simulations by different groups have generally converged . almost all simulations agree that the radial period of the sgr dwarf should be about 3/4 gyr : in this work we find a period for our best - fit models of 0.85/0.88/0.87 gyr , in reasonable agreement with previous estimates of 0.76 gyr ( velazquez & white 1995 , ibata et al . 1997 ) , 0.7 gyr ( ibata & lewis 1998 ) , 0.55 - 0.75 gyr ( johnston et al . 1999 ) , 0.85 gyr ( helmi & white 2001 ) , and 0.74 gyr ( mart ' inez - delgado et al . there is a little more spread in the estimates proposed by different groups for the perigalacticon and apogalacticon distances of the dwarf s orbit : previous estimates include ( respectively ) 10 kpc and 52 kpc ( velazquez & white 1995 ) , 15 kpc and 60 kpc ( ibata & lewis 1998 ) , 15 kpc and 70 kpc ( g ' omez - flechoso , fux & martinet 1999 ) , 13 kpc and 41 kpc ( johnston et al . 1999 ) , 16 kpc and 60 kpc ( ibata et al . 2001 ) , and 12 kpc and 60 kpc ( mart ' inez - delgado et al . 2004 ) . with the 2mass database it is possible to measure the apogalacticon of leading tidal debris directly , and we match this constraint best by using models for sgr that have orbits with perigalacticon and apogalacticon distances of 10 - 16/14/14 - 19 kpc and 56 - 58/59/56 - 59 kpc respectively . we note , however , that the distance scale assumed for the m giants in paper i is not yet secure , and that the estimated size of sgr s orbit may scale accordingly . among those areas in which common values among the disruption models presented by various groups have not yet been found , perhaps foremost is the @xmath149 component of the galactic @xmath134 velocity of the sgr dwarf . some simulations ( e.g. , ibata et al . 1997 ) have simply set @xmath150 km s@xmath22 ( thereby assuming a polar orbit ) since this component was so poorly known . now that we have an accurate measurement of sgr s orbital pole ( paper i ) , we are able to predict the direction of its motion more precisely . based on our best - fit model , we predict that the proper motion of the sgr dwarf should be @xmath151 mas yr@xmath22 and @xmath152 mas yr@xmath22 in the solar rest frame km s@xmath22 relative to the lsr , for which we adopt a rotation velocity of 220 km s@xmath22 ( 3.1.5b ) . ] . the direction of this proper motion prediction ( @xmath153 ) is expected to be fairly robust within potentials with each choice of @xmath54 . however , the amplitude of the proper motion will depend on the exact form of the galactic potential , and hence should be revised once other fundamental galactic parameters such as @xmath52 and @xmath24 are known more precisely . conversely , as more accurate measurements of sgr s proper motion become available it will be possible to refine constraints on the galactic rotation curve . a second area of debate concerns the present bound mass of the sgr dwarf , for which estimates range from @xmath154 ( mart ' inez - delgado et al . 2004 ) to @xmath155 ( ibata et al . helmi & white ( 2001 ) find an intermediate value for a purely stellar satellite model with initial mass @xmath156 . as demonstrated in 3.2.1 , we find that a range of final masses @xmath157 - @xmath158 yield tidal tails whose thickness and velocity dispersion are consistent with m giant measurements in oblate , spherical , and prolate models of the galactic potential . using figure [ mplot ] we conclusively rule out models with a mass far outside this range ( such as that of mart ' inez - delgado et al . 2004 ) , since models with very high or low masses will not be able to produce tidal tails with the observed thickness and dispersion . visual inspection of the figures in mart ' inez - delgado et al . ( 2004 ) appears to contradict this statement . however , these authors simulation embeds the model satellite in a 40,000 particle live halo , which is probably responsible for the width of the debris stream : earlier work ( johnston , spergel & haydn 2002 ) has found that significant heating of a sgr - like debris stream can occur in a simulation using a live halo , even in a halo model realized with @xmath159 particles . as another consequence of the smaller satellite mass used in their model , mart ' inez - delgado et al . ( 2004 ) predict leading debris at @xmath160 ( corresponding to @xmath161 ) to be composed of stars which have been unbound from the satellite for 5 gyr or more ( since debris from lower - mass satellites takes longer to spread along the orbit ) , in contrast to the roughly 2 gyr found by our own analysis . as demonstrated in figure [ simplot1 ] , the sgr m giants which have an estimated age of 2 - 3 gyr are visible to at least this point in the leading tidal stream . as mart ' inez - delgado et al . ( 2004 ) point out ( and we discuss in paper i ) , stellar populations formed in the densest central regions of the satellite should not be immediately reflected in the tidal streams , and it will take some time for these stars to be present in any quantity in the outer regions of the satellite . hence , it is unlikely that the m giant population became mixed into the outer regions of sgr within a small fraction of a gyr , and we consider the mean age estimate of 5 gyr for this section of the tidal stream to be too high . in paper iii we showed that only galactic potentials with oblate halos could reproduce the precession of the orbital plane apparent in the leading _ vs _ trailing data sets . in contrast , @xcite demonstrated that only galactic potentials with prolate halos could reproduce the velocity trends in the leading debris . in this paper ( 3.1 ) we explore a much wider variety of galactic potentials than has been considered previously but fail to find a single orbit that can fit both the velocity trends and sense of precession . our conclusion is that the assumption of non - evolution of the orbit over the time - period that the debris explores is incorrect . since simulated debris in the region with troublesome velocities is cyan and green ( lost 2 and 3 orbits ago respectively ) , we estimate the timescale over which the evolution has taken place to be @xmath162 gyrs . we can get some idea of the physical scale of the evolution necessary by looking at the difference between the orbits in prolate , spherical and oblate potentials that is responsible for the difference in the velocity trend . figure [ orbxz.fig ] plots the orbits shown in figure [ orb.fig ] in galactic coordinates with the region corresponding to the leading debris velocity data shown as bold along each curve . as the potential moves from prolate to oblate , the orbit passes progressively nearer the sun and line - of - sight velocities more closely reflect the full motion along the orbit . this explains why the simulated line of sight velocities in this region become more extreme with the oblateness of the potential . figure [ orbxz.fig ] also suggests that observed debris velocities in the leading region might be accounted for even in an oblate or spherical potential if the pericenter of sgr s orbit has decreased by a factor of order unity within the last 2 - 3 gyrs ( from visual inspection of the figure ) since such a decrease in pericenter of the sgr _ dwarf _ over time could shift older sgr _ debris _ out to greater distances from the sun corresponding to the greater pericenter of the dwarf on the passage on which the debris became unbound . three factors could contribute to this evolution : an encounter with a large lump in the milky way potential , either dark or luminous ( e.g. such as the large magellanic cloud , see * ? ? ? * for a full description of this idea ) : we consider this unlikley since we would expect the signature of such an event to be a sudden change in sgr s orbit , and a corresponding sudden change in the velocities along its debris , rather than the smooth trends seen . global evolution of the galactic potential : we also consider this unlikely since : ( i ) the evolution would have to be very large in order to bring the pericenter inwards by a factor of two in such a short amount of time ; and ( ii ) any global evolution would affect both sgr s and the debris orbits similarly . dynamical friction : if we re - arrange equation [ 7 - 27 ] from @xcite we can find the mass necessary @xmath163 for a circular orbit at @xmath164 kpc ( i.e. to represent an orbit with of order unity larger pericenter than sgr today ) to decay to the center of the galaxy over a time period @xmath165 gyrs in a galaxy with a flat rotation curve and @xmath166 km s@xmath22 : @xmath167 @xcite estimate @xmath168 for the combined large and small magellanic clouds . since we expect @xmath169 , and know the current mass of sgr to be 2 - 5 @xmath120 , we expect @xmath170 to be the relevant range for our own estimate and hence @xmath171 . moreover , we consider this only a lower limit on the necessary mass since sgr s orbit is not circular . ( see * ? ? ? * ; * ? ? ? * for a general discussion of dynamical friction acting on sgr over a hubble time . ) although dynamical friction seems like the most favourable explanation for the orbit evolution it does require sgr to be an order of magnitude more massive just 2 gyrs ago and debris lost at that time in our mass - follows - light models would have a correspondingly larger dispersion in velocity ( by a factor of order @xmath172 ) and distances . since the observed velocity dispersion in the debris in the discrepant , leading portion of the stream is actually quite similar to that seen in our simulations ( @xmath173 km s@xmath22 , see fig . [ simplot1 ] ) this suggests that , in order to fit the data , in addition to dropping our assumption of a single orbit for sgr , we will also have to move beyond modelling sgr as a _ single component _ system . hence , while the mean trend in the leading streamer will tell us how much total mass needs to have been lost from sgr , the low dispersion offers the additional opportunity of constraining how much more tightly bound the light matter is compared to the dark matter . a study of these combined effects is in progress . in this paper we have presented the first model of the tidal tails of the sgr dwarf galaxy based upon a coherent , all - sky picture of the system in both position and radial velocity , as represented by m giants selected from the 2mass database . we summarize our conclusions as follows : shape and evolution of the galactic potential in a companion paper ( paper iii ) we have shown that _ oblate _ ( @xmath1 0.90 ) models of the galactic halo potential best reproduce the difference in orbital poles between leading and trailing m giant tidal debris , while in this paper ( see also @xcite ) we find that _ prolate _ ( @xmath1 1.25 ) models are required to reproduce the trend of observed m giant leading debris velocities . although we explore a wide variety of galactic potentials we fail to find a single orbit that can simultaneously reproduce the observed orbital pole precession and leading debris velocity trend , and conclude that some evolution of the orbit of sgr has occured over the past few gyr . mass of the sgr dwarf the present bound mass of the sgr dwarf has been restricted to the range @xmath157 - @xmath158 , constrained by the width and velocity dispersion of the trailing m giant tidal tail . taking @xmath131 as the luminosity of sgr , this gives a range of possible values for the mass - to - light ratio of sgr from @xmath174 - 36 . although all of our models maintained cores of bound material , we would expect similar dispersions to be seen in the debris if a similar amount of recently unbound ( i.e. on the current pericentric passage ) mass were present within the same distance scale . orbit of sgr the sgr orbit in our best - fit models ( @xmath1 0.90/1.0/1.25 ) has a pericenter of 10 - 16/14/14 - 19 kpc , an apocenter of 56 - 58/59/56 - 59 kpc and a radial time period of 0.85/0.88/0.87 gyr . these values depend on the distance scale adopted for the m giants and the exact form of the galactic potential . proper motion of sgr for our best - fit models ( @xmath1 0.90/1.0/1.25 ) , the tidal tails of the dwarf as traced by 2mass m giants are best reproduced by a satellite situated at @xmath175 ( 16.2 , 2.3 , -5.9 ) kpc with velocity tangential to the line - of - sight @xmath121 275 - 280/265 - 270/255 - 260 km s@xmath22 corresponding to space velocities @xmath176 ( 238 , -42 , 222)/(235 , -40 , 213)/(231 , -37 , 198 ) km s@xmath22 , i.e. @xmath136(230 , 75 , 222)/(227 , 73 , 213)/(224 , 69 , 197 ) km s@xmath22 , @xmath137 ( 171 , 272 , -65)/(171 , 263 , -63)/(171 , 247 , -59 ) km s@xmath22 , and proper motion @xmath151 mas yr@xmath22 and @xmath152 mas yr@xmath22 in the solar rest frame . these velocities are dependent on the model assumed for the galactic potential , and will scale roughly with choice of @xmath94 and other galactic parameters . solar neighborhood debris our best - fit models orbiting in oblate ( @xmath177 ) and spherical ( @xmath178 ) potentials predict that the sun is currently bathing in a stream of debris from sgr , passing both inside and outside the solar circle . however , models orbiting in prolate ( @xmath1 1.25 ) potentials are inconsistent with this prediction , suggesting that conclusive proof of the presence or absence of sgr debris in the solar neighborhood could prove a useful tool for discriminating between models of the galactic potential . the authors would like to thank m.f . skrutskie for helpful discussion , and d. mart ' inez - delgado and m.a . g ' omez - flechoso for clarification on their satellite model and for making available a pre - publication copy of their latest work . srm acknowledges support from space interferometry mission key project nasa / jpl contract 1228235 , nsf grant ast-0307851 , a david and lucile packard foundation fellowship , and the f.h . levinson fund of the peninsula community foundation . kvj s contribution was supported through nasa grant nag5 - 9064 and nsf career award ast-0133617 . kundu , a. , majewski , s.r . , rhee , j. , rocha - pinto , h.j . , polak , a.a . , slesnick , c.l . , kunkel , w.e . , johnston , k.v . , patterson , r.j . , geisler , d. , gieren , w. , seguel , j. , smith , v.v . , palma , c. , arenas , j. , crane , j.d . , & hummels , c.b . 2002 , , 576 , 125 majewski , s. r. , kunkel , w.e . , law , d.r . , polak , a.a . , rocha - pinto , h.j . , crane , j.d . , frinchaboy , p.m. , hummels , c.b . , johnston , k.v . , patterson , r.j . , rhee , j. , skrutskie , m.f . & weinberg , m.d . 2004a , , 128 , 245 c|c|c|c||c|c parameter / property & description & value(s ) tested & constrained@xmath179 by : & acceptable values@xmath180 & value adopted + + @xmath25 & scale length of & 0 - 20 kpc & 4,5 & 1 - 20 kpc & 13/12/11 kpc + & the galactic halo & & & & + @xmath54 & flattening of the galactic & 0.8 - 1.4 & 6,7 & 0.8 - 1.4 & 0.90/1.0/1.25 + & dark halo potential & & & + @xmath94 & circular velocity & 180 - 240 km s@xmath22 & 4,5 & 180 - 240 km s@xmath22 & 220 km s@xmath22 + & at @xmath52 & & & + @xmath53 & contribution of disk & 0.25 - 1.00 & 4,5 & 0.25 - 1.00 & 1.00 + & to rotation curve & & & & + + & tangential & & & & + @xmath49 & velocity of sgr & 200 - 400 km s@xmath22 & 3,4,5,7 & 230 - 330 km s@xmath22 & 280/270/254 km s@xmath22 + @xmath181 & sgr line of sight velocity & fixed & 2 & 171 @xmath128 1 km s@xmath22 & 171 km s@xmath22 + + @xmath39 & distance of sgr & 22 - 28 kpc & 4,5 & 22 - 28 kpc & 24 kpc + & from the sun & & & + @xmath182 & galactic longitude and & fixed & 1 & & @xmath183 + & latitude of the sgr dwarf & & & + @xmath184 & distance of the sun from & 7.0 - 9.0 kpc & 4,5 & 7.0 - 9.0 kpc & 7.0 kpc + & the galactic center & & & +

we analyze multi - band photometry of a subsample of low mass stars in the associations ori ob1a and 1b discovered during the cida orion variability survey , which have ages of 7 - 10 myr and 3 - 5 myr , respectively . we obtained photometry at mt hopkins for 6 classical t tauri stars ( ctts ) and 26 weak t tauri stars ( wtts ) in ori ob1a , and for 21 ctts and 2 wtts in ori ob1b . we also obtained band photometry for 14 ctts at mt . hopkins , and 10 m and 18 m photometry with oscir at gemini for 6 ctts ; of these , all 6 were detected at 10 m while only one was detected at 18 m . we estimate mass accretion rates from the excess luminosity at , and find that they are consistent with determinations for a number of other associations , with or without high mass star formation . the observed decrease of mass accretion rate with age is qualitatively consistent with predictions of viscous evolution of accretion disks , although other factors can also play a role in slowing accretion rates . we compare the excesses over photospheric fluxes in , and , and with the younger sample of taurus and find an overall decrease of disk emission from taurus to ori ob1b to ori ob1a . this decrease implies that significant grain growth and settling towards the midplane has taken place in the inner disks of ori ob1 . we compare the sed of the star detected at both 10 m and 18 m with disk models for similar stellar and accretion parameters . we find that the low m fluxes of this ori ob1b star can not be due to the smaller disk radius expected from viscous evolution in the presence of the fuv radiation fields from the ob stars in the association . instead , we find that the disk of this star is essentially a flat disk , with little if any flaring , indicating a a significant degree of dust settling towards the midplane , as expected from dust evolution in protoplanetary disks . 0 et al . et al .

astro-ph0410522

most star formation studies are based on a few nearby associations without massive stars , as taurus and chameleon or in highly populated clusters with significant high mass formation , like the orion nebula cluster ( onc ) . in recent years , several teams have begun systematic studies of the populations of the ob associations . these associations cover large volumes in the sky and include populations covering the entire mass range . ob associations have been subject of studies with hipparcos ( brown et al . 1997 ; de zeeuw et al . 1999 ) , which has helped pinpoint their distances and mean ages . in a number of associations , the populations display a gradient of ages , from @xmath8 1 myr to a few @xmath8 10 myr , which has been interpreted as the result of triggered star formation ( blaauw 1964 ) . this range of ages makes the ob associations the most suitable laboratory for studies of protoplanetary disk evolution , because it encompasses the time span in which giant planets are expected to form ( pollack et al . 1996 ; alibert et al . 2004 ) , and significant solid evolution is expected to take place , as indicated by meteoritic evidence ( podoseck & cassen 1996 ; wood 2004 ) . we are carrying out a photometric and spectroscopic survey of @xmath9 128 degrees in the orion ob1 association in order to identify the low mass population . initial results of this survey have been presented by briceo et al . ( 2001 , 2004 = paper i ) . in paper i , we identified 197 new low mass members in the subassociations ori ob1a and 1b , spanning the mass range from @xmath8 0.2 to 1.4 @xmath10 . we found ages of @xmath8 3 - 5 myr for ori ob1b and @xmath8 7 - 10 myr for ori ob1a , confirming age determinations based on the ob stars ( blaauw 1964 ; warren & hesser 1977 , 1978 ; brown et al . we also found that ori ob1b could be identified with a population associated with a ring of low density gas and dust , probably formed by a supernova event , with most of the classical t tauri stars ( i.e. , stars accreting from circumstellar disks ; ctts ) near the ring , similar to the case of @xmath11 ori ( dolan & mathieu 1999 , 2001 , 2002 ) . in contrast , ori ob1a is in a region devoid of gas and dust , in agreement with its older age . we also found that the number of ctts decreases sharply between ori ob1b and 1a , indicating the few disks remain by @xmath8 10 myr , in agreement with findings from young clusters , some containing high mass stars , and low mass associations as tw hya ( haisch , lada , & lada 2003 ; hillenbrand , carpenter , and meyer 2004 , in preparation ; sicilia - aguilar et al . 2004 , in preparation ; muzerolle et al . 2000 , 2001 ) . in this work , we present multiwavelength photometry of a subset of the stars in paper i , including stars in ori ob1a , 1b and 1c . in particular , we have obtained @xmath3 photometry of a significant number of ctts in these subassociations , from which we have determined mass accretion rates for comparison with disk evolution theories . we have also obtained @xmath1 band and mid - infrared ( ir ) measurements at 10@xmath2 m and 18@xmath2 m of a small subset of stars in ori ob1 , which has allowed us to make assessments of the state of the dust in their disks in comparison with younger populations . in 2 we present the photometry , and in 3 , we examine the properties of the disks in ori ob1a and 1b in comparison with the taurus populations . we discuss the implications of these results for disk evolution in [ sec_discussion ] . the targets were selected from the sample of the orion variability survey presented in paper i , which were originally found on the basis of variability in the @xmath12 band using the quest camera on the 1 m schmidt telescope on the venezuela national observatory . follow - up spectroscopy using the fast spectrograph ( fabricant et al . 1998 ) on the 1.5 m telescope of the fred lawrence whipple observatory ( flwo ) at mt . hopkins , and the hydra multi - fiber spectrograph ( barden & armandroff 1995 ) on the wyin/3.5 m telescope at kitt peak , confirmed their membership to ori ob1 ( cf . paper i ) . we used the 4shooter ccd array on the 1.2 m telescope of the fred l. whipple observatory on mt . hopkins during two observing runs to obtain @xmath0 photometry of our list of stars . the first run was during the nights of november 29 through december 4 , 2002 , although observations could be done only in the second half of the night of december 3 and all night on december 4 . the second run was during october 15 - 20 , 2003 . all six nights were clear . the seeing was between 1.2 and 2.5 throughout the observations . the 4shooter camera contains four @xmath13 loral ccds separated by 45 and arranged in a @xmath14 grid . after binning @xmath14 during readout , the plate scale was @xmath15 . in order to achieve more uniform and consistent measurements , we placed all the target standard stars on the same ccd detector ( chip 3 ) . we obtained @xmath16 photometry for 65 stars in ori ob1a , 1b , and 1c , and @xmath3-band photometry for 59 of these stars . in particular , out of 36 stars in ori ob1a , we obtained @xmath3 photometry for all 6 ctts and 26 wtts ; for the 27 stars in ori ob1b , we obtained @xmath3 photometry for the 2 wtts and 21 ctts ; finally , we obtained @xmath3 photometry for the 2 ctts observed in ori ob1c . the basic processing of the flwo data was done with iraf routines in the standard way . for both runs , the @xmath3-band data were flat - fielded using sky flats taken at dusk / dawn . instrumental magnitudes were obtained using the apphot package in iraf . though orion is at a moderately low galactic latitude @xmath17 , our fields are not crowded in the relevant magnitude range , so aperture photometry is adequate . we used an aperture radius of 15 pixels for the source , @xmath18 the typical 3.5 pixel fwhm of our observations , large enough to include all the light from the standard stars as well as some observations made under mediocre seeing . we used a sky annulus with of 15 and 20 pixels for the inner and outer radii , respectively . the instrumental magnitudes were calibrated in the johnson @xmath19 and cousins @xmath20 system with observations of landolt ( 1992 ) standard fields sa92 , sa95 , sa113 and pg0231 + 051 at various air masses every night ; each field contains stars with a range of colors similar to our target objects . our overall photometric errors are dominated by the transformation uncertainty , which are 0.05 mag in @xmath3 and 0.02 magnitudes for @xmath21 . the nightly variation in the zero points was 0.15 , 0.07 , 0.06 , 0.04 , 0.05 magnitudes for @xmath22 respectively , suggesting reasonably good photometric stability throughout both runs . the resultant photometry is listed in table [ tab_photometry ] , where stars are identified by their running numbers in paper i. two exposures in some bands were obtained for star cvso 35 in ori ob1a and for stars cvso 146 and 190 in ori ob1b , as indicated in table [ tab_photometry ] . we obtained @xmath1-band photometry for a subsample of 14 stars using the dual channel infrared camera stelircam ( tollestrup & willner 1998 ) on the 1.2 m telescope at flwo during the nights of december 18 - 19 , 2002 . stelircam consists of two @xmath23 pixel insb detector arrays . each is fed from a dichroic mirror that separates wavelengths longer and shorter than @xmath24 m into two independent imaging channels ( red and blue channels ) for simultaneous observations on the sky . three separate magnifications can be selected by rotating cold lens assemblies into the beam . the magnification is the same in both wavelength channels . for our observations we used the medium lens yielding a scale of @xmath25 and a @xmath26 field of view , together with the l - barr filter ( @xmath27 m ) in the red channel . to convert the instrumental magnitudes of stelircam to the standard system , stars bd+0.1694 , g77 - 31 , gl 105.5 , gl 406 , hd 1160 , hd 18881 , hd 225023 and hd 40335 from the list of elias et al . ( 1982 ) were observed each night at various air masses . both the orion and standard stars were observed with a @xmath28 square pattern with 20 dither between positions . the integration time at each dither position was 10 s ( @xmath29 coadditions ) . all data were first linearized using frames obtained at various exposure times and routines developed in the interactive data language ( idl ) by t. megeath at cfa . we then used standard iraf routines to proceed with the data reduction . average dark frames were constructed from darks taken at the beginning and end of each night s observations . these average dark frames were then subtracted from each image . sky frames were individually made for each observation by median - combining the nine unregistered frames in one dither pattern . the sky frames were subtracted from each image within the dither set . the high and low airmass frames were grouped into pairs , and the low airmass frame in each pair was subtracted from the high airmass frame . these subtracted pairs were then normalized and the pairs combined using a median statistic to create the flat field . typically only the target star was visible in the @xmath1-band images , with a fwhm ranging from @xmath30 to 3 pixels . thus , we measured instrumental magnitudes with the iraf apphot package , using an aperture radius of 6 pixels , and a sky annulus with inner and outer radii of 10 and 20 pixels respectively , for both orion and standard stars . after obtaining the zero points and color terms ( which were small ) we arrived at an @xmath1-band @xmath31 error = 0.07 mags , mostly dominated by measurement errors the @xmath1-band photometry is presented in table [ tab_photometry ] , where we also include 2mass @xmath32 magnitudes for the stars from paper i. during the nights of december 3 , 5 and 9 , 2003 we obtained 10@xmath2 m and 18@xmath2 m photometry of a subsample of 6 stars in ori ob1a and b. this subsample was selected among those ctts with @xmath33 and with large excesses in the @xmath32 diagram . the observations were obtained in service mode using oscir on gemini north at mauna kea . oscir is a mid - infrared ( @xmath34 m ) imager and low / medium - resolution ( r=100 - 1000 ) spectrograph , which uses a @xmath35 pixel si : as ibc detector optimized for the wavelength range @xmath34 m . on gemini , oscir has a plate scale of about 0.084 arcsec / pixel and a total field of view in its imaging configuration of @xmath36 . we used the n - wide filter ( @xmath37 m and @xmath38 m ) and the ihw18 filter ( @xmath39 m and @xmath40 m ) . in the rest of this article we will refer to these filters as the @xmath41 m and @xmath42 m filters . typical exposure times for the orion targets were 60s at @xmath41 m and 300s at @xmath42 m . sky conditions were mostly clear through the observations . all the observations were performed using the standard technique of chopping and nodding , with a chop throw of 15 in declination . for calibration of the photometry the standard star @xmath43 cma ( sirius ) was observed on all nights at both wavelengths . flux density estimates for sirius were calculated using the seds published by cohen et al . ( 1999 ) . for sirius , integration times were of the order of 6s in both @xmath41 m and @xmath42 m . the data were processed through the standard pipeline for oscir data . chop pairs are subtracted : beam a - beam b or beam b - beam a for each set , depending on the nod position . then these differences are averaged to form a final image . because we were dealing with point sources , per recommendation of the gemini staff no flat fielding was performed . the oscir array is very flat over the small number of pixels within the point spread function . experience shared between gemini and university of florida is that the errors associated with the photometric calibration ( @xmath44 ) are much larger than any flat field correction . the @xmath41 m and @xmath42 m images showed only one object per frame , when detected at all . we used apphot to obtain aperture photometry of all targets . images at @xmath41 m showed typical fwhm@xmath45 pixels ( or @xmath46 arcsec ) , so we used an aperture radius of 15 pixels and inner and outer sky annulus radii of 15 and 25 pixels respectively . because most of the orion stars were not detected at @xmath42 m , we used consecutive pairs of @xmath41 m and @xmath42 m @xmath43 cma images to determine average offset positions between the two filters , and used these to place the aperture at the predicted position in the @xmath42 m images of the orion stars . in this way we measured upper limits to the flux in this band . the mid - ir measurements are listed in table [ tab_oscir ] . all stars were detected at 10@xmath2 m but only one measurement was secured at 18@xmath2 m . we list 3-@xmath47 upper limits in table 2 for the non detections . the galactic coordinates of stars in the photometric sample are plotted in figure [ fig_galactic ] compared to the entire sample in paper i. the stars are projected against the map of integrated @xmath48co emissivity from bally et al . it can be seen that our sample is representative of that in paper i. it includes the ctts and a number of wtts in the `` clump '' at ( @xmath49,@xmath50 ) in ori ob1a . it also includes a number of ctts in ori ob1b , located mostly in the inside of the molecular ring of higher extinction ( cf . paper i ) . some of these stars are probably associated with ngc 2023 and ngc 2024 . here we adopt the boundaries of ori ob1b from paper i , shown in figure [ fig_galactic ] . however , there is some uncertainty in the actual membership of stars located near the assumed boundary between ori ob1a and 1b at @xmath51 , @xmath52 to @xmath53 . in addition , with the adopted boundaries , 2 stars of the sample fall in ori ob1c , in the region of high molecular gas density ; however , they are so close to the uncertain boundary that we include them in ori ob1b for the rest of the analysis . we took the spectral types of the sample stars from paper i and recalculate here stellar properties from the simultaneous multiwavelength photometry to minimize effects of variability . we used the average distances of 330 pc for ori ob1a and 440 for ori ob1b , as discussed in paper i. the extinctions @xmath54 in table [ tab_prop ] are calculated from the @xmath55 color , using the cardelli , clayton , & mathis ( 1989 ) extinction law with @xmath56 = 3.1 . the mean extinctions for 1a , 1b , 1c , and the entire sample are 0.5 , 0.6 , 2 , and 0.6 , respectively . using individual extinctions , we recalculated stellar luminosities @xmath1 and radii @xmath57 . location in the hr diagram and comparison with the siess et al . ( 2000 ) tracks yielded stellar masses @xmath58 and ages . these quantities , shown in table [ tab_prop ] , in general agree within 50% with those of paper i. figures [ figviuv ] and [ figviub ] show photometry from table [ tab_photometry ] plotted in the @xmath59 vs. @xmath55 and @xmath60 vs. @xmath55 diagrams , as well as the location of the main sequence colors taken from kenyon & hartmann ( 1995 , kh95 ) . ctts and wtts are shown separately , and the ctts in ori ob1a are marked . note that the two observations for stars cvso 35 , 146 , and 190 are plotted as different points . the wtts in the sample are consistent with main sequence colors , reddened by @xmath8 0.5 - 0.6 magnitudes of extinction , as expected from the average @xmath54 of the sample . the ctts have similar @xmath55 as the wtts , but show large excesses relative to the wtts in @xmath59 and @xmath60 . these excesses are expected to arise from emission of the accretion shock formed as material accreting from the disk strikes the stellar surface ( hartigan et al . 1991 ; hartigan , edwards , & ghandour 1995 ; calvet & gullbring 1998 ) . in figures [ figviuv ] and [ figviub ] we also plot a sample of stars in taurus taken from kh95 , equally separated in wtts and ctts . the age of the taurus association has been analyzed in detail by hartmann ( 2003 ) , who showed that most of the stars have ages of @xmath8 1 - 2 myr . with the ages for the subassociations found in paper i , i.e. @xmath8 3 - 5 myr for ori ob1b and @xmath8 7 - 10 myr for ob1a , the taurus population is younger than ori ob1b by @xmath8 2 - 3 myr , and than ori ob1a by @xmath8 6 - 8 myr , so it provides an initial reference point to study evolutionary effects . the orion ob1 and taurus stars encompass a similar range of @xmath55 in figures [ figviuv ] and [ figviub ] , indicating that we are probing the same range of stellar masses . the samples of ori ob1b and taurus show a similar range in @xmath59 and @xmath60 excesses relative to photospheric colors ; the lowest excesses correspond to stars in ori ob1a , some of them compatible with measurements for the wtts . in figure [ fighkub ] we show the ori ob1 sample in the @xmath60 vs. @xmath4 diagram , indicating the standard main sequence colors with values from kh95 . we have used the transformations from carpenter ( 2001 ) to convert the 2mass @xmath61 colors to the cit system . a sample of taurus stars , with colors from kh95 is also shown . it is apparent that the excesses in @xmath4 relative to photospheric colors in ori ob1b are smaller than those found in the taurus sample . moreover , the excesses in @xmath4 in ori ob1a , the oldest of the three populations shown , are even smaller than in ori ob1b . this effect is more apparent in the @xmath60 vs. @xmath5 diagram , shown in figure [ figklub ] , using the @xmath5 colors of the ctts in ori ob1 in table [ tab_photometry ] . we use here @xmath62 for the orion ob1 stars . colors for the taurus stars and for the main sequence are from kh95 . the color excesses in @xmath5 relative to the main sequence in the ori ob1 stars are clearly smaller than those in taurus . finally , we locate the 6 stars observed with gemini / oscir in the @xmath6 vs. @xmath5 diagram in figure [ figknkl ] , and compare it to the taurus sample . the ori ob1 stars are located at the low end of the taurus ctts distribution , and even tend to populate the clear gap between ctts and tts noticed by kh95 ( the three taurus stars inside the gap are binaries , kh95 ) . these comparisons clearly indicate a decrease of ir emission at @xmath63 m with age . a significant decrease of @xmath4 and @xmath5 with with age is also found in a sample of clusters and groupings by hillenbrand , carpenter , & meyer ( 2004 , in preparation ) ; sicilia - aguilar et al . ( 2004 , in preparation ) observe a similar near - ir flux deficit relative to taurus in spitzer / irac colors of the @xmath8 5 myr cluster tr 37 . ir emission at @xmath63 m arises in the inner disks in t tauri stars , @xmath64 few au , which are sufficiently hot to produce such emission ( meyer et al . 1997 ; dalessio et al . 1998 ; 1999 ; 2001 ) . thus , the observed decrease of fluxes with age must be related to physical phenomena occurring in these inner regions of the disks as a result of disk evolution . we will explore possible implications of these results in [ sec_discussion ] . analyses of color - color diagrams in [ sec_colcol ] indicate a clear difference between emission of the disks in ori ob1 and in taurus ; disks in the orion ob1 stars emit less flux at near - infrared wavelengths than typical disks in taurus . the comparison is seen clearly in figure [ fig_seds ] , where we plot the spectral energy distributions ( seds ) of the stars observed with gemini oscir , including the measurement and upper limits at 18 @xmath2 m from table [ tab_oscir ] . optical and near - ir fluxes have been calculated from magnitudes in table [ tab_photometry ] and corrected for reddening in table [ tab_prop ] . these seds are compared with the median sed of ctts in taurus ( dalessio et al . 2001 ) , scaled to the flux at @xmath65 of each star . the error bars in the median show the first quartiles of the distribution ( i.e. , 50% of the taurus stars fall within the error bars ) . the seds of the orion stars show excesses relative to the photosphere , consistent with the fact that the sample observed with gemini was selected on the basis of being bright at @xmath66 and have large @xmath4 excesses . however , even with this selection bias , the ori ob1 disks are fainter than @xmath8 50 - 75 % of the taurus sample in the mid - ir . we interpret the excesses in @xmath60 and @xmath59 relative to photospheric fluxes in the ctts in ori ob1a and 1b as due to emission from the accretion shock on the stellar surface ( hartigan et al . 1991 ; hartigan , edwards , & ghandour 1995 ; gullbring et al . 1998 ; calvet & gullbring 1998 ) . in the framework of this interpretation , the excess luminosity above photospheric colors is a measure of the accretion luminosity released as matter falls from the disk onto the star , following the magnetic field lines that disrupt the disk ( cf . hartmann 1988 ) . we used the de - reddened @xmath3 photometry and spectral types to obtain the excess luminosity in u above the intrinsic photospheric fluxes , @xmath67 ; this luminosity was used to estimate the accretion luminosity @xmath68 , using the calibration of gullbring et al . these estimates are listed in table [ tab_prop ] . the largest uncertainty in the determination of @xmath67 , and thus @xmath69 , comes from uncertainties in the determination of the extinction @xmath54 , and it amounts to @xmath70% . using @xmath71 , ( assuming a radius of @xmath72 for the magnetospheric radius ; calvet & gullbring 1998 ) , with the mass and radius determined from the location in the hr diagram , we can infer the mass accretion rate @xmath73 of the objects . these rates are listed in table [ tab_prop ] . in addition to the uncertainties in @xmath68 , the formal uncertainty in @xmath73 includes errors in radius and mass , which are of order 15 - 20% ; thus , the mass accretion rates are constrained within a factor of 2 . given the uncertainties in the original calibration between @xmath67 and @xmath69 , this uncertainty can be a factor of 2 larger ( gullbring et al . as the errors are uncorrelated , the total error amounts to a factor of 3 . in figure [ fig_distmdot ] we show histograms with the number of objects per bin of log @xmath73 , each of size 0.5 , for the three populations , ori ob1a at the top , 1b at the middle , and taurus at the bottom . the mass accretion rates in taurus are taken from gullbring et al . ( 1998 ) and hartmann et al . these measurements of @xmath73 were obtained from @xmath68 using masses estimated with the dantona & mazitelli ( 1994 ) evolutionary tracks ; we have recalculated them using masses for the taurus stars estimated with the siess et al . ( 2000 ) evolutionary tracks , for consistency with the work presented here . the actual values of @xmath73 in bins with @xmath74 are highly uncertain . at these levels , emission due to stellar surface activity becomes an important or dominant contributor to the u excess , as comparison with the wtts in the sample shows . in these cases , the determined values are upper limits to the true @xmath73s . the age of the three populations shown increases from bottom to top . in all three cases , there is a large spread in the ranges of values of @xmath73 . however , as time proceeds , the number of rapid accretors decreases , only the low accretors remain in 1a . one `` continuum '' star is found in ori ob1b ( see table [ tab_prop ] ) ; we did not have enough spectral resolution to distinguish spectral features and thus determine its spectral type and derived properties . this continuum star does not appear in figure [ fig_distmdot ] ; if its properties were similar to those of continuum stars in taurus , then it could have @xmath75 few @xmath76 ( gullbring et al . 1998 ; calvet & gullbring 1998 ; gullbring et al . however , even including this star , the relative number of rapid accretors in ori ob1b would be smaller than in taurus . we compared the distributions of log @xmath73 in the three regions using a two - sided ks test . the results indicated that the probability that the ori ob1b and taurus samples are drawn from the same underlying distribution is only 3% ; the same is true for ori ob1b vs. ori ob1a . for a further test , we have added the continuum star in ori ob1b to the distribution , assuming a a mass accretion rate of @xmath77 ; even in this case , the probability that the ori ob1b sample is drawn from the same distribution as taurus is only marginally larger , 9% . thus we conclude that the distributions of accretion rates are significantly different between the three regions . figure [ fig_mdot ] plot the derived mass accretion rates and ages on ctts in ori ob1a and 1b in table [ tab_prop ] . we also plot in figure [ fig_mdot ] similar data for taurus , @xmath78 ophiucus and chameleon ( hartmann et al . 1998 ) , and the tw hya association ( twa , muzerolle et al . 2000 , 2002 ) . all masses and mass accretion rates of these latter objects have been re - calibrated using the siess et al . ( 2000 ) evolutionary tracks . the mass accretion rates of the disks in ori ob1 are consistent with those in other associations , and show a clear decrease with age . the star cvso 41 in ori ob1a has an apparent age of 88 myr ( table [ tab_prop ] ) and still shows a fairly high value of @xmath73 . the spectral type of this star is k2 , so it could have a large age uncertainty . it is known that pre - main - sequence stars in this spectral type range tend to have higher apparent ages than cooler objects ( hillenbrand 1997 ) ; this behavior has been attributed to uncertainties in the birthline location ( hartmann 2003 ) . the solid line in figure [ fig_mdot ] shows the decay of @xmath73 with time expected from viscous evolution for a disk with initial mass @xmath79 and viscous parameter @xmath80 , from hartmann et al . the new values of @xmath73 help confirm a decrease of @xmath73 with age qualitatively consistent with viscous disk evolution . however , viscous evolution can not explain the decreasing fraction of accreting objects with age ( muzerolle et al . 2000 ; briceo et al . 2004 ) ; other factors ( i.e. , inner disk clearing associated with planet formation , calvet et al . 2002 ; rice et al . 2003 ; dalessio et al . 2004 ) must play a role in slowing accretion onto the central star . in our comparison of disk properties in three populations with ages @xmath8 1 - 2 myr ( taurus ) , 3 - 5 myr ( ori ob1b ) , and 7 - 10 myr ( ori ob1a ) , we have found a clear decrease of disk emission with age . this decrease is apparent at all wavelengths with expected substantial disk contribution to the emission , from @xmath65 to 18@xmath2 m . the disk emission in the ir depends on the temperature of the dust . the two main heating mechanisms active on disks around young stars are viscous dissipation and stellar irradiation ( kenyon & hartmann 1987 , kh87 ; dalessio et al . 1998 , 1999 , 2001 , viscous dissipation depends on the mass accretion rate throughout the disk ; for the range of stellar parameters of stars in ori ob1 , and for @xmath81 , which characterize these stars ( [ sec_mdots ] ) , viscous heating is important only near the midplane of the innermost disk , @xmath82 1 au ( dalessio et al . 1999 , 2001 ) . thus , absorption of stellar radiation by dust in the disks around the ori ob1 stars mostly determines their temperature and emission . one important development in recent years is the recognition that the near - ir fluxes in accretion disks around pre - main sequence stars are dominated by emission from a sharp transition between dust and gas at the dust destruction radius . at this transition , the optically thick dust disk is high enough to form a `` wall '' , which is frontally illuminated by the star . opacities in the inner gas disk are due to molecules and are much lower than the dust opacities for typically low mas accretion rates ; thus , the gas disk is optically thin , enabling frontal illumination of the wall ( muzerolle et al . the existence of this wall was proposed for herbig ae / be stars ( natta et al . 2001 ; tuthill , monnier & danchi 2001 ; dullemond , dominik , & natta et al . 2001 ) to explain the peculiar `` bump '' at @xmath83 m observed in the seds of these stars ( hillenbrand et al . interferometric measurements of inner disk structures in herbig ae / be stars confirmed this prediction ( millan - gabet et al . 1999 ; monnier & millan - gabet 2002 ) . a similar wall emission was then reported in the low mass ctts by muzerolle et al . in contrast to their higher mass counterparts , ctts can have accretion luminosities of the same order as the stellar luminosities , thus increasing the total energy output and moving outwards the disk location where the temperature is equal to the dust destruction temperature for objects with high values of @xmath68 . the rapid accretors are therefore expected to have the wall located at larger radius than objects with low @xmath73 , increasing the total emitting area of the wall and thus the near - ir excess ( muzerolle et al . 2003 ; dalessio et al . 2003 ) . this prediction has been confirmed by the keck interferometer measurement of an inner structure at the expected location of the dust destruction radius in a high accretion ctts ( colavita et al . 2003 ; dalessio et al . 2003 ) . according to this model , the near - ir excess depends on the strength of the wall emission relative to photospheric fluxes . this emission is approximately that of a black - body at the dust destruction temperature , @xmath84 1400k ( muzerolle et al . 2003 ) , times the solid angle subtended by the wall . this solid angle is essentially that of a cylinder of given height , and radius equal to the dust destruction radius , with an inclination dependence which makes the emission maximum at intermediate lines of sight ( 60@xmath85 - 80@xmath85 ) where the wall has maximum exposure ( dullemond et al . 2001 ; dalessio et al . the dust destruction radius is essentially @xmath86 ( dalessio et al . the height of the wall @xmath87 is estimated as the height where the radial optical depth from the star becomes @xmath8 1 ( see muzerolle et al . 2004 ) . we can examine the decrease of near - ir excesses with age in terms of these ideas . as they age , pre - main sequence low mass stars move down along hayashi tracks essentially at constant @xmath88 while decreasing in luminosity . as @xmath73 decreases , @xmath68 becomes much smaller than @xmath89 , so the dust destruction radius becomes fixed relative to the stellar radius , @xmath90 const . if the height of the wall is a fixed multiple of the disk scale height at @xmath91 , which in turn is typically @xmath92 , then the emitting area of the wall , @xmath93 , and thus its total energy output , would decrease in the same proportion as the stellar output , namely @xmath94 . in this case , no decrease of near - ir excess relative to the photosphere with time would be expected . since @xmath95 const , the most natural way to explain the observed decrease is that the height of the wall @xmath87 decreases with time . for this to happen , the opacity at the upper levels of the wall has to decrease , which implies that dust grains at the top of the wall significantly grow and/or settle towards the midplane as age increases . grain growth and/or settling towards the midplane in the disk regions outside the wall also result in low mid - ir fluxes . the disk is heated by stellar + accretion energy captured at the disk surface . thus , heating of the disk depends on the shape of this surface ; the more `` flared '' the surface is , the more energy it can capture ( kenyon & hartmann 1987 , kh87 ) ; disks with highly flared surfaces are hotter and thus emit more . to be more specific , the height of the surface @xmath96 at a given radius is given by the condition that the radial optical since the density decreases with vertical height above the midplane , @xmath96 decreases as the opacity of the dust in the upper layers decreases . small grains have high opacity at wavelengths characteristic of the stellar radiation , in the optical and near - ir . they produce a highly flared surface , which results in high energy capture , and hot and bright disks . in contrast , large grains have low opacity at short wavelengths . they produce flatter surfaces , which result in less energy capture and cooler and fainter disks ( dalessio et al . 2001 , d01 ) . alternatively , the opacity at upper layers may be reduced because grains are settling to the midplane . this also results in flatter surfaces and fainter disks in the infrared ( dullemond & dominik 2004 ; dalessio et al . 2004 , in preparation ) . both growth and settling are expected to occur together . dust particles , affected by gas drag , collide and stick together ; as the particles grow , they settle towards the midplane , falling under the effect of gravity ( weidenschilling & cuzzi 1993 ; weidenschilling 1997 ) . these theories have many unknown parameters , and so far predict time scales for complete settling @xmath82 1 myr ; however , the timescales depend upon the magnitude of turbulence in disks , which could retard or prevent settling . a number of claims of grain growth in disks have been made over the years , starting with the interpretation of slopes of the fluxes in the millimeter , which are expected to become flatter as the grain size increases ( beckwith & sargent 1991 ) . assuming a dust mixture and a dust size distribution of the form @xmath97 between minimum and maximum sizes @xmath98 and @xmath99 , d01 find that a much better fit to the median sed of taurus can be achieved with dust particle growth to a maximum size @xmath99 of 1 mm ( @xmath100=3.5 ) than with interstellar grains , with @xmath101 m . grain growth and settling has been claimed to explain the seds of some ctts in taurus . one well studied case is that of the ctts lkca 15 , which has a sed much flatter than the taurus median in the ir and high emission at submillimeter and millimeter wavelengths ( qi et al . 2003 ) ; the interpretation of this sed requires significant grain growth and settling ( chiang et al . 2001 ; dalessio et al . 2003 , 2004 ; bergin et al . 2004 ) . in figures [ figviuv ] , [ figviub ] , [ fighkub ] , [ figklub ] , and [ figknkl ] we have marked the location of lkca 15 ; disks in ori ob1 have similar colors to those of lkca 15 in all bands , suggesting that significant dust evolution has also taken place in these older disks . in figure [ fig_est109 ] we compare the sed of cvso 109 , the only object for which we have detections up to 18@xmath2 m , to fluxes for lkca 15 from kh95 and to the taurus median , both scaled at the @xmath65 of cvso 109 . note that lkca 15 is a k5 star ( kh95 ) , so it is brighter in the optical than cvso 109 , which is an m0 ( table [ tab_prop ] ) . the seds of lkca 15 and cvso 109 are significantly lower than the taurus median , although the flux deficit is much larger in cvso 109 . to try to quantify the degree of dust evolution in cvso 109 , we have calculated models with uniformly distributed gas and dust following the procedures of d01 . we have used stellar and accretion parameters from table [ tab_prop ] , and have varied the parameters @xmath99 and @xmath100 which characterize the dust size distribution . figure [ fig_est109 ] shows our best model . to fit the observed fluxes , we require @xmath99 = 10 cm ; this is higher than the maximum sizes required to explain the median sed in taurus , @xmath102 1 mm ( d01 ) , or the sed of the @xmath8 10 myr old star tw hya , @xmath102 1 cm ( calvet et al . 2002 ; wilner et al . 2004 ) . moreover , we require @xmath103 , a value consistent with significant dust coagulation ( miyake & nakagawa 1995 ) ; modeling of ctts disks can generally be carried out successfully with the assumption of @xmath104 3.5 , a value characteristic of the interstellar medium ( cf . d01 , chiang et al . 2001 ) . our modeling is only indicative , in that we have assumed that the dust is uniformly distributed with the gas . even with this assumption , our comparison implies that the disk of cvso 109 is essentially a flat disk , with little if any flaring . thus , our modeling indicates that the dust in the disk of cvso 109 has strongly settled towards the midplane , as expected from theories of dust evolution in protoplanetary disks ( weidenschilling 1997 ) . models including settling in addition to observations at longer wavelengths are required to actually constrain the size distribution of the solids in the midplane . a possible alternative to consider is that disks in ori ob1 are smaller than their taurus counterparts because their outer parts have been photoevaporated by the external ionizing fields of the high mass stars in the associations . a smaller disk would have less emission at long wavelengths since the low temperature regions would be removed . models for the viscous evolution of disks subject to the effects of photoevaporating radiation fields from the accretion flows and external radiation fields have been calculated by clarke , gendrin , & sotomayor ( 2001 ) and matsuyama , johnston & hartmann ( 2003 ) . since the distance from the ionizing stars are of the order of a few pc in ob associations , fuv external photons are expected to dominate over euv photons in the photoevaporation of disks ( johnstone et al . 1998 ) . for this case , matsuyama et al . ( 2003 ) find that that the outer portions of the disk could be removed on time scales which range from 1 to several myr , depending on the viscosity parameter . after significant photoevaporation , the disk edge is fixed by the fuv gravitational radius , which for parameters as those of the ori ob1b stellar sample is @xmath8 30 au . to determine the effects of a reduction of disk radius on the sed , and in particular to see if a smaller disk radius can explain the low mid - infrared fluxes without extreme grain growth , we have calculated disk models with @xmath99 = 1 mm and @xmath100 = 3.5 , characteristic of the taurus population , for disk radii 100 au , 50 au , and 10 au . these seds are shown in figure [ figsmalldisk ] . models are calculated at a high inclination to the line of sight ( @xmath105 ) to maximize the contrast between the star and the disk . even small disks produce too much long wavelength flux compared to the the observations , if @xmath99 = 1 mm as in taurus . this implies again that the disk has to be essentially flat , which require that grains grow to larger sizes and/or have settled . we have detected clear signatures of disk evolution in samples of accreting stars in the ori ob1a and 1b associations , with ages 7 - 10 myr and 3 - 5 myr , respectively . the mass accretion rates of these stars are consistent with viscous evolution of accretion disks . in addition , we find a significant overall decrease of infrared emission with age . comparison with disk models indicates that grain growth and settling , processes expected from dust evolution , are responsible for this decrease . future observations of larger samples of ori ob1 stars with spitzer will provide more detail on the nature of the dust evolution process . acknowledgments . we thank the referee for many useful suggestions . this work was supported by nsf grant ast-9987367 and nasa grant nag5 - 10545 . pd acknowledges grants from papiit , dgapa ( unam ) and conacyt , mxico . this research 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 paper is based on observations obtained with the mid - infrared camera oscir , developed by the university of florida with support from the national aeronautics and space administration , and operated jointly by gemini and the university of florida infrared astrophysics group . this research has made use of the simbad database , operated at cds , strasbourg , france . wood , j. 2004 . in the search for other worlds ( eds . s. s. holt , d. deming ) , 14th annual october astrophysics conference in maryland ( oct . 13 - 14 2003 , college park md ) , preprint at http://cfa-www.harvard.edu/ jwood/ lccccccccc [ fotometria ] 1 & 17.12 & 1.36 & -0.36 & 1.26 & 2.81 & 12.82 & 0.68 & 0.25 & + 3 & 16.31 & 1.47 & 0.99 & 1.06 & 2.25 & 12.69 & 0.67 & 0.26 & + 5 & 16.09 & 1.48 & 0.81 & 1.08 & 2.23 & 12.50 & 0.68 & 0.22 & + 10 & 15.41 & 1.42 & 1.27 & 0.93 & 1.91 & 12.31 & 0.68 & 0.24 & + 11 & 15.74 & 1.53 & 0.96 & 1.08 & 2.33 & 12.10 & 0.66 & 0.23 & + 15 & 15.95 & 1.44 & 0.46 & 0.90 & 2.12 & 12.57 & 0.74 & 0.19 & + 17 & 16.16 & 1.51 & & 1.07 & 2.34 & 12.52 & 0.72 & 0.20 & + 19 & 15.89 & 1.49 & 1.54 & 1.14 & 2.51 & 11.92 & 0.69 & 0.21 & + 22 & 16.33 & 1.51 & 0.86 & 1.06 & 2.29 & 12.70 & 0.66 & 0.22 & + 24 & 16.31 & 1.49 & & 1.06 & 2.25 & 12.79 & 0.72 & 0.14 & + 25 & 15.46 & 1.42 & & 0.95 & 1.91 & 12.42 & 0.69 & 0.15 & + 29 & 16.25 & 1.51 & & 1.14 & 2.53 & 12.38 & 0.67 & 0.24 & + 34 & 15.15 & 1.36 & 1.16 & 0.96 & 1.97 & 12.28 & 0.99 & 0.23 & + 34 & 15.15 & 1.36 & 1.16 & 0.96 & 1.97 & 12.72 & 1.34 & 0.23 & + 35 & 14.55 & 1.41 & 0.73 & 0.99 & 1.95 & 11.46 & 0.70 & 0.41 & + & 14.84 & 1.30 & 0.76 & 0.92 & 1.93 & 11.46 & 0.70 & 0.41 & + 36 & 16.07 & 1.45 & & 1.09 & 2.35 & 12.38 & 0.69 & 0.21 & + 38 & 16.24 & 1.57 & 0.99 & 1.07 & 2.27 & 12.79 & 0.73 & 0.23 & + 38 & 16.37 & 1.26 & 1.12 & 1.07 & 2.46 & 12.79 & 0.73 & 0.23 & + 39 & 15.98 & 1.41 & 0.90 & 1.07 & 2.45 & 12.30 & 0.68 & 0.20 & + 39 & 16.11 & 1.22 & 0.66 & 1.10 & 2.46 & 12.30 & 0.68 & 0.20 & + 40 & 15.23 & 1.35 & 1.07 & 0.90 & 1.81 & 12.36 & 0.72 & 0.26 & + 41 & 15.23 & 0.97 & 0.28 & 0.73 & 1.54 & 12.41 & 0.70 & 0.33 & + 42 & 15.96 & 1.43 & 1.12 & 1.03 & 2.15 & 12.68 & 0.68 & 0.17 & + 43 & 16.27 & 1.42 & 0.98 & 1.09 & 2.46 & 12.50 & 0.74 & 0.16 & + 44 & 14.74 & 1.21 & 0.96 & 0.73 & 1.43 & 12.30 & 0.66 & 0.08 & + 46 & 15.15 & 1.42 & & 0.94 & 1.93 & 12.02 & 0.69 & 0.21 & + 46 & 15.19 & 1.46 & 0.91 & 0.94 & 1.92 & 12.02 & 0.69 & 0.21 & + 47 & 15.09 & 1.38 & 1.12 & 0.93 & 1.86 & 11.28 & 0.84 & 0.50 & 9.36 + 48 & 14.07 & 1.16 & 0.70 & 0.78 & 1.44 & 11.50 & 0.68 & 0.22 & 10.45 + 54 & 15.80 & 1.31 & 1.18 & 0.91 & 1.95 & 12.86 & 0.70 & 0.14 & + 55 & 15.60 & 1.43 & 1.16 & 1.06 & 2.27 & 11.93 & 0.69 & 0.24 & + 55 & 15.69 & 1.54 & 0.99 & 1.01 & 2.26 & 11.93 & 0.69 & 0.24 & + 57 & 15.67 & & & 0.91 & 1.91 & 12.65 & 0.72 & 0.19 & + 58 & 14.93 & 0.89 & -0.30 & 0.84 & 1.65 & 12.05 & 0.78 & 0.56 & 9.97 + 72 & 17.98 & 1.64 & 0.87 & 0.94 & 2.22 & 14.49 & 0.71 & 0.21 & + 75 & 16.72 & 0.92 & 0.41 & 0.44 & 1.04 & 13.00 & 0.76 & 0.51 & + 77 & 16.37 & 1.42 & & 1.07 & 2.26 & 12.81 & 0.72 & 0.23 & + 81 & 16.22 & 1.45 & 0.87 & 1.17 & 2.54 & 12.24 & 0.66 & 0.23 & + 90 & 14.61 & 0.38 & -0.97 & 0.65 & 1.40 & 12.13 & 0.91 & 0.67 & 9.73 + 104 & 14.22 & 0.79 & -0.08 & 0.66 & 1.44 & 11.78 & 0.75 & 0.58 & + 107 & 14.78 & 0.84 & -0.38 & 0.81 & 1.73 & 11.55 & 0.81 & 0.45 & 9.85 + 109 & 13.97 & 0.88 & -0.35 & 0.80 & 1.69 & 11.01 & 0.80 & 0.42 & 9.37 + 121 & 14.24 & 1.18 & 0.23 & 0.87 & 1.69 & 11.70 & 0.92 & 0.64 & + 124 & 15.91 & 1.50 & 0.89 & 1.11 & 2.41 & 12.07 & 0.70 & 0.25 & + 126 & 15.82 & 1.44 & 0.99 & 0.92 & 1.85 & 12.85 & 0.71 & 0.17 & + 133 & 14.88 & 1.15 & 1.01 & 0.76 & 1.59 & 12.19 & 0.61 & 0.15 & + 143 & 15.25 & 1.34 & 0.41 & 1.03 & 2.19 & 11.79 & 0.78 & 0.21 & + 146 & 13.92 & 1.19 & 0.14 & 0.84 & 1.63 & 11.30 & 0.86 & 0.42 & 9.65 + & 14.01 & & & & 1.52 & 11.30 & 0.86 & 0.42 & 9.65 + 152 & 13.98 & 1.12 & 0.81 & 0.71 & 1.40 & 11.58 & 0.82 & 0.50 & + 153 & 16.00 & 1.43 & -0.64 & 1.07 & 2.29 & 12.44 & 0.72 & 0.23 & + 155 & 14.35 & 1.33 & 0.63 & 0.92 & 1.85 & 11.48 & 0.66 & 0.33 & + 157 & 17.87 & 0.70 & -0.98 & 1.14 & 2.09 & 14.01 & 0.59 & 0.35 & + 164 & 16.71 & 1.55 & 0.30 & 1.31 & 3.03 & 11.80 & 0.69 & 0.32 & + 165 & 13.68 & 1.12 & 0.45 & 0.76 & 1.48 & 11.07 & 0.83 & 0.41 & 9.24 + 168 & 17.33 & 1.24 & -0.25 & 1.27 & 2.82 & 12.78 & 0.60 & 0.28 & + 171 & 14.84 & 1.46 & 0.64 & 0.84 & 1.69 & 11.96 & 0.70 & 0.30 & + 176 & 15.65 & 1.21 & -0.35 & 1.14 & 2.46 & 11.71 & 0.91 & 0.50 & 9.84 + 177 & 15.62 & 0.71 & & 0.96 & 1.92 & 12.30 & 0.77 & 0.33 & 10.98 + 178 & 15.66 & 1.48 & 0.64 & 1.11 & 2.28 & 11.47 & 0.77 & 0.34 & + 181 & 15.63 & 1.22 & & 0.87 & 2.00 & 12.30 & 0.74 & 0.19 & + 184 & 15.34 & 1.86 & & 1.41 & 2.75 & 10.40 & 1.27 & 0.91 & + 185 & 15.28 & 1.61 & & 1.08 & 2.18 & 10.88 & 0.88 & 0.49 & 8.77 + 190 & 14.04 & 1.10 & 0.28 & 0.83 & 1.72 & 11.04 & 1.08 & 0.77 & 8.49 + & 14.76 & & & & 1.89 & 11.04 & 1.08 & 0.77 & 8.49 + 192 & 15.99 & 1.26 & -0.14 & 1.17 & 2.37 & 11.56 & 1.01 & 0.55 & 99.00 + 193 & 15.15 & 1.29 & 0.04 & 1.15 & 2.29 & 11.34 & 1.22 & 0.66 & 99.00 + lcccccc [ mid - irfotometria ] 47 & 9.418 & 0.081 & @xmath106 7.627 & 0.383 & -11.779 @xmath107 0.03 & @xmath64 -11.654 + 58 & 8.448 & 0.388 & @xmath106 7.517 & 0.444 & -11.391@xmath107 0.15 & @xmath64 -11.576 + 90 & 8.294 & 0.282 & @xmath106 9.533 & 2.773 & -11.330@xmath107 0.12 & @xmath64 -11.792 + 107 & 8.109 & 0.276 & @xmath106 9.092 & 1.866 & -11.256@xmath107 0.12 & @xmath64 -11.764 + 109 & 7.935 & 0.120 & 6.437 & 0.165 & -11.186@xmath107 0.05 & -11.492@xmath107 0.1 + 165 & 6.921 & 0.093 & @xmath106 6.983 & 0.279 & -10.781@xmath107 0.05 & @xmath64 -11.710 + lccccccccccc [ prop ] 1 & m3&3470 & 0.18 & 1.17 & 0.33 & 0.81 & 4.02 & 0.01 & 0.09 & 1a & ctts + 3 & m3&3470 & 0.16 & 1.11 & 0.33 & 0.00 & 4.68 & 0.00 & 0.00 & 1a & wtts + 5 & m2&3580 & 0.21 & 1.20 & 0.38 & 0.22 & 4.12 & 0.00 & 0.01 & 1a & wtts + 10 & m0&3850 & 0.27 & 1.17 & 0.58 & 0.27 & 7.20 & 0.00 & 0.00 & 1a & wtts + 11 & m2&3580 & 0.33 & 1.49 & 0.39 & 0.46 & 2.51 & 0.00 & 0.02 & 1a & wtts + 15 & m1&3720 & 0.21 & 1.11 & 0.47 & 0.39 & 6.42 & 0.00 & 0.03 & 1a & wtts + 17 & m3&3470 & 0.19 & 1.20 & 0.33 & 0.00 & 3.63 & 0.00 & 0.06 & 1a & wtts + 19 & m3&3470 & 0.34 & 1.60 & 0.34 & 0.10 & 2.23 & & & 1a & wtts + 22 & m2&3580 & 0.18 & 1.12 & 0.38 & 0.36 & 5.18 & 0.00 & 0.01 & 1a & wtts + 24 & m2&3580 & 0.17 & 1.06 & 0.38 & 0.27 & 5.97 & 0.00 & 0.01 & 1a & wtts + 25 & m0&3850 & 0.24 & 1.11 & 0.58 & 0.27 & 8.09 & & & 1a & wtts + 29 & m2&3580 & 0.29 & 1.40 & 0.39 & 0.94 & 2.73 & & & 1a & wtts + 34 & m0&3850 & 0.29 & 1.21 & 0.58 & 0.41 & 6.61 & 0.00 & 0.01 & 1a & wtts + 34 & m0&3850 & 0.19 & 0.98 & 0.59 & 0.41 & 10.55 & 0.00 & 0.01 & 1a & wtts + 35 & k7&4060 & 0.72 & 1.72 & 0.77 & 0.85 & 2.88 & 0.02 & 0.17 & 1a & ctts + & k7&4060 & 0.71 & 1.70 & 0.77 & 0.80 & 2.90 & 0.02 & 0.13 & 1a & ctts + 36 & m3&3470 & 0.21 & 1.28 & 0.34 & 0.00 & 2.95 & 0.03 & 0.36 & 1a & wtts + 38 & m2&3580 & 0.17 & 1.06 & 0.38 & 0.31 & 5.88 & 0.00 & 0.00 & 1a & wtts + 38 & m2&3580 & 0.19 & 1.13 & 0.38 & 0.77 & 4.98 & 0.00 & 0.02 & 1a & wtts + 39 & m2&3580 & 0.30 & 1.41 & 0.39 & 0.74 & 2.70 & 0.00 & 0.04 & 1a & wtts + 39 & m2&3580 & 0.30 & 1.42 & 0.39 & 0.77 & 2.68 & 0.01 & 0.08 & 1a & wtts + 40 & m0&3850 & 0.24 & 1.10 & 0.58 & 0.02 & 8.18 & 0.00 & 0.00 & 1a & ctts + 41 & k2&4900 & 0.37 & 0.84 & 0.89 & 1.30 & 88.18 & 0.06 & 0.19 & 1a & ctts + 42 & m2&3580 & 0.17 & 1.08 & 0.38 & 0.02 & 5.71 & 0.00 & 0.00 & 1a & wtts + 43 & m2&3580 & 0.25 & 1.29 & 0.39 & 0.77 & 3.03 & 0.00 & 0.02 & 1a & wtts + 44 & k6&4205 & 0.27 & 0.98 & 0.82 & 0.00 & 22.97 & 0.00 & 0.01 & 1a & wtts + 46 & m0&3850 & 0.36 & 1.34 & 0.58 & 0.32 & 4.74 & 0.00 & 0.03 & 1a & wtts + 46 & m0&3850 & 0.35 & 1.34 & 0.58 & 0.29 & 4.80 & 0.00 & 0.02 & 1a & wtts + 47 & k5&4350 & 0.97 & 1.73 & 1.10 & 1.21 & 4.96 & 0.01 & 0.04 & 1a & ctts + 48 & k4&4590 & 1.25 & 1.77 & 1.28 & 0.71 & 6.60 & 0.05 & 0.22 & 1b & ctts + 54 & m0&3850 & 0.17 & 0.92 & 0.58 & 0.36 & 14.57 & 0.00 & 0.01 & 1a & wtts + 55 & m2&3580 & 0.37 & 1.58 & 0.39 & 0.31 & 2.30 & 0.00 & 0.01 & 1a & wtts + 55 & m2&3580 & 0.37 & 1.58 & 0.39 & 0.29 & 2.31 & 0.00 & 0.01 & 1a & wtts + 57 & m0&3850 & 0.20 & 1.00 & 0.59 & 0.27 & 9.95 & 0.00 & 0.00 & 1a & ctts + 58 & k7&4060 & 0.61 & 1.58 & 0.78 & 0.12 & 4.02 & 0.07 & 0.45 & 1b & ctts + 72 & m6&3050 & 0.04 & 0.74 & 0.11 & 0.00 & 6.20 & 0.00 & 0.01 & 1b & ctts + 75 & m1&3720 & 0.23 & 1.16 & 0.47 & 0.00 & 5.82 & 0.00 & 0.04 & 1b & ctts + 77 & k6&4205 & 0.28 & 0.99 & 0.82 & 1.88 & 21.91 & & & 1a & wtts + 81 & m3&3470 & 0.45 & 1.86 & 0.34 & 0.17 & 1.76 & 0.00 & 0.02 & 1b & wtts + 90 & k7&4060 & 0.55 & 1.50 & 0.79 & 0.00 & 5.04 & 0.29 & 1.77 & 1b & ctts + 104 & k7&4060 & 0.76 & 1.76 & 0.77 & 0.00 & 2.76 & 0.10 & 0.75 & 1b & ctts + 107 & k7&4060 & 1.02 & 2.04 & 0.77 & 0.32 & 2.12 & 0.13 & 1.09 & 1b & ctts + 109 & m0&3850 & 1.48 & 2.73 & 0.56 & 0.00 & 0.80 & 0.16 & 2.52 & 1b & ctts + 121 & k3&4730 & 1.31 & 1.70 & 1.32 & 1.49 & 8.06 & 0.40 & 1.67 & 1b & ctts + 124 & m3&3470 & 0.51 & 1.97 & 0.34 & 0.00 & 1.58 & 0.00 & 0.02 & 1b & wtts + 126 & m0&3850 & 0.16 & 0.89 & 0.58 & 0.12 & 16.16 & 0.00 & 0.00 & 1a & wtts + 133 & k6&4205 & 0.32 & 1.06 & 0.84 & 0.27 & 18.14 & 0.00 & 0.01 & 1a & wtts + 143 & m1&3720 & 0.82 & 2.18 & 0.47 & 0.55 & 1.16 & 0.02 & 0.37 & 1b & ctts + 146 & k6&4205 & 1.32 & 2.17 & 0.91 & 0.37 & 2.01 & 0.11 & 0.81 & 1b & ctts + & k6&4205 & 1.23 & 2.09 & 0.92 & 0.10 & 2.19 & 0.12 & 0.88 & 1b & ctts + 152 & k3&4730 & 1.21 & 1.64 & 1.30 & 0.78 & 8.65 & 0.03 & 0.14 & 1b & ctts + 153 & m2&3580 & 0.23 & 1.26 & 0.39 & 0.36 & 3.43 & 0.02 & 0.16 & 1a & wtts + 155 & k6&4205 & 1.30 & 2.14 & 0.91 & 0.90 & 2.06 & 0.07 & 0.53 & 1b & ctts + 157 & c & & & & & & & & & 1b & ctts + 164 & m3&3470 & 1.06 & 2.85 & 0.35 & 1.82 & 0.45 & 0.05 & 1.19 & 1b & ctts + 165 & k6&4205 & 1.48 & 2.29 & 0.91 & 0.00 & 1.72 & 0.05 & 0.37 & 1b & ctts + 168 & m4&3370 & 0.25 & 1.46 & 0.29 & 0.00 & 2.49 & 0.00 & 0.05 & 1b & ctts + 171 & k5&4350 & 0.82 & 1.60 & 1.08 & 0.81 & 6.41 & 0.02 & 0.10 & 1b & ctts + 176 & m3&3470 & 0.71 & 2.32 & 0.35 & 0.00 & 1.06 & 0.02 & 0.43 & 1b & ctts + 177 & k6&4205 & 0.64 & 1.50 & 0.94 & 1.07 & 6.21 & 0.01 & 0.04 & 1b & ctts + 178 & m1&3720 & 1.17 & 2.60 & 0.47 & 0.77 & 0.81 & 0.01 & 0.25 & 1b & ctts + 181 & m0&3850 & 0.51 & 1.61 & 0.57 & 0.48 & 2.63 & 0.02 & 0.14 & 1b & ctts + 184 & k1&5080 & 9.67 & 4.01 & 2.35 & 4.35 & 2.03 & & & 1b & ctts + 185 & k7&4060 & 2.54 & 3.22 & 0.81 & 1.40 & 0.79 & & & 1b & ctts + 190 & k6&4205 & 1.78 & 2.52 & 0.90 & 0.58 & 1.26 & 0.13 & 1.17 & 1b & ctts + & k6&4205 & 1.99 & 2.66 & 0.90 & 0.99 & 1.00 & 0.14 & 1.38 & 1b & ctts + 192 & k6&4205 & 1.68 & 2.44 & 0.90 & 2.14 & 1.40 & 0.36 & 3.08 & 1c & ctts + 193 & k6&4205 & 1.96 & 2.63 & 0.90 & 1.95 & 1.02 & 0.48 & 4.49 & 1c & ctts + co emissivity from bally et al . ( 1987 ) is shown in a halftone grey scale , covering the range from 0.5 to 20 @xmath108 . the isocontour for @xmath109 1 from schlegel et al . ( 1998 ) is also shown . the dotted lines show the boundaries of ori ob1b adopted in paper i and this work . the dashed lines show the limits of the survey in paper i. ]

a novel approach to electronic correlations and magnetism of crystals based on realistic electronic structure calculations is reviewed . in its simplest form it is a combination of the `` local density approximation '' ( lda ) and the dynamical mean field theory ( dmft ) approaches . using numerically exact qmc solution to the effective dmft multi - orbital quantum - impurity problem , a successful description of electronic structure and finite temperature magnetism of transition metals has been achieved . we discuss a simplified perturbation lda+dmft scheme which combines the t - matrix and fluctuation - exchange approximation ( tm - flex ) . we end with a discussion of cluster generalization of the non - local dmft scheme and its applications to the magnetism and superconductivity of high - t superconductors .

cond-mat0211076

the theory of electronic structure and magnetism of solids historically was split into two distinct parts , namely , the model investigations of many - body effects and the calculations of the energy spectra and properties of specific compounds in the framework of density functional ( df ) scheme @xcite . recently , within the dynamical mean - field theory ( dmft , for a review see ref . @xcite ) the correlation effects have been incorporated into realistic electronic structure calculations [ 4 - 10 ] . this method has been successfully applied to a number of classical problems of solid state physics such as the finite - temperature magnetism of iron - group metals @xcite , @xmath1 transition in plutonium @xcite , electronic structure of doped mott insulators @xcite . in contrast with standard df theory , in this new approach known as `` lda+dmft '' @xcite or `` lda++ '' @xcite the total energy of the system ( or , more accurately , the thermodynamic potential @xmath2 ) is considered as a functional of the green function instead of the denstity matrix @xcite . to stress this new feature more explicitely we will use the term `` spectral density functional ( sdf ) '' . the analytical properties of the green function garantee that the knowlegde of the spectral density is equivalent to the knowledge of the time - dependent green function whereas the density matrix is just static value of the latter @xcite . here we will describe the basic ideas of the sdf method , both in the framework of a standard dmft and from a more general point of view , discuss possible cluster generalizations of the dmft , and consider the applications of sdf to correlation effects and magnetism in transition metals . a most economical approach to unifying the various dynamical mean - field approximations in use , is provided by the effective action construction @xcite . the idea is to select a set of variables which is relevant to the physics of the problem and to write down a functional of the relevant variables . the extremum of this functional yields the values of those variables in equilibrium , and the value of this functional at stationarity gives the free energy of the system in equilibrium . density functional theory is the simplest example of this construction @xcite , here the total energy of the solid is expressed in terms of the density of the electrons . another well tested example is the spin density functional theory in which the total energy is expressed in terms of the spin and charge densities . the construction of an explicit expression of the exact effective - action functional is usually not available except for the case of very simple examples and the success of the method relies on the availability of good approximations to this functional . the lda and the lsda approaches have been extraordinarily successful for weakly correlated systems . the dynamical mean - field approach to model hamiltonians of strongly correlated systems on a lattice can also be brought to such an effective action perspective . more important this construction can be easily generalized to incorporate the so - called extended dynamical mean - field approach ( e - dmft ) @xcite , by constructing an effective action for both the local green function and the local density - density and local spin - spin autocorrelation function . the effective - action approach also allows the formulation of the e - dmft approach in the case where spatial and spin or charge symmetries are broken . @xcite . the effective - action approach , allows us a simple combination of the density functional theory and the dmft . by legendre transformation techniques one can construct a functional of the density and of the local spectral function of the heavy orbitals whose extremization would yield the exact density local spectral function of the heavy orbitals , and the total energy of the system . again the exact form of this functional are not known in explicit form , but useful approximations to it are available , and can be used to study interesting problems . this realistic dmft , lda+dmft approach or lda++ approach , was first implemented ignoring the coupling between the density and the local spectral function . this amounts to performing first a lda calculation to derive a tight - binding model hamiltonian , and then performing a dmft calculation for the spectra of the dmft hamiltonian . this is very close in spirit to the philosphy of model hamiltonian calculations . recently a full implementation of the self - consistent determination of the density and the local spectra was carried out by s. savrasov @xcite , and this is now closer in spirit to traditional first - principles electronic structure calculations . the effective action approach , can be generalized to clusters , if short - range correlations need to be taken into account . in the context of model hamiltonians , a functional of the restriction of the green function to a given cluster is defined . the extremization of this functional give rise to cluster dynamical mean field equations . there is no difficulty in constructing hybrid funtionals for the self - consistent determination of both the density and the cluster green functions , namely a cdmft+lda method . in the next section we motivate the approximate dmft form of those functionals from a perspective of a reduced fermionic description . in the sdf approach two - electron subsytems are introduced , one of them is described by the standard df theory ( usually this is the subsystem of @xmath3-electrons ) and dynamical interelectron correlations are taken into account for the another one ( usually this is the subsystem of @xmath4- or @xmath5-electrons ) . further simplifications can be connected with the local form of interelectron interactions ( only hubbard - type on - site electron correlations are considered ) , using some approximations like dmft or its cluster generalizations @xcite , etc . from a general point of view all these approaches can be considered as specific cases of a `` coarse graining '' ( reduced description ) ideology @xcite when all the variables describing the system can be separated into the `` gross '' and `` slave '' variables ; the only assumption that a closed set of the equations of motion for the gross variables exist is sufficient to find an explicit form of these equations @xcite . here we demonstrate the coarse graining procedure for fermionic degrees of freedom for the system of interacting electrons in a crystal . let us start with the functional integral over the grassman creation and annihilation electron fields @xmath6 , where the `` measure '' is @xmath7 , @xmath8 + s_{int}\left [ c^{+},c\right ] \label{action}\ ] ] here @xmath9 is the chemical potential , @xmath10 is the inverse temperature , and @xmath11 is the `` one - particle '' part of the hamiltonian . in the sdf approach @xmath12 where @xmath13 is the kohn - sham self - consistent potential @xcite . in the spirit of a reduced description approach @xcite we introduce `` gross '' variables @xmath14 ( more explicitly , @xmath15 where @xmath16 are site indices ; e.g. , in a standard dmft @xmath17 ) . the effective action for the gross variables @xmath18 is defined by introducing a delta - functional @xmath19 @xcite into the functional integral . @xmath20\end{aligned}\ ] ] unfortunately , this functional integral can not be estimated exactly and approximations are needed . here we consider only the simplest approximation to the derivation of the effective action , we replace s@xmath21 \rightarrow s_{int}^{cg}\left [ d^{+},d\right ] $ ] but compensate for the omission of the interaction terms away from those in the gross variables , namely in the medium by adding a medium self - energy in the standard dmft @xmath22 in other words , the dmft approach treats exactly the interaction terms _ only in _ those part of the interaction lagrangian which can be written in terms of gross variables , while _ the rest of the interaction terms are handled in a gaussian approximation _ by replacing the interaction terms by a medium self - energy . now the effective measure for the gross variables is given by @xmath23 \label{eff1}\end{aligned}\ ] ] where @xmath24 .$ ] passing to the matsubara frequencies @xcite one has @xmath25 c\left ( i\omega _ { n}j\right ) \label{matsubara}\end{aligned}\ ] ] in the standard dmft scheme the correlations between @xmath26-electrons on a central site @xmath27 are treated exactly while the correlations in the medium are treated in the gaussian approximation . it is sometimes stated in the literature that this assumption violates the translational invariance of the system . the effective - action approach , is meant to construct a functional of the selected variables , so by construction it is not possible to discuss within this framework the issue of translation invariance . the `` self - consistency condition '' does the best possible job within a gaussian approximation to restore the equivalence of the central site and the medium . we show below that this reasoning in a cluster setting leads to the `` cellular dmft '' approach @xcite . in principle , the correlations between the gross and slave variables ( e.g. , the long - range part of the coulomb interaction ) also can be taken into account in a close analogy with the classical spin models @xcite ; the corresponding modification of the dmft approach will be considered elsewhere . in accordance with the general scheme of the non - equilibriun statistical operator method @xcite , we add an `` auxiliary field '' conjugated to the green function of the gross variables , namely , the term @xmath28 . these fields will restore the translational invariance for the fermionic green function ( see below , eq.([equal ] ) ) . calculating the gaussian integral over @xmath29 one + d^{+}\bar{\sigma } d\right)*\nonumber \\ \int d\lambda d\lambda ^{+}\exp \left [ i\lambda d^{+}+i\lambda ^{+}d-\lambda ^{+}f\cdot g_{med}\cdot f^{\ast } \lambda \right ] \label{eff2}\end{aligned}\ ] ] where the medium green function reads @xmath31 ^{-1}. \label{medium}\ ] ] at last , calculating the gaussian integral over @xmath32 we obtain @xmath33 = s_{int}^{cg}\left [ d^{+},d\right ] -t\sum_{\omega _ { n}}d^{+}\left ( i\omega _ { n}\right ) { \cal g}_{0}^{-1}\left ( i\omega _ { n}\right ) d\left ( i\omega _ { n}\right ) \label{eff3}\ ] ] where @xmath34 the self - consistency condition that determines @xmath35 requires the equality of the local green function computed from the reduced description and from the medium agree , i.e. , @xmath36 further , @xmath37 ( because of the replacement @xmath38 \rightarrow s_{int}^{cg}\left [ d^{+},d\right ] $ ] ) and @xmath39 on the other hand , @xmath40 satisfies the dyson equation @xmath41 and we have @xmath42 where @xmath43 is a `` coarse - grained '' self - energy . the most natural choise is @xmath44 to obtain the main equation of the dmft @xcite we have to choise @xmath5 as a projection operator on the central site . in this case eq.([bath1 ] ) will take the desired form @xmath45 ^{-1}+\sigma \left ( i\omega _ { n}\right ) \label{dmft}\ ] ] where the quasimomentum @xmath46 runs the brillouine zone , @xmath47 is the fourier transform of the hamiltonian @xmath11 projected into the subspace of @xmath26-electrons . to consider posiible cluster generalizations one can choose let @xmath48 where @xmath49 is the projection operator on the cluster and @xmath50 is an unitary transformation of the variables in this cluster . thus our result reads : @xmath51 ^{-1}+u\sigma u^{+ } \label{cluster}\ ] ] where @xmath52 runs the `` new '' supercell brillouin zone , @xmath53 and @xmath54 are the sizes of the crystal and cluster , correspondingly . in terms of @xmath55@xmath56 ^{-1}+\sigma _ { cg } \label{clus1}\ ] ] as it was shown in ref . @xcite this equation for general basis set can be very useful in the optimisation of interaction problem within the cluster or cellular dmft scheme . another version of the cluster dmft @xcite will be described below . in a standard df theory the thermodynamic potential for non - correlated conduction-``c '' electrons @xmath57 is represented as a functional of the electron density @xmath58 which is , generally speaking , a matrix in spin indices . formally it can be represented as a thermodynamic potential of the kohn - sham quasiparticles @xcite , @xmath59 , minus the contribution of the so called `` double counted '' terms , @xmath60 : @xmath61 \\ \omega _ { dc}^{c } & = & \int v_{ks}({\bf r})\rho ( { \bf r})d{\bf r-}\int v_{ext}(% { \bf r})\rho ( { \bf r})d{\bf r-}\frac{1}{2}\int \frac{\rho ( { \bf r})\rho ( % { \bf r}^{\prime } ) } { |{\bf r}-{\bf r}^{\prime } |}d{\bf r}d{\bf r}^{\prime } -e_{xc}[\rho ] \nonumber \label{densfun}\end{aligned}\ ] ] where @xmath62 is the sum over matsubara frequencies @xmath63 @xmath64 @xmath65 @xmath66 is the temperature , and @xmath67 are site numbers ( @xmath68 ) , orbital quantum numbers ( @xmath69 ) and spin projections @xmath70 , respectively , @xmath71 is the external potential , @xmath72 $ ] is the exchange - correlation energy , and the kohn - sham effective potential is defined as @xmath73}{\delta \rho ( { \bf r})}. \label{ks}\ ] ] in contrast with the standard density functional theory , the sdf approach deals with the real dynamical quasiparticles for correlated `` d - electrons '' defined via local green functions rather than with kohn - sham `` quasiparticles '' which are , strictly speaking , only auxiliary states to calculate the total energy . therefore , instead of working with the thermodynamic potential @xmath2 as a _ density _ functional we have to start from its general expression in terms of an exact green function @xcite @xmath74 \right\ } \nonumber \\ \omega _ { dc}^{d } & = & tr\sigma g-\phi \label{sdf}\end{aligned}\ ] ] where @xmath75 and @xmath35 are an exact green function , its bare value and self - energy , @xmath76 is the luttinger generating functional ( sum of the all connected skeleton diagrams without free legs ) , respectively . a complete sdf thermodynamic potential is equal to @xmath77 we have to keep in mind also the dyson equation @xmath78 and the variational identity @xmath79 when neglecting the quasiparticle damping , @xmath80 will be nothing but the thermodynamic potential of free fermions but with exact quasiparticle energies . formal analogies between eqs.([densfun ] ) and ( [ sdf ] ) , ( [ ks ] ) and ( [ var ] ) are obvious : the self - energy plays the role of the kohn - sham potential ( without the external potential ) and the green function plays the role of the density matrix . as an example of this correspondence one can prove @xcite in the framework of the sdf an useful identity known as the `` local force theorem '' basically in the same way as it has been done within df theory @xcite . for both parts of the thermodynamic potential , @xmath57 and @xmath81 the local force theorem is based on extremum properties with respect to the variation of the density matrix and green function , respectively . further we will consider the contribution @xmath82 ( omitting the index @xmath83 for brevity ) . in principle , fermionic reduced description scheme allows us to combined conduction `` c '' and correlated `` d '' part of the total sdf : @xmath84 @xcite . let us discuss the problem of calculation of effective exchange interactions ( @xmath85 ) in correlated systems . in principle the @xmath85 parameters are not well defined for arbitrary magnetic systems , and the traditional way to study spin excitations related to the calculation of non - local frequency - dependent spin - susceptibility @xcite . in this case the auxiliary space - time dependent magnetic field is added to the the hamiltonian : @xmath86 and the second derivative of the free - energy with respect to the magnetic field gives the interacting spin - susceptibility : @xmath87 , where @xmath88 is an empty - loop susceptibility and @xmath89 is the vertex correction @xcite . here we consider a simple approximation of `` rigid spin rotation '' of spectral density for a small angle : @xmath90 where @xmath91 is a general direction of constrained effective spin - dependent potential on site @xmath68 and @xmath92 is a rotation vector . in this case it is useful to write explicitly the spinor structure of the self - energy and green functions : @xmath93 where @xmath94 , @xmath95 @xmath96 are pauli matrices , @xmath97 and @xmath98 . we suppose that the bare green function @xmath99 does not depend on spin direction and all the spin - dependent terms including the hartree - fock terms are incorporated in the self - energy . in the rigid spin approximation we assume that the unit vector @xmath91 does not depend on the energy and orbital indices and represents the direction of the average local magnetic moment on the site @xmath68 . note , that the thermodynamic potential @xmath2 should be considered as a constrained sdf which depends on @xmath91 as on external parameters ( cf . then the variation of the thermodynamic potential with respect to small spin - rotation can be written as @xmath100 where @xmath101 is the variation without taking into account the change of the self - consistent potential ( i.e. self - energy ) and @xmath102 is the variation due to this change of @xmath35 . taking into account eq . ( [ var ] ) it can be easily shown ( cf . @xcite ) that @xmath103 and hence @xmath104 \label{var4}\ ] ] which is an analog of the `` local force theorem '' in the density functional theory @xcite . in the case of rigid spin rotation the corresponding variation of the thermodynamic potential can be written as @xmath105 where the torque @xmath106 is equal to @xmath107 \label{torque}\ ] ] based on the expansion of this expression ( [ torque ] ) in a sum of pairwise contributions one can obtain @xcite useful formula for the effective magnetic interactions : @xmath108 and , correspondingly , for the stiffness tensor of a ferromagnet : @xmath109 where @xmath110 is the magnetic moment per unit cell . these results generalize the lsda expressions of ref . @xcite to the case of correlated systems . note that passing from eq.([torque ] ) to eq.([jij ] ) is not accurate , since the exchange parameters are connected with the second variations of the @xmath111-potential and use of the local force theorem can not be justified . ( [ jij ] ) corresponds to the `` empty loop '' approximation neglecting the vertex corrections . at the same time , for the stiffness the latter are absent and eq.([d ] ) appears to be exact provided that the self - energy and three - leg vertex are local ( as in the dmft ) @xcite . the fact that vertex corrections to the spin stiffness are absent within dmft , is also suggested by the analogy between electric and spin transport developed in ref . the spin - wave stiffness can be obtained from the zero - frequency limit of the spin conductivity . within dmft , the charge current - current correlation function does not require vertex corrections @xcite and can therefore be obained directly from the convolution of the one electron spectra . these arguments are independent of the spin structure , and can therefore be used for the spin - wave stiffness . in order to elucidate the approximation behind the expression for the exchange parameters ( eq . ( [ jij ] ) ) , we consider the energy of a spiral magnetic configuration with the rigid rotation of the spinor - electron operators by the polar angles @xmath112 and @xmath113 : @xmath114 where @xmath115 assuming that @xmath116 = const and @xmath117 = @xmath118 where @xmath119 is the site - lattice vector . since we take into account only on - site correlation effects the interaction term in the hamiltonian is invariant under that transformation , and the change of the hamiltonian is @xmath120 } } } = \delta _ { 1}h+\delta _ { 2}h \\ \delta _ { 1}h=\sin ^{2}{\frac{{\theta } } { { 2}}}{\sum\limits_{k}{tr_{m\sigma } } } % { \left [ { \left ( { t\left ( { \bf k}{+}{\bf q}\right ) -t\left ( { \bf k}\right ) } % \right ) c_{{\bf k}}^{+}c}_{{\bf k}}\right ] } \\ \delta _ { 2}h={\frac{{1}}{{2}}}\sin \theta { \sum\limits_{ij}{tr_{m}}}{\left [ { % t_{ij}c_{i\downarrow } ^{+}c_{j\uparrow } } \right ] } \left ( { \exp \left ( { i{\bf % qr}_{i}}\right ) -\exp \left ( { i{\bf qr}_{j}}\right ) } \right ) \end{array}\ ] ] consider further the case of small @xmath112 , we can calculate change of the total energy to lowest order in @xmath112 corresponds to the first order in @xmath121 and the second order in @xmath122 : @xmath123 } n_{k}-}}}itr_{m}{% \int { { \frac{{d^{4}k}}{{\left ( { 2\pi } \right ) ^{4}}}}{\left [ { t\left ( { \bf k}% { + } { \bf q}\right ) -t\left ( { \bf k}\right ) } \right ] } } } \ast \\ & & { { \gamma \left ( k , q\right ) g_{\downarrow } \left ( k+q\right ) { \left [ { % t\left ( { \bf k}{+}{\bf q}\right ) -t\left ( { \bf k}\right ) } \right ] } % g_{\uparrow } \left ( k\right ) \}}},\end{aligned}\ ] ] where @xmath124 , @xmath125 is a four - vector with component ( * q,*0 ) , and @xmath126 is the three - leg vertex . our main approximation is to neglect of the vertex corrections ( @xmath126 = 1 ) . in this case the previous equation takes the following form : @xmath127 \ast } } } \label{deltae}\\ & & { { g_{\downarrow } \left ( { \bf k}{+}{\bf q}\right ) { \left [ { g_{\downarrow } ^{-1}\left ( k+q\right ) -g_{\uparrow } ^{-1}\left ( k\right ) + t\left ( { \bf k}{+% } { \bf q}\right ) -t\left ( { \bf k}\right ) } \right ] } g_{\uparrow } \left ( { k}% \right ) { \ } } } } \nonumber\end{aligned}\ ] ] using the following consequence of the dyson equation : @xmath128 one can rewrite eq . ( [ deltae ] ) in the form : @xmath129 } $ ] with the exchange integrals corresponding to eq . ( [ jij ] ) . we conclude that the expression for @xmath85 is accurate if the vertex corrections can be neglected . note that in the limit of small * q * this can be justified rigorously , provided that the self - energy and three - leg scalar vertex are local . therefore , the expression for the stiffness constant of the ferromagnet ( eq . ( [ d ] ) ) appears to be exact in the framework of dmft @xcite . it should be stressed that the exchange integrals discussed here are just static characteristics connected with energy of inhomogeneous spin configurations . they determine the frequencies of spin excitation in itinerant - electron systems only under adiabatic ( rigid - spin ) approximation . for more general consideration one should calculate @xmath130- and @xmath131- dependent spin susceptibility . now we describe the applications of the sdf approach to a classical problem of finite - temperature magnetism of the iron - group transition metals . despite a lot of attempts starting from seminal works by heisenberg and frenkel ( for review of early theories see e.g. @xcite ) we have yet no complete quantitative theory describing their magnetic and spectral properties . the reason is that to describe the properties of fe , co , and ni one has to solve the problem of taking into account moderately strong electronic correlations where approaches developed both for weakly correlated systems such as normal - group metals and to higly correlated systems such as mott insulators are not , generally speaking , reliable.there were many attempts to introduce by some way correlation effects in band structure calculations of these metals @xcite , not referring to numerous purely model works . but the question of applicability of specific approximations such as lowest order perturbation theory @xcite , moment method @xcite or three - body faddeev equations @xcite is not clear and one needs a reliable approach which would be checked carefully for model systems and demonstrate its applicability for moderately correlated systems . it has been demonstrated in ref . @xcite that the _ ab initio _ dynamical mean - field theory does give a very successful description of both correlation effects in the electron energy spectra and the finite - temperature magnetic properties of fe and , especially , ni . here we present the corresponding results . we start with the lda hamiltonian in the tight - binding orthogonal lmto representation @xmath132 @xcite , where @xmath133 describes the orbital basis set containing @xmath134- , @xmath135- and @xmath136- states , and @xmath137 runs over the brillouin zone ( bz ) . the interactions are parameterized by a matrix of screened local coulomb interactions , @xmath138 , and a matrix of exchange constants , @xmath139 , which are expressed in terms of two screened hubbard parameters , @xmath50 and @xmath140 , describing the average coulomb repulsion and the interatomic ferromagnetic exchange , respectively . we use the values @xmath141 @xmath142 ev for fe ( ni ) and the same value of the interatomic exchange , @xmath143 ev for both fe and ni , a result of constrained lda calculations @xcite . these parameters , which are consistent with those of many studies result in a very good description of the physical properties of fe and ni . as was discussed above , the dmft maps the many - body system onto a multi - orbital quantum impurity , i.e. a set of local degrees of freedom in a bath described by the weiss field function @xmath144 . the impurity action ( here @xmath145 and @xmath146 $ ] is a vector of grassman variables ) is given by : @xmath147 + \nonumber \\ & & \frac{1}{2}\sum_{m , m^{\prime } , \sigma } \int_{0}^{\beta } d\tau \lbrack u_{mm^{\prime } } n_{\sigma } ^{m}n_{-\sigma } ^{m^{\prime } } + ( u_{mm^{\prime } } -j_{mm^{\prime } } ) n_{\sigma } ^{m}n_{\sigma } ^{m^{\prime } } ] \label{path}\end{aligned}\ ] ] it describes the spin , orbital , energy and temperature dependent interactions of a particular magnetic @xmath134-atom with the rest of the crystal and is used to compute the local greens function matrix : @xmath148e^{-s_{eff}}{\bf c}(\tau ) { \bf c}^{+}(\tau ^{\prime } ) \label{pathint}\ ] ] ( @xmath149 is the partition function ) and the impurity self - energy @xmath150 . the weiss field function is required to obey the self - consistency condition ( [ dmft ] ) , which can be specified for a given case as @xmath151^{-1 } \label{bzi}\ ] ] the local matrix @xmath152 is the sum of two terms , the impurity self energy and a so - called `` double counting '' correction , @xmath153 which is meant to subtract the average electron - electron interactions already included in the lda hamiltonian . for metallic systems we proposed the general form of dc - correction : @xmath154 . this is motivated by the fact that the static part of the correlation effects are already well described in the density functional theory . only the @xmath4-part of the self - energy is presented in our calculations , therefore @xmath155 for @xmath156- and @xmath157- states as well as for non - diagonal @xmath158 , @xmath157 contributions . in order to describe the finite temperature ferromagnetism of transition metals we use the _ non _ spin - polarized lda hamiltonian @xmath159 and accumulate _ all _ temperature - dependent spin - splittings in the self - energy matrix @xmath160 . we used the impurity qmc scheme ( see appendix ) for the solution of the multiband dmft equations @xcite . the hirsch discrete hubbard - stratonovich transformations @xcite introduces @xmath161 auxiliary ising fields @xmath162 where @xmath163 is the orbital degeneracy of the @xmath4-states and calculate @xmath164 by an exact integration of the fermion degrees of freedom in the functional integral ( eq.([pathint ] ) ) @xcite . in order to sample efficiently all the spin configurations in the multi - band qmc scheme , it is important to use `` global '' spin - flips : @xmath165\rightarrow \lbrack -s_{mm^{\prime } } ^{-\sigma -\sigma ^{\prime } } ] $ ] in addition to the local moves of the auxiliary fields . the number of qmc sweeps was of the order of 10@xmath166 . a parallel version of the dmft program was used to sample the 45 ising fields for @xmath134-orbitals . we used 256 @xmath46-points in the irreducible part of the bz for the k integration . 10 to 20 dmft iterations were sufficient to achieve convergence far from the curie point . due to the cubic symmetry of the bcc - fe and fcc - ni lattices the local green function is diagonal in the basis of real spherical harmonics . the spectral functions for real frequencies were obtained from the qmc data by applying the maximum entropy method @xcite . our results for the local spectral function for iron and nickel are shown in figs.[dosfe ] and [ dosni ] , respectively . the sdf approach describes well all the qualitative features of the density of states ( dos ) , which is especially non - trivial for nickel . our qmc results reproduce well the three main correlation effects on the one particle spectra below @xmath167 @xcite : the presence of a famous 6 ev satellite , the 30% narrowing of the occupied part of @xmath4-band and the 50% decrease of exchange splittings compared to the lda results . note that the satellite in ni has substantially more spin - up contributions in agreement with photoemission spectra @xcite . the exchange splitting of the d - band depends very weakly on temperature from t=0.6 t@xmath168 to t=0.9 t@xmath168 . correlation effects in fe are less pronounced than in ni , due to its large spin - splitting and the characteristic bcc - structural dip in the density of states for spin - down states near fermi level , which reduces the density of states for particle hole excitations . now we discuss the applications of the sdf approach to the description of the finite - temperature magnetic properties of iron and nickel . while density functional theory can in principle provide a rigorous description of the thermodynamic properties , at present there is no accurate practical implemention available . as a result the finite - temperature properties of magnetic materials are estimated following a simple suggestion @xcite , whereby constrained dft at @xmath169 is used to extract exchange constants for a _ classical _ heisenberg model , which in turn is solved using approximation methods ( e.g. rpa , mean field ) from classical statistical mechanics of spin systems @xcite . the most recent implementation of this approach gives good values for the transition temperature of iron but not of nickel @xcite . while these localized spin models give , by construction , at high temperatures a curie - weiss like magnetic susceptibility , as observed experimentally in fe and ni , they encounter difficulties in predicting the correct values of the curie constants@xcite . the uniform spin susceptibility in the paramagnetic state , @xmath170 , was extracted from the qmc simulations by measuring the induced magnetic moment in a small external magnetic field . it includes the polarization of the impurity weiss field by the external field @xcite . the dynamical mean field results account for the curie - weiss law which is observed experimentally in fe and ni . as the temperature increases above @xmath171 , the atomic character of the system is partially restored resulting in an atomic like susceptibility with an effective moment : @xmath172 the temperature dependence of the ordered magnetic moment below the curie temperature and the inverse of the uniform susceptibility above the curie point are plotted in fig . [ mchi ] together with the corresponding experimental data for iron and nickel@xcite . the lda+dmft calculations describes the magnetization curve and the slope of the high - temperature curie - weiss susceptibility remarkably well . the calculated values of high - temperature magnetic moments extracted from the uniform spin susceptibility are @xmath173 @xmath174 for fe ( ni ) , in good agreement with the experimental data @xmath175 @xmath176 for fe ( ni)@xcite . 0.5 cm 0.5 cm 0.9 cm 0.9 cm we have estimated the values of the curie temperatures of fe and ni from the disappearance of spin polarization in the self - consistent solution of dmft problem and from the curie - weiss law in eq.([effective ] ) . our estimates @xmath177 @xmath178 are in reasonable agreement with experimental values of @xmath179 @xmath180 for fe ( ni ) respectively@xcite , considering the single - site nature of the dmft approach , which is not able to capture the reduction of @xmath167 due to long wavelength spin - waves . these effects are governed by the spin - wave stiffness . since the ratio of the spin - wave stiffness ( @xmath181 ) to @xmath167 , @xmath167/@xmath182 is nearly a factor of 3 larger for fe than for ni@xcite ( a is the lattice spacing ) , we expect the dmft @xmath167 to be much higher than the observed curie temperature in fe than in ni . note that this is a consequence of the long - range oscilating character of exchange interactions in iron compared to short - range ferromagnetic exchange interactions in nickel @xcite . quantitative calculations demonstrating the sizeable reduction of @xmath167 due to spin waves in fe in the framework of a heisenberg model were performed in ref @xcite . 0.5 cm within dynamical mean field theory one can also compute the local spin susceptibility defined by @xmath183 where @xmath184 is the gyromagnetic ratio and @xmath185 is single - site spin operator and @xmath186 are pauli matrices . it differs from the @xmath187 susceptibility by the absence of spin polarization in the weiss field of the impurity model . eq.([local ] ) can not be probed directly in experiments but it is easily computed in dmft - qmc . its behavior as function of temperature gives a very intuitive picture of the degree of correlations in the system . in a weakly correlated system we expect eq.([local ] ) to be nearly temperature independent , while in a strongly correlated system we expect a leading curie - weiss behavior at high temperatures @xmath188 where @xmath189 is an effective local magnetic moment . in the heisenberg model with spin @xmath190 , @xmath191 and for well - defined local magnetic moments ( e.g. , for rare earth magnets ) this quantity should be temperature independent . for the itinerant electron magnets , @xmath189 is temperature - dependent , due to a variety of competing many - body effects such as kondo screening , the induction of local magnetic moment by temperature @xcite and thermal fluctuations which disorders the moments @xcite . all these effects are included in the dmft calculations . the @xmath192-dependence of the correlation function @xmath193 results in the temperature dependence of @xmath194 and is displayed in the inserts on the figs.[dosfe],[dosni ] . iron can be considered as a magnet with very well - defined local moments above @xmath167 ( the @xmath192-dependence of the correlation function is relatively weak ) , whereas nickel is more of an itinerant electron magnet ( stronger @xmath192 -dependence of the local spin - spin autocorrelation function ) . the comparison of the values of the local and the @xmath187 susceptibility gives a crude measure of the degree of short - range order which is present above @xmath171 . as expected , the moments extracted from the local susceptibility eq.([local ] ) are a bit smaller ( 2.8 @xmath195 for iron and 1.3 @xmath196 for nickel ) than those extracted from the uniform mangetic susceptibility . this reflects the small degree of the short - range correlations which remain well above @xmath167 @xcite . the high - temperature lda+dmft clearly show the presence of a local - moment above @xmath171 . this moment , is correlated with the presence of high energy features ( of the order of the coulomb energies ) in the photomeission . this is also true below @xmath167 , where the spin dependence of the spectra is more pronounced for the satellite rigion in nickel than for that of the quasiparticle bands near the fermi level ( fig . [ dosni ] ) . this can explain the apparent discrepancies between different experimental determinations of the high - temperature magnetic splittings @xcite as being the results of probing different energy regions . the resonant photoemission experiments @xcite reflect the presence of local - moment polarization in the high - energy spectrum above the curie temperature in nickel , while the low - energy arpes investigations @xcite results in non - magnetic bands near the fermi level . this is exactly the dmft view on the electronic structure of transition metals above @xmath167 . fluctuating moments and atomic - like configurations are large at short times , which results in correlation effects in the high - energy spectra such as spin - multiplet splittings . the moment is reduced at longer time scales , corresponding to a more band - like , less correlated electronic structure near the fermi level . the qmc method described above is probably the most accurate way of solving the effective impurity problem in the dmft . however , it is rather cumbersome and expensive computationally ; besides that , it deals with the `` truncated '' two - indices interaction matrix ( see eq.([path ] ) ) instead of the complete four - indices one . therefore a scheme has been proposed in ref . @xcite based on a multiband spin - polarized generalization of the `` fluctuating exchange '' ( flex ) approximation by bickers and scalapino @xcite . the original formulation of the flex approximation treats both particle - hole ( ph ) and particle - particle ( pp ) channels on an equal footing . but their roles in magnetism are completely different . the interaction of electrons with spin fluctuations in ph channel leads to the most relevant correlation effects @xcite whereas pp processes are important for the renormalizations of the effective interactions in spirit of the @xmath66-matrix approach ( `` ladder aprroximation '' ) by galitskii @xcite and kanamori @xcite . therefore we used in ref . @xcite a `` two - step '' procedure when , at first , the bare matrix vertex is replaced by a @xmath66-matrix , and , secondly , ph channel processes with this effective interaction are taken into account explicitly . this approximation has high enough accuracy both for the hubbard model and for real systems with moderate correlations @xmath197 where @xmath50 is the hubbard on - sire repulsion snergy and @xmath198 is the bandwidth ( see @xcite and refs therein ) . however , specific form of the approximation used in @xcite can be improved further by taking into account the spin - dependence of the @xmath66-matrix . here we present the formulation of this t - matrix - flex approximation . let us start , as in refs . @xcite , with the general many - body hamiltonian for a crystal in the lda+u scheme @xcite : @xmath199 where @xmath200 are the site number @xmath201 and orbital @xmath202 quantum numbers , @xmath203 is the spin projection , @xmath6 are the fermion creation and annihilation operators , @xmath204 is the effective single - particle hamiltonian from the lda , corrected for the double - counting of average interactions among correlated electrons as it was described above , and the coulomb matrix elements are defined in the standard way @xmath205 where we define for briefness @xmath206 etc . following ref . @xcite we take into account the ladder ( @xmath66-matrix ) renormalization of the effective approximation : @xmath207 further we rewrite the perturbation theory in terms of this effective interaction matrix . at first , we take into account the `` hartree '' and `` fock '' diagrams with the replacement of the bare interaction by the @xmath66-matrix @xmath208 note that @xmath209 @xmath210 contains exactly all the second - order contributions . now we have to consider the contribution of particle - hole excitations to sigma . similar to @xcite we will replace in the corresponding diagrams the bare interaction by the static limit of the @xmath66-matrix . however , we improve the approximation @xcite by taking into account its spin dependence . when considering the particle - hole channel we replace in the hamiltonian ( [ hamilu ] ) @xmath211 which is the solution of eq.([tmatrix ] ) at @xmath212 eq . ( [ harfock ] ) is exact in the limit of low electron ( or hole ) density which is important for the criterion of magnetism e.g. in the case of nickel ( with almost completely filled @xmath4- band ) . now we rewrite the effective hamiltonian ( [ hamilu ] ) with the replacement @xmath213 by @xmath214 in @xmath215 . to consider the correlation effects due to ph channel we have to separate density ( @xmath4 ) and magnetic ( @xmath133 ) channels as in @xcite @xmath216 @xmath217 then the interaction hamiltonian can be rewritten in the following matrix form @xmath218 where * means the matrix multiplication with respect to the pairs of orbital indices , e.g. @xmath219 the supervector d defined as @xmath220 and the effective interactions have the following form : @xmath221 \label{effpotent}\end{aligned}\ ] ] to calculate the ph contribution to the electron self - energy we first have to write the expressions for the generalized susceptibilities , both transverse @xmath222 and longitudinal @xmath223 the corresponding expressions are the same as in @xcite but with another definition of the interaction vertices . one has @xmath224 ^{-1}*\gamma ^{\uparrow \downarrow } ( i\omega ) \ , , \label{xi+-}\ ] ] where @xmath225 is an `` empty loop '' susceptibility and @xmath226 is its fourier transform . the corresponding longitudinal susceptibility matrix has a more complicated form : @xmath227 ^{-1}*\chi _ 0^{\parallel } ( i\omega ) , \label{xipar}\ ] ] and the matrix of bare longitudinal susceptibility : @xmath228 in the @xmath229- , @xmath230- , @xmath231- , and @xmath232- channels ( @xmath233 in the supermatrix indices ) . an important feature of these equations is the coupling of longitudinal magnetic fluctuations and of density fluctuations . it is absent in one - band hubbard model due to the absense of the interaction of electrons with parallel spins . for this case eqs . ( [ xi+-],[xipar ] ) coinsides with the well - known result @xcite . now we can write the particle - hole contribution to the self - energy . according to @xcite one has @xmath234 with p - h fluctuation potential matrix : @xmath235 , \label{wpp}\ ] ] where the spin - dependent effective potentials are defined as @xmath236 \ast v^{\parallel } \ ] ] @xmath237 \ast v^{\parallel } \\ ] ] @xmath238 \ast v_{m}^{\perp } \ ] ] @xmath239 \ast v_{m}^{\perp } .\ ] ] where @xmath240 differ from @xmath241 by the replacement of @xmath242 in eq.([xi0par ] ) . we have subtracted the seconf - order contributions since they have already been taken into account in eq.([harfock ] ) . our complete expression for the self energy is @xmath243 this expression takes into account accurately spin - polaron effects because of the interaction with magnetic fluctuations @xcite , the energy dependence of @xmath66-matrix which is important for describing the satellite effects in ni @xcite , contains the exact second - order terms in @xmath244 and is rigorous ( because of the first term ) for almost filled or almost empty bands . in spirit of the dmft approach we have to use @xmath245 instead of @xmath246 in all the expressions when calculating the self - energy on a separated central site . it should be noted that this tm - flex scheme is not conserved ( or `` @xmath247-derivable '' ) therefore one need to inforce the luttinger theorem by introducing the @xmath248 for the bath green function as in iterative preturbation theory@xcite we have started from the spin - polarized lsda band structure of ferromagnetic nickel within the tb - lmto method @xcite in the minimal @xmath249 basis set and used numerical orthogonalization to find the @xmath204 part of our starting hamiltonian . we take into accounts the coulomb interactions only between @xmath4-states . semiempirical analysis of the appropriate interaction value gives @xmath250 ev . the difficulties with choosing the correct value of @xmath50 are connected with complicated screening problems , definitions of orthogonal orbitals in the crystal , and contributions of the intersite interactions . in the quasiatomic ( spherical ) approximation the full @xmath50-matrix for the @xmath251shell is determined by the three parameters @xmath252 and @xmath253 or equivalently by effective slater integrals @xmath254 , @xmath255 and @xmath256 @xcite . for example , @xmath257 , @xmath258 and we use the simplest way of estimating @xmath253 or @xmath256 keeping the ratio @xmath259 equal to its atomic value 0.625 @xcite . note that the value of intra - atomic ( hund ) exchange interaction @xmath140 is not sensitive to the screening and approximately equals 0.9 ev in different estimations @xcite . 0.0 cm 0.0 cm the resulting dos for ni ( fig . [ tmflex ] ) shows that spin - polarized tm - flex calculations approximatelly reproduce the satellite structure and reduction of the band width in satisfactory agreement with exact qmc - result(fig . [ dosni ] ) as it was stressed above the sdf approach does not necessarily connect with the standard dmft scheme ; one can use more general choise of fermion coarse - grained variables , e.g. , considering the case of an effective cluster instead of an effective impurity . it is especially important for the problems where intersite correlations are involved from the beginning such as @xmath4-wave superconducting pairing @xcite or charge ordering @xcite . here we consider , following ref . @xcite , one of the cluster generalizations of the dmft , and its application to the problem of magnetism and high - temperature superconductivity ( htsc ) of copper - oxide compounds . the microscopic theory of high - temperature superconducting cuprates is still far from a final understanding @xcite . one of the most important recent experimental achievements was the discovery of the pseudo - gap ( pg ) phenomenon above the superconducting transition temperature @xcite and existence of a sharp 41-mev resonance below t@xmath0 related to some collective antiferromagnetic excitations @xcite . thus , an interplay of an antiferromagnetism ( afm ) and @xmath4-wave superconductivity ( d - sc ) in cuprates could be a natural way of discussing different htsc phenomena . this require a quantitative electronic structure theory including two different type of the order parameters : afm and d - sc . within such approach one can in principle analyze the phase diagram of htsc compounds and resolve the long - standing problem of competition between antiferromagnetism and d - wave superconductivity in cuprates @xcite . a standard theoretical tool for cuprates electronic structure consists of the two - dimensional hubbard model @xcite . we start with the extended - hopping hubbard model on the square lattice : @xmath260 where @xmath261 is an effective hopping and @xmath262 local coulomb interactions . we have chosen the nearest - neighbor hopping @xmath263 ev and the next nearest hopping @xmath264 for the model of la@xmath265sr@xmath266cuo@xmath267@xcite . the total band width is @xmath198=2 ev and all coulomb parameters set to be @xmath50=1.2 ev ( @xmath268 ) . let us introduce the `` super - site '' as an 2@xmath2692 square plaquet . the numeration of the atoms in the super - site is shown in the fig.[afmsc ] . it is useful to introduce the superspinor @xmath270 where @xmath271 ( the spin indices are not shown ) . taking into account the spin degrees of freedom , this is the 8-component superspinor creation operator . then the crystal green function for the hubbard model can be rewritten as @xmath272 ^{-1 } \label{freeclust}\ ] ] where @xmath273 is the effective hopping supermatrix with self - energy corrections . for simplicity we will write all the formulas in the nearest - neighbor approximations : @xmath274 where @xmath275 , @xmath276 is the lattice constant , and each element is a 2@xmath2692 matrix in spin space . within the cluster - dmft approach we introduce the intra - atomic self - energy @xmath277 and the inter - atomic self - energies @xmath278 @xmath279 , and both functions are of intra - site nature in the sense of our super - site : @xmath280 for the small 2@xmath2692 cluster it is usefull to introduce the translationally invariant ( * k*-dependent ) self - energy and rewrite @xmath281 where @xmath282 the effective hamiltonian is defined through the renormalized energy dependent hoppings : @xmath283 @xmath284 @xmath285 . the functions @xmath286 @xmath287 @xmath288 are found self - consistently within the cluster dmft scheme and for the @xmath4-wave superconduction state @xmath289 it is straightforward to generalize this scheme for a next - nearest neighbor hopping as well as the long - range green function and the self - energy . in this case we can renormalized also the second - nearest hopping : @xmath290 for the 2@xmath2692 cluster , where @xmath291 ( or @xmath292 ) is the non - local self - energy in the @xmath293 direction . 0.5 cm 0.4 cm in the cluster version of the dmft scheme one can write the matrix equation for the bath green function matrix @xmath144 which describe an effective interaction with the rest of crystal : @xmath294 where the local cluster green function matrix is equal to @xmath295 @xmath296 , and the summation is run over the brillouin zone of the square lattice . if instead of eq.([trinv ] ) we use eq.([hkom ] ) , it would corresponds to a free cluster dmft scheme or the simplest case of the so - called cellular - dmft . note that we use a translationally invariant self - energy obtained from the cluster dmft scheme ( eq.([trinv ] ) ) or the simplest version of so - called cellular dmft approach @xcite . we believe that for a given choise of the small ( 2@xmath2692 ) cluster , the renormalization of the hopping parameter by @xmath278 @xmath279 terms is physically essential . the present `` matrix '' form of a cluster dmft with the self - energy which is not periodic inside the cluster allow us to study a multicomponent ordered state . unfortunately , in contrast with the so - called dynamical cluster approximation ( dca ) @xcite or cellular - dmft approach @xcite ( see above , section 1 ) , we were unable to prove the casuality of this approach for arbitrary band structure and interaction parameters . however , the casuality of the green function for the _ realistic _ choise of the parameters has been checked numerically . in this case we have the standard dmft problem with four `` orbital '' states per super - site . it has been solved by the multi - orbital qmc technique described above ( section 3 ) . we used the generalized nambu technique @xcite to analyze the coexistence of the magnetic ordering and superconductivity . let us introduce the superspinor @xmath297 and the anomalous averages describing the ( collinear ) antiferromagnetism @xmath298 and the superconductivity @xmath299 @xmath300 . the generalization of the hirsch - fye qmc - algorithm @xcite for the superconducting problem @xcite have been used . in the 4-spinor case a discrete hubbard - stratonovich transformation has the following form : @xmath301 = \label{hirsuper } \nonumber \\ \frac{1}{2}\sum_{\sigma = \pm 1}\exp [ \lambda _ { i}\sigma ( \psi _ { 1i}^{+}\psi _ { 1i}-\psi _ { 2i}^{+}\psi _ { 2i}-\psi _ { 3i}^{+}\psi _ { 3i}+\psi _ { 4i}^{+}\psi _ { 4i } ) ] \end{aligned}\ ] ] where @xmath302 $ ] . since we take into account only the singlet pairing , we obtain the following nonzero elements of the d - sc energy gap parameters : @xmath303 . one can chose @xmath304 to be real and therefore symmetric : @xmath305 . separating normal and anomalous parts of the green function we have @xmath306 where @xmath307 and @xmath308 are the matrices in spin and `` orbital '' space . it is convenient to expand the anomalous green function in pauli matrices @xmath309 and use the symmetry properties @xcite : @xmath310 then a @xmath311 spinor formalism is reduced to @xmath312 one in the collinear antiferromagnetic case with the d - wave superconductivity with the following spin - matrix form of the local green function for the super - site : @xmath313 and the qmc formalism for the antiferromagnetic superconducting state is equivalent to the previous non - magnetic one @xcite . using the discretization of @xmath314 $ ] interval with @xmath53-time slices : @xmath315 ( @xmath10 is an inverse temperature ) the @xmath316- and @xmath317- green functions become the matrices of @xmath318 dimension , where @xmath319 is the number of atoms in the cluster . after fourier transform to the matsubara frequencies the green function matrix has the following form : @xmath320 in superconducting states the self - energy defined as @xcite : @xmath321 and the inverse crystal green function matrix is equal to : @xmath322 where @xmath323 is the translationally invariant anomalous part of the self - energy @xmath324 similar to eq.([hkom ] ) . -0.5 cm -0.4 cm the two - component order parameters state which includes neel antiferromagnetism and d - wave superconductivity ( fig.[afmsc]a ) lowered the symmetry of the effective cluster - dmft problem . a self - consistent dmft cluster problem with afm and d - sc general order parameters have been solved within the qmc scheme for 8x8 matrix green function with l=64 time slices . the resulting two order parameters for @xmath325 ev@xmath326 ( t=190 k ) and @xmath327 presented in fig.[afmsc]c together with the generic htsc phase diagram ( fig.[afmsc]b ) as function of the hole doping . in this case the ordered magnetic moment is directly related with imaginary - time green function @xmath328 : @xmath329 and for the d - sc order parameter we chose a positive value of superconducting imaginary time green function @xmath330 . it is important , that we find no serious sign - problem for all qmc calculations with various doping level , probably due to `` stabilized '' antiferromagnetic dynamical mean fields acting on the atoms in our 2@xmath3312 cluster . note that the afm cluster - dmft solution exists for a much higher doping concentration than experimental afm ordered state and describes a dynamical mean - field version of afm - spin fluctuations related to pseudogap phenomena ( the pg - region on fig.[afmsc]b ) . the maximum of d - sc order parameter corresponds to a doping level of about 15@xmath332 in agreement with the generic htsc phase diagram . the d - sc order parameter is zero close to the undoped region ( @xmath333=0 ) , due to presence of a large afm - gap . when the afm - gap is closed ( @xmath334 ) the d - sc states develope but for @xmath335 the d - gap decrease again since afm spin - fluctuations around ( @xmath336 ) point disappear @xcite . the precise characteristic of the phase diagram including the interactions between the afm and d - sc order parameters demands an extensive cluster - dmft calculations for different temperatures and doping . we would like to note that the existing of d - sc cluster - dmft solution for such high temperatures does not necessary means that the superconducting transition temperature is larger then 190k in our model . a crude estimation shows that the d - sc solution disappears at @xmath66=300 k for @xmath333=0.15 and the afm solutions for x=0 become unstable at the temperature just above 1000 k. this could be the sign of a `` local '' afm solution and a local d - wave solution , like local moments in magnetic systems @xcite . due to the multiscale nature of the problem under considerations , essentially different energies connected with local moment formation , long - range magnetic order , local d - wave pairs within the 2@xmath3312 plaquet , and finally coherent superconductivity , it is difficult to distinguish a real long - range ordering from slow dynamical fluctuations in our qmc simulations . we plane to separate these energy scales analytically and estimate superconductiong transition temperature in a future publication . the role of next - nearest hopping is to lower of the van - hove singularity @xcite which increases the density of state at the fermi level for the hole - doped case and favores the d - sc solution for a moderate correlation strength . there is also a change in the spin - fluctuation spectrum related with the broadening of afm - peak near the ( @xmath336 ) point due to formation of so - called extended van - hove singularities with increasing of the @xmath337 . we show one of the afm - dsc solution on the fig.[gtau ] with the next nearest - neighbor hopping for the 10% doping level and @xmath338 ev@xmath326 . the resulting local magnetic moment is @xmath110=0.28 @xmath195 and the d - sc order parameter @xmath339=0.036 . one can see that the superconduction order parameter is really of the d@xmath340 symmetry since diagonal elements ( @xmath341 ) as well as the next nearest - neighbors elements ( @xmath342 ) are all equal to zero and only the nearest neighbor superconducting green functions ( @xmath343 ) are non - zero and change the sign for @xmath344 and @xmath345 components . the normal local green function ( @xmath346 ) ( plotted for the spin - up atom in the fig.[gtau ] ) as well as ( @xmath347 ) are spin - split , while the nearest - neighbor green function ( @xmath348 ) has no spin - splitting due to afm spin symmetry ( see fig.[afmsc ] ) . the absence of magnetic polarization in the non - diagonal g - function along the x(y)-directions suppress the magnetic pair - breaking and makes the afm - dsc coexistence possible . the spectral density functional ( sdf ) approach allows us to study correlation effects in solids based on realistic electronic structure calculations . among all possible applications we have chosen the magnetism of transition metals and high - temperature superconductivity . from these two examples one can see already all the main advantages of the new approach in comparison with standard density functional theory . first , we can describe now the spectral density transfer phenomena ( e.g. , the formation of 6 ev satellite in ni ) , the quasiparticle damping and other effects connected with the frequency dependence of the self - energy ; they are absent completely not only in the df approach but also in the hartree - fock , lda+u , or self - interaction corrections approximation ( see , e.g. , @xcite ) . second , we can describe adequately the contribution of the bose degrees of freedom ( e.g. , spin fluctuations ) to the electronic structure and thermodynamic properties . in the df - based calculations the temperature is really taken into account only via the thermal expansion and the fermi distribution function @xcite . it was the main reason for the failure of the standard band theory for the description of finite - temperature effects in magnetic metals . we show that the sdf gives a satisfactory solution of this problem . most of real applications of the sdf approach are connected with the single - site dynamical mean field theory . at the same time , for a number of problems this can be insufficient and some generalizations to take into account the non - local effects are necessary , for example , cluster ones . the first attempts of such generalizations are already leading to some interesting results in the theory of high - t@xmath0 superconductivity @xcite but a lot of additional work is required in the area of cluster dynamical mean field theories to reach the level of understanding that was reached in single site dmft . the work was supported by the netherlands organization for scientific research ( nwo project 047 - 008 - 16 ) and partially supported by russian basic research foundation ( grant 00 - 15 - 96544 ) . gk is supported by the nsf under grantnsf dmr 0096462 , by the division of basic sciences of the department of energy under grant de - fg02 - 99er45761 , and by the onr under grant n000149910653 . the qmc important sampling scheme allowed us to integrate over the ising fields with the abs(@xmath362 $ ] ) as a stochastic weight@xcite . for a single spin - flip @xmath363 the determinant ratio is calculated as following : @xmath364/\det [ \widehat{g^{\prime } } ] % & = & r_{i}r_{j}-r_{ij } \\ r_{i } & = & 1+[1-g_{ii}(\tau , \tau ) ] \delta _ { i}(\tau ) \\ r_{j } & = & 1+[1-g_{jj}(\tau , \tau ) ] \delta _ { j}(\tau ) \\ r_{ij } & = & g_{ij}(\tau , \tau ) \delta _ { j}(\tau ) g_{ji}(\tau , \tau ) \delta _ { i}(\tau ) \end{aligned}\ ] ] @xmath365\delta _ { i}(\tau ) /r_{_{i}}(\tau ) g_{ij_{2}}(\tau , \tau _ { 2 } ) \\ g_{i_{1}j_{2}}^{new}(\tau _ { 1},\tau _ { 2 } ) & = & g_{i_{1}j_{2}}^{\prime } ( \tau _ { 1},\tau _ { 2})+[g_{i_{1}j}^{\prime } ( \tau _ { 1},\tau ) -\delta _ { i_{1}j}\delta _ { \tau _ { 1},\tau } ] \delta _ { j}(\tau ) /r_{_{j}}(\tau ) g_{jj_{2}}^{\prime } ( \tau , \tau _ { 2})\end{aligned}\ ] ] a. i. lichtenstein and m. i. katsnelson , in : _ band ferromagnetism . ground state and finite - temperature phenomena _ ( lecture notes in physics , springer , berlin , 2001 ) , ed . by k. baberschke , m. donath , and w. nolting , p. 75 . h. kajueter , rutgers university ph.d . thesis ( 1996 ) ; h. kajueter and g. kotliar , rutgers university preprint ( 1996 ) ; q. si and j.l . smith , phys . lett * 77 * , 3391 ( 1997 ) ; r. chitra and g. kotliar , phys . lett . * 84 * , 3500 ( 2000 ) . e. kisker , k. schrder , m. campagna , and w. gudat , phys . * 52 , * 2285 ( 1984 ) ; a. kakizaki , j. fujii , k. shimada , a. kamata , k. ono , k.h . park , t. kinoshita , t. ishii , and h. fukutani , phys . lett . * 72 * , 2781 ( 1994 ) . j. r. schrieffer , _ theory of superconductivity _ ( benjamin , new york , 1964 ) ; s. v. vonsovsky , yu . a. izyumov , and e. z. kurmaev , _ superconductivity of transition metals , their alloys and compounds _ ( springer , berlin , 1982 ) .

we study time - inhomogeneous markov chains with finite state spaces using nash and logarithmic - sobolev inequalities , and the notion of-stability . we develop the basic theory of such functional inequalities in the time - inhomogeneous context and provide illustrating examples . and . .

1104.1560

this article is part of a series of works where we study quantitative merging properties of time inhomogeneous finite markov chains . time inhomogeneity leads to a great variety of behaviors . moreover , even in rather simple situations , we are at a loss to study how a time inhomogeneous markov chain might behave . here , we focus on a natural but restricted type of problem . consider a sequence of aperiodic irreducible markov kernels @xmath1 on a finite set @xmath2 . let @xmath3 be the invariant measure of @xmath4 . assume that , in a sense to be made precise , all @xmath4 and all @xmath3 are similar and the behavior of the time homogeneous chains driven by each @xmath4 separately is understood . can we then describe the behavior of the time inhomogeneous chain driven by the sequence @xmath1 ? to give a concrete example , on @xmath5 , consider a sequence of aperiodic irreducible birth and death chain kernels @xmath4 , @xmath6 with @xmath7 and with reversible measure @xmath3 satisfying @xmath8 , for all @xmath9 . what can we say about the behavior of the corresponding time inhomogeneous markov chain ? remarkably enough , there is very little known about this question . what can we expect to be true ? what can we try to prove ? let @xmath10 denote the distribution , after @xmath11 steps , of the time inhomogeneous chain described above started at @xmath12 . it is not hard to see that such a chain satisfies a doeblin type condition that implies @xmath13 in the absence of a true target distribution and following @xcite , we call this property _ merging_. of course , this does not qualify as a quantitative result . extrapolating from the behavior of each kernel @xmath4 taken individually , we may hope to show that , if @xmath14 then @xmath15 the aim of this paper and the companion paper @xcite is to present techniques that apply to this type of problem . the simple minded problem outlined above is actually quite challenging and we will not be able to resolve it here without some additional hypotheses . however , we show how to adapt techniques such as singular values , nash and log - sobolev inequalities to time inhomogeneous chains and provide a variety of examples where these tools apply . in @xcite , we discussed singular value techniques . here , we focus on nash and log - sobolev inequalities . the examples treated here ( as well as those treated in @xcite ) are quite particular despite the fact that one may believe that the techniques we use are widely applicable . whether or not such a belief is warranted is a very interesting and , so far , unanswered question . this is deeply related to the notion of @xmath0-stability that is introduced here and in @xcite . the examples we present here and in @xcite are about the only existing evidence of successful quantitative analysis of time inhomogeneous markov chains . a more detailed introduction to these questions is in @xcite . the references @xcite discuss singular value techniques in the case of time inhomogeneous chains that admit an invariant distribution [ all kernels @xmath4 in the sequence @xmath16 share a common invariant distribution ] . time inhomogeneous random walks on finite groups provide a large collection of such examples ( see also @xcite for a particularly interesting example : semirandom transpositions ) . the papers @xcite are also concerned with quantitative results for time inhomogeneous markov chains . in particular , the techniques developed in @xcite are closely related to ours and we will use some of their results concerning the modified logarithmic sobolev inequality . references on the basic theory of time inhomogeneous markov chains are @xcite . for a different perspective , see also @xcite . a short review of the relevant aspects of the time inhomogeneous markov chain literature , including the use of `` ergodic coefficients '' can be found in @xcite . the vast literature on the famous simulated annealing algorithm is not very relevant for our purpose but we refer to @xcite for a recent discussion . the paper @xcite concerned with filtering and genetic algorithms describes problems that are related in spirit to the present work . let @xmath2 be a finite set equipped with a sequence of kernels @xmath17 such that , for each @xmath11 , @xmath18 and @xmath19 . an associated markov chain is a @xmath2-valued random process @xmath20 such that , for all @xmath11 , @xmath21 the distribution @xmath22 of @xmath23 is determined by the initial distribution @xmath24 and given by @xmath25 where @xmath26 is defined inductively for each @xmath11 and each @xmath27 by @xmath28 with @xmath29 ( the identity ) . if we interpret the @xmath30 s as matrices , then this definition means that @xmath31 . this paper is mostly concerned with the behavior of the measures @xmath10 as @xmath11 tends to infinity . in the case of time homogeneous chains where all @xmath32 are equal , we write @xmath33 . our main interest is in ergodic like properties of time inhomogeneous markov chains . in general , one does not expect @xmath34 to converge toward a limiting distribution . instead , the natural notion is that of merging of measures as discussed in @xcite . fix a sequence of markov kernels as above . we say the sequence is merging if for any @xmath35 , @xmath36 if the sequence @xmath16 is merging then , for any two starting distributions @xmath37 , the measures @xmath34 and @xmath38 are merging , that is , @xmath39 . since we assume the set @xmath2 is finite , merging is equivalent to @xmath40 . hence , we also refer to this property as `` total variation merging . '' total variation merging is also referred to as weak ergodicity in the literature and there exists a body of work concerned with understanding when weak ergodicity holds . see , for example , @xcite . a main tool used to show weak ergodicity is that of contraction coefficients . furthermore , in @xcite , birkhoff s contraction coefficient is used to study _ ratio ergodicity _ which is equivalent to what we will later call relative - sup merging . however , it should be noted that even for time homogeneous chains birkhoff coefficients and related methods fail to provide useful quantitative bounds in most cases . our goal is to develop quantitative results in the context of time inhomogeneous chains in the spirit of the work of aldous , diaconis and others . in these works , precise estimates of the mixing time of ergodic chains are obtained . typically , a family of markov chains indexed by a parameter , say @xmath41 , is studied . loosely speaking , as the parameter @xmath41 increases , the complexity and size of the chain increases and one seeks bounds that depend on @xmath41 in an explicit quantitative way . see , for example , @xcite . efforts in this direction for time inhomogeneous chains are in @xcite . still , there are only a very small number of results and examples concerning the quantitative study of merging as defined above for time inhomogeneous markov chains so that it is not very clear what kind of results should be expected and what kind of hypotheses are reasonable . we refer the reader to @xcite for a more detailed discussion . the following definition is useful to capture the spirit of our study . it indicates that the simplest case we would like to think about is the case when the sequence @xmath4 is obtained by deterministic but arbitrary choices between a finite number of kernels @xmath42 . we say that a set @xmath43 of markov kernels on @xmath2 is merging in total variation if for any sequence @xmath44 with @xmath45 for all @xmath46 , we have @xmath47 in the study of ergodicity of finite markov chains , the convergence toward the target distribution is measured using various notions of distance between probability measures . these include the total variation distance @xmath48 the chi - square distance ( w.r.t . @xmath49 . note the asymmetry between @xmath50 and @xmath49 . ) @xmath51 and the relative sup - distance ( again , note the asymmetry ) @xmath52 these will be used here to measure merging . in the quantitative theory of ergodic time homogeneous markov chains , the notion of mixing time plays a crucial role . for time inhomogeneous chain , we propose to consider the following definitions . fix @xmath53 . given a sequence @xmath16 of markov kernels on a finite set @xmath2 , we call max total variation merging time the quantity @xmath54 fix @xmath53 . we say that a set @xmath43 of markov kernels on @xmath2 has max total variation @xmath55-merging time at most @xmath56 if for any sequence @xmath1 with @xmath45 for all @xmath46 , we have @xmath57 , that is , @xmath58 of course , merging can be measured in ways other than total variation . also merging is a bit less flexible than mixing in this respect since there is no reference measure . one very natural and much stronger notion than total variation is relative sup - distance . for time inhomogeneous chains , total variation merging does not necessarily imply relative - sup merging as defined below . see @xcite . we say a sequence @xmath16 of markov kernels on a finite set @xmath2 is merging in relative - sup if for all @xmath35 @xmath59 with the convention that @xmath60 and @xmath61 for @xmath62 . fix @xmath53 , we call relative - sup merging time the quantity @xmath63 we say a set @xmath43 of markov kernels on @xmath2 is merging in relative - sup if any sequence @xmath1 with @xmath64 for all @xmath46 is merging in relative - sup . fix @xmath53 . we say that @xmath43 has relative - sup @xmath55-merging time at most @xmath56 if for any sequence @xmath1 with @xmath45 for all @xmath46 , we have @xmath65 , that is , @xmath66 the following problem is open . it is a quantitative version of the problem stated at the beginning of the introduction . [ pb - bdu ] let @xmath5 and @xmath67 . let @xmath68 be the set of all birth and death chains @xmath69 on @xmath70 with @xmath71 $ ] if @xmath72 , and reversible measure @xmath73 satisfying @xmath74 , @xmath9 . 1 . prove or disprove that there exists a constant @xmath75 independent of @xmath41 such that @xmath68 has total variation @xmath55-merging time at most @xmath76 . 2 . prove or disprove that there exists a constant @xmath75 independent of @xmath41 such that @xmath68 has relative - sup @xmath55-merging time at most @xmath77 . this problem is open ( in most cases ) even if one considers a sequence @xmath1 drawn from a set @xmath78 of two kernels . observe that the hypothesis that the invariant measures @xmath3 are all comparable to the uniform plays some role . how to harvest the global hypothesis of comparable stationary distributions @xmath3 is not entirely clear . see theorem [ th - bdu ] below for a partial solution . if @xmath79 and @xmath80 are not comparable , it is possible for @xmath81 and @xmath82 to have the same mixing time yet for @xmath78 to have a merging time of a higher order . assume that @xmath83 and @xmath84 are two biased random walks with equal drift , one drift to left , the other to the right . despite the fact that each of these random walks has a relative - sup mixing time of order @xmath41 , the inhomogeneous chain driven by the sequence @xmath85 has a relative - sup merging time of order @xmath86 , see @xcite . in this section , we consider a property , @xmath0-stability , that plays a crucial role in the techniques we develop to provide quantitative bounds for time inhomogeneous markov chains . this property was introduced and discussed in @xcite . it is a straightforward generalization of the property of sharing the same invariant measure . unfortunately , it is hard to check . fix @xmath87 . a sequence of markov kernels @xmath88 on a finite set @xmath2 is @xmath0-stable if there exists a measure @xmath24 such that @xmath89 where @xmath34 . if this holds , we say that @xmath88 is @xmath0-stable with respect to the measure @xmath24 . a set @xmath43 of markov kernels is @xmath0-stable with respect to a measure @xmath24 if any sequence @xmath1 such that @xmath45 for all @xmath46 is @xmath0-stable with respect to @xmath24 . if all @xmath4 share the same invariant distribution @xmath73 then @xmath16 is @xmath90-stable with respect to @xmath73 . suppose a set @xmath43 of aperiodic irreducible markov kernels is @xmath0-stable with respect to a measure @xmath24 . let @xmath73 be an invariant measure for some @xmath91 . then we must have @xmath92 hence , @xmath43 is also @xmath93-stable with respect to @xmath73 and any two invariant measures @xmath94 for kernels @xmath95 must satisfy @xmath96 the following theorem which relates to a special case of problem [ pb - bdu ] illustrates the role of @xmath0-stability . [ th - bdu ] let @xmath5 . let @xmath68 be the set of all birth and death chains @xmath69 on @xmath70 with @xmath97 \qquad\mbox{if } |x - y|\le1\ ] ] and reversible measure @xmath73 satisfying @xmath74 , @xmath9 . let @xmath16 be a sequence of birth and death markov kernels on @xmath70 with @xmath98 . assume that @xmath16 is @xmath0-stable with respect to the uniform measure on @xmath70 , for some constant @xmath87 independent of @xmath41 . then there exists a constant @xmath99 ( in particular , independent of @xmath41 ) such that the relative - sup merging time for @xmath16 on @xmath70 is bounded by @xmath100 this will be proved later in a stronger form in section [ section - nash ] . in @xcite the weaker conclusion @xmath101 was obtained using singular value techniques . here , we will use nash inequalities to obtain @xmath102 . it is possible that the set @xmath68 is @xmath0-stable with respect to the uniform measure for some @xmath0 . indeed , it is tempting to conjecture that this is the case although the evidence is rather limited ( see also the discussion in @xcite ) . if this is true , then theorem [ th - bdu ] solves problem [ pb - bdu ] . however , we do not know how to approach the problem of proving @xmath0-stability for @xmath68 . while the assumption of @xmath0-stability in theorem [ th - bdu ] is quite strong , sections 4.2 and 5 of @xcite give specific examples of families @xmath68 for which it holds . further , we note that the question of whether or not @xmath0-stability holds is extremely natural and interesting in itself . one key idea in the study of markov chains is to associate to a markov kernel @xmath103 the operator @xmath104 . in the case of time homogeneous chains , one uses the basic fact that this operator acts on @xmath105 with norm @xmath90 when @xmath73 is an invariant measure . in the case of time inhomogeneous chains , it is crucial to consider @xmath103 as an operator between @xmath106 spaces with different measures in the domain and target spaces . the following simple observation is key . given a measure @xmath50 and a markov kernel @xmath103 on a finite set @xmath2 , set @xmath107 . fix @xmath108 and consider @xmath103 as a linear operator @xmath109 then @xmath110 this follows from jensen s inequality . see , for example , @xcite . we will use the notation @xmath111 whenever we need to emphasize the fact that @xmath103 is viewed as an operator between @xmath112 and @xmath113 for some @xmath114 . when the context is clear , we will drop the subscript @xmath50 as was done above . given a sequence of markov kernels @xmath16 , fix a starting measure @xmath24 and set @xmath115 . we will assume that @xmath116 for all @xmath11 . note that if @xmath117 and @xmath30 are all irreducible then @xmath116 for all @xmath118 . we are interested in the behavior of @xmath119 for @xmath120 , a classical argument involving the duality between @xmath106 and @xmath121 where @xmath122 , yields @xmath123\biggr|\dvtx \|f\|_{\ell^q(\mu_n)}\le1 \biggr\}\ ] ] and one checks that the function @xmath124 is nonincreasing ( see @xcite ) . of course , @xmath125 and , if @xmath126 , @xmath127 in particular , @xmath128 and @xmath129 further , if @xmath130 then @xmath131 to see the last inequality , note that if @xmath132 with @xmath133 then @xmath134 in @xcite , we developed basic inequalities for @xmath135 , @xmath136 based on singular value decompositions . the basic fact here is that , if @xmath50 is a probability measure on @xmath2 , @xmath103 a markov kernel and @xmath107 then @xmath137 where @xmath138 , @xmath139 , are the singular values of @xmath140 in nonincreasing order , that is the square root of the eigenvalues of @xmath141 where @xmath142 is the adjoint of @xmath143 . the @xmath144 s form an orthonormal basis for @xmath145 and are eigenfunctions of @xmath146 , @xmath144 being associated with @xmath147 . of course , the @xmath148 s can also be viewed as the eigenvalues of @xmath149 . in any case , a crucial fact for us here is that @xmath150 , the second largest singular value of @xmath151 , is also the norm of @xmath152 , that is , @xmath153 given a sequence @xmath16 of markov kernels on @xmath2 and a positive measure @xmath24 , set @xmath34 and let @xmath154 be the second largest singular value of @xmath155 . noting that @xmath156 we obtain @xmath157 this inequality seems very promising and this is rather misleading . there is very little hope to compute or estimate the singular values @xmath158 , even if we have a good grasp on the kernel @xmath4 . the reason is that @xmath154 depends very much on the unknown measure @xmath159 . this is similar to the problem one faces when studying an irreducible aperiodic time homogeneous finite markov chain for which one is not able to compute the stationary measure ( although this case is rarely discussed , it is the typical case ) . for positive examples and a more detailed discussion , see @xcite . given a reversible markov kernel @xmath69 with reversible measure @xmath73 on a finite set @xmath2 , the associated dirichlet form is @xmath160 this definition is essential for the techniques considered in this paper . to illustrate this , we note that the singular value @xmath161 associated to a markov kernel @xmath103 and a positive probability measure @xmath50 is the square root of the second largest eigenvalue of @xmath162 , @xmath107 . this operator is associated with the markov kernel @xmath163 which is reversible with respect to @xmath164 and has associated dirichlet form @xmath165 hence , using the classical variational formula for eigenvalues , we have @xmath166 where @xmath167 . the use of nash inequalities to study the convergence of ergodic ( time homogeneous ) finite markov chains was developed in @xcite ( section 7 of @xcite discusses time homogeneous chains that admits an invariant measure ) . we refer the reader to that paper for background on this technique . in this section , we observe that it can be implemented in the context of time inhomogeneous chains . we start with some basic material . [ def - opnorm ] let @xmath2 be a state space equipped with a markov kernel @xmath103 and probability measures @xmath50 and @xmath49 . if @xmath168 then @xmath169 if @xmath170 and @xmath171 are conjugate exponents , that is , if @xmath172 , then @xmath173 the following proposition is well known in a much more general context . [ pro - dualnorm ] let @xmath103 be a markov kernel . let @xmath174 be the markov operator on @xmath2 with adjoint @xmath175 with respect to the inner product @xmath176 if @xmath177 , @xmath172 and @xmath178 then @xmath179 let now @xmath16 be a sequence of markov kernels on @xmath2 . fix a positive probability measure @xmath24 and set @xmath34 as usual . consider @xmath180 , its adjoint @xmath181 and @xmath182 . the operator @xmath183 is given by the markov kernel @xmath184 this kernel is reversible with reversible measure @xmath185 . we let @xmath186 be the associated dirichlet form on @xmath187 . [ thm - nash ] referring to the setup and notation introduced above , let and assume that there are constants @xmath188 such that for @xmath189 the following nash inequalities hold @xmath190\\[-8pt ] & & \hspace*{13.3pt } { } + \frac{1}{n}\|f\|_{\ell^2(\mu _ m)}^2\biggr)\|f\|_{\ell^1(\mu_m)}^{1/d } .\nonumber\end{aligned}\ ] ] then , for @xmath191 , @xmath192 where @xmath193 . let @xmath44 be a sequence of markov kernels on @xmath2 such that the nash inequalities ( [ eqn - nash ] ) hold . pick a function @xmath194 such that @xmath195 . for @xmath196 define @xmath197 note that for any @xmath198 , @xmath199 is nonincreasing . indeed , using the contraction property ( [ cont ] ) , we have @xmath200 moreover , note that for any @xmath201 @xmath202 where @xmath203 and @xmath204 are the constants in ( [ eqn - nash ] ) . this follows by applying the nash inequality to the function @xmath205 . corollary 3.1 of @xcite then yields that @xmath206 where @xmath193 . in particular , if @xmath191 , @xmath207 from proposition [ pro - dualnorm ] it follows that , for @xmath191 , @xmath208 next we bound @xmath209 for @xmath191 . consider the quantity @xmath210 where @xmath211 let @xmath212 , so that @xmath213 . we have @xmath214 note that for all @xmath215 @xmath216 this follows from the fact that for any function @xmath194 @xmath217 by ( [ cont ] ) , we have @xmath218 the last inequality follows from the fact that @xmath219 so we have @xmath220 and it follows that for all @xmath191 @xmath221 by duality , we get that @xmath222 next , we use the riesz thorin interpolation theorem , see @xcite , page 179 , which gives us the desired result . the next results show how theorem [ thm - nash ] together with the singular value technique of section [ sec - sv ] yields merging results . [ thm - nash2 ] referring to the above setup and notation , let @xmath223 and assume that there are constants @xmath188 such that for @xmath189 the nash inequalities @xmath224\\[-8pt ] & & { } \times\|f\|_{\ell^1(\mu_m)}^{1/d}\nonumber\end{aligned}\ ] ] hold . let @xmath225 be the second largest singular value of @xmath226 , that is , the square root of the second largest eigenvalue of @xmath227 . then , for @xmath228 , @xmath229 , we have @xmath230 moreover , for any @xmath231 , @xmath232 , we have @xmath233 we have @xmath234 where @xmath22 is understood as the expectation operator @xmath235 . moreover , for any @xmath236 , @xmath237 because @xmath238 . hence , for @xmath232 , @xmath239 using @xmath240 , gives ( [ n+s ] ) . to obtain the stronger result ( [ sup - n+s ] ) , write @xmath241 and @xmath242 the stated bound ( [ sup - n+s ] ) follows . just as we did for singular values , let us emphasize that the powerful looking results stated in this theorem are actually extremely difficult to apply . again , the point is that the dirichlet form @xmath243 , the space @xmath244 , and the singular values @xmath225 all involve the unknown sequence of measures @xmath34 , @xmath245 the following subsection gives similar but more applicable results under additional hypotheses involving the notion of @xmath0-stability . we state two results that parallel theorems 5.9 and 5.10 of @xcite . [ th - nash - stab ] fix @xmath246 . let @xmath16 be a sequence of irreducible markov kernels on a finite set @xmath2 . assume that @xmath16 is @xmath0-stable with respect to a positive probability measure @xmath24 . for each @xmath46 , set @xmath247 and let @xmath248 be the second largest singular value of @xmath249 as an operator from @xmath250 to @xmath251 . let @xmath252 . let @xmath223 and assume that there are constants @xmath188 such that for @xmath189 the nash inequalities @xmath253\\[-8pt ] & & { } \times\|f\|_{\ell^1(\mu^0_m)}^{1/d}\nonumber\end{aligned}\ ] ] holds . then , for @xmath228 , @xmath229 , we have @xmath254\\[-8pt ] & & { } \times \prod_{m+1}^n \biggl(1-\frac{1-\sigma(k_i,\mu_0)^2}{c^2}\biggr)^{1/2}.\nonumber\end{aligned}\ ] ] moreover , for any @xmath231 , @xmath232 , we have @xmath255 first note that since @xmath256 $ ] , we have @xmath257 $ ] . consider the operator @xmath183 with kernel @xmath258 by assumption @xmath259,\ ] ] where the term in brackets on the right - hand side is the kernel of @xmath260 . this kernel has second largest eigenvalue @xmath261 . a simple eigenvalue comparison argument yields @xmath262 further , comparison of measures and dirichlet form yields the nash inequality @xmath263 together with theorem [ thm - nash2 ] , this gives the stated result . the next result is based on a stronger hypothesis . [ th - nash - stab2 ] fix @xmath246 . let @xmath264 be a family of irreducible aperiodic markov kernels on a finite set @xmath2 . assume that @xmath264 is @xmath0-stable with respect to some positive probability measure @xmath24 . let @xmath16 be a sequence of markov kernels with @xmath64 for all @xmath46 . let @xmath3 be the invariant measure of @xmath4 . let @xmath265 where @xmath266 . let @xmath267 be the second largest singular value of @xmath4 as an operator on @xmath268 . let @xmath223 and assume that there are constants @xmath188 such that for @xmath189 the nash inequalities @xmath269\\[-8pt ] & & { } \times\|f\|_{\ell^1(\pi_m)}^{1/d}.\nonumber\end{aligned}\ ] ] then , for @xmath228 , @xmath229 , we have @xmath270\\[-8pt ] & & { } \times\prod_{m+1}^n \biggl(1-\frac{1-\sigma_1(k_i)^2}{c^4}\biggr)^{1/2}.\nonumber\end{aligned}\ ] ] moreover , for any @xmath231 , @xmath232 , we have @xmath271 note that the hypothesis that @xmath264 is @xmath0-stable implies @xmath272 $ ] for all @xmath273 . consider again the operator @xmath183 and its kernel @xmath258 by assumption @xmath274\\ & \ge & c^{-2 } \pi_i(x)\tilde{p}_i(x , y).\end{aligned}\ ] ] a comparison argument similar to the one used in the previous proof yields the desired result . this section describes applications of the nash inequality technique to several examples . all these examples are of the following general type . \(1 ) there is a basic reversible model @xmath275 on a space @xmath70 ( growing with @xmath41 ) that is well understood because : * we have good grasp on the second largest singular value @xmath276 of @xmath275 . * the model @xmath275 satisfies a good nash inequality , that is , an inequality of the form @xmath277 with @xmath278 independent of @xmath41 and @xmath279 . here , @xmath280 implies that there exist constants @xmath281 such that @xmath282 . * together , the nash inequality and second largest singular value estimate yield the mixing time estimate @xmath283 where @xmath75 is independent of @xmath41 . \(2 ) we are given a sequence @xmath16 or a set @xmath68 of markov kernels on @xmath70 which satisfies : * @xmath16 or @xmath68 is @xmath0-stable with respect to a measure @xmath24 which is either equal or at least comparable to @xmath73 . * the markov kernels @xmath4 or the elements of @xmath68 are all bounded perturbations of @xmath103 in the sense that @xmath284 is bounded away from @xmath285 and away from @xmath286 for all @xmath287 . in particular , @xmath288 if and only if @xmath289 . under such circumstances , theorem [ th - nash - stab ] ( or theorem [ th - nash - stab2 ] ) applies and yields the conclusion that the time inhomogeneous markov chain associated with the sequence @xmath4 under investigation has a relative - sup merging time @xmath290 bounded by @xmath291 for some constant @xmath292 independent of @xmath41 . the most obvious basic model is , perhaps , the simple random walk on @xmath293 ( with some holding if @xmath41 is even to avoid periodicity ) . this model has @xmath294 and satisfies the desired nash inequality with @xmath295 . the first subsection presents applications to a perturbation of this model . in this example , @xmath296 is a finite circle . it will be convenient to enumerate the points in @xmath70 by writing @xmath297 if @xmath298 and @xmath299 if @xmath300 . the simple random walk in @xmath2 has kernel @xmath301 and reversible measure @xmath302 . for any @xmath303 , define the perturbation kernel @xmath304 for @xmath305 , the markov kernel @xmath306 is a perturbation of @xmath69 . see figure [ fig1 ] . for any fixed @xmath307 , set @xmath308\}.\ ] ] we shall see below that @xmath309 is @xmath0-stable . let @xmath310 be the set of all probability measures on @xmath70 which satisfy the following two properties : 1 . for all @xmath9 , there exist constants @xmath311 such that @xmath312 and @xmath313 2 . for all @xmath9 we have that @xmath314 . note that we always have @xmath315 ( since @xmath316 ) and , in the case when @xmath298 , @xmath317 . [ thm - cstab ] let @xmath318 defined above , then for any @xmath319 we have that @xmath320 . let @xmath318 and @xmath321 , @xmath322 $ ] . we show that @xmath323 has the properties required to be in @xmath310 . \(1 ) any measure @xmath324 can be written as @xmath325 where @xmath326 is the ( nonprobability ) measure @xmath327 . a simple calculation yields that @xmath328 since @xmath312 , we obtain that @xmath329 the fact that @xmath330 implies that @xmath331 satisfies property @xmath332 in the definition of @xmath333 . to see that @xmath334 also satisfies this property , we note that @xmath335 it now follows that @xmath336 has property @xmath332 in the definition of @xmath310 since @xmath337 . \(2 ) we consider the measure @xmath323 . for @xmath338 property @xmath339 of @xmath310 follows easily from the fact that @xmath314 and @xmath340 for @xmath341 , we note that @xmath342 similarly @xmath343 the proof now follows from the fact that @xmath344 as proved in part ( 1 ) above . the family @xmath345 is @xmath346-stable with respect to any @xmath347 . claim [ thm - cstab ] implies that for any sequence @xmath44 such that @xmath348 and any measure @xmath347 we have @xmath349 for all @xmath118 . note that for any measure @xmath350 we have that @xmath351 hence , @xmath352 when @xmath298 , the kernels @xmath353 yield periodic chains on @xmath70 . in this case , we will study the merging properties of @xmath354 that is , the so - called lazy version of @xmath345 . we set @xmath355 for any @xmath318 , we consider the kernel @xmath356 which is the kernel of @xmath357 where @xmath358 . this is @xmath285 unless @xmath359 and we compare it to @xmath360 which is @xmath361 if @xmath362 , @xmath363 if @xmath364 , @xmath365 if @xmath366 and @xmath285 otherwise . the definitions of @xmath353 and @xmath310 yield @xmath367 this yields @xmath368 whereas the stability property implies that the relevant measures @xmath369 and @xmath370 satisfy @xmath371 in the case when @xmath300 , we may work directly with the kernels @xmath372 as they are not periodic . an analysis similar to that above will give versions of ( [ eqn - lazycirc - comp - dir ] ) and ( [ eqn - lazycirc - comp - meas ] ) for @xmath353 . applying the line of reasoning explained at the beginning of this section and using theorem [ th - nash - stab2 ] , we get the following result . [ thm - lazycirc ] fix @xmath133 . for any @xmath373 the total variation @xmath374-merging time of the family @xmath375 on @xmath376 [ resp . , @xmath345 on @xmath377 is at most @xmath378 for some constant @xmath379 . in fact , we can choose @xmath380 such that @xmath381 for any sequence @xmath382 [ resp . , @xmath383 . in @xcite , nash inequalities are used to study certain birth and death chains on @xmath384 with reversible measures which belong to one of the following two families : @xmath385 and @xmath386 here , we consider @xmath387 to be a fixed parameter and are interested in what happens when @xmath41 tends to infinity . from this perspective , the normalizing constants @xmath388 are comparable and behave as @xmath389 set @xmath390 here , all @xmath391 must be understood for fixed @xmath392 and the implied comparison constants depend on @xmath392 . let @xmath393 ( resp . , @xmath394 ) be the markov kernel of the metropolis chain with basis the symmetric simple random walk on @xmath70 with holding @xmath395 at all points except at the end points where the holding is @xmath396 , and target @xmath397 , ( resp . , @xmath398 ) . let @xmath399 , @xmath400 be the corresponding spectral gaps . let @xmath401 , @xmath402 be the relative - sup mixing times of these chains . it is proved in @xcite that @xmath403 whereas @xmath404 \log_+ 1/\eta\bigr ) .\end{aligned}\ ] ] note that @xmath405 these results are based on the nash inequalities satisfied by these chains . namely , letting @xmath406 or @xmath407 and @xmath408 or @xmath409 , there are constants @xmath410 such that @xmath411 with @xmath412 . see @xcite . in cite @xcite , the authors consider the class of birth and death chains @xmath69 on @xmath413 that are symmetric with respect to the middle point , that is , satisfy @xmath414 , @xmath415 , @xmath416 , @xmath417 . for any such chain @xmath69 , let @xmath49 be the reversible measure . it satisfies @xmath418 . consider the perturbation set @xmath419\},\qquad \varepsilon\in[0,q_0),\ ] ] where @xmath420 , @xmath421 and @xmath422 otherwise . these perturbations at the middle vertex have reversible measure @xmath423 that satisfy @xmath424 the main point of this construction is the following . fix @xmath69 , @xmath49 as above and @xmath425 . the set @xmath426 is @xmath0-stable with respect to @xmath427 with @xmath428 . in order to apply this results to our example @xmath429 , we observe that @xmath430 and @xmath431 now , theorem [ th - nash - stab2 ] yields the following result . fix @xmath387 and set @xmath432 , @xmath433 . 1 . there exists a constant @xmath75 independent of @xmath41 such that , for any sequence @xmath16 with @xmath434 , we have @xmath435 2 . there exists a constant @xmath75 independent of @xmath41 such that , for any sequence @xmath16 with @xmath436 , we have @xmath437 this section develops the technique of logarithmic sobolev inequality for time inhomogeneous finite markov chains . it should be noted that the logarithmic sobolev technique has been mostly applied in the literature in the context of continuous time chains . in @xcite , miclo tackled the problem of adapting this technique to discrete time ( time homogeneous ) chains . there are two different ways to use logarithmic sobolev inequality for mixing estimates . one , the most powerful , provides results for relative - sup merging and is based on hypercontractivity . the other is based on entropy and only produces bounds for total variation merging . we will discuss and illustrate both approaches below in the context of time inhomogeneous chains . the entropy approach is already treated in @xcite . recall that , for any positive probability distribution @xmath50 , a markov kernel @xmath103 can be thought of as a contraction @xmath438 the adjoint @xmath439 has kernel @xmath440 set @xmath441 . we define the logarithmic sobolev constant @xmath442 where the @xmath443 relative entropy @xmath444 of a function @xmath194 with respect to the measure @xmath49 is defined by @xmath445 proof of proposition [ pro - opnorm - ls ] to prove proposition [ pro - opnorm - ls ] is suffices to only consider positive functions . for @xmath457 , we begin by writing @xmath458\\[-8pt ] & & { } + \|f\|_{\ell^q(\mu')}-\|f\| _ { \ell^2(\mu')}.\nonumber\end{aligned}\ ] ] the difference of the last two terms on the right - hand side is controlled by lemma [ lem3 ] . to control the first two terms , we will use the concavity result @xmath459 it follows that @xmath460 set @xmath461 following the notation of lemma [ lem4 ] , fix @xmath462 and set @xmath463 , @xmath464 and @xmath465 . if @xmath466 , then @xmath467 and so @xmath468 fix @xmath12 and integrate with respect to the measure @xmath469 to get @xmath470 we also have @xmath471 hence , @xmath472 integrating with respect to @xmath50 gives us that @xmath473 it follows from lemma [ lem3 ] , ( [ eqn1-ls ] ) and ( [ eqn2-ls ] ) that @xmath474 in @xcite , it is noted that for all @xmath475 and @xmath452 we have @xmath476 . so if @xmath447q_0 $ ] then @xmath477q_0 $ ] . hence , @xmath478 since @xmath479 is the logarithmic sobolev constant , we get our desired result , @xmath480 [ cor - sob ] let @xmath481 be a sequence of markov kernels on a finite set @xmath2 and @xmath24 be an initial distribution on @xmath2 . set @xmath34 . consider @xmath482 and @xmath483 . let @xmath484 be the logarithmic sobolev constant of @xmath183 . then for any @xmath446 and @xmath485 , we have that @xmath486 [ thm - ls ] let @xmath2 be a finite set equipped with a sequence of markov kernels @xmath481 and an initial distribution @xmath24 . let @xmath34 . consider @xmath491 and @xmath483 . let @xmath484 be the logarithmic sobolev constant of @xmath183 . set @xmath492 then for @xmath493 , we have that @xmath494 fix @xmath12 , and let @xmath495 . if @xmath496 , @xmath497 . indeed , for any @xmath498 and @xmath499 we have that @xmath500 moreover , if @xmath501 is thought of as the expectation operator @xmath502 , @xmath503 , then @xmath504 . let @xmath505 set @xmath506 and @xmath507 to be the conjugate exponent of @xmath508 so that @xmath509 . by duality , we have @xmath510 by assumption , we have that @xmath511 , it now follows from corollary [ cor - sob ] that @xmath512 [ thm - ls - c - stab ] fix @xmath246 . let @xmath2 be a finite set equipped with a sequence of irreducible markov kernels , @xmath1 . assume that @xmath1 is @xmath0-stable with respect to a positive probability measure @xmath24 . for each @xmath46 , set @xmath247 and let @xmath513 be the second largest singular value of the operator @xmath514 and @xmath515 the logarithmic sobolev constant for the operator @xmath516 . if @xmath517 then for @xmath518 we have that @xmath519 first , we note that @xmath520 $ ] . let @xmath183 be the markov kernel described in the statement of theorem [ thm - ls ] . by the same arguments as in theorem [ th - nash - stab ] , we get that for all @xmath462 @xmath521 a simple comparison argument similar to those used in the proof of theorem [ th - nash - stab ] ( see also @xcite ) yields that @xmath522 the first inequality implies that @xmath523 where @xmath524 is defined in the proof of theorem [ thm - ls ] . using the results of theorem [ thm - ls ] and the second inequality above gives the desired result . let @xmath246 . let @xmath264 be a family of irreducible aperiodic markov kernels on a finite set @xmath2 . assume that @xmath264 is @xmath0-stable with respect to some positive probability measure @xmath24 . let @xmath1 be a sequence of markov kernels with @xmath64 for all @xmath46 . let @xmath3 be the invariant measure of @xmath4 . let @xmath525 be the second largest singular value for the operator @xmath526 . let @xmath515 be the logarithmic sobolev constant for the operator @xmath527 where @xmath528 is the adjoint of @xmath526 . if @xmath529 then for @xmath518 we have that @xmath530 to control the relative - sup merging time by this method , we need an additional hypothesis . in the case of the @xmath443 distance , we only required a control over the logarithmic sobolev constant of the kernel @xmath483 . in this case , we will also need to control the logarithmic sobolev constant of @xmath533 where @xmath528 is the adjoint of the operator @xmath4 from @xmath187 to @xmath534 . [ thm - ls - sup ] let @xmath2 be a finite set equipped with a sequence of markov kernels @xmath481 and an initial distribution @xmath24 . let @xmath34 and @xmath483 and @xmath533 where @xmath528 is the adjoint of @xmath4 with respect to the measure @xmath185 . let @xmath484 and @xmath535 be the logarithmic sobolev constants of @xmath183 and @xmath536 , respectively . if @xmath537 and @xmath538 then for any @xmath539 , @xmath540 where @xmath541 . proof of theorem [ thm - ls - sup ] write @xmath543 and @xmath544 note that @xmath545 so we just need to bound the remaining terms in the right - hand side of the inequality above . to bound @xmath546 set @xmath547 and write @xmath548 it follows from corollary [ cor - sob ] that @xmath549 by assumption , we have that @xmath550 so we get @xmath551 to bound @xmath552 set @xmath506 and write @xmath553 it follows from corollary [ cor - sob ] that @xmath554 since @xmath555 we get @xmath556 . fix @xmath246 . let @xmath2 be a finite set equipped with a sequence of markov kernels @xmath88 . assume that @xmath88 is @xmath0-stable with respect to a positive probability measure @xmath24 . for each @xmath46 , set @xmath247 and @xmath557 . let @xmath558 be the second largest singular value of the operator @xmath514 . let @xmath515 be the logarithmic sobolev constant of the operator @xmath516 where @xmath528 is the adjoint of @xmath514 . let @xmath559 be the logarithmic sobolev constant of the operator @xmath560 where @xmath528 is the adjoint of @xmath561 . if @xmath562 and @xmath563 then for any @xmath564 @xmath565 where @xmath566 . note that @xmath520 $ ] and @xmath567 $ ] . let @xmath183 and @xmath536 be the markov kernels described in theorem [ thm - ls - sup ] with kernels @xmath568 similar reasoning to that of theorem [ thm - ls - c - stab ] gives @xmath569 where @xmath528 above is the adjoint of @xmath570 . this implies that @xmath571 where @xmath572 is defined in theorem [ thm - ls - sup ] . in the case of @xmath536 , equation ( [ eqn - ch - p - ls - sup ] ) gives @xmath573 where @xmath528 is the adjoint of the operator @xmath574 . a simple comparison argument yields @xmath575 and so @xmath576 where @xmath577 is defined in theorem [ thm - ls - sup ] . the desired result now follows from theorem [ thm - ls - sup ] . fix @xmath246 . let @xmath264 be a family of irreducible aperiodic markov kernels on @xmath2 . assume that @xmath264 is @xmath0-stable with respect to some positive probability measure @xmath24 . let @xmath88 be a sequence with @xmath578 for all @xmath579 . let @xmath3 be the invariant measure of @xmath4 and @xmath267 the second largest singular value of the operator @xmath4 acting on @xmath580 . let @xmath515 and @xmath559 be the logarithmic sobolev constants of the operators @xmath527 and @xmath581 where @xmath582 is the adjoint of @xmath583 . if @xmath562 and @xmath584 then for any @xmath564 @xmath585 where @xmath566 . first , note that @xmath532 $ ] . equation ( [ eqn - p - ls - sup ] ) implies that @xmath586 to bound @xmath535 , we use ( [ eqn - ch - p - ls - sup ] ) to get that for all @xmath462 @xmath587 this implies that @xmath588 . it follows that @xmath589 where @xmath542 is defined in theorem [ thm - ls - sup ] . applying theorem [ thm - ls - sup ] now gives us the desired result . denote by @xmath590 the @xmath591-dimensional hypercube , we say that @xmath462 are neighbors , or @xmath592 if @xmath593 where @xmath594 is the @xmath46th coordinate of @xmath595 . the simple random walk on @xmath2 is driven by the kernel @xmath596 it is easy to check that @xmath50 , the uniform measure on @xmath2 , is stationary for @xmath103 . fix @xmath53 and consider the following perturbed version of @xmath103 . @xmath597 for @xmath53 , set @xmath598 \}.\ ] ] the example of time inhomogeneous markov chains associated to @xmath599 above is related to the binomial example in @xcite . see remark [ rem - bino ] below . 1 . for all @xmath595 with @xmath603 we have @xmath604 . 2 . for all @xmath605 there exists constants @xmath606 such that @xmath607 and for any @xmath12 with @xmath608 we have @xmath609 3 . for all @xmath605 we have @xmath610 . \(2 ) for @xmath46 such that @xmath618 , consider an element @xmath12 such that @xmath608 . then @xmath619\\ & = & \frac{1}{4^n}+\frac{1}{2n}\bigl(a_{\nu , i+1}|x|+a_{\nu , i-1}(2n-|x|)\bigr).\end{aligned}\ ] ] a similar computation as above yields that for an element @xmath12 with @xmath620 we have @xmath621 finally , we check that cases for elements @xmath12 with @xmath628 . consider an @xmath12 such that @xmath629 , then @xmath630 when @xmath631 , then @xmath632 as desired . we can now concluded that @xmath633 . \(3 ) from the calculations in part ( 2 ) , we know that for @xmath12 with @xmath634 and @xmath618 and @xmath635 we have @xmath636 and @xmath637 respectively . it follows from the fact that for all @xmath46 , @xmath610 that for both cases above @xmath638 . when @xmath639 , we have that for @xmath12 with @xmath640 @xmath641 a similar calculation yields @xmath642 . the proof now follows from the fact that @xmath633 . the kernels @xmath648 drive periodic chains that will alternate between points with an even number of @xmath90 s and odd number of @xmath90 s . so we will study following random walk driven by the kernel @xmath649 where @xmath650 is the identity . set @xmath651\}.\ ] ] [ clm - hyper - sing - ls ] let @xmath1 be a sequence of markov kernels such that @xmath652 for all @xmath579 . let @xmath644 be a positive measure , and let @xmath34 . set @xmath483 where @xmath528 is the adjoint of @xmath482 . let @xmath653 and be the second largest singular value of @xmath654 . let @xmath484 be logarithmic sobolev constant of @xmath183 . then @xmath655 where @xmath656 . let @xmath657 and @xmath370 be the uniform measure on @xmath658 . let @xmath659 . using the @xmath660-stability of the sequence @xmath661 , we get that @xmath662 a simple comparison yields @xmath663 further comparison gives that @xmath664 it is well known that for @xmath665 ( the simple random walk ) we have @xmath666 . this implies that @xmath667 . the singular value inequality in claim [ clm - hyper - sing - ls ] now follows from ( [ hyper - sigma ] ) . for the rest of the proof , we note that lemma 2.5 of @xcite tells us that @xmath668 , and so we get @xmath669 . the logarithmic sobolev inequality now follows from ( [ hyper - ls ] ) and the fact that @xmath670 . [ th - hyper ] for any @xmath53 there exists a constant @xmath671 such that the total variation merging time of the sequence @xmath1 with @xmath652 for all @xmath672 is bounded by @xmath673 moreover , we can chose @xmath671 such that @xmath674 [ rem - bino ] the theorem above is closely related to the example in section 5.2 of @xcite which studies a time inhomogeneous chain on @xmath675 resulting from perturbations of a birth and death chain with binomial stationary distribution . both @xcite and theorem [ th - hyper ] give the correct upper bound on the merging time yet @xcite requires knowledge about the entire spectrum of the operators driving the chain while the theorem above uses logarithmic sobolev techniques . let @xmath49 and @xmath676 be two probability measures on @xmath2 . define the relative entropy between @xmath50 and @xmath49 as @xmath677 it is well known that @xmath678 . let @xmath481 be a sequence of markov kernels on @xmath2 , @xmath24 be some initial distribution on @xmath2 and @xmath34 . it follows by the triangle inequality that for any @xmath462 @xmath679 let @xmath680 be the largest constant such that for any probability measure @xmath50 @xmath681 let @xmath107 and @xmath682 be the adjoint of @xmath683 . set @xmath684 in @xcite , the contraction constant @xmath392 is related to the so - called modified logarithmic sobolev constant @xmath685 [ pro - dlm ] there exists a universal constant @xmath686 such that for any markov kernel @xmath103 and any probability measure @xmath50 , @xmath687 where @xmath688 and @xmath689 is the adjoint of the operator @xmath683 , @xmath107 . the proof of proposition 5.1 in @xcite uses the fact that there exists some @xmath691 such that for all @xmath692 @xmath693 where @xmath694 let @xmath695 . we will show that for all @xmath692 then @xmath696 . by differentiating @xmath194 we get @xmath697 and @xmath698 in particular , for @xmath699 $ ] we have @xmath700 . this along with the fact that @xmath701 implies that there exists only one @xmath702 such that @xmath703 . it follows that @xmath194 is decreasing on @xmath704 $ ] and @xmath194 is increasing on @xmath705 $ ] . since @xmath706 , then for @xmath699 $ ] we have that @xmath696 . for @xmath707 , we note that @xmath708 which implies that @xmath709 . the fact that @xmath710 implies @xmath711 . the desired result follows from the fact that the proof of proposition 5.1 in @xcite shows that @xmath712 [ pro - entropy ] let @xmath2 be a finite state space equipped with a sequence of markov kernels @xmath88 and an initial distribution @xmath24 . let @xmath34 and @xmath533 where @xmath528 is the adjoint of @xmath482 . set @xmath713 then for any @xmath462 @xmath714 where @xmath715 is given in [ pro - dlm ] [ pro - rho - ls ] . in this section , we present two examples where the modified logarithmic sobolev inequality technique yields the correct merging time while the regular logarithmic sobolev inequality technique does not . let @xmath719 be the symmetric group equipped with the uniform probability measure @xmath370 . let @xmath720 be the the kernel of transpose @xmath46 with random , that is , @xmath721 $ , \cr 0 , & \quad otherwise . } \ ] ] let @xmath722 be the associated lazy chain . it is known that the lazy chain has a mixing time of of @xmath723 . more precisely , @xmath724 see , for example , @xcite . the results of @xcite show that the modified logarithmic sobolev constant for @xmath725 is bounded by @xmath726 set @xmath727 . since all @xmath725 are reversible with respect to the uniform distribution @xmath370 , the set @xmath264 is @xmath90-stable with respect to @xmath370 . using the methods of @xcite ( see also @xcite ) , one can prove that for any sequence @xmath16 with @xmath64 for all @xmath579 we have @xmath728 the inequality above is due to the fact that the @xmath725 are driven by probability measures so the @xmath443 distance bounds the @xmath729 distance and the eigenvectors in theorem 3.2 of @xcite drop out to give @xmath730 one can then group the singular values in the equality above since the @xmath725 s are all are images of each other under some inner automorphism of @xmath731 which implies @xmath732 for all @xmath733 . for a more detailed discussion , see @xcite . we now consider two variants of this example that can not be treated using the singular values techniques of @xcite or the logarithmic sobolev inequality technique of sections [ subsec - hyper][subsec - hypercube ] but where the modified logarithmic sobolev inequality does yield a successful analysis . this technique can be applied to the two examples in this section because of the following three reasons : 1 . any sequence @xmath1 of interest can be shown to be @xmath0-stable with respect to some well chosen initial distribution ; 2 . all the kernels @xmath4 driving the time inhomogeneous process are directly comparable to the @xmath725 s and , 3 . due to @xmath332 and the laziness of the @xmath725 s we can successfully estimate the modified logarithmic sobolev constants @xmath734 to be of order @xmath735 . hence , @xmath736 is the set of all symmetric edge perturbations of kernels in @xmath264 . as we require symmetry , the uniform distribution is invariant for all the kernels in @xmath736 . now , what can be said of the merging properties of sequences @xmath16 with @xmath741 ? unlike @xmath264 , the kernels in @xmath736 are not invariant under left multiplication in @xmath731 . so the eigenvectors of theorem 3.2 in @xcite do not drop out , and we only get @xmath742 singular value comparison yields @xmath743 which gives @xmath744 this indicates merging after order @xmath745 steps instead of the expected order @xmath746 steps . for any sequence @xmath16 with @xmath747 for all @xmath579 set @xmath748 where @xmath528 is the adjoint of the operator @xmath749 . a simple comparison argument gives @xmath750 for some @xmath751 $ ] . further comparison yields @xmath752 . lemma 2.5 of @xcite implies that @xmath753 so @xmath754 is of order at least @xmath735 . hence , there exists some constant @xmath755 independent of @xmath11 such that @xmath756 in particular , for some constant @xmath671 we get @xmath757 . to obtain a result for the relative - sup norm , one can use the ( nonmodified ) logarithmic sobolev technique as the modified logarithmic sobolev technique only gives bounds in total variation . it is known that the logarithmic sobolev constant for top to random is of order @xmath758 , see @xcite , leading to results that are off by a factor of @xmath759 . this technique yields the best available result , @xmath760 we now consider a second variation on the transpose cyclic to random example . let @xmath761 , @xmath762 and consider the markov kernel @xmath763$. } \ ] ] in words , @xmath103 is obtained from @xmath764 by adding extra holding probability at @xmath715 , making @xmath715 `` sticky . '' next , if @xmath765 is the cycle @xmath766 , let @xmath767 in words , @xmath4 is @xmath725 with some added holding at @xmath768 . we would like to consider the merging properties of the sequence @xmath16 . unlike the previous example , the uniform probability is not invariant under @xmath4 . however , this type of construction is considered in @xcite . let @xmath769 so that @xmath770 . it is proved that @xmath16 is @xmath771-stable with respect to the probability measure @xmath772 , where @xmath773 is the invariant probability measure of the markov kernel @xmath774 . from the analysis in @xcite , section 5 , one can see that @xmath775 applying the singular value techniques used in section 5 of @xcite would give us an upper bound on the relative sup merging time of order @xmath745 . set @xmath533 where @xmath528 is the adjoint of the operator @xmath776 . since @xmath777 , for @xmath778 we can write @xmath779 it follows by comparison that @xmath780 . we can successfully estimate @xmath781 due to lemma 2.5 of @xcite which implies @xmath782 . so we have that @xmath754 is at least @xmath783 . proposition [ pro - entropy ] gives us that @xmath784 where @xmath715 is as in proposition [ pro - rho - ls ] . so for some constant @xmath785 , we get @xmath786

in this paper , we study the degrees of freedom ( dof ) of a new network information flow model called symmetric multi - relay mimo y channel , where three user nodes , each with antennas , communicate via geographically separated relay nodes , each with antennas . for this model , we establish a general dof achievability framework based on linear precoding and post - processing methods . the framework poses a nonlinear problem with respect to user precoders and post - processors , as well as relay precoders . to solve this problem , we adopt an _ uplink - dowlink asymmetric _ strategy , where the user precoders are designed for signal alignment and the user post - processors are used for interference neutralization . with the user precoder and post - processor designs fixed as such , the original problem then reduces to a problem of relay precoder design . to address the solvability of the system , we propose a general method for solving matrix equations . this method is also useful to the dof analysis of many other multiway relay networks . together with the techniques of antenna disablement and symbol extension , an achievable dof of the symmetric multi - relay mimo y channel is derived for an arbitrary setup of . we show that , for , the optimal dof is achieved for . we also show that for symmetric multi - relay mimo y channel , our design outperforms the conventional uplink - downlink symmetric design considerably . multiway relay channel , network coding , signal alignment , partial antenna disablement , symbol extension

1511.08887

wireless relaying , as a promising technique to extend the coverage and increase the system capacity , have attracted much research interest in recent years @xcite . physical - layer network coding ( pnc ) @xcite is an advanced relaying technique that significantly enhances spectral efficiency by allowing a relay node to decode and forward message combinations rather than individual messages . pnc was first proposed as a two - way relaying mechanism , where two users exchange information via a relay node . later , the idea of pnc was extended to support efficient communications over multiway relay channels ( mrc ) @xcite , where multiple users exchange data with the help of a single relay . efficient pnc design has been studied for various data exchange models , including pairwise data exchange @xcite , full data exchange @xcite , and clustered pairwise / full data exchange @xcite . multiple - input multiple - output ( mimo ) techniques have also been incorporated into pnc - aided relay networks to achieve spatial multiplexing @xcite . the capacity of general mimo mrc remains a challenging open problem @xcite . existing work @xcite was mostly focused on analyzing the degrees of freedom ( dof ) that characterizes the capacity slope at high signal - to - noise ratio ( snr ) . various signaling techniques have been developed to intelligently manipulate signals and interference based on the ideas of pnc and interference alignment @xcite . particularly , the authors in @xcite studied the dof of the mimo y channel , where three users exchange data in a pairwise manner with the help of a single relay . to derive the dof of this model , a key difficulty is how to jointly optimize the linear processors , including the precoders at the user transmitters , the precoder at the relay , and the post - processers at user receivers . this problem was elegantly solved in @xcite by optimal design of the signal space seen at the relay , where the user precoders and post - processors are constructed by _ pairwise signal alignment _ and _ uplink - downlink symmetry _ , and the relay precoder by appropriate orthogonal projections . similar ideas have also been used to derive the dof of other multiway relay models @xcite . in the work on mimo mrc mentioned above , a major limitation is that a single relay node is employed to serve multiple user nodes simultaneously . this implies that the relay node is usually the performance bottleneck of the overall network @xcite . as such , some recent work began to explore the potential of deploying more relay nodes for enhancing the network capacity . for instance , the authors in @xcite derived an achievable dof of the two - relay mimo mrc in which two pairs of users exchange messages in a pairwise manner via two relays . later , the work in @xcite improved the dof result in @xcite by using the techniques of pairwise signal alignment and uplink - downlink symmetric design . the extension to the case of more than two user pairs was also considered in @xcite . however , the dof characterization of the multi - relay mimo mrc is still at a very initial stage . on one hand , the achievability proof of @xcite relies on some recent progresses on the solvability of linear matrix systems , and is difficult to be extended for other multi - relay mrc models , especially those with more than two relay nodes . on the other hand , for a multi - relay mrc , as the received signals at geographically separated relays can not be jointly processed , manipulating the relay signal space is far more difficult than the single - relay case . also , there is no guarantee that the optimal single - relay techniques , such as pairwise signal alignment and uplink - downlink symmetric design , are still optimal for a multi - relay mrc . in this paper , we analyze the dof of the symmetric multi - relay mimo y channel , where three user nodes , each with @xmath0 antennas , communicate with each other via @xmath1 relay nodes , each with @xmath2 antennas . compared with the mimo y channel in @xcite , a critical difference is that our new model contains an arbitrary number of relays , rather than only a single relay . following @xcite , we formulate a general dof achievability problem for the multi - relay mimo y channel based on linear processing techniques , involving the design of user precoders , relay precoders , and user post - processors . the main contributions of this paper are as follows . * in contrast to the conventional uplink - downlink symmetric design , we propose a new uplink - downlink asymmetric approach to solve the dof achievability problem . specifically , in the proposed approach , only user precoders are designed based on signal space alignment ; the user post - processors are designed for interference neutralization . furthermore , the uplink - downlink asymmetry also allows the relays to deactivate a certain number of receiving ( not transmitting ) antennas to facilitate the signal space alignment at relays . * given the designed user precoders and post - processors , the original problem boils down to a linear system on the relay precoders with certain rank constraints . to tackle the solvability of this system , we establish a novel method to solve linear matrix equations with rank constraints . we emphasize that this method can generally be used to analyze the dof of other multi - relay mrcs with various data exchange models . * based on the above new techniques , we derive an achievable dof of the symmetric multi - relay mimo y channel with an arbitrary configuration of @xmath6 . our dof result significantly outperforms the dof of the single - relay mimo y channel , implying that introducing multiple relays greatly enhances the relaying capability of the network . also , our achievable dof is considerably higher than the achievable dof with the uplink - downlink symmetric approach in @xcite . furthermore , we establish the optimality of the derived dof for @xmath5 ( with @xmath7 ) . _ notation _ : we use bold upper and lower case letters for matrices and column vectors , respectively . @xmath8 denotes the @xmath9 dimensional complex space . @xmath10 and @xmath11 represent the @xmath12 zero matrix and the @xmath13-dimensional identity matrix , respectively . for any matrix @xmath14 , @xmath15 denotes the vectorization of @xmath14 formed by stacking the columns of @xmath14 into a single column vector . moreover , @xmath16 represents the kronecker product operation . consider a symmetric multi - relay mimo y channel as shown in fig . [ fig:1 ] , where three user nodes , each equipped with @xmath0 antennas , exchange information with the help of @xmath1 relay nodes , each with @xmath2 antennas . pairwise data exchange is employed , i.e. , every user delivers two independent messages , one to each of the other two users . we assume that the information delivering is half - duplex , i.e. , nodes in the network can not transmit and receive signals simultaneously in a single frequency band . every round of data exchange consists of two phases , namely , the uplink phase and the downlink phase . the two phases have equal duration @xmath17 , where @xmath17 is an integer representing the number of symbols within each phase interval . in the uplink phase , the users transmit signals to the relays simultaneously . the received signal at each relay is written as @xmath18 where @xmath19 denotes the channel matrix from user @xmath20 to relay @xmath21 ; @xmath22 is the transmitted signal of user @xmath20 ; @xmath23 is the received signal at relay @xmath21 ; @xmath24 is the additive white gaussian noise ( awgn ) matrix at relay @xmath21 , with the elements independently drawn from @xmath25 . note that @xmath26 is the noise power at relay @xmath21 . the power constraint for user @xmath20 is @xmath27 , where @xmath28 is the maximum transmission power allowed at user @xmath20 . in the downlink phase , the relays broadcast signals to the users . the received signal at each user is represented as @xmath29 where @xmath30 is the channel matrix from relay @xmath21 to user @xmath20 ; @xmath31 is the transmitted signal from relay @xmath21 ; @xmath32 is the received signal at user @xmath20 ; @xmath33 is the awgn matrix at user @xmath20 , with elements independently drawn from @xmath34 . here , @xmath35 is the noise power at user @xmath20 . the power constraint of relay @xmath21 is given by @xmath36 , where @xmath37 is the power budget of relay @xmath21 . the elements of channel matrices @xmath38 and @xmath39 are drawn from a continuous distribution , implying that the channel matrices are of full column or row rank , whichever is smaller , with probability one . we assume that channel state information ( csi ) is globally known at every node in the model , following the convention in @xcite . moreover , for notational convenience , we interpret the user index by modulo 3 so that , e.g. , user 3 is the same as user 0 . the goal of this paper is to analyze the degrees of freedom of the symmetric multi - relay mimo y channel described above . for convenience of discussion , we assume symmetric power constraints , i.e. , @xmath40 and @xmath41 . let @xmath42 be the message from user @xmath20 to @xmath43 , where @xmath44 is the corresponding information rate , for @xmath45 and @xmath46 . note that @xmath44 is in general a function of power @xmath47 and @xmath48 , denoted by @xmath49 . an information rate @xmath49 is said to be achievable if the error probability of decoding message @xmath50 at receiver @xmath43 approaches zero as @xmath51 . an achievable dof of user @xmath20 to user @xmath43 is defined as @xmath52 intuitively , @xmath53 can be interpreted as the number of independent spatial data streams that user @xmath20 can reliably transmit to user @xmath43 . then the total achievable dof is given by @xmath54 the optimal total dof of the considered model is defined as the superemum of @xmath55 . in this paper , we assume a symmetric dof setting with @xmath56 . then the total dof can be represented as @xmath57 . we now present a dof upper bound by assuming full cooperation among the relays . under this assumption , the system model in and reduces to a single - relay mimo y channel , with @xmath58 antennas at the relay . therefore , the dof capacity of such a single - relay mimo y channel naturally serves as a dof upper bound of the model considered in and . from @xcite , this upper bound is given by [ upperbound ] @xmath59 or equivalently @xmath60 in this paper , an achievable dof of the symmetric multi - relay mimo y channel is derived by linear processing techniques . the message @xmath50 , @xmath61 , is encoded into @xmath62 , with @xmath63 independent spatial streams in @xmath17 channel uses . the transmitted signal of user @xmath20 is given by @xmath64 where @xmath65 is the linear precoding matrix for @xmath66 . for each relay , the transmitted signal is represented by @xmath67 where @xmath68 is the precoding matrix of relay @xmath21 . with , and , we can express the received signal of user @xmath20 in as @xmath69 in the above , @xmath70 consists of four signal components : the desired signal , the self interference , the other interference and the noise . since user @xmath20 perfectly knows the csi and the self message @xmath71 , the self - interference term in can be pre - cancelled before further processing . each user @xmath20 is required to decode @xmath72 spatial streams , @xmath63 from each of the other two users . to this end , there must be an interference - free subspace with dimension @xmath72 in the receiving signal space of user @xmath20 . more specifically , denote by @xmath73 a projection matrix , with @xmath74 being the projected image of @xmath70 in the subspace spanned by the row space of @xmath75 . then , to ensure the decodability of @xmath76 , @xmath77 , at user @xmath20 , we should appropriately design @xmath78 , @xmath79 and @xmath80 to satisfy two sets of requirements , as detailed below . on one hand , @xmath74 should be interference free . that is , the following interference neutralization requirement should be met : [ zeroforcing ] @xmath81 on the other hand , user @xmath20 needs to decode @xmath72 spatial streams from the projected signal @xmath82 . to ensure the decodability , the desired signal in after projection should be of rank @xmath72 . define @xmath83 \nonumber\\ \mathbf{w}_{k,1 } & = \left[\mathbf{h}_{k,0}\mathbf{u}_{0,1},\mathbf{h}_{k,2}\mathbf{u}_{2,1}\right ] \nonumber\\ \mathbf{w}_{k,2 } & = \left[\mathbf{h}_{k,0}\mathbf{u}_{0,2},\mathbf{h}_{k,1}\mathbf{u}_{1,2}\right ] \nonumber.\end{aligned}\ ] ] then the effective channel for the messages desired by user @xmath20 can be expressed as @xmath84 . to ensure the decodability of @xmath72 spatial streams at each user @xmath20 , we have the following rank requirements : @xmath85 given an antenna setup @xmath86 and a target dof @xmath63 , if there exist suitable matrices @xmath87 satisfying with probability one for random @xmath88 , then a total dof @xmath89 is achieved by the proposed linear processing scheme . thus , the key issue is to analyze the solvability of the system with respect to @xmath90 , which is the main focus of the rest of this paper . in general , to check the achievability of a certain dof @xmath63 , we need to jointly design @xmath87 to meet . this is a challenging task since the equations in are nonlinear with respect to @xmath87 . to tackle this problem , we start with a conventional approach based on the idea of uplink - downlink symmetry . uplink - downlink symmetry has been widely used in precoding design for mimo multiway relay channels @xcite . particularly , the recent work in @xcite used this symmetry to analyze the dof of the mimo multipair two - way multi - relay channel . in this subsection , we follow the idea in @xcite and describe an uplink - downlink symmetric approach to solving . then , for an arbitrary number of relays , we derive an achievable dof with respect to @xmath91 . we show that there is a significant dof gap between this result and the dof upper bound in , implying the inadequacy of the precoding design based on uplink - downlink symmetry for multiway multi - relay channels . to start with , we split each projection matrix @xmath75 equally into two parts as @xmath92\ ] ] where @xmath93 is the projection matrix for the message @xmath66 . then the interference neutralization conditions and for user @xmath94 can be expanded as [ zeroforcing_symg ] @xmath95 here , interference neutralization " refers to a special transceiver design strategy for interference cancellation such that a common source of interference from different paths cancels itself at a destination . the interference neutralization conditions for user 1 and user 2 are similar to . thus , for all the 3 users , there are 12 matrix equations for interference neutralization . moreover , the rank requirements remain the same as - . we next establish @xmath1 achievable dof points of the symmetric multi - relay mimo y channel , namely , @xmath96 and @xmath97 , @xmath98 , by following the approach in @xcite and @xcite . the first dof point is derived by deactivating @xmath99 of the @xmath1 relays . in this case , the model reduces to a single - relay mimo y channel with the antenna number of the relay equal to @xmath2 . then , the precoding design in @xcite can be applied directly . from the result of @xcite , a total dof of @xmath100 can be achieved for single - relay mimo y channel with @xmath101 . that is , @xmath102 is achievable . we now consider the dof point @xmath103 . note that the remaining dof points can be straghtforwardly obtained by deactivating @xmath104 relays , for @xmath105 . following @xcite , we apply signal alignment techniques for the design of @xmath106 to reduce the number of linearly independent constraints for interference neutralization . first consider the uplink signal alignment design . we align the signals of users @xmath20 and @xmath43 at each relay @xmath21 by designing @xmath78 to satisfy @xmath107 the alignment implies that @xmath108 and @xmath109 span a common subspace , i.e. @xmath110 note that for the single - relay case , it usually suffices to use for signal alignment @xcite . however , for the multi - relay case here , we rely on a more strict constraint for signal alignment , so as to reduce the number of linearly independent equations for interference neutralization . we now consider the downlink signal alignment . from , we see that the uplink equivalent channel matrix @xmath111 is of the same size as the transpose of downlink equivalent channel matrix @xmath112 , for @xmath113 . this structural symmetry implies that any beamforming design in the uplink phase can directly carry over to the downlink phase . specifically , we design the downlink receiving matrices @xmath114 to satisfy @xmath115 from the rank - nullity theorem , to ensure the existence of full - rank @xmath106 satisfying and , the following conditions must be met : @xmath116 with and , the four interference neutralization conditions of user @xmath94 in reduces to [ zeroforcing_sym ] @xmath117 similarly , the linearly independent interference neutralization conditions for users 1 and 2 are given by @xmath118 in the above , is a linear system of @xmath79 with @xmath119 equations and @xmath120 unknown variables . the system has a non - zero solution of @xmath121 provided @xmath122 . together with , we obtain a dof point @xmath123 . note that by deactivating @xmath104 relays , we immediately obtain the remaining the dof points @xmath124 , for @xmath125 . we now apply the antenna disablement lemma @xcite to the above @xmath1 dof points , yielding a continuous dof curve of uplink - downlink symmetric design for an arbitrary value of @xmath91 : @xmath126 where @xmath127 and the @xmath128-function is defined as @xmath129 it is worth noting that the dof in is not necessarily achievable , since we still need to check whether the solution of @xmath121 to satisfies the rank requirements - . in general , the dof in serves as an upper bound for the uplink - downlink symmetric design described above . in fact , this bound is enough to illustrate the limitation of the uplink - downlink symmetric design . specifically , we compare the dof in with the upper bound in . we see that for a sufficiently large @xmath91 ( say , @xmath130 ) , @xmath55 in is proportional to @xmath131 while @xmath55 in increases linearly with @xmath1 . this dof gap implies the inefficiency of the uplink - downlink symmetric design . we now propose a new and more efficient approach to the precoding design for the symmetric multi - relay mimo y channel , so as to achieve a higher dof than the result in . the novelty of our approach is as follows . first , instead of following the conventional uplink - downlink symmetry , our approach performs signal alignment only in the uplink phase . this means that , in the design of user precoding and post - processing , only @xmath78 are used for the purpose of aligning signals to reduce the number of linearly independent equations in - ; the post - processors @xmath80 are used for nulling interference rather than for signal alignment . this introduces additional freedom for system design . further , to exploit the advantage of this uplink - downlink asymmetric design , we allow a certain number of _ receiving _ ( but not transmitting ) antennas at each relay to be deactivated . with this treatment , signal alignment can be performed in a larger range of @xmath91 . with the above design of @xmath78 and @xmath132 , the problem reduces to the design of the relay precoders @xmath79 that satisfy the linear system in - and the rank requirements in - . we then establish a general method to address this solvability problem . we emphasis that this method can be used to analyze the solvability of any linear system with rank constraints , and so , will find applications to the dof analysis of many other multiway relay networks . the main result of this paper is presented below , with the proof given in section iv . for the symmetric multi - relay mimo y channel , any total dof @xmath133 is achievable , where @xmath134 for @xmath135 , any total dof @xmath133 is achievable , where @xmath136 by setting @xmath137 , corollary 1 follows directly from theorem 1 . for @xmath4 , the result in theorem 1 is given piecewisely in the following corollaries . for @xmath138 or @xmath139 , any total dof @xmath133 is achievable , where @xmath140 by setting @xmath141 and 3 , the result in follows from theorem 1 . for @xmath142 , any total dof achievable total dof @xmath133 is achievable , where @xmath143 by setting @xmath144 , the result of follows from theorem 1 . for @xmath4 , the optimal total dof of the symmetric multi - relay mimo y channel is given by @xmath145 the achievable dof in corollary 2 and corollary 3 coincide with the dof upper bound in for @xmath146 \cup \big[\frac{3k+\sqrt{9k^2 - 12k}}{6},\infty\big]$ ] . thus the optimal total dof is achieved in these ranges . [ fig:21 ] and fig . [ fig:22 ] illustrate the total dof in corollary 2 ( @xmath147 ) and corollary 3 ( @xmath148 ) , respectively , together with the dof in and the dof upperbound . we see that our approach significantly outperforms the conventional approach based on the uplink - downlink symmetric design . particularly , as @xmath149 , the corresponding gap increases unboundedly . for @xmath150 , the dof of our approach coincides with the dof upper bound given in , indicating that the optimal dof of the considered model is achieved . however , for the uplink - downlink symmetric design , the optimal dof is achieved only for @xmath151 $ ] , implying that adding additional relay nodes can not achieve the optimal dof in a broader range of @xmath91 . and @xmath139.,width=604 ] .,width=604 ] we also see that the achievable dof of the conventional approach fails to achieve the dof upper bound for large @xmath91 , implying that the uplink - downlink symmetric approach can not make full use of all the relay antennas . moreover , recall that the dof by the conventional approach is obtained based on @xmath1 dof points . increasing @xmath1 by one generates one new dof point while the other dof points remain unchanged . for example , in fig . [ fig:21 ] , the dof is unchanged for @xmath152 $ ] by increasing @xmath1 from 2 to 3 . in general , adding the relays to a system with @xmath1 relays can not improve the dof of the conventional approach for @xmath153 $ ] . we now study the asymptotic behavior of the dof as @xmath1 goes to infinity . it is readily seen from corollary 3 that , for any fixed @xmath2 , the total dof increases linearly in @xmath1 . what is more interesting is to understand the asymptotic dof behavior by fixing the total number of relay antennas , denoted by @xmath154 . this limiting process illustrates how the degree of cooperation among relay antennas affects the dof of the system . we have the following asymptotic bound . for the symmetric multi - relay mimo y channel , as @xmath149 with fixed @xmath154 , the asymptotic bound of @xmath55 is given by @xmath155 by dividing both sides of by @xmath154 , we obtain @xmath156 with respect to @xmath157 : @xmath158 letting @xmath159 , we obtain . in fig . [ fig:3 ] , we set the total antenna number @xmath160 as a constant , and compare the achievable dof for different values of @xmath1 . our target is to see the impact on dof when a single @xmath58-antenna relay into @xmath1 relay is split into @xmath1 relay nodes . fig . [ fig:3 ] illustrates the dof results for @xmath161 and @xmath162 , where the case of @xmath163 is also known as the dof upper bound in . we see that as @xmath1 increases , the dof monotonically decreases and approaches the asymptotic bound in corollary 5 . fig . [ fig:4 ] plots the total dof versus the number of relays for various values of @xmath91 . clearly , @xmath55 can not exceed @xmath100 , the case in which all the user antennas are fully utilized . so , it is interesting to explore the number of relay nodes required to achieve @xmath164 at a fixed value of @xmath91 . from fig . [ fig:4 ] , the minimum number of relays required to achieve @xmath164 is @xmath165 for @xmath166 , respectively . [ fig:4 ] also shows that , if a small amount of inefficiency can be tolerated in utilizing user antennas , then the required number of relays can be considerably reduced . for example , only 5 relays are required to achieve @xmath167 for @xmath168 , while that number is 27 in achieving @xmath169 . it is also worth noting that there are transitional flat dof curves for the cases of @xmath170 and 3 . this flatness corresponds to @xmath171 ( that is independent of @xmath1 ) . , for @xmath172 and @xmath162.,width=604 ] this section is dedicated to the proof of theorem 1 . for @xmath135 , theorem 1 has been proved in @xcite . thus , it suffices to only consider the case of @xmath4 . based on the linear processing techniques in section ii - b , our goal is to jointly design @xmath173 to satisfy the condition . we propose two types of signal alignment methods for @xmath174 to generate redundant equations in - , as detailed below . signal alignment i is described as follows . for each @xmath175 , @xmath176 and @xmath177 are aligned to satisfy @xmath178\mathbf{u}_{j , j+1 } & = \left [ \begin{array}{c } \mathbf{h}_{0,j+1 } \\ \vdots \\ \mathbf{h}_{k-1,j+1 } \end{array } \right]\mathbf{u}_{j+1,j } , \quad \mathrm{for}\quad\!\ ! j = 0,1,2,\end{aligned}\ ] ] or equivalently @xmath179}_{2m\times d } = \mathbf{0}\end{aligned}\ ] ] where @xmath180 \in \mathbb{c}^{kn \times 2m}.\end{aligned}\ ] ] from the rank - nullity theorem , there exist full - rank @xmath176 and @xmath177 satisfying with probability one , provided that @xmath181 with , we see that , , are identical to , , , respectively . ignoring the redundant equations , we rewrite - as [ new_zeroforcing ] @xmath182 for completeness , we repeat - as follows : @xmath183 what remains is to determine the value of @xmath184 to ensure the existence of @xmath79 and @xmath80 satisfying . based on the proposed signal alignment in , for user @xmath20 , the other interference term in is rewritten as @xmath185 with dimension of @xmath63 . to neutralize the above interference at every user , we first design @xmath186 to partially neutralize the interference in , and then design @xmath187 to null the remaining part . specifically , for @xmath188 , we split the beamforming matrix @xmath176 as @xmath189,\end{aligned}\ ] ] where @xmath190 and @xmath191 with @xmath192 defined as @xmath193 we design @xmath79 to neutralize the interference corresponding to @xmath194 for each user @xmath20 . that is , @xmath79 need to satisfy [ pneutralize ] @xmath195 with , the received signal at user @xmath20 in still contains a @xmath72-dimensional desired signal and the @xmath196-dimensional interference corresponding to @xmath197 . with @xmath192 in , we have @xmath198 , implying that the antennas at each user are already fully utilized . thus , the rank requirements - can be alternatively represented as [ prank ] @xmath199\right ) & = m\\ \label{pneutralize5 } \mathrm{rank}\left(\sum^{k-1}_{k=0}\mathbf{g}_{1,k}\mathbf{f}_k\left[\mathbf{w}_{k,1},\mathbf{h}_{k,2}\mathbf{u}^{(r)}_{2,0}\right]\right ) & = m\\ \label{pneutralize6 } \mathrm{rank}\left(\sum^{k-1}_{k=0}\mathbf{g}_{2,k}\mathbf{f}_k\left[\mathbf{w}_{k,2},\mathbf{h}_{k,0}\mathbf{u}^{(r)}_{0,1}\right]\right ) & = m.\end{aligned}\ ] ] we design @xmath75 to null the remaining interference corresponding to @xmath197 for each user . specifically , we require that @xmath200 satisfies @xmath201 for @xmath202 . note that the left null space of @xmath203 has at least @xmath204 dimensions . therefore , there always exists @xmath73 with full row rank satisfying . we now show that the above @xmath79 and @xmath80 satisfy . we start with : @xmath205 \\ \label{constructv2_1 } = & \left[\mathbf{0},\quad\!\!\ ! \mathbf{v}_0\sum_{k=0}^{k-1}\mathbf{g}_{1,k}\mathbf{f}_k\mathbf{h}_{k,1}\mathbf{u}^{(r)}_{1,2}\right ] \\ \label{constructv2_2 } = & \quad \!\!\!\!\mathbf{0}\end{aligned}\ ] ] where and follow from and , respectively . similarly , conditions and are also satisfied . for , we have @xmath206\right)\\ \label{constructv3_2 } = & \quad \!\!\!\mathrm{rank}\left(\mathbf{v}_0\sum_{k=0}^{k-1}\mathbf{g}_{0,k}\mathbf{f}_k\left[\mathbf{w}_{k,0 } , \mathbf{h}_{k,1}\mathbf{u}^{(r)}_{1,2}\right]\right)=2d,\end{aligned}\ ] ] where follows from ; is due to the fact that @xmath207 and @xmath208\in \mathbb{c}^{m\times m}$ ] is of full column rank from . similarly , and hold . consequently , to prove that a certain point @xmath184 is achievable for signal alignment i , it suffices to show that the signal alignment condition holds and there exist @xmath79 satisfying and . we have the following lemma : [ lemmadof1 ] for @xmath209 and @xmath210 , there exist @xmath78 and @xmath79 satisfying , and with probability one . see appendix i - a . note that the dof @xmath63 in lemma [ lemmadof1 ] is not necessarily an integer . a frequently used technique to achieve a rational dof is symbol extension in which an extended mimo system is constructed with multiple channel uses . the details can be found in appendix ii . the derivation in the preceding subsection is based on the signal alignment . however , this alignment imposes the requirement of that may not always be met . in this subsection , we generalize the signal alignment in by allowing the signal to be aligned in a subspace of the @xmath2-dimensional receiving signal space seen at each relay . let @xmath211 be the dimension of the subspace for signal alignment . of course , @xmath63 should be chosen to ensure @xmath212 . then , each relay deactivates @xmath213 antennas for signal reception , with the number of active antennas intact for broadcasting . that is , the number of active antennas of each relay is @xmath214 in the uplink phase and @xmath2 in the downlink phase . this implies that the proposed uplink - downlink precoding design is asymmetric in nature . a convenient way to select a subspace of dimension @xmath214 is to deactivate the last @xmath213 antennas at each relay . then the precoding matrix of each relay @xmath21 can be decomposed as @xmath215 where @xmath216 and @xmath217 \in \mathbb{c}^{n'\times n}$ ] . by the deactivation , the effective channel matrix from user @xmath20 to relay @xmath21 is @xmath218 and the processing matrix of relay @xmath21 becomes @xmath219 . as analogous to , we aim to design @xmath220 satisfying @xmath221\mathbf{u}_{j , j+1 } & = \left [ \begin{array}{c } \tilde{\mathbf{h}}_{0,j+1 } \\ \vdots \\ \tilde{\mathbf{h}}_{k-1,j+1 } \end{array } \right]\mathbf{u}_{j+1,j},\end{aligned}\ ] ] or equivalently @xmath222}_{2m\times d } = \mathbf{0}\end{aligned}\ ] ] where @xmath223 \in \mathbb{c}^{kn ' \times 2 m } , \quad \!\ ! j = 0,1,2.\end{aligned}\ ] ] since @xmath224 by the definition of @xmath214 in , there exist full - rank matrices @xmath176 and @xmath177 satisfying with probability one . similarly to signal alignment i , with and , - reduces to [ new_zeroforcing ] @xmath225 and - can be rewritten as @xmath226 where @xmath227 then , what remains is to determine @xmath228 that ensures the existence of @xmath229 and @xmath80 satisfying . we now consider the construction of @xmath230 to satisfy . similarly to signal alignment i , @xmath229 are designed to neutralize the interference corresponding to @xmath231 , while @xmath80 are designed to null the remaining interference corresponding to @xmath232 . more specifically , as analogous to and , we require @xmath229 to satisfy [ pneutralize ] @xmath233 @xmath234\right ) & = m\\ \label{pneutralize5 ' } \mathrm{rank}\left(\sum^{k-1}_{k=0}\mathbf{g}_{1,k}\tilde{\mathbf{f}}_k\left[\tilde{\mathbf{w}}_{k,1},\tilde{\mathbf{h}}_{k,2}\mathbf{u}^{(r)}_{2,0}\right]\right ) & = m\\ \label{pneutralize6 ' } \mathrm{rank}\left(\sum^{k-1}_{k=0}\mathbf{g}_{2,k}\tilde{\mathbf{f}}_k\left[\tilde{\mathbf{w}}_{k,2},\tilde{\mathbf{h}}_{k,0}\mathbf{u}^{(r)}_{0,1}\right]\right ) & = m.\end{aligned}\ ] ] moreover , @xmath200 is of full row rank and satisfies @xmath235 for @xmath202 . it can be readily shown that , with @xmath192 in and @xmath214 in , there always exist @xmath80 satisfying . thus , to prove the achievability of a certain dof point @xmath228 , it suffices to show that there exist @xmath229 satisfying . we have the following result . [ lemmadof2 ] for @xmath236 and @xmath237 , there exist @xmath78 and @xmath229 satisfying and with probability one . see appendix i - b . from lemmas [ lemmadof1 ] and [ lemmadof2 ] , we see that both signal alignment i and ii can not be realized when @xmath91 is relatively small . we next focus on relatively small values of @xmath91 and show that @xmath169 ( or equivalently , @xmath238 ) is achievable for @xmath239 . to start with , we set [ uv ] @xmath240\\ \mathbf{u}_{j , j-1 } & = \left [ \begin{array}{c } \mathbf{0}_{d } \\ \mathbf{i}_{d } \end{array } \right ] , \quad\!\!\mathrm{for } \quad\!\ ! j = 0,1,2.\end{aligned}\ ] ] then , - reduce to a linear system of @xmath79 with @xmath120 unknown variables and @xmath241 equations . the system allows non - zero solutions of @xmath79 provided @xmath242 , i.e. , @xmath243 . to prove the achievability of @xmath244 , it suffices to show that there exist @xmath79 satisfying all the conditions in . we have the following result . [ lemmadof3 ] for @xmath245 and @xmath244 , there exist @xmath121 , together with @xmath246 in , satisfying with probability one . see appendix i - c . in the preceding subsections , we have established an achievable dof for @xmath247 $ ] . we now follow the antenna disablement approach @xcite to establish achievable dofs for other ranges of @xmath91 . specifically , for @xmath248 , we disable @xmath249 antennas at each relay . then , from lemma [ lemmadof2 ] , we see that any dof @xmath250 can be achieved , where @xmath251 is the number of active antennas at each relay . the only issue is that @xmath252 may be not an integer . this issue can be solved by the technique of symbol extension described in appendix ii . similarly , with symbol extension and antenna disablement , @xmath253 is achievable for @xmath254 . combining lemmas [ lemmadof1 ] and [ lemmadof2 ] , we conclude that any dof @xmath63 satisfying @xmath255 is achievable , where @xmath256 with @xmath89 , we immediately obtain . this completes the proof of theorem 1 . in this paper , we developed a new formalism to analyze the achievable dof of the symmetric multi - relay mimo y channel . specifically , we adopted the idea of uplink - downlink asymmetric design and proposed a new approach to tackle the solvability problem of linear systems with rank constraints . in the proposed design , we also incorporated the techniques of signal alignment , antenna disablement , and symbol extension . an achievable dof for an arbitrary configuration of @xmath257 was derived . we believe that the formalism developed in this paper can be used to analyze the dof of other mimo mrcs with general user and relay settings , as well as with various data exchange models . this will be an interesting research topic worthy of future endeavor . we first show that the signal alignment can be performed for @xmath258 and @xmath210 . to see this , we have @xmath259 , implying @xmath260 further , as @xmath210 , we obtain @xmath261 therefore , is met , and so there exist @xmath78 satisfying with probability one . we next show that @xmath262 holds for @xmath63 chosen sufficiently close to @xmath263 . for @xmath264 $ ] , @xmath265 and @xmath266 , we obtain @xmath267 where the last step holds by noting @xmath268 for @xmath269 . on the other hand , as @xmath270 we can always choose @xmath63 close to @xmath271 to ensure @xmath272 . we henceforth always assume that @xmath63 is appropriately chosen such that @xmath273 . what remains is to show that given @xmath78 satisfying , there exist @xmath79 satisfying and . to start with , we rewrite using the kronecker product as @xmath274 where @xmath275\in \mathbb{c}^{3d'm\times kn^2}\ ] ] @xmath276 \in \mathbb{c}^{kn^2 \times 1}.\ ] ] then , the problem can be recast as finding a solution of satisfying . to proceed , we present a useful lemma below . [ provelemma1 ] for @xmath277 and @xmath258 , assume that there exists a certain choice of @xmath278 satisfying a. @xmath279 for @xmath280 where @xmath281 is defined in ; b. @xmath282 ; c. , and holds . then for random @xmath283 together with full - column - rank @xmath78 satisfying , there exist @xmath121 satisfying and with probability one . we show that for random @xmath283 and full - column - rank @xmath78 satisfying , @xmath284 in is of full row rank with probability one . to see this , we first note that @xmath284 is a wide matrix since @xmath285 second , from the channel randomness , we have @xmath286 and @xmath287 . then @xmath288 . each @xmath289 block row of @xmath284 consists of @xmath1 submatrices in the form of @xmath290 . therefore , from the channel randomness , the rank of each @xmath291 block row is given by @xmath292 , where the last step utilizes the fact that @xmath293 . further , by noting that the three block rows of @xmath284 are statistically independent of each other , we conclude that @xmath284 is of full row rank ( @xmath294 ) with probability one . using gaussian elimination , the general solution of can be written as @xmath295 where @xmath296 are free variables and @xmath297 is the total number of free variables @xmath296 . note that @xmath298 , for @xmath299 , span the right null space of @xmath284 . clearly , gaussian elimination involves the following arithmetic operations : addition , subtraction , multiplication and division . this implies that each element of @xmath300 is a rational function of @xmath301 and @xmath302 . thus , with proper scaling , we can always express each @xmath303 as a finite - degree polynomial with respect to @xmath301 and @xmath302 . following similar arguments , since @xmath304 are design to satisfy @xmath305 and each @xmath306 is of full row rank with probability one , matrices @xmath176 and @xmath177 can be represented as polynomials with respect to @xmath307 , where @xmath308 are free variables with @xmath309 . now consider the following determinant product : @xmath310\right)\!\!.\end{aligned}\ ] ] note that @xmath311 if and only if the @xmath312 matrix @xmath313 $ ] is of full rank for @xmath175 , or equivalently , holds . therefore , to prove lemma [ provelemma1 ] , it suffices to show that for random @xmath314 , there exist @xmath79 given by satisfying @xmath315 with probability one . to this end , we first note that @xmath316 is a finite - degree polynomial with respect to @xmath317 , i.e. , @xmath316 can be represented as @xmath318 where @xmath319 is a polynomial of entries of @xmath320 and @xmath301 , @xmath321 is a monomial of @xmath296 and @xmath322 , and @xmath323 is a finite integer . by assumption , there exist particular @xmath278 satisfying ( a)-(c ) , implying that there exist particular @xmath324 such that @xmath325 . thus , @xmath316 is a non - trivial polynomial , i.e. , at least one @xmath319 is not always zero . therefore , for randomly generated @xmath283 , @xmath316 is a non - trivial polynomial of @xmath326 with probability one . in other words , for random @xmath283 , we can find @xmath327 satisfying @xmath328 with probability one . therefore , there exist @xmath79 satisfying and with high probability . we emphasize that the full - rank conditions ( a ) and ( b ) in lemma [ provelemma1 ] are indispensable . too see this , suppose that certain @xmath329 satisfy ( a ) and ( c ) , but ( b ) does not hold . then @xmath284 is rank - deficient , and so is no longer a general solution of ( since the latter now involves more free variables ) . therefore , the subsequent arguments in the proof of lemma [ provelemma1 ] does not hold any more . to prove lemma [ lemmadof1 ] , what remains is to show the existence of @xmath330 satisfying the conditions ( a)-(c ) in lemma [ provelemma1 ] . to this end , we set @xmath331 for @xmath332 , and choose @xmath333 to be random matrices with the elements independently drawn from a continuous distribution . then , we choose full - rank matrices @xmath334 to satisfy @xmath335}_{m \times kn } \underbrace{\left[\begin{array}{c}\mathbf{h}_{0,j+1}\mathbf{u}^{(l)}_{j+1,j+2}\\ \mathbf{h}_{1,j+1}\mathbf{u}^{(l)}_{j+1,j+2 } \\ \vdots \\ \mathbf{h}_{k-1,j+1}\mathbf{u}^{(l)}_{j+1,j+2 } \end{array}\!\!\right]}_{kn\times d ' } = \mathbf{0}.\end{aligned}\ ] ] it can be verified that @xmath336 for @xmath265 and @xmath337 $ ] . then @xmath338 implying that the null space of the @xmath339 matrix in has at least @xmath192 dimensions . thus full - rank @xmath334 exist with probability one . for given @xmath334 and @xmath340 , we can determine the values of @xmath320 and @xmath341 ( not uniquely ) . with @xmath305 , @xmath342 are also determined . we now show that the above constructed @xmath343 satisfy ( a)-(c ) in lemma [ provelemma1 ] with probability one . from the randomness of @xmath301 and the construction of @xmath344 , it is easy to verify that ( a ) and ( b ) hold with probability one . moreover , since @xmath331 , condition is exactly . thus holds . to check , we take as an example . note that @xmath345 and @xmath346 are determined by @xmath347 and @xmath348 in , respectively . thus , @xmath349 are not a function of @xmath350 . by construction , @xmath349 and @xmath351 are independent . therefore , @xmath352\right ) = \min\{m , n\ } = n\ ] ] with probability one , where the first equality follows by noting @xmath353 , @xmath354 and @xmath355 \in \mathbb{c}^{n \times m}$ ] , and the second equality by noting @xmath356 for @xmath337 $ ] . moreover , as @xmath357 , @xmath358\right ) = \mathrm{min}\{kn , m\ } = m\ ] ] with probability one , and so is met . following similar arguments , and are also satisfied with probability one . to summarize , the above constructed @xmath359 satisfy conditions ( a)-(c ) in lemma [ provelemma1 ] with probability one , which concludes the proof of lemma [ lemmadof1 ] . the proof of lemma [ lemmadof2 ] closely follows that of lemma [ lemmadof1 ] . we briefly present the proof in the following . we first show that the signal alignment in can be performed for @xmath360 and @xmath237 . from the discussion in section iv - c , it suffices to show that for @xmath63 close to @xmath361 , @xmath214 defined in satisfies @xmath212 . too see this , note that @xmath362 implying that @xmath363 for any @xmath237 . further , since @xmath364 , we have @xmath365 therefore , we can always choose @xmath63 close to @xmath361 to ensure @xmath366 . moreover , following the proof of lemma [ lemmadof1 ] , we can readily verify that for @xmath360 , if @xmath63 is chosen sufficiently close to @xmath361 , we have @xmath367 . we next show that with @xmath368 satisfying , there exist @xmath229 satisfying with probability one . as analogous to lemma [ provelemma1 ] , we have the following result . note that the proof of lemma [ provelemma2 ] follows that of lemma [ provelemma1 ] step by step , and is omitted for brevity . [ provelemma2 ] for @xmath237 and @xmath360 , assume that there exists a certain choice of @xmath369 satisfying a. @xmath370 for @xmath280 where @xmath371 is defined in ; b. @xmath372 , where @xmath373\in \mathbb{c}^{3d'm\times kn'n};\end{aligned}\ ] ] c. and holds . then for random @xmath374 together with full column - rank @xmath78 satisfying , there exist @xmath375 satisfying with probability one . what remains is to show the existence of @xmath376 satisfying the conditions ( a)-(c ) in lemma [ provelemma2 ] . to this end , we set @xmath377 \in \mathbb{c}^{n \times n'}\end{aligned}\ ] ] for @xmath378 , and choose @xmath379 to be random matrices with the elements independently drawn from a continuous distribution . then , we choose full - rank matrices @xmath380 to satisfy @xmath381}_{m \times kn ' } \underbrace{\left[\begin{array}{c}\tilde{\mathbf{h}}_{0,j+1}\mathbf{u}^{(l)}_{j+1,j+2}\\ \tilde{\mathbf{h}}_{1,j+1}\mathbf{u}^{(l)}_{j+1,j+2 } \\ \vdots \\ \tilde{\mathbf{h}}_{k-1,j+1}\mathbf{u}^{(l)}_{j+1,j+2 } \end{array}\!\!\right]}_{kn'\times d ' } = \mathbf{0}.\end{aligned}\ ] ] for @xmath360 and @xmath237 , we have @xmath382 implying that the null space of the @xmath383 matrix in has at least @xmath192 dimensions . thus full - rank @xmath380 exist with probability one . for given @xmath380 and @xmath384 , we determine the values of @xmath385 and @xmath341 ( not uniquely ) . with @xmath386 , @xmath342 are also determined . we show that the above constructed @xmath387 satisfy ( a)-(c ) in lemma [ provelemma2 ] with probability one . note that by construction of @xmath229 , @xmath388 is simply the first @xmath214 columns of @xmath389 , for @xmath175 and @xmath390 . from the randomness of @xmath391 and the construction of @xmath392 , we have ( a ) , ( b ) , and - hold with probability one . we next take as an example to check - . since @xmath393 and @xmath394 are determined by @xmath395 and @xmath396 in , respectively , @xmath397 are not a function of @xmath398 . moreover , @xmath397 and @xmath399 are independent of each other by construction . therefore , @xmath400\right ) = \min\{m , n'\ } = n'.\ ] ] as @xmath401 , we obtain @xmath402\right ) = \mathrm{min}\{kn',m\ } = m\ ] ] with probability one . thus holds , and so do and by following similar arguments . this concludes the proof of lemma [ lemmadof2 ] . following the arguments in proving lemma [ lemmadof1 ] in appendix i - a , the general solution of - with respect to @xmath79 can be represented as finite - degree polynomials of the elements of @xmath314 . then as analogous to lemma [ provelemma1 ] , we have the following result for the solvability of . we omit the proof for brevity . [ provelemma3 ] for @xmath403 and @xmath244 , assume that there exists a certain choice of @xmath404 satisfying and @xmath405 , where @xmath406\in \mathbb{c}^{3m^2\times kn^2}.\end{aligned}\ ] ] then for random @xmath283 together with @xmath407 in , there exist @xmath121 satisfying with probability one . what remains is to construct @xmath408 satisfying and @xmath405 . we basically follow the construction in the proof of lemma [ lemmadof1 ] . let @xmath409 , for @xmath410 , and @xmath301 be random matrices with elements independently drawn from a continuous distribution . then we design @xmath320 with full column rank satisfying @xmath411}_{m \times kn } \underbrace{\left[\begin{array}{cc } \mathbf{h}_{1,j+1}\mathbf{u}_{j+1,j+2 } & \mathbf{h}_{1,j+2}\mathbf{u}_{j+2,j+1}\\ \mathbf{h}_{2,j+1}\mathbf{u}_{j+1,j+2 } & \mathbf{h}_{2,j+2}\mathbf{u}_{j+2,j+1 } \\ \vdots & \vdots\\ \mathbf{h}_{k-1,j+1}\mathbf{u}_{j+1,j+2 } & \mathbf{h}_{k-1,j+2}\mathbf{u}_{j+2,j+1 } \end{array}\!\!\right]}_{kn\times m } = \mathbf{0 } , \quad\!\ ! j = 0,1,2.\end{aligned}\ ] ] from the rank - nullity theorem , there exist full - rank @xmath412 satisfying with probability one , provided @xmath413 , ( which is true for @xmath414 ) . note that from , @xmath415 and @xmath415 are the left @xmath416 columns and the right @xmath416 columns of @xmath417 for every @xmath175 and @xmath418 . thus , we can combine them together to form @xmath417 with full column rank . it can be readily verified that , together with @xmath419 in , the above constructed @xmath408 satisfy - . moreover , since @xmath389 is randomly generated , - and @xmath405 are met with probability one , which concludes the proof of lemma [ lemmadof3 ] . with the techniques of symbol extension and antenna disablement , we show that all the results in section iv hold for any rational dof @xmath63 . we focus on signal alignment i in section iv - b . the treatments for other signal alignment methods are similar . let @xmath420 satisfying @xmath421 @xmath422 . further , assume that both @xmath423 and @xmath63 are rational numbers satisfying @xmath424 . let @xmath425 be a positive integer that both @xmath426 and @xmath427 are integers . then , we extend the channel by @xmath425 times and disable @xmath428 antennas at each user end . with some abuse of notation , we represent the extended channel of each link in a block - diagonal form as where @xmath431 is the channel from user @xmath20 to relay @xmath21 with in the @xmath432-th channel use with @xmath433 and @xmath434 is the channel from relay @xmath21 to user @xmath20 in the @xmath432-th channel use with @xmath435 our goal is to show that any dof @xmath427 with @xmath436 is achievable for the extended channel , implying that an average dof @xmath63 is achieved per channel use . to this end , we need to design @xmath437 and @xmath438 , for @xmath410 and @xmath202 , to satisfy condition ( with the rank @xmath72 in - replaced by @xmath439 ) . this can be done by closely following the arguments in appendix i - a , with the only main difference being that @xmath320 and @xmath301 take the block diagonal forms in . to proceed , we rephrase lemma [ provelemma1 ] for the extended channel as below . what remains is to construct @xmath446 satisfying ( a)-(d ) in lemma [ symbollemma ] . similarly to appendix i , we set @xmath447 for @xmath378 . further , we set @xmath301 to be random block - diagonal matrices of the form and @xmath340 to be random matrices with the elements independently drawn from a continuous distribution . then , we construct full - rank matrices @xmath334 satisfying @xmath448}_{lm^ * \times lkn } \underbrace{\left[\begin{array}{c}\mathbf{h}_{0,j+1}\mathbf{u}^{(l)}_{j+1,j+2}\\ \mathbf{h}_{1,j+1}\mathbf{u}^{(l)}_{j+1,j+2 } \\ \vdots \\ \mathbf{h}_{k-1,j+1}\mathbf{u}^{(l)}_{j+1,j+2 } \end{array}\!\!\right]}_{lkn\times ld ' } = \mathbf{0}.\end{aligned}\ ] ] by and the arguments therein , we readily see that full - rank @xmath334 exist with probability one . with the obtained @xmath449 , we can determine the values of @xmath320 and @xmath341 , with @xmath320 of the form . with @xmath305 , @xmath342 are also determined . from the discussions in appendix i - a , the above constructed @xmath446 satisfy the conditions ( a)-(d ) in lemma [ symbollemma ] . so far , we have shown that with symbol extension and antenna disablement , any rational dof @xmath424 is achievable for @xmath440 , @xmath422 . following similar treatments , we can show that the dof results in subsections iv - d and iv - e hold for rational values of @xmath63 . we omit the details for brevity . 100 m. gastpar and m. vetterlj , on the capacity of wireless networks : the relay case , " in _ proc . 2002 ieee infocom _ , 1577 - 1586 . editors : y. hua , d. w. bliss , s. gazor , y. rong , and y. sung , theories and methods for advanced wireless relays : issue i , " _ ieee j. sel . areas commun . _ , vol . 30 , no . 8 , sept . 2012 . s. zhang , s. liew , and p. p. lam , hot topic : physical - layer network coding , " in _ proc . acm mobicom06 _ , los angeles , usa , sept . 2006 . d. gndz , a. yener , a. goldsmith , and h. v. poor , the multiway relay channel , " _ ieee trans . inf . theory _ , 51 - 63 , jan . 2013 . k. liu , m. tao , z. xiang , and x. long , generalized signal alignment : on the achievable dof for multi - user mimo two - way relay channels , " _ ieee trans . 61 , no . 6 , pp . 3365 - 3386 , june 2015 . z. xiang , m. tao , j. mo , and x. wang , degrees of freedom for mimo two - way x relay channel , " _ ieee trans . on signal process . _ , vol . 61 , no . 7 , pp . 1711 - 1720 , july 2013 . b. yuan , x. liao , f. gao , and x. luo , achievable degrees of freedom of the four - user mimo y channel , " _ ieee commun . 6 - 9 , jan . 2014 . r. wang , x. yuan , and m. tao , degrees of freedom of mimo multiway relay channel with clustered pairwise exchange , " _ ieee j. select . area . 337 - 351 , dec . 2014 . y. tian and a. yener , degrees of freedom for the mimo multi - way relay channel , " _ ieee trans . inf . theory _ , 5 , pp . 2495 - 2511 , may 2014 . x. yuan , mimo multiway relaying with clustered full data exchange : signal space alignment and degrees of freedom , " _ ieee trans . wireless commun . _ , vol . 13 , no . 12 , pp . 6795 - 6808 , may 2015 x. yuan , t. yang , and i. b. collings , multiple - input multiple - output two - way relaying : a space - division approach , " _ ieee trans . inf . theory _ 6421 - 6440 , oct . 2013 . c. wang , beyond one - way communication : degrees of freedom of multiway relay mimo interference networks , " submitted to _ ieee trans . inf . theory _ , jan . 2014 , [ online ] . available : http://arxiv.org/abs/1401.5582 . a. chaaban , a. sezgin , and a. s. avestimehr , approximate sum - capacity of the y - channel , " _ ieee trans . inf . theory _ , 9 , pp . 5723 - 5740 , sept . n. lee and r. w. heath , degrees of freedom for the two - cell two - hop mimo interference channel : interference - free relay transmission and spectrally efficient relaying protocol , " _ ieee trans . inf . theory _ 2882 - 2896 , may 2013 . s. chen and r. s. cheng , on the achievable degrees of freedom of a @xmath1-user mimo interference channel with a mimo relay , " _ ieee trans . wireless commun . 4118 - 4128 , aug . r. wang and x. yuan , mimo multiway relaying with pairwise data exchange : a degrees of freedom perspective , " _ ieee trans . on signal process . , 5294 - 5307 , july 2014 . n. lee , j. b. lim , and j. chun , degrees of freedom of the mimo y channel : signal space alignment for network coding , _ ieee trans . inf . 56 , no . 7 , pp . 3332 - 3342 , july 2010 . a. chaaban , k. ochs , and a. sezgin , the degrees of freedom of the mimo y - channel , " in _ proc . ieee isit _ , istanbul , turkey , july 2013 . a. a. zewail , m. nafie , y. mohasseb , and h. el - gamal , achievable degrees of freedom region of mimo relay networks using detour schemes , " in _ proc . of ieee int . ( icc ) _ , sydney , australia , 2014 . v. r. cadambe and s. a. jafar , interference alignment and degrees of freedom of the @xmath1-user interference channel , " _ ieee trans . inf . theory _ , 54 , no . 8 , pp . 3425 - 3441 , aug . k. lee , n. lee , and i. lee , achievable degrees of freedom on mimo two - way relay interference channels , " _ ieee trans . wireless commun . 4 , pp . 1472 - 1480 , april 2013 . r. wang , x. yuan , and r. w. yeung , distributed mimo multiway relaying : joint signal alignment and interference neutralization , " in _ proc . ( icc ) _ , london , england , june 2015 .

exploiting a 3 + 1 analysis of the mars - simon tensor , conditions on a vacuum initial data set ensuring that its development is isometric to a subset of the kerr spacetime are found . these conditions are expressed in terms of the vanishing of a positive scalar function defined on the initial data hypersurface . applications of this result are discussed .

0805.4505

the kerr spacetime is one of the most important exact solutions to the einstein vacuum equations . its relevance stems from the uniqueness theorems for black holes which state that under rather general conditions the kerr spacetime is the only asymptotically flat , stationary , vacuum black hole solution see e.g. the introduction of @xcite for a critical review of the issue of black hole uniqueness and the involved assumptions . although , the kerr spacetime is very well understood from a spacetime perspective , the same can not be said if one adopts a 3 + 1 point of view which would be the case if one tries to numerically calculate the spacetime from some cauchy initial data . as soon as one moves away from a 3 + 1 gauge which is adapted to the stationary and axial symmetries which can occur in some applications , in particular in numerical ones an analysis of the kerr spacetime and initial data sets thereof becomes very complicated . the explicit nature of the kerr solution makes it tempting to perform detailed calculations in order to , say , verify a particular property of the spacetime . this approach usually leads to very lengthy expressions which can be very hard to analyse . in a sense one could say that exact solutions contain too much information . in this case it can be more convenient to adopt more abstract approaches and , accordingly , it may prove useful to have at hand a characterization of kerr initial data . the question of providing an invariant characterization of initial data sets for the schwarzschild spacetime has been addressed in @xcite . in particular , the analysis of @xcite provides an algorithmic characterization of schwarzschild data . that is , a procedure is provided to verify whether a given initial data set for the einstein field equations will render a development which is isometric to a portion of the schwarzschild spacetime . one of the most important algebraic properties of the kerr spacetime is that its weyl tensor is of petrov type d. the close relation between vacuum spacetimes with a weyl tensor of petrov type d and killing spinors has been exploited in @xcite to provide a characterization of initial data sets whose developments will be of petrov type d. this characterization relies on one being able to decide whether a set of overdetermined partial differential equations has solutions for a given initial data set . accordingly , such a characterization is not algorithmic . although not explicitly stated in @xcite , from that analysis it should be possible to obtain a characterization of kerr initial data by adding some global conditions . the characterization of initial data sets discussed in @xcite has followed the general strategy of starting from a given tensorial ( respectively , spinorial ) _ spacetime _ characterization of the class of spacetimes under consideration . necessary conditions on the initial data set are obtained by performing a 3 + 1 decomposition of the spacetime characterization . given a set of necessary conditions on the initial data , it is then natural to address the question of sufficiency . this is , usually , the most challenging part of the procedure as one has to discuss the evolution of complicated tensorial objects . the idea behind this is to show that if the necessary conditions are satisfied on some open subset of the initial hypersurface , then one can possibly , under some additional assumptions recover the spacetime characterization on the development of the open subset on the initial hypersurface from which one started . in this article a particular characterization of kerr initial data is addressed . our starting point is a certain spacetime characterization of the kerr spacetime given in @xcite . this characterization was developed with the aim of providing an alternative way of proving the uniqueness of kerr spacetime among the class of stationary , asymptotically flat black holes . this expectation has been recently fulfilled in @xcite , where a proof of the uniqueness of kerr which does not make assumptions on the analyticity of the metric has been given . at the heart of the spacetime characterization given in @xcite cfr . theorem [ kerr - characterization] is a certain tensor , _ the mars - simon tensor _ , whose construction requires the existence of a timelike killing vector . the mars - simon tensor is a spacetime version of the _ simon tensor _ , a tensor , defined in the quotient manifold of a stationary spacetime , which characterizes the kerr spacetime see @xcite . following the general strategy for the construction of characterizations of initial data sets out of spacetime characterizations , necessary conditions for kerr initial data are deduced from a 3 + 1 splitting of the mars - simon tensor . accordingly , one assumes that the spacetime one is working with has a timelike killing vector . this requirement can be encoded in the initial data by requiring that the data has a suitable _ killing initial data _ ( kid ) . the mars - simon tensor has the same symmetries as the weyl tensor , and hence its 3 + 1 decomposition can be given in terms of its _ electric _ and _ magnetic _ parts . in order to discuss the propagation of the mars - simon tensor we make use of a framework for discussing the propagation tensorial fields using _ superenergy - type _ tensors this framework has been discussed in e.g. @xcite . it should be pointed out that the characterization discussed in this article is not algorithmic . that is , like the one for type d initial data discussed in @xcite it depends on being able to decide whether a certain overdetermined system of partial differential equations admits a solution . the article is structured as follows : in section [ preliminaries ] our main conventions are fixed and relevant aspects of the 3 + 1 formalism are discussed . section [ weyl : c ] discusses the properties and causal propagation of weyl candidates i.e . tensors with the same symmetries of the weyl tensor . section [ section : mstensor ] is concerned with the properties of the mars - simon tensor . section [ cp - ms ] discusses the causal propagation of the mars - simon tensor . section [ section : application ] applies the previous discussion to the construction of a characterization of kerr initial data . our main result is provided in theorem [ theorem : kerr_data ] . in section [ schwarzschild ] we particularize to the case of the schwarzschild spacetime , where , with the aim of the results of @xcite it is possible to obtain an initial data characterization which is algorithmic . some concluding remarks are provided in section [ conclusions ] . finally , some technical details , too lengthy to be presented in the main text are presented in the appendix a. all the tensor computations of this paper have been performed with the software _ _ xact _ is a suite of mathematica packages which has among its many features the capability to efficiently canonicalize tensor expressions by the use of powerful algorithms based on permutation group theory @xcite . currently no other software , either free or commercial , is capable to handle the tensor computations needed in this paper . let @xmath0 denote a smooth orientable spacetime . the following conventions will be used : plain greek letters @xmath1 denote abstract indices and boldface latin characters @xmath2 will be used for component indices . the signature of the metric tensor @xmath3 will be taken to be @xmath4 , while @xmath5 , @xmath6 and @xmath7 denote , respectively the riemann , ricci and weyl tensors of @xmath3 . the tensor @xmath8 is the volume element which is used to define the hodge dual of any antisymmetric tensor denoted by attaching a star @xmath9 to the tensor symbol . sometimes we will need to work with complex tensors in which case the complex conjugation of a tensor is denoted by an overbar . the operator @xmath10 symbolizes the lie derivative with respect to the vector field @xmath11 . let @xmath12 be a unit timelike vector , @xmath13 defined on @xmath14 . then any tensor or tensorial expression can be decomposed with respect to @xmath12 and the way to achieve it is the essence of the orthogonal splitting ( also known as 3 + 1 formalism ) which is described in many places of the literature see e.g. @xcite ) . we review the parts of this formalism needed in this work . the spatial metric is defined by @xmath15 and it has the algebraic properties @xmath16 , @xmath17 . we shall call a covariant tensor @xmath18 _ spatial _ with respect to @xmath19 if it is invariant under @xmath20 i.e. if @xmath21 with the obvious generalization for any mixed tensor . this property implies that the inner contraction of @xmath12 with @xmath22 ( taken on any index ) vanishes . the orthogonal splitting of a tensor expression consists in writing it as a sum of terms which are tensor products of the unit normal and spatial tensors of lesser degree or the same degree in which case the unit normal is absent . in order to find the orthogonal splitting of expressions containing covariant derivatives we need to introduce the _ spatial derivative _ @xmath23 which is an operator whose action on any tensor field @xmath24 , @xmath25 is given by @xmath26 from equation ( [ spatialcd ] ) it is clear that @xmath27 is spatial . the results just described hold for an arbitrary unit timelike vector @xmath12 but in our framework we only need to consider integrable timelike vectors which are characterized by the condition @xmath28}=0 $ ] ( frobenius condition ) . in this case there exists a foliation of @xmath14 such that the vector field @xmath12 is orthogonal to the leaves of the foliation . we shall denote by @xmath29 , @xmath30 , the family of leaves of this foliation and @xmath31 is called the _ initial data hypersurface _ it is assumed that @xmath32 . the tensor @xmath19 plays the role of the _ first fundamental form _ for any of the leaves while the symmetric tensor @xmath33 defined by @xmath34 can be identified with the _ second fundamental form_. combining the previous definition with the frobenius condition we easily derive @xmath35 where @xmath36 is the acceleration of @xmath12 . by using these quantities one is in principle able to work out the orthogonal splitting of any tensorial expression . we supply below the explicit results of calculations which will be needed repeatedly in the sequel . * orthogonal splitting of the volume element : @xmath37 here @xmath38 is the _ spatial volume element _ which is a fully antisymmetric spatial tensor . * orthogonal splitting of an antisymmetric tensor . if @xmath39 is an arbitrary antisymmetric tensor then its orthogonal splitting takes the form @xmath40}-q_\rho\varepsilon_{\mu\nu}^{\phantom{\mu\nu}\rho},\ ] ] where @xmath41 , @xmath42 are clearly spatial tensors . it is straightforward from the previous equation to find an analogous formula for the orthogonal splitting of @xmath43 . if we work with the _ self - dual _ part of @xmath39 , denoted by @xmath44 and defined by @xmath45 then its orthogonal splitting takes the form @xmath46},\ \mathcal{p}_\mu\equiv p_\mu+\mbox{i}\ q_\mu . \label{decompose_h}\ ] ] * orthogonal splitting of the weyl tensor : @xmath47\nu}- l_{\nu[\lambda}e_{\rho]\mu}- n_{[\lambda}b_{\rho]\tau}\epsilon^\tau_{\phantom{\tau}\mu\nu } -n_{[\mu}b_{\nu]\tau}\epsilon^\tau_{\phantom{\tau}\lambda\rho } \right),\label{weyl-1}\ ] ] where @xmath48 denote the weyl tensor _ electric _ and _ magnetic _ parts respectively , and @xmath49 . the tensors @xmath50 and @xmath51 are symmetric , traceless , and spatial . equation ( [ weyl-1 ] ) also holds for any rank-4 tensor possessing the same algebraic properties as the weyl tensor . such tensors will be designated collectively as _ weyl candidates _ see section [ weyl : c ] . we can also choose to work with the self - dual weyl tensor @xmath52 which is given by @xmath53 using ( [ weyl-1 ] ) and the analogous formula for the orthogonal splitting of @xmath54 one can obtain an expression for the orthogonal splitting of @xmath52 @xmath55\mu}+ l_{\mu[\rho}\mathcal{e}_{\lambda]\nu}+\mbox{i}\ \varepsilon_{\rho\lambda\alpha}n_{[\mu}\mathcal{e}_{\nu]}^{\phantom{\nu]}\alpha}+ \mbox{i}\ \varepsilon_{\mu\nu\alpha}n_{[\rho}\mathcal{e}_{\lambda]}^{\phantom{\nu]}\alpha } ) , \label{decompose - weyl}\ ] ] where @xmath56 again a similar formula to ( [ decompose - weyl ] ) holds for any weyl candidate . * let @xmath57 be any spatial vector . then the orthogonal splitting of @xmath58 is @xmath59 note that since @xmath60 is spatial then @xmath61 is also spatial . we briefly review the standard formulation of the cauchy problem for the vacuum einstein equations . in this formulation one considers a 3-dimensional connected riemannian manifold @xmath62 we use small plain latin letters @xmath63 for the abstract indices of tensors on this manifold and an isometric embedding @xmath64 . the map @xmath65 is an isometric embedding if @xmath66 where @xmath67 realises the pullback , @xmath68 , of tensor fields from @xmath14 to @xmath69 . the metric @xmath70 defines a unique affine connection @xmath71 ( levi - civita connection ) by means of the standard condition @xmath72 the riemann tensor of @xmath71 is @xmath73 and from it we define its ricci tensor by @xmath74 and its scalar curvature @xmath75 in @xmath69 indices are raised and lowered with @xmath70 and its inverse @xmath76 . let @xmath62 be a riemannian manifold and suppose that there exists a symmetric tensor field @xmath77 on it which verifies the conditions ( vacuum constraints ) @xmath78 where @xmath79 . provided that @xmath70 and @xmath77 are suitably smooth there exists an isometric embedding @xmath65 of @xmath69 into a globally hyperbolic , vacuum solution @xmath80 of the einstein field equations . the set @xmath81 is then called a vacuum initial data set and the spacetime @xmath80 is the data development . furthermore the spacelike hypersurface @xmath82 is a cauchy hypersurface in @xmath83 . [ vacuum - data ] a statement of this theorem containing precise regularity conditions on @xmath70 , @xmath77 is formulated in reference @xcite . under the conditions of the theorem [ vacuum - data ] , we may construct a foliation of @xmath83 with @xmath12 as the timelike unit vector which is orthogonal to the leaves and , in the above notation , set @xmath84 . it is then clear that @xmath85 . other key properties are @xmath86 using these properties and the ricci identity for @xmath12 one can show that @xmath87 another property which is needed later on is the following one : for any spatial tensor @xmath88 , @xmath89 with respect to @xmath12 we have the property @xmath90 fundamental for our discussion will be tensors of rank 4 having the same algebraic properties as the weyl tensor . more precisely , one defines a _ weyl candidate _ as any rank 4 tensor @xmath91 fulfilling the properties @xmath92\gamma\delta}=c_{\alpha\beta\gamma\delta}= c_{\gamma\delta\alpha\beta } , \quad c^{\alpha}_{\phantom{\alpha}\alpha\gamma\delta}=0 , \quad c_{\alpha[\beta\gamma\delta]}=0 . \label{weyl_candidatep}\ ] ] from ( [ weyl_candidatep ] ) we easily deduce that @xmath93 and indeed @xmath94 is also a weyl candidate . given a weyl candidate @xmath91 , its _ weyl current _ @xmath95 is defined by @xmath96 the tensor @xmath97 is trace - free , antisymmetric in the last pair of indices and it has the property @xmath98}=0 $ ] . from ( [ weyl_current ] ) we deduce the following identities @xmath99\mu\nu}= -\frac{1}{3}\eta_{\alpha\beta\gamma\sigma}j^{*\sigma}_{\phantom{*\sigma}\mu\nu},\quad \nabla_{\beta}j^{\beta}_{\phantom{\beta}\gamma\delta}= r_{[\gamma}^{\phantom{[\gamma}\sigma\rho\mu}c_{\delta]\sigma\rho\mu } , \label{weyl_bianchi}\ ] ] where @xmath100 . equations ( [ weyl_current])-([weyl_bianchi ] ) have counterparts involving @xmath101 which we will not discuss . also , if we combine ( [ weyl_current ] ) and ( [ weyl_bianchi ] ) we get a wave equation for the weyl candidate : @xmath102\delta\beta\alpha}-2r_{\chi\phantom{\alpha}\rho}^{\phantom{\chi}\alpha \phantom{\rho}\delta}c_{\beta\gamma\alpha\delta}+ 2r_{[\rho}^{\phantom{[\rho}\alpha}c_{\chi]\alpha\beta\gamma}+ 2r_{\beta\phantom{\alpha}[\rho}^{\phantom{\beta}\alpha\phantom{[\rho}\delta } c_{\chi]\delta\gamma\alpha } \nonumber\\ \fl \hspace{3cm}+\eta_{\chi\rho\lambda\pi}\nabla^{\pi } j^{*\lambda}_{\phantom{*\lambda}\beta\gamma}+2\nabla_{[\chi}j_{\rho]\beta\gamma}. \label{box_weyl}\end{aligned}\ ] ] the previous considerations hold regardless of whether the weyl candidate is real or _ complex_. indeed , from any real weyl candidate @xmath103 we may construct a complex weyl candidate @xmath104 by means of @xmath105 a weyl candidate constructed in such a way is _ self - dual _ is given the name of `` self - dual '' . ] @xmath106 reciprocally , if a complex weyl candidate @xmath104 fulfils ( [ self - dual ] ) then its real part @xmath107 and its imaginary part @xmath108 are also weyl candidates related via @xmath109 besides the basic algebraic properties of a weyl candidate just explained we need to introduce a further concept which will play a crucial role in the sequel . [ causal - prop ] let @xmath0 be a spacetime and consider a weyl candidate @xmath110 defined in the whole of @xmath14 . we say that the weyl candidate @xmath110 propagates causally on @xmath14 if for any embedded spacelike hypersurface @xmath111 the condition @xmath112 implies @xmath113 on @xmath114 . given a set @xmath115 , a point @xmath116 belongs to the future cauchy development @xmath117 if any past - inextendible causal curve containing @xmath118 intersects @xmath115 . there is a similar notion of @xmath119 and the set @xmath120 is called the _ cauchy development _ of @xmath115 . it is well - known that the interior of @xmath114 is globally hyperbolic and since @xmath115 is an embedded spacelike hypersurface then it will be a cauchy hypersurface for @xmath114 see e.g. @xcite for an elementary introduction to these concepts of causality theory . the simplest example of a weyl candidate which propagates causally is the weyl tensor @xmath7 of a vacuum spacetime see e.g. @xcite . it is possible to address the causal propagation of a weyl candidate by following the general techniques given in @xcite see also @xcite in which the causal propagation of any tensor is analysed . these ideas are reviewed next . an essential object to study the causal propagation of a weyl candidate is its _ bel - robinson tensor_. given a real weyl candidate @xmath91 , we define its bel - robinson tensor by @xmath121 the previous definition has been described in a more general framework in @xcite as the _ superenergy tensor _ of the weyl candidate @xmath91 . in our particular context we prefer to call this tensor the bel - robinson tensor of @xmath91 in analogy with the bel - robinson tensor constructed outof the weyl tensor @xcite . the bel - robinson tensor of a weyl candidate has the following properties see @xcite for detailed proofs of points ( [ first ] ) , ( [ second ] ) and ( [ third ] ) ; point ( [ fourth ] ) follows by taking the covariant divergence of ( [ b - r ] ) and then using repeatedly ( [ weyl_current ] ) and ( [ weyl_bianchi ] ) . [ sproperties ] if @xmath122 is the bel - robinson tensor of the weyl candidate @xmath110 then 1 . [ first ] @xmath123 , @xmath124 . [ second]generalized dominant property : if @xmath125 , @xmath126 , @xmath127 , @xmath128 are causal future - directed vectors then @xmath129 . this property admits an alternative formulation : let @xmath130 be any orthonormal frame with @xmath131 the timelike vector . then the following inequality holds @xmath132 where the component indices refer to the frame @xmath133 . [ third ] @xmath134 a timelike vector @xmath135 such that @xmath136 . [ fourth]the covariant divergence of @xmath122 is given by @xmath137 next , we explain how the properties summarized in theorem [ sproperties ] are useful in the study of the causal propagation of @xmath138 . we follow closely @xcite in our exposition . let @xmath117 be the future cauchy development of a closed achronal hypersurface @xmath115 and pick up any event @xmath139 . furthermore , let @xmath140 denote the causal past of the event @xmath141 . the set @xmath142 is compact and it contains the set @xmath143 . the boundary of @xmath144 is @xmath145 where @xmath146 is the future cauchy horizon of @xmath147 . also , since the interior of @xmath117 is globally hyperbolic we deduce that it can be foliated by a family of spacelike hypersurfaces @xmath29 , @xmath148 , where @xmath149 , @xmath150 , @xmath151 and @xmath152 . the figure [ figure1 ] shows a schematic view of this geometric construction . [ figure1 ] , @xmath147 , @xmath117 and @xmath146 see the main text for further details . ] let @xmath153 be the timelike normal 1-form to the leaves of @xmath154 and define the quantity @xmath155 here , @xmath156 is the positive measure on the spacetime @xmath14 constructed from the volume form @xmath157 ( any measure defined from a volume form will be understood as positive ) and @xmath158 is the measure on the hypersurface @xmath159 obtained from its volume form @xmath160 . the scalar @xmath161 is called the _ superenergy density _ and is everywhere non - negative see point ( [ second ] ) of theorem [ sproperties] which entails @xmath162 . also , from the definition of @xmath163 we get @xmath164 where @xmath165 is defined by @xmath166 again point ( [ second ] ) of theorem [ sproperties ] implies that @xmath167 is also a non - negative quantity and hence @xmath168 . now , according to the gauss theorem we have @xmath169 where @xmath170 and @xmath171 denote the measures on the hypersurfaces @xmath147 and @xmath146 respectively induced by the volume forms @xmath172 , and @xmath173 . in this last expression @xmath174 is a causal vector which is orthogonal to the null hypersurface @xmath146 , points outward to @xmath144 ( therefore it is future - directed ) and is such that @xmath175 . note that these three conditions do not fix the vector @xmath174 univocally . by combining ( [ dwdt ] ) and ( [ gauss ] ) we deduce @xmath176 now , we need to estimate the right hand side of ( [ estimate1 ] ) . according to point ( [ second ] ) of theorem [ sproperties ] , the quantity @xmath177 is positive and therefore @xmath178 on the other hand , if @xmath110 vanishes on @xmath115 then @xmath179 and @xmath180 next , we set up an orthonormal frame @xmath130 with @xmath181 and we consider the set of components @xmath182 in such a frame . since these are continuous functions we deduce that in the compact set @xmath144 the estimate @xmath183 , @xmath184 holds for some constant @xmath185 . hence @xmath186 where in the last step , point ( [ second ] ) of theorem [ sproperties ] was used . from here , we deduce @xmath187 using the above estimates in ( [ estimate1 ] ) we get @xmath188 now it only remains to estimate @xmath189 and to that end ( [ covdivbr ] ) is used . the actual estimate will depend on the particular expression of the weyl current . for example , the simplest case arises when @xmath190 where no further estimates are needed and grnwall s lemma entails @xmath191 which in turn implies that the scalar @xmath192 vanishes in @xmath144 . from the arbitrariness of @xmath141 we deduce that the superenergy density is zero in @xmath117 and by a similar argument as the above one can show that this is also the case for @xmath119 . hence point 3 of theorem [ sproperties ] implies that @xmath113 on @xmath193 . consequently the tensor @xmath110 propagates causally whenever the weyl current is zero . _ point ( [ third ] ) of theorem [ sproperties ] entails @xmath112 if and only if @xmath194 . accordingly , the causal propagation of a weyl candidate can also be formulated in terms of the vanishing of the superenergy density at the hypersurface @xmath115 . [ remark : se_density ] _ we start this section by recalling some basic background about vacuum spacetimes possessing a killing vector see e.g. @xcite for a complete account . let @xmath0 be a vacuum solution of einstein field equations and let @xmath195 be a killing vector , @xmath196 . this condition enables us to define a 2-form , the _ killing form _ , by @xmath197 } = \nabla_\mu\xi_\nu . \label{killing-2form}\ ] ] elementary manipulations lead to @xmath198 the _ self - dual killing form _ @xmath199 is then @xmath200 with @xmath201 the hodge dual of @xmath202 . from the previous definition it is straightforward to obtain the property ( self - duality ) @xmath203 we shall write @xmath204 . the killing form is said to be degenerate if @xmath205 . other basic algebraic properties of @xmath199 are @xmath206\rho}=0,\quad \mathcal{f}_{\mu\nu}\overline{\mathcal{f}}^{\mu\nu}=0 . \label{f2}\ ] ] from a killing vector @xmath195 we may define its associated _ ernst 1-form _ by @xmath207 where @xmath208 is the norm of the killing vector and @xmath209 denotes its _ twist form_. it is well - known that in a vacuum spacetime @xmath210 is closed i.e . @xmath211}=0 $ ] . thus locally if we are in a simply connected region there exists a scalar field @xmath212 , the _ ernst potential _ , such that @xmath213 . the self - dual weyl tensor and the self - dual killing form are related by @xmath214 a property which is easily obtained from ( [ covdivf ] ) . finally , we introduce the tensor @xmath215 defined by @xmath216 the tensor @xmath215 has the obvious symmetries @xmath217\lambda\rho}= \mathcal{i}_{\lambda\rho\mu\nu}$ ] . this tensor can be regarded as a metric in the space of self - dual 2-forms . we have now all the ingredients to define the _ mars - simon tensor_. this is a four - rank tensor @xmath218 whose expression is given by @xmath219 from the previous expression we deduce that in a vacuum spacetime admitting a killing vector , the mars - simon tensor is determined up to a choice for the ernst potential . we make a choice such that @xmath220 on @xmath14 . our interest in the mars - simon tensor lies in the following result see @xcite for a proof . let @xmath0 be a smooth , vacuum spacetime with a killing vector @xmath195 . let @xmath199 denote the associated self - dual killing form . if there is a non - vanishing real constant @xmath221 such that the conditions @xmath222 hold on a non - empty @xmath223 then @xmath80 is locally isometric to the kerr spacetime . [ kerr - characterization ] condition ( [ kerr_condition1 ] ) of theorem [ kerr - characterization ] can be reformulated in the form @xmath224 where the scalar @xmath225 is defined for any killing vector @xmath226 and any real constant @xmath221 by the expression @xmath227 a very important property of the scalar @xmath225 is that it fulfils the differential equation @xmath228 to see it , differentiate both sides of ( [ mars - simon : modified ] ) and use ( [ covdivfs ] ) to replace the covariant derivatives of @xmath199 . the result is @xmath229 next use ( [ mars - simon : tensor ] ) to write @xmath230 in terms of @xmath231 and replace all the occurrences of @xmath232 by means of ( [ 2formelement ] ) to obtain @xmath233 we use here ( [ self - duality ] ) and ( [ ernst1 ] ) with the result @xmath234 from which ( [ cdxi ] ) is deduced . in @xcite it is proved that the mars - simon tensor @xmath235 is a weyl candidate . furthermore , it was shown that its weyl current is a linear function of @xmath235 in which case the wave equation ( [ box_weyl ] ) is homogeneous in @xmath235 . the derivation of these results are discussed as some of the intermediate calculations will be needed in section [ cp - ms ] . the mars - simon tensor is a weyl candidate . [ mars - weyl ] _ proof . _ from ( [ mars - simon : tensor ] ) it is clear that @xmath236\alpha\lambda}=\mathcal{s}_{\mu\nu[\alpha\lambda]}= \mathcal{s}_{\alpha\lambda\mu\nu}$ ] . using ( [ f2 ] ) one can show that @xmath237 and therefore the mars - simon tensor is traceless . to finish our proof we need to show that the mars - simon tensor fulfils the cyclic identity . to that end we compute @xmath238}$ ] getting @xmath239}=\frac{6}{1-\sigma } \mathcal{f}_{\mu[\nu}\mathcal{f}_{\rho\lambda]}-\frac{\mbox{i}\ \mathcal{f}^2}{2(1-\sigma ) } \eta_{\mu\nu\rho\lambda}.\ ] ] the right hand side of this expression vanishes by virtue of the identity @xmath240}=\frac{1}{12}\mbox{i}\ \mathcal{f}^2 \eta_{\mu\nu\rho\lambda},\ ] ] which in turn is a consequence of the algebraic property @xmath241}= \mathcal{f}_{[\mu\nu}\mathcal{f}_{\rho\lambda]}$ ] and the fact that in four dimensions @xmath242}$ ] is proportional to @xmath243 . a direct computation shows that the mars - simon tensor is self - dual @xmath244 therefore @xmath245 and @xmath246 are then related by duality as explained in section [ weyl : c ] : @xmath247 any property of @xmath248 will admit an equivalent counterpart formulated in terms of either its real or imaginary part . in particular , this shows that both @xmath245 and @xmath246 are weyl candidates and any of them contains the same information as the original mars - simon tensor . [ covdivms ] the covariant divergence of the mars - simon tensor is given by @xmath249\beta\alpha\rho}+\mathcal{f}_{\beta}^{\phantom{\beta}\rho } \mathcal{s}_{\gamma\delta\alpha\rho})+2g_{\beta[\gamma}\mathcal{s}_{\delta]\alpha\lambda\rho}\mathcal{f}^{\lambda\rho}\right)\xi^\alpha . \label{covdivmarssimon}\ ] ] [ cdms ] _ proof . _ to calculate @xmath250 we start from ( [ mars - simon : tensor ] ) and take the covariant divergence of @xmath218 . next we apply the relations ( [ covdivfs ] ) , @xmath251 , @xmath252 and @xmath253 . finally we expand all the occurrences of @xmath254 using ( [ 2formelement ] ) with the result @xmath255 } ) \mathcal{f}^2 ) \nonumber \\ & & + \frac{1}{1-\sigma}((2\mbox{i}\ \eta_{\beta\gamma\delta\alpha } \overline{\mathcal{w}}_{\rho\phantom{\alpha}\pi\kappa}^{\phantom{\rho}\alpha } \xi^\rho+4g_{\beta[\delta}\overline{\mathcal{w}}_{\gamma]\alpha\pi\kappa}\xi^\alpha)\mathcal{f}^{\pi\kappa}-12\mathcal{f}_{\beta}^{\phantom{\beta}\rho}\overline{\mathcal{w}}_{\gamma\delta\alpha\rho}\xi^\alpha)\end{aligned}\ ] ] the dependence on the ernst 1-form can be removed from this expression if we write @xmath210 in terms of @xmath199 and @xmath195 by means of ( [ ernst1 ] ) . the final step is to rearrange the terms in the resulting expression in order to render it in the form of ( [ covdivmarssimon ] ) . this is accomplished by using ( [ mars - simon : tensor ] ) to write @xmath256 in terms of the mars - simon tensor , expanding again all the occurrences of @xmath254 by means of ( [ 2formelement ] ) and using ( [ f2 ] ) , ( [ self - duality ] ) where necessary . equation ( [ covdivmarssimon ] ) enables us to define the weyl current of the mars - simon tensor as @xmath257\beta\alpha\rho}+\mathcal{f}_{\beta}^{\phantom{\beta}\rho } \mathcal{s}_{\gamma\delta\alpha\rho})+2g_{\beta[\gamma}\mathcal{s}_{\delta]\alpha\sigma\rho}\mathcal{f}^{\sigma\rho}\right)\xi^\alpha . \label{ms - weylcurrent}\ ] ] since @xmath258 is linear in the mars - simon tensor , we conclude in view of ( [ box_weyl]) that @xmath259 fulfils a homogeneous wave equation . this interesting property was used in @xcite to show that the _ domain of outer communication _ of an asymptotically flat smooth stationary spacetime is , under certain additional conditions , isometric to the kerr domain of outer communication i.e . the _ uniqueness of kerr_. using @xmath260 together with the self - duality of the mars - simon tensor one deduces @xmath261 that is to say , @xmath262 is also self - dual with respect to the block of indices @xmath263 . in our forthcoming calculations it is necessary to use the orthogonal splitting of @xmath248 with respect to a unit timelike vector field @xmath12 . since @xmath248 is a weyl candidate we can use ( [ decompose - weyl ] ) to find such a orthogonal splitting . the resulting split is given by @xmath264\mu}+ l_{\mu[\rho}\mathcal{t}_{\lambda]\nu}+\mbox{i}\ \varepsilon_{\rho\lambda\alpha}n_{[\mu}\mathcal{t}_{\nu]}^{\phantom{\nu]}\alpha}+ \mbox{i}\ \varepsilon_{\mu\nu\alpha}n_{[\rho}\mathcal{t}_{\lambda]}^{\phantom{\nu]}\alpha } ) , \label{decompose - ms}\ ] ] where @xmath265 the tensor @xmath266 is symmetric , spatial and traceless . alternatively , the orthogonal splitting of the mars - simon tensor can be calculated directly from ( [ mars - simon : tensor ] ) . to that end one needs to find the orthogonal splitting of the different quantities which appear in ( [ mars - simon : tensor ] ) . the orthogonal splitting of @xmath267 is given in ( [ decompose - weyl ] ) and the orthogonal splitting of @xmath199 is calculated by means of ( [ decompose_h ] ) and is given by @xmath268 } , \label{decompose - f}\ ] ] where @xmath269 one also needs the orthogonal splitting of @xmath270 which is easily calculated from ( [ 2formelement ] ) and ( [ decompose_eta ] ) so that @xmath271\nu}+2n_{[\nu}h_{\mu][\lambda}n_{\rho]}+\mbox{i}\ ( n_{[\mu}\varepsilon_{\nu]\rho\lambda}+n_{[\rho}\varepsilon_{\lambda]\mu\nu } ) . \label{decompose_i}\ ] ] we insert relations ( [ decompose - weyl ] ) , ( [ decompose - f ] ) and ( [ decompose_i ] ) into the formula ( [ mars - simon : tensor ] ) and then equal the resulting expression to ( [ decompose - ms ] ) . the result of that is the relation @xmath272 which shall be used later on . also needed later on is the relation @xmath273 which is derived from equation ( [ decompose - f ] ) . the killing vector @xmath195 can be decomposed in the form @xmath274 where @xmath275 , @xmath276 are called the _ killing lapse _ and the _ killing shift _ respectively . we need to find a formula relating @xmath277 to @xmath275 , @xmath278 . to that end , we start by taking the covariant derivative of both sides of @xmath279 and replace @xmath280 by means of ( [ gradnormaltoextrinsic ] ) . the result is @xmath281 next we use in this expression the property @xmath282 and equation ( [ decompose_cdt ] ) to replace @xmath283 and @xmath284 respectively yielding @xmath285 the antisymmetric part of this expression gives the orthogonal splitting of the 2-form @xmath202 which is @xmath286}+n_{[\mu}d_{\nu]}y+d_{[\mu}y_{\nu]}- n_{[\mu}\mathcal{l}_{\vec{n}}y_{\nu]}. \label{ot - f}\ ] ] also the symmetric part of ( [ decompose_cdxi ] ) is the killing condition @xmath287 . explicitly one has @xmath288 from which we deduce @xmath289 which renders the relation ( [ ot - f ] ) in the form @xmath290\rho}y^\rho+d_{[\mu}y_{\nu]}+2n_{[\mu}d_{\nu]}y . \label{decompose_f}\ ] ] from this expression we may compute the orthogonal splitting of @xmath291 which is @xmath292\alpha\beta}d^\alpha y^\beta , \label{decompose_fdual}\ ] ] where equation ( [ decompose_eta ] ) was used along the way . inserting equations ( [ decompose_f ] ) and ( [ decompose_fdual ] ) in ( [ mcalf ] ) and applying ( [ define - e ] ) on the resulting expression we deduce @xmath293 which is the required relation . in this section a proof of the following result is provided . the mars - simon tensor @xmath248 propagates causally . [ ms - propagation ] _ proof . _ according to ( [ re - im ] ) it is enough to show that either @xmath245 or @xmath246 propagates causally . we choose to work with the former , so let us set @xmath294 and define @xmath295 . equation ( [ covdivmarssimon ] ) entails @xmath296 hence , @xmath297 is the weyl current of the weyl candidate @xmath298 . to study the causal propagation of @xmath298 we follow the general procedure explained in section [ causal - prop - weyl ] . we denote by @xmath299 the bel - robinson tensor constructed from the weyl candidate @xmath298 . to prove our result , we need to find a good estimate for the quantity @xmath300 see equation ( @xmath301 ) . one can write @xmath302 where ( [ covdivbr ] ) was used . in order to work out the right hand side of equation ( [ q - est ] ) we calculate the orthogonal splitting of @xmath298 and @xmath303 with respect to @xmath12 . the tensors @xmath298 and @xmath304 are related to @xmath248 and @xmath262 by means of the relations @xmath305 and therefore their orthogonal splittings can be calculated once those of @xmath259 and @xmath262 are known . the orthogonal splitting of @xmath259 is given in equation ( [ decompose - ms ] ) and the orthogonal splitting of @xmath262 can be calculated from equation ( [ ms - weylcurrent ] ) given that we know the orthogonal splitting of all the quantities which appear in the definition of @xmath262 see ( [ ot - mcalj ] ) and ( [ ot - mcalz])-([ot - mcalp ] ) in appendix [ ot - j ] for the precise formulae . inserting these orthogonal splittings in ( [ q - est ] ) we get @xmath306\nonumber\\ & & \hspace{5cm}-6\mbox{i}\ \varepsilon_{\mu\nu\rho}\mathcal{t}^{\lambda\nu } \overline{\mathcal{t}}_{\lambda}^{\phantom{\sigma}\rho}y\left(\frac{\mathcal{e}^\mu}{\sigma-1}-\frac{\overline{\mathcal{e}}^\mu } { \overline{\sigma}-1}\right ) . \label{step1}\end{aligned}\ ] ] now , using the equations ( [ ots1])-([ots2 ] ) of appendix [ ot - br ] we can deduce the property @xmath307 which when combined with ( [ step1 ] ) yields @xmath308 . \label{step2}\ ] ] equation ( [ step2 ] ) is the key to get the required estimate . to obtain it , let us introduce an orthonormal frame @xmath130 with @xmath309 . since @xmath310 , @xmath276 are spatial vectors they can be written as a linear combination of the vectors @xmath311 . therefore , point ( [ second ] ) of theorem [ sproperties ] entails the estimates @xmath312\leq m_1 b_{\mu\nu\alpha\beta}n^\mu n^\nu n^\alpha n^\beta,\\ -\mbox{re}\left[\frac{12}{1-\sigma}b_{\mu\nu\alpha\beta } \mathcal{e}^\mu n^\alpha n^\beta y^\nu\right]\leq m_2 b_{\mu\nu\alpha\beta}n^\mu n^\nu n^\alpha n^\beta,\end{aligned}\ ] ] for some scalar continuous functions @xmath313 , @xmath314 defined on @xmath144 . using these estimates in equation ( [ step2 ] ) we obtain our final estimate @xmath315 where @xmath316 is the upper bound of @xmath317 in the compact set @xmath144 and thus ( [ gronwall ] ) can be written in the form @xmath318 since @xmath179 , then an application of grnwall s lemma enables us to conclude that @xmath319 , @xmath320 which , according to the considerations made at the end of section [ causal - prop - weyl ] proves the causal propagation . _ we stress that the fact that @xmath248 satisfies a wave equation and the property ( [ covdivmarssimon ] ) do not , by themselves , suffice to show the causal propagation of this tensor . if we impose that @xmath248 and some directional derivative of this quantity vanish on a spacelike hypersurface @xmath115 then using the wave equation and standard results on hyperbolic partial differential equations allow us to conclude that @xmath248 vanishes on a neighbourhood of @xmath115 . in doing so , extra conditions needed to be imposed so that @xmath248 is indeed zero in @xmath114 . this approach leads , in principle , to a weaker result than the one given in theorem [ ms - propagation ] . _ consider now any vacuum spacetime admitting a killing vector @xmath195 and use @xmath195 to construct the mars - simon tensor @xmath248 and as above , define its real part @xmath298 and the bel - robinson tensor @xmath299 of @xmath298 . for any unit timelike vector @xmath12 , define the positive scalar @xmath321 one can find an explicit expression for @xmath322 if we use the property see equation ( [ ots0 ] ) in [ ot - br] @xmath323 - 12\left|\frac{\mathcal{e}^\mu\mathcal{e}_\mu}{\sigma-1}\right|^2 + 36\left|\frac{\mathcal{e}^\mu\overline{\mathcal{e}}_\mu}{\sigma-1}\right|^2 , \label{se - density}\ ] ] which shows that @xmath322 can be expressed in terms of @xmath324 , @xmath277 and @xmath212 exclusively . when @xmath322 vanishes on a cauchy hypersurface , the causal propagation of @xmath248 entails the following result . let @xmath0 be a globally hyperbolic vacuum spacetime and suppose that @xmath14 admits a killing vector @xmath195 . if there are a cauchy hypersurface @xmath111 and a real constant @xmath221 such that @xmath325 and @xmath326 where @xmath327 is the unit timelike normal to @xmath115 , then @xmath14 is locally isometric to the kerr spacetime . [ positive - f ] _ proof . _ if @xmath328 is zero on @xmath115 then by ( [ phi ] ) @xmath329 . therefore , remark [ remark : se_density ] leads us to @xmath330 using that the mars - simon tensor propagates causally , one deduces that @xmath331 on @xmath14 . in particular , this implies that ( [ cdxi ] ) is rendered in the form @xmath332 which can be integrated to yield @xmath333 for some complex constant @xmath334 . the condition @xmath326 implies that @xmath335 and hence @xmath225 vanishes on @xmath14 . theorem [ kerr - characterization ] can now be applied and we conclude that @xmath14 is locally isometric to the kerr spacetime . we explain in this section how to use previous results to find conditions which ensure that the data development of a vacuum initial data set is isometric to a subset of the kerr spacetime . the basic idea behind is to express the result of theorem [ positive - f ] in terms of conditions on a vacuum initial data set @xmath336 . let @xmath336 be a vacuum initial data set and assume that there exist two scalar fields @xmath337 , @xmath338 , a vector field @xmath339 and a real constant @xmath221 , all defined on @xmath69 , fulfilling the following conditions @xmath340-\frac{3}{4}m^4|1-\tilde{\sigma}|^6 + 36\left|\frac{\mathcal{e}^i\overline{\mathcal{e}}_i}{\tilde{\sigma}-1}\right|^2=0 , \label{scalar0}\\ \fl d_j\tilde{\sigma}=2\tilde{y}k_{jl}\tilde{y}^l-2\ \mbox{i}\ \varepsilon_{jml } ( k_{k}^{\phantom{k}l}\tilde{y}^k\tilde{y}^m+\tilde{y}^m d^l\tilde{y})+2\tilde{y}d_j\tilde{y}+2\tilde{y}^l d_{[l}\tilde{y}_{j]},\label{dsigma}\\ \fl d_{(i}\tilde{y}_{j)}+\tilde{y } k_{ij}=0 , \label{kid1 } \\ \fl d_i d_j\tilde{y}+\mathcal{l}_{\tilde{y}^l } k_{ij } = \tilde{y}(r_{ij } + k k_{ij } - 2 k_{il } { k^{l}_{\phantom{l}j } } ) , \label{kid2}\end{aligned}\ ] ] where @xmath341 then the data development @xmath0 of @xmath336 is locally isometric to an open subset of the kerr spacetime . [ theorem : kerr_data ] [ killing_existence]_the conditions ( [ kid1 ] ) and ( [ kid2 ] ) are the _ killing initial data _ ( kid ) equations . it is well - known that if both a solution @xmath342 and the initial data set @xmath336 are _ suitably _ smooth , then the development of the initial data @xmath336 possesses a killing vector such that the pull - back of its restriction to @xmath69 coincides with the given killing data see for example , @xcite , and in particular @xcite for an explicit set of smoothness conditions on the killing data . alternatively , if the killing data is _ transversal _ , that is @xmath343 , then one can make use of the concept of killing development to ensure the existence of a spacetime with a killing vector associated to the killing data under question see @xcite . again , under suitable regularity conditions it can be shown that the killing development always includes the cauchy development of @xmath336 see again @xcite : the killing development is unique and maximal among the class of spacetimes containing the relevant cauchy data and a killing vector with non - closed orbits ; hence , if the cauchy development has a timelike killing vector , then it must be a subset of the killing development . _ _ proof . _ let @xmath14 be the data development of the vacuum initial data set @xmath81 and let @xmath195 be the killing vector mentioned in remark [ killing_existence ] . as usual consider a foliation @xmath29 , @xmath148 , of @xmath14 with @xmath84 and let @xmath12 be the unit timelike normal to the leaves of this foliation . let us adopt the formalism and notation explained in section [ section : mstensor ] with @xmath195 being the killing vector . then the orthogonal splitting of @xmath195 is @xmath344 . as explained in remark [ killing_existence ] we have the property @xmath345 which combined with equations ( [ ei ] ) and ( [ def - mce ] ) entails @xmath346 . the orthogonal splitting of @xmath210 is calculated on one hand from equation ( [ ernst1 ] ) using the expressions ( [ decompose - f ] ) , ( [ def - mce ] ) and the property @xmath344 . the result is @xmath347 where @xmath348}.\label{def - smu}\end{aligned}\ ] ] on the other hand , the orthogonal splitting of @xmath210 can be calculated from the relation @xmath349 yielding @xmath350 combining the previous equation with the pull - back of equation ( [ def - smu ] ) under @xmath65 and ( [ dsigma ] ) we deduce @xmath351 upon a suitable choice of the ernst potential @xmath212 , this equation can be integrated on @xmath69 to give @xmath352 . next consider the mars - simon tensor @xmath248 constructed with the killing vector @xmath195 and calculate its orthogonal splitting with respect to @xmath12 equation ( [ decompose - ms ] ) . using ( [ tgoestoe ] ) and our previous results , we get @xmath353 expanding this product and using conditions ( [ e2])-([scalar0 ] ) we conclude @xmath354 moreover , equation ( [ e2 ] ) implies @xmath355 see ( [ ot - f2 ] ) . thus , theorem [ positive - f ] applies and therefore we conclude that @xmath0 must be locally isometric to an open subset of the kerr spacetime . theorem [ theorem : kerr_data ] asserts that under certain conditions the development of a vacuum initial data set is isometric to an open subset of the kerr spacetime but nothing is said as to the resulting development being one of the specializations of the kerr spacetime : the schwarzschild solution . in this section we show the form which theorem [ theorem : kerr_data ] takes in that case . this complements previous work of the present authors about the same issue @xcite . to state our result we need a preliminary lemma whose proof can be found in @xcite . let @xmath336 be a vacuum initial data set and assume that there exist a scalar @xmath337 and a vector field @xmath339 , all defined on @xmath69 , fulfilling the following conditions @xmath356}+ 2 \tilde{y}_{[i}d_{j]}\tilde{y } + 2 \tilde{y}_{[i}k_{j]l}\tilde{y}^l = 0,\label{kids-1}\\ & & \tilde{y}_{[i}d_{j}\tilde{y}_{k]}=0\label{kids-2}.\end{aligned}\ ] ] then there exists an integrable killing vector @xmath195 in the data development of @xmath336 . [ lemma : static_killing ] with the aid of this lemma we can now derive conditions guaranteeing that the development of a vacuum initial data set is isometric to an open subset of the schwarzschild spacetime let @xmath336 be a vacuum initial data set and assume that there exist two scalar fields @xmath337 , @xmath338 and a vector field @xmath339 , all defined on @xmath69 , fulfilling the conditions of theorem [ theorem : kerr_data ] with ( [ kid1])-([kid2 ] ) replaced by ( [ kids-1])-([kids-2 ] ) . then the data development @xmath14 is locally isometric to an open subset of the schwarzschild spacetime . [ theorem : schwarzschild_data ] _ proof . _ the first step in the proof of this theorem is to use lemma [ lemma : static_killing ] to show that the data development @xmath14 admits an integrable killing vector @xmath195 . the remaining part of the proof follows the same pattern as that of theorem [ theorem : kerr_data ] and therefore we reach the conclusion that @xmath14 is isometric to an open subset of a specialization of the kerr spacetime possessing an integrable killing vector which , as is well known , is the schwarzschild spacetime . _ it is worth recalling that the analysis of @xcite has provided an explicit formula for a _ timelike kid candidate _ in terms of concomitants of the initial data @xmath357 . hence , a combination of theorem [ theorem : schwarzschild_data ] with the such timelike kid candidate renders an algorithmic characterization of schwarzschild initial data alternative to the one given in @xcite . _ the main result obtained in this article , theorem [ theorem : kerr_data ] provides a characterization of initial data for the kerr spacetime by first requiring that a set of overdetermined elliptic differential equations the kid equations admit a solution ; and then in turn requiring that certain objects constructed out of the solutions to the kid equation satisfy very particular relations . this characterization fails to be algorithmic in two ways . first , it is in practise a non - trivial problem to verify whether a certain initial data set admits a kid or not . a pointwise procedure for doing this has been discussed in @xcite . alternatively , one could resort to a non - time symmetric generalization of the geometric invariant characterizing static time symmetric initial data sets constructed in @xcite which requires solving a fourth - order elliptic partial differential equation . in any case and this leads to the second way the result fails to be algorithmic even if one could ascertain the existence of a kid on the initial data , one requires to know the kid in an explicit manner in order to be able to verify the conditions ( [ e2 ] ) , ( [ scalar0 ] ) and ( [ dsigma ] ) . it is plausible that from an hypothetical characterization of the kerr spacetime in terms of , say , concomitants of the weyl tensor analogous to that for the schwarzschild spacetime given in @xcite one would be able to construct , following the ideas of @xcite , kid candidates for the initial data set @xmath336 . if such a kid candidate were available , then theorem [ theorem : kerr_data ] would constitute a powerful tool to analyse the behaviour of dynamical black hole spacetimes , both from an analytic and a numerical point of view . we would like to thank jmm senovilla , m mars , w simon and cm losert - vk for useful remarks on the manuscript . javk is supported by an epsrc advanced research fellowship . agp gratefully acknowledges financial support from the spanish `` ministerio de ciencia e innovacin '' under the postdoctoral grant ex-2006 - 0092 . given the symmetries of the weyl current @xmath262 it is immediate to realise that its orthogonal splitting must take the form @xmath358\mu}+n_\mu n_{[\nu}\mathcal{v}_{\rho]}+ \mathcal{p}_{\mu\nu\rho}. \label{ot - mcalj}\ ] ] it takes a greater effort , however , to find the explicit expression of the spatial tensors @xmath359 , @xmath360 , @xmath361 , @xmath362 . these expressions are found by inserting into equation ( [ ms - weylcurrent ] ) the orthogonal splitting of @xmath248 shown in ( [ decompose - ms ] ) , the orthogonal splitting of @xmath199 shown in ( [ decompose - f ] ) and the relation @xmath344 . a computer calculation yields the results @xmath363\alpha\rho}+3\mathcal{e}_{[\mu}\mathcal{t}_{\nu]\rho}y^\rho+y_{[\nu}\mathcal{t}_{\mu]\alpha}\mathcal{e}^\alpha-\mbox{i}\ \varepsilon_{\mu\nu\rho}\mathcal{e}^\alpha\mathcal{t}_\alpha ^{\phantom{\alpha}\rho}y),\label{ot - mcalz}\\ & & \hspace{-2cm}\mathcal{x}_{\mu\nu}\equiv \frac{1}{1-\sigma}\bigg(\mathcal{e}^\alpha\big(4\mbox{i } y(\mathcal{t}_{(\alpha}^{\phantom{(\alpha}\rho}\varepsilon_{\nu)\mu\rho}-\varepsilon_{\nu\alpha\rho}\mathcal{t}_{\mu}^{\phantom{\mu}\rho})+6\mathcal{t}_{\nu\mu}y_\alpha-2\mathcal{t}_{\mu\alpha}y_{\nu}-4\mathcal{t}_{\nu\alpha}y_\mu+4h_{\nu\mu}\mathcal{t}_{\alpha\rho}y^\rho \big ) \nonumber\\ & & \hspace{1cm}+6\mathcal{e}_\nu\mathcal{t}_{\mu\alpha}y^\alpha\big ) \bigg),\label{ot - mcalx}\\ & & \hspace{-2cm}\mathcal{v}_\mu\equiv\frac{\mathcal{e}^\alpha}{1-\sigma}\left(4\mbox{i}\ y^\rho(\varepsilon_{\mu\lambda(\rho}\mathcal{t}_{\alpha)}^{\phantom{\alpha)}\lambda}+\varepsilon_{\alpha\rho\lambda}\mathcal{t}_{\mu}^{\phantom{\mu}\lambda})-2y\mathcal{t}_{\mu\alpha}\right),\label{ot - mcalv}\\ & & \hspace{-2cm}\mathcal{p}_{\mu\nu\rho}\equiv\frac{2}{1-\sigma}\bigg(2\ \mbox{i}\ \varepsilon{}_{\nu}{}_{\rho}{}_{\lambda } \ \mathcal{e}{}_{\mu } \mathcal{t}{}_{\alpha}{}^{\lambda } y{}^{\alpha } + 2\mathcal{e}{}_{[\rho } ( 3y\mathcal{t}{}_{\nu]\mu } + \mbox{i } \ \varepsilon_{\nu]\mu\lambda}\mathcal{t}{}_{\alpha}{}^{\lambda } \ y{}^{\alpha } ) \nonumber\\ & & \hspace{1cm}+\mathcal{e}{}^{\alpha}\bigg(-2\mbox{i } \big(\varepsilon{}_{\rho}{}_{\alpha}{}_{\lambda}\mathcal{t}{}_{[\nu}{}^{\lambda } \ y{}_{\mu ] } + \varepsilon{}_{\nu}{}_{\alpha}{}_{\lambda } \mathcal{t}{}_{[\mu}{}^{\lambda } y{}_{\rho ] } + 2\varepsilon{}_{\mu}{}_{\alpha}{}_{\lambda}\mathcal{t}{}_{[\nu}{}^{\lambda } y{}_{\rho]}\nonumber \\ & & \hspace{1 cm } + y{}^{\lambda } ( \varepsilon{}_{\nu}{}_{\rho}{}_{\alpha } \mathcal{t}{}_{\mu}{}_{\lambda}+\varepsilon{}_{\mu}{}_{\alpha}{}_{[\nu } \mathcal{t}{}_{\rho]}{}_{\lambda})\big)\nonumber\\ & & \hspace{1cm}+2h{}_{\mu}{}_{[\nu } \big(\mathcal{t}{}_{\rho]}{}_{\alpha } y + \mbox{i}\ y{}^{\lambda } ( \varepsilon{}_{\rho]}{}_{\lambda}{}_{\delta } \mathcal{t}{}_{\alpha}{}^{\delta } + \ \mathcal{t}{}_{\rho]}{}^{\delta}\varepsilon{}_{\alpha}{}_{\lambda}{}_{\delta}+ \ \varepsilon{}_{\rho]}{}_{\alpha}{}_{\delta } \mathcal{t}{}_{\lambda}{}^{\delta})\big)\bigg)\bigg).\label{ot - mcalp}\end{aligned}\ ] ] in @xcite the orthogonal splitting of the bel - robinson tensor was calculated and the different parts resulting from such splitting were studied in detail . the calculations are similar for the bel - robinson tensor of a weyl candidate and therefore we only need to adapt these results for the particular case of @xmath298 . in this case the orthogonal splitting of @xmath299 reads see section ( 5.1 ) of @xcite @xmath365 where the different parts of this splitting are given by @xmath366 where @xmath367 and @xmath368 are respectively the _ electric _ and _ magnetic _ parts of the weyl candidate @xmath369 . from ( [ br - decomposition ] ) and ( [ br - parts ] ) we get @xmath370 where the relations @xmath371 need to be used .

in an earlier paper we presented nuclear x - ray flux densities , measured with _ rosat _ , for the b2 bright sample of nearby low - luminosity radio galaxies . in this paper we construct a nuclear x - ray luminosity function for the b2 radio galaxies , and discuss the consequences of our results for models in which such radio galaxies are the parent population of bl lac objects . based on our observations of the b2 sample , we use monte carlo techniques to simulate samples of beamed radio galaxies , and use the selection criteria of existing samples of bl lac objects to compare our simulated results to what is observed . we find that previous analytical results are not applicable since the bl lac samples are selected on beamed flux density . a simple model in which bl lacs are the moderately beamed ( ) counterparts of radio galaxies , with some random dispersion ( decades ) in the intrinsic radio - x - ray relationship , can reproduce many of the features of the radio - selected and x - ray - selected bl lac samples , including their radio and x - ray luminosity functions and the distributions of their radio - to - x - ray spectral indices . in contrast , models in which the x - ray and radio emission have systematically different beaming parameters can not reproduce important features of the radio - galaxy and bl lac populations , and recently proposed models in which the radio - to - x - ray spectral index is a function of source luminosity can not in themselves account for the differences in the slopes of the radio and x - ray - selected bl lac luminosity functions . the redshift distribution and number counts of the x - ray - selected emss sample are well reproduced by our best models , supporting a picture in which these objects are beamed fri radio galaxies with intrinsic luminosities similar to those of the b2 sample . however , we can not match the redshift distribution of the radio - selected 1-jy sample , and it is likely that a population of frii radio galaxies is responsible for the high - redshift objects in this sample , in agreement with previously reported results on the sample s radio and optical - emission - line properties . galaxies : active x - rays : galaxies bl lacertae objects : general galaxies : jets

astro-ph0209099

the extreme properties of bl lac objects are explained in terms of relativistic beaming of the emission from a jet oriented close to the line of sight . this model implies the existence of a substantial ` parent population ' of sources whose jets are less favourably aligned , and it is widely accepted that this is the population of low - power radio galaxies ( browne 1983 ; urry & padovani 1995 ) . properties which are isotropic and unaffected by beaming should be similar in bl lac objects and low - power radio galaxies . this is broadly supported by observations of extended radio emission ( e.g. , antonucci & ulvestad 1985 , kollgaard et al . 1992 , perlman & stocke 1993 , 1994 ) , host galaxies ( e.g. , ulrich 1989 , abraham , mchardy & crawford 1991 ; wurtz , stocke & yee 1996 ; falomo 1996 ) and cluster environments ( pesce , falomo & treves 1995 ; smith , odea & baum 1995 ; wurtz , stocke & ellingson 1997 ) . in a series of papers ( padovani & urry 1990 ; padovani & urry 1991 ; urry , padovani & stickel 1991 ) , urry and co - workers made this model quantitative by predicting the luminosity function of bl lac objects based on that of radio galaxies . they used the analysis of urry & schafer ( 1984 ) and urry & padovani ( 1991 ) to calculate the expected luminosity function of a population of beamed objects given a parent ( unbeamed ) luminosity function . with the data on radio - galaxy and bl lac populations then available they were able to show reasonable agreement between the predictions of the model and the observed luminosity functions . they found that the luminosity function of radio - selected bl lac objects from the 1-jy sample ( stickel et al.1991 ) was consistent with that of radio galaxies from the 2-jy sample ( wall & peacock 1985 ) if the radio - emitting plasma in the cores had a bulk lorentz factor @xmath2 . the x - ray luminosity function and number counts of x - ray selected bl lac objects were consistent with the luminosity function of fri radio galaxies observed with _ einstein _ by fabbiano et al . ( 1984 ) if the x - ray emitting plasma has a somewhat lower bulk lorentz factor , @xmath3 . the issue of the relationship between radio galaxies and x - ray and radio - selected bl lac objects is only partially resolved by this work . the best - studied sample of x - ray selected bl lacs is the emss sample ( wolter et al . 1991 , rector et al . 2000 ) and the best radio - selected sample is the 1-jy sample ( stickel et al . 1991 , rector & stocke 2001 ) . when these samples are compared , a number of differences emerge . some perhaps as many as half of the 1-jy objects have radio structures , luminosities and emission - line properties similar to those of frii radio galaxies ( antonucci & ulvestad 1984 , kollgaard et al.1992 , rector & stocke 2001 ) while the emss bl lacs are all fri - like ( perlman & stocke 1993 ) . the luminosity functions of the two samples are also different ; the radio - selected objects show evidence for positive evolution ( i.e. sources were more numerous or more powerful at higher redshift ) while x - ray - selected sources appear to be _ negatively _ evolving . urry et al . were forced by the data then available to use an x - ray - selected sample of bl lacs for their comparison with the x - ray luminosity function of radio galaxies , and a radio - selected sample when considering the radio luminosity function . they were thus unable to say anything about the relation between radio galaxies and the two bl lac populations . radio - selected bl lacs are often considered to be more extreme than x - ray selected objects ; they have more prominent radio cores ( perlman & stocke 1993 , laurent - muehleisen et al . 1993 , rector & stocke 2001 ) , higher optical polarization with less stable position angle ( jannuzi , smith & elston 1994 ) and a lower optical starlight fraction ( stocke et al . 1985 ) . on the other hand , their environments and host galaxies are similar to x - ray selected sources ( wurtz , stocke & yee 1996 , wurtz et al . this has led to suggestions that radio - selected objects are more strongly affected by doppler boosting than x - ray selected objects , and so are being observed at smaller angles to the line of sight . since radio - selected and x - ray selected objects have similar x - ray luminosities , this requires that the x - ray - emitting regions be less strongly beamed , or even isotropic ( maraschi et al . 1986 ; celotti et al . 1993 ) ; the idea of weaker beaming is consistent with the difference in the doppler beaming factors found by urry et al . this in turn implies that x - ray - selected bl lacs should be the more numerous population , since they can be seen at larger angles to the line of sight , which celotti et al . ( 1993 ) argue is consistent with the x - ray luminosity functions of the two populations . models of this kind can either rely on differences in velocity between x - ray and radio - emitting regions ( e.g. , ghisellini & maraschi 1989 ) or differences in opening angle of the jet ( celotti et al . 1993 ) . however , these models do not account for the differences in evolutionary properties between the two samples . in section [ beaming ] of this paper we shall discuss whether they are consistent with observations of radio galaxies and with the x - ray and radio luminosity functions of the two classes . a description of bl lac objects in terms of the selection waveband does not necessarily reflect the underlying physics . an alternative approach is to refer to high - energy peaked bl lacs ( hbl ) and low - energy peaked bl lacs ( lbl ) ( e.g. , giommi & padovani 1994 , padovani & giommi 1995 ) , distinguishing between the two classes by their radio / x - ray flux ratios ; a typical dividing line is that hbl have @xmath4 , or , equivalently , @xmath5 , where @xmath6 is defined here and throughout the paper in the sense @xmath7 and @xmath8 denotes the radio - to - x - ray two - point spectral index . by selecting bright sources at one or the other waveband we may simply be picking up objects with extreme radio / x - ray flux ratios , which suggests various possible schemes for unifying the two populations ( e.g. , giommi & padovani 1994 , fossati et al . 1997 ) and is consistent with the discovery of ` intermediate ' bl lacs in deeper surveys ( laurent - muehleisen et al . 1999 ) . from multi - wavelength observations it has been found that the spectral energy distributions ( seds ) of bl lacs can be represented by a low - frequency peak ( assumed to be synchrotron radiation ) and a high - frequency peak ( perhaps inverse - compton radiation ) . there is evidence that the positions of these peaks shift to higher frequencies at lower bolometric luminosities ( sambruna , maraschi & urry 1996 , fossati et al . however , giommi , menna & padovani ( 1999 ) do not find the increase in the numbers of hbl at fainter radio fluxes that is expected in such a model . to try to resolve some of the outstanding issues in bl lac unification it is productive to learn more about the presumed parent population of many or all of the bl lac objects , namely the low - power radio galaxies . work to date has been hampered by the lack of a well - defined , low - frequency - selected sample of radio galaxies with well - known radio and x - ray properties . in a previous paper ( canosa et al . 1999 ) we presented results for the 40 members ( @xmath9 per cent ) of the b2 bright sample of radio galaxies which had been observed in pointed _ rosat _ observations . the high spatial resolution of _ rosat _ allowed us to separate the nuclear emission from extended emission due to the hot - gas environment of the radio source . in this paper we use these data to derive a new x - ray luminosity function for the nuclei of b2 radio sources , largely free from contamination from thermal emission from the sources hot - gas environments . we use numerical techniques to extend earlier work , first asking how well a simple beaming model can be used to link the x - ray and radio luminosity functions of radio galaxies , and then investigating the expected relationship between radio galaxies and bl lac objects when selection bias is taken into account . throughout the paper we use a cosmology in which @xmath10 and @xmath11 . the b2 bright sample , described by colla et al . ( 1975 ) , consists of 52 radio sources with 408-mhz radio flux @xmath12 jy identified with galaxies with photographic magnitudes @xmath13 . ( the flux limit is declination - dependent ; see colla et al . for details . ) the sample is flux - complete for all but the largest sources . the sky coverage of the sample is 1.154 sr [ cf . colla ( 1975 ) ; the figure quoted by ulrich ( 1989 ) appears to be in error ] . of the 52 objects in the sample , two ( b2 0207 + 38 and 1318 + 34 ) are starburst galaxies , and will not be considered further in this paper . one ( b2 1833 + 32 ) is the broad - line frii radio galaxy 3c382 , which we also exclude from further consideration because frii radio sources do not participate in most radio galaxy - bl lac unification models , and two ( b2 1101 + 38 and 1652 + 39 ) are the well - known bl lac objects mrk 421 and mrk 501 , which we did not discuss in canosa et al . ( 1999 ) , but which will be included in the sample we discuss here in order that it should be unbiased with respect to orientation . the remainder are more or less typical fri radio sources , though the sample includes some peculiar objects such as b2 1626 + 39 ( 3c338 ) . ulrich ( 1989 ) has shown that the b2 bright sample is well matched in _ unbeamed _ properties , like extended radio luminosity , to samples of nearby bl lac objects . [ cols= " < , < , > , > , > , > , > , > " , ] 408-mhz flux densities and @xmath14 ( the low - frequency spectral index ) have been corrected to the scale of baars et al.(1977 ) as described in the text . 5-ghz flux densities for the cores are taken from the literature . references for these are as follows : ( 1 ) giovannini et al . ( 1988 ) ( 2 ) fanti et al . ( 1987 ) ( from 1.4 ghz ) ( 3 ) venturi et al . ( 1993 ) ( 4 ) fomalont , private communication ( 5 ) stickel et al . ( 1991 ) ( 6 ) measured from maps supplied by r. morganti ( 7 ) cotton et al . ( 1995 ) ( 8) antonucci & ulvestad ( 1985 ) ( extrapolated from 1.5 ghz ) . 1-kev nuclear flux densities are taken from canosa et al . ( 1999 ) , except for those for mrk 421 and 501 , which are discussed in the text , and for b2 1553 + 24 , for which canosa et al.quoted an incorrect x - ray flux density , although the count rate they quote is correct . because an unbiased selection on unbeamed properties is important , we need to consider whether some of the sample sources exceed the 408-mhz flux limit solely because of a beamed radio component . we may estimate this by subtracting the 5-ghz core flux density from the total 408-mhz flux , on the assumption that the core spectrum is flat between these two frequencies , and seeing whether the result lies below the b2 position - dependent flux limit . on this basis , 4 sources ( b2 0648 + 27 , 1108 + 27 , 1144 + 35 and 1553 + 24 ) possibly have too low a 408-mhz flux to be in the sample through extended emission alone . of these , the core of 1144 + 35 is known to dominate the extended structure , and to be variable ( giovannini et al . 1999 ) . 0648 + 27 is also core - dominated ( e.g. , antonucci 1985 ) . we exclude these two sources from the sample . the other two are reasonably normal members of the sample which happen to lie at the flux limit , so that the case for excluding them ( given the large uncertainties in core spectrum ) is uncertain laing et al.(1999 ) , on a similar basis , chose not to exclude these two sources from their sample of jetted objects . our revised , unbiased sample then consists of 47 objects , whose properties are listed in table [ sample ] . we have nuclear x - ray information for all but 8 of these . we are now in a position to construct a luminosity function for the extended radio emission from our unbiased sample . for a source which has a detected radio core at 5 ghz , we subtract its flux density from the 408-mhz flux density before making the luminosity calculation ; for sources without detected cores ( where the upper limit on core flux density is typically a small fraction of the total flux density ) we just use the total 408-mhz flux density , and we also do this for 1108 + 27 and 1553 + 24 . for the purposes of calculating luminosities we correct the 408-mhz flux densities to the baars et al . ( 1977 ) scale from the kellerman ( 1964 ) scale used in b2 the catalogue ( colla et al.1970 ) , using the factor 1.074 suggested by baars et al .. spectral indices are also corrected to the baars et al . scale in the same way . to generate a differential luminosity function we use the @xmath15 method , taking into account both the optical and radio selection criteria ; the optical selection criterion is the important one in 31/47 objects . the radio flux limits and flux densities used to determine @xmath16 are the original values , without core subtraction . we assign errors to the bins in the luminosity function using the method of wolter et al . ( 1994 ) , having verified by simulation that the error estimates derived from this method are close to the real statistical errors . our differential luminosity function for the extended flux from the sample is plotted in fig . [ 408-lf ] . it is consistent with the luminosity function derived for the sample by de ruiter et al . . we can similarly derive 5-ghz and 1-kev luminosity functions for the nuclei of the sample , using the method described by wolter et al.(1994 ) . the values of @xmath16 ( or @xmath17 in their notation ) are derived from the selection band ( i.e. the 408-mhz flux densities and optical magnitudes ) while the luminosities are those in the observation band . the 5-ghz radio core flux densities include 5 upper limits ; this is a sufficiently small fraction of the data that we are not going to be grossly in error if we assume that the true values of the core fluxes for these sources are equal to the limits ( i.e. treat limits as detections at the limit level ) . doing this leads to the luminosity function plotted in fig . ( throughout the paper we assume that the radio spectral index @xmath6 of the cores is 0 . ) however , the 1-kev flux densities include 12 upper limits and 8 unobserved sources . in most cases the upper limits are the result of an ambiguity in the separation of nuclear and extended x - ray emission . since we have demonstrated that the unobserved sources have no known bias with respect to the rest of the sample ( canosa et al . 1999 ) , we can simply scale the luminosity function computed for the other sources by a factor 47/39 . but the large number of x - ray limits ( nearly one third of the total observed sources ) , could seriously affect the derived luminosity function . to assess the effect of the limits , we computed the luminosity function in four different ways : ( a ) with the limits treated as detections ; ( b ) with the non - detected sources omitted altogether and the results scaled ; ( c ) with the limits treated as detections at 0.1 times the limit level ; and ( d ) with the limits replaced by the values that the x - ray cores ` should ' have according to the radio - x - ray core correlation reported in canosa et al . the resulting luminosity functions were similar , with slopes and normalizations of straight - line fits to the data being consistent within the statistical uncertainties , so our results here are insensitive to the treatment of the upper limits . accordingly , we treat the limits as detections in order to generate fig . [ 1-lf ] , which has also been scaled to take account of the unobserved sources . ( throughout the paper we assume that the x - ray spectral index of cores is @xmath18 . ) the result is the first nuclear x - ray luminosity function to be computed for a sample of radio galaxies . since the analysis of canosa et al . ( 1999 ) , we have obtained the first _ chandra _ images of fri radio galaxies from the b2 sample ( worrall et al . 2001 ; hardcastle et al . these show that in some of these objects there is extended x - ray emission on scales of a few arcseconds , in addition to a point - like nucleus and the larger - scale group or cluster emission that we have discussed elsewhere ( canosa et al . 1999 ; worrall & birkinshaw 2000 ) . this small - scale extended emission would have been unresolved to _ rosat _ , and therefore contaminates the nuclear 1-kev flux densities we quote in this paper . but it does not dominate the total soft - x - ray flux in the b2 objects we have observed with _ chandra _ so far , and we know from the correlation between radio , x - ray and optical nuclear emission in these objects ( canosa et al . 1999 ; hardcastle & worrall 1999 , 2000 ) that a significant fraction of the _ rosat_-observed 1-kev emission must arise in a beamed nucleus . it is therefore reasonable to use the current values to test the predictions of unified models , provided we bear in mind the possible distorting effects of this contaminating extended emission . [ chandra - caveat ] to test unified models using the new b2 luminosity function we must have comparison samples of bl lac objects . the two comparison samples we shall consider in this paper are the 1-jy sample of radio - selected bl lacs [ we use the 34-object sample of stickel et al . ( 1991 ) , because of the good availability of x - ray information , but with the redshifts tabulated by rector & stocke ( 2001 ) ] and the emss sample of x - ray selected objects [ we use the 41-object ` d40 ' sample of rector et al.(2000 ) ] . flux densities are now available for almost all objects in both these samples at both 5-ghz radio and x - ray frequencies ( urry et al . 1996 , fossati et al . 1998 and references therein ; rector et al.and references therein ) . so , unlike urry and co - workers , we are not limited to comparing the radio luminosity function of the b2s to the radio luminosity function of radio - selected bl lacs , and so on ; we can compare the luminosity functions of both kinds of bl lac in both wavebands to the radio - galaxy observations . accordingly , we have constructed radio and x - ray luminosity functions for the two samples in the same way as for the radio - galaxy data . in both cases , we calculate @xmath16 using the energy band in which the source was selected . we take account of the optical flux selection criterion for the 1-jy sample ( @xmath19 mag on sky survey plates ; stickel et al.1991 ) and the variable sensitivity of the emss ( gioia et al.1990 ) in the manner described by morris et al . ( 1991 ) . in both cases , we include modest luminosity evolution as an exponential function of look - back time ( positive for the radio - selected objects , negative for the x - ray - selected objects ) , in the manner described by wolter et al.(1994 ) , to obtain @xmath20 ; the effect of this assumption on the results is quite small . for the sources with only lower redshift limits , we have used the limit as though it were the actual redshift of the source . for the few sources without known redshifts ( 5 in the 1-jy sample , 2 in the emss sample ) , we have used the median redshift for the detected objects in the sample from which they were drawn ( 0.55 for the 1-jy sample , 0.30 for the emss sample ) . we have constructed luminosity functions on the assumption that the nuclear components in both radio and x - ray are dominant in both classes of source . the results are plotted in figs [ 5-lfbl ] and [ 1-lfbl ] . they are , as we would expect , consistent on the whole with other determinations of the luminosity functions for these samples ( e.g. , stickel et al . 1991 , wolter et al . 1994 ) . [ 5-lfbl ] [ 1-lfbl ] urry & schafer ( 1984 ) discuss the effect of relativistic beaming on a differential luminosity function which is intrinsically a simple power law ( i.e. , one which has the form @xmath21 , where @xmath22 is the luminosity , @xmath23 has dimensions volume@xmath24 luminosity@xmath24 , and @xmath25 is the slope of the luminosity function . ) they show that the observed luminosity function is a broken power law . the low - luminosity slope is given by @xmath26 , where @xmath27 is the power to which the doppler factor is raised in the beaming calculation ; for a jet , @xmath28 , where @xmath6 is the jet spectral index . at high luminosities , the slope of the observed luminosity function is the same as that of the intrinsic luminosity function . the luminosity functions for the radio and x - ray cores of the b2 objects shown in figs [ 5-lf ] and [ 1-lf ] are for beamed components of the radio source . the unbeamed 408-mhz radio luminosity function in fig . [ 408-lf ] can be treated as a single power law , since the well - known steepening ( e.g. , machalski & godlowski 2000 ) in the low - frequency radio source luminosity function occurs around @xmath29 w hz@xmath24 . if we make the further assumption that the _ intrinsic _ core radio and x - ray fluxes of the b2s are proportional to the extended 408-mhz flux , the urry & schafer analysis can be applied to these data . since most sources are at large angles to the line of sight , we are predominantly seeing the low - luminosity end of the beamed luminosity function . urry & schafer s result leads us to expect slopes of @xmath30 and @xmath31 for the 5-ghz and 1-kev core luminosity functions , respectively . the actual slopes ( @xmath32 and @xmath33 respectively ) are somewhat steeper than , but consistent with , these expectations , supporting our assumption above about the relationship between total luminosity and intrinsic core luminosity . since the slope of the unbeamed , ` parent ' luminosity function should be equal to that of the 408-mhz data , @xmath34 , which is similar to the predicted slope of the beamed luminosity function , we would not expect to see a break in the slope of the beamed luminosity function at high luminosities , and indeed there is no evidence for such a break in the luminosity functions for the beamed emission from the b2s . what would we expect to see on this model for the bl lac samples if the b2 sources were their parent population ? the bl lac objects are clearly at the luminous end of the beamed luminosity function , and therefore we should recover the slope of the parent luminosity function . since this is @xmath35 , a simple application of the urry & schafer model would cause us to expect the slopes of the differential luminosity functions of bl lac objects that are beamed versions of the b2 radio galaxies to be @xmath35 ( this result is independent of the value of @xmath27 ) . the observed slopes are much steeper than this , corresponding to a parent population with an unbeamed slope @xmath36 . since the slope of the luminosity function of higher - luminosity radio galaxies is more similar to 2 , we might be able to account for this by saying that these bl lacs are beamed versions of more luminous sources than the b2s . however , the extended radio luminosities of the bl lacs are mostly fri - like , and the luminosity functions in radio and x - ray of the two populations overlap . either the b2s , in spite of these similarities , are not the parent population of these objects , or the analysis of urry & schafer is not applicable . urry & schafer predicted the luminosity functions of beamed populations using simple mathematical arguments . an alternative approach is to use monte carlo simulations to generate observations of such sources , based on the information we have on the parent population , and to generate luminosity functions from this simulated population . in this section we adopt that approach . the initial model we adopt is a simple one . each monte - carlo source is assigned a low - frequency luminosity drawn from the power - law fit to the 408-mhz luminosity function of the b2 radio galaxies ( fig.[408-lf ] ) , assuming that luminosities lie between @xmath37 and @xmath38 w hz@xmath24 . this luminosity range matches the b2 observations , and its lower bound is close to the lower limit on the luminosity of agn found when the radio - galaxy luminosity function is constructed for large samples ( condon 1989 ; machalski & godlowski 2000 ) . however , it is not clear whether the lack of radio galaxies below this luminosity indicates a real physical lower limit on radio - galaxy luminosity , or whether it simply becomes impossible to distinguish them from the much more numerous starbursts ; for example , sadler et al . ( 2001 ) , using an optical approach to classification , find no sign of a turnover , with a luminosity function going as low as a few @xmath39 w hz@xmath24 at 408 mhz . we comment below ( section [ matching ] ) on the effects on unified schemes of allowing the radio - galaxy luminosity function to extend to lower luminosities . the rest - frame luminosity of the core in both radio and x - rays is assumed to be a fixed fraction @xmath40 of the total low - frequency luminosity ; the observed core luminosity depends on @xmath40 and on the beaming lorentz factor and the ( randomly chosen ) angle to the line of sight . this implies that the ratio of core luminosity to extended luminosity , known as the ` core prominence ' , for a given source is a function only of the doppler factor @xmath41^{-1}$ ] , where @xmath42 is the lorentz factor , @xmath43 , and @xmath44 is the angle to the line of sight and of the ` intrinsic prominence ' , the core prominence in its rest frame . this simple model has been used previously to constrain core lorentz factors ( e.g. , morganti et al . 1995 ; hardcastle et al.1999 ) . later we will consider the effects of introducing some dispersion into the relation between rest - frame core and total low - frequency luminosity . to use this model to generate luminosity functions , we need to constrain @xmath40 and @xmath42 . we do this by examining the distribution of radio and x - ray core prominences in the b2 sample , using the total 408-mhz luminosity as the normalizing quantity . this dispersion is quite broad in both bands ( fig . [ prom ] ) , so that large values of @xmath42 are required . we use the approach of hardcastle et al . ( 1999 ) , who used the k - s test to investigate the consistency between the probability distribution implied by the data and that of a beaming model for a given value of @xmath42 and intrinsic prominence . for the lorentz factors ( @xmath45 ) required from observations of superluminal motion in bl lac objects , fitting in small samples results in a degeneracy between @xmath40 and @xmath42 ; typically , the best - fitting values are constrained to lie along a line in @xmath40-@xmath42 space which can be shown , by integrating the doppler beaming expression over all angles , to be @xmath46\right \}^{-(2+\alpha ) } \approx \bar f_{\rm o}\ ] ] where @xmath47 is the mean observed prominence at the frequency of interest . only with samples of @xmath48 sources could there be enough sources in the highly beamed tail of the distribution to set upper and lower limits on @xmath42 . using the present sample , we find that the data are best fit with @xmath49 for both the radio and x - ray . in what follows , we choose values of @xmath42 which obey this constraint , and corresponding values of @xmath40 derived from our fits . we simulated observations of bl lac sources with sky coverage and flux limits similar to those of the real samples . we began by generating samples matched to the b2 sample itself ; we verified that we obtained 408-mhz , 5-ghz and 1-kev luminosity functions which were consistent with the observations ( and with the analytical results of urry & schafer ) and that the luminosity functions obtained were unaffected if we extended the range of simulated luminosity beyond that observed in the b2 sample . to simulate the optical magnitude limit , we assigned optical absolute magnitudes to the simulated sources based on their radio luminosities and the observed weak correlation between radio luminosity and absolute magnitude . next we simulated 1-jy and emss samples . the simulated 1-jy sample was selected on the basis of radio core flux density and did not use the optical magnitude limit of the real sample ; this is acceptable , since the optical limit in the real data only dominates the @xmath16 calculation in 3/34 objects . the simulated emss sample was selected on x - ray core flux density and included the effects of the variable sensitivity of the emss . in both cases , we assume that all objects which meet the flux density selection criterion are bl lac objects , given that they have been generated from the b2 parent population and that they are all strongly core - dominated as a result of the selection criterion ( we find that , even for the lowest @xmath42 values , almost all the selected objects are more strongly core - dominated than the simulated or real b2 sources , which of course include some bl lac objects themselves , and that their core prominences at 5 ghz are comparable to those of real bl lac objects ) . this assumption may be an important source of error in the simulations , and we return to it below ( section [ conclusion ] ) . we also assume no evolution in the fri parent population , since the amount of evolution is known to be small at low redshifts . we discuss the possible effects of evolution in section [ conclusion ] . several interesting results emerge from these simulations . most strikingly , the slopes of the simulated core luminosity functions for both radio- and x - ray selected objects are systematically steeper than the values predicted by the models of urry & schafer , and more consistent with the measured slopes ( figs [ 5-lfbl ] and [ 1-lfbl ] ) . it appears that the analysis of urry & schafer does not apply _ if the beamed sample is selected on beamed rather than unbeamed flux _ ; it is only applicable if the selection flux is unbeamed ( as for the b2 objects ) . p2.9cmrrrrrrrrr model & & & + & number&slope&intercept&number&slope&intercept&number&slope&intercept + @xmath50 , no scatter&@xmath51 @xmath52&@xmath53 @xmath54&@xmath55 @xmath56&@xmath57 @xmath58&@xmath59 @xmath60&@xmath61 @xmath62&@xmath63 @xmath52&@xmath64 @xmath65&@xmath66 @xmath67 + @xmath68 , no scatter&@xmath69 @xmath70&@xmath53 @xmath54&@xmath71 @xmath72&@xmath73 @xmath74&@xmath75 @xmath76&@xmath77 @xmath78&@xmath79 @xmath52&@xmath80 @xmath81&@xmath82 @xmath83 + @xmath84 , no scatter&@xmath85 @xmath70&@xmath86 @xmath87&@xmath88 @xmath89&@xmath90 @xmath91&@xmath92 @xmath93&@xmath94 @xmath95&@xmath96 @xmath97&@xmath98 @xmath99&@xmath100 @xmath101 + @xmath102 , 0.25 dex scatter&@xmath51 @xmath103&@xmath104 @xmath105&@xmath106 @xmath107&@xmath108 @xmath52&@xmath109 @xmath110&@xmath111 @xmath112&@xmath113 @xmath74&@xmath114 @xmath60&@xmath115 @xmath116 + @xmath102 , 0.4 dex scatter&@xmath117 @xmath74&@xmath118 @xmath119&@xmath120 @xmath121&@xmath122 @xmath74&@xmath123 @xmath81&@xmath124 @xmath125&@xmath51 @xmath103&@xmath126 @xmath127&@xmath128 @xmath129 + @xmath102 , 0.5 dex scatter&@xmath130 @xmath52&@xmath86 @xmath81&@xmath88 @xmath131&@xmath132 @xmath133&@xmath134 @xmath99&@xmath135 @xmath101&@xmath136 @xmath103&@xmath137 @xmath138&@xmath139 @xmath140 + @xmath68 , 0.25 dex scatter&@xmath130 @xmath133&@xmath141 @xmath87&@xmath142 @xmath143&@xmath144 @xmath52&@xmath64 @xmath54&@xmath145 @xmath146&@xmath147 @xmath133&@xmath148 @xmath149&@xmath150 @xmath151 + @xmath68 , 0.4 dex scatter&@xmath122 @xmath52&@xmath152 @xmath87&@xmath153 @xmath143&@xmath154 @xmath155&@xmath156 @xmath93&@xmath157 @xmath158&@xmath159 @xmath70&@xmath160 @xmath127&@xmath161 @xmath162 + @xmath68 , 0.5 dex scatter&@xmath51 @xmath74&@xmath163 @xmath99&@xmath164 @xmath165&@xmath166 @xmath167&@xmath168 @xmath93&@xmath169 @xmath158&@xmath170 @xmath91&@xmath171 @xmath172&@xmath173 @xmath174 + @xmath68 , luminosity - dependent @xmath8&@xmath85 @xmath52&@xmath118 @xmath81&@xmath175 @xmath176&@xmath73 @xmath70&@xmath168 @xmath54&@xmath177 @xmath146&@xmath178 @xmath179&@xmath180 @xmath181&@xmath182 @xmath183 + details of the models used are given in the text . for each model the mean result of a number of simulations is quoted , together with the standard deviation of that parameter in parentheses . slopes and intercepts are for the radio luminosity function only ; the intercept is in units of @xmath184gpc@xmath185 w@xmath24 hz@xmath186 . since the b2 objects are selected on unbeamed flux , their numbers in simulated samples are expected to be independent of beaming parameters , and are included here for completeness only . these results are insensitive to the choice of cosmological parameters . for comparison , the total numbers of b2 , 1-jy and emss objects in the real data are 47 , 34 , and 41 respectively , and the slopes of their radio core luminosity functions are @xmath187 , @xmath188 , and @xmath189 respectively . we can now try to vary the parameters of these simulations to obtain a better match to the observed data . there are several features of the observed samples which we would like our simulated samples to match , including : 1 . the total number of sources 2 . the slope and range of the luminosity functions 3 . the redshift distribution , and 4 . the distribution of @xmath8 . in what follows we concentrate on reproducing the 5-ghz radio luminosity functions , as the data in this waveband have smaller uncertainties than the x - ray data and are less strongly affected by factors such as variability the inputs we are using necessarily make the simulated b2 sample match the data on criteria ( i ) , ( ii ) and ( iii ) , and this match is relatively insensitive to beaming parameters , since the b2s are not selected on beamed emission . the simulated luminosity functions for both x - ray and radio - selected bl lac objects in our models overlap with the luminosity function for the ( simulated ) b2 objects , as in the real data . the degree to which they overlap in luminosity , and the number of sources in the simulated bl lac samples , is quite a strong function of the beaming lorentz factor @xmath42 ( and its associated @xmath40 values ) . for @xmath50 , the simulations slightly underpredict the observed numbers of sources in both bl lac samples ( table [ numbers ] ) , and they also predict considerably lower minimum and maximum luminosities than are observed . with @xmath84 , we over - predict the observed numbers of sources , but do a better job of predicting their luminosity bounds . the best results come from simulations with intermediate values of @xmath42 , for example @xmath190 ( fig . [ sim3 ] ) . this simulation predicts approximately the right total number of bl lacs in the two surveys , and it also matches the luminosity range and the slope of the luminosity function , so that criteria ( i ) and ( ii ) are satisfied . it does not produce enough high - luminosity , high - redshift 1-jy bl lac objects , so that we can not claim to have satisfied criterion ( iii ) for this sample ; the redshift distributions of the real and simulated data are shown in fig.[redshifts ] . however , it is known that at least some of the 1-jy sample are too luminous in extended emission to be fris and have frii - like radio structure ; such objects can not be generated in our simulations . criterion ( iii ) seems to be satisfied for the simulated emss samples ( fig . [ redshifts ] ) . the numbers of bl lacs produced in the simulations do not depend very strongly on the adopted lower limit on the total . this tells us that relatively few of the simulated bl lac objects come from intrinsically low - luminosity radio galaxies , on the reasonable assumption that these objects do not have more strongly beamed cores than the b2s . this lends support to the idea that the b2 radio galaxies ( together , in the case of the 1-jy radio sources , with more luminous radio galaxies ) are indeed representative of the parent population for these samples . these simulations fail to reproduce the observed distinction between the luminosity functions of the x - ray and radio - selected bl lac objects . this is not surprising : in the model we have considered so far , the radio and x - ray cores are both a fixed fraction of the total low - frequency luminosity , with the result that selection in the radio is effectively identical to selection in the x - ray . in other words , the radio - to - x - ray spectral index of the simulated sources , @xmath8 , is approximately fixed , which is unrealistic . [ alphas ] shows the distribution of @xmath191 for the real and simulated data . criterion ( iv ) is not met either for the simulated b2s or for the simulated bl lac samples . in the next section of the paper we discuss ways of reproducing the observed differences in the slopes of the luminosity functions in different bands . a simple way of producing a distinction between the two bl lac samples is to assume that there is some dispersion , uncorrelated with intrinsic luminosity , in the intrinsic ( rest - frame ) relationship between x - ray and radio core flux , reflecting source - to - source variation in , for example , the structure of the jet or the size and age of the extended emission . this is distinct from , but in practice has very similar effects to , introducing an uncorrelated dispersion in the beaming lorentz factors for the radio and x - ray components . since the most luminous sources at any given redshift are most likely to be selected , we expect sources selected in the radio to be typically more radio - luminous than sources in the x - ray , and vice versa . we can use our simulations to investigate how great this dispersion should be . a starting point is the estimated logarithmic dispersion about the regression line in the radio - galaxy radio - x - ray core luminosity relationship ( canosa et al . 1999 ) , which was about 0.2 decades . this is an underestimate of the dispersion in the _ rest - frame _ relationship , because common correlation of the x - ray and radio core luminosities with the intrinsic luminosity of the source tends to stretch out the observed correlation . but even 0.2 dex of dispersion begins to recover the differences between the luminosity functions of the two bl lac samples . larger dispersions produce larger differences between the slopes of the luminosity functions and the distributions of @xmath8 , but also produce more sources ( table [ numbers ] ) . the models which come closest to satisfying criteria ( i ) , ( ii ) and ( iv ) above for the radio luminosity function seem to be those which combine a low @xmath42 with a moderate dispersion ; the luminosity functions for a representative model , with @xmath102 and a dispersion of 0.4 dex in the intrinsic radio - x - ray relation , are plotted in fig . [ sim4 ] ( compare fig . [ 5-lfbl ] ) and the distribution of spectral indices for this model is plotted in fig.[alphas ] . these models also replicate reasonably well the slope and range of the x - ray luminosity function for the simulated emss sample . however , we are not able to reproduce the observed steep slope of the 1-jy x - ray luminosity function ( fig . [ 1-lfbl ] ) . it is of interest to ask whether an asymmetrical dispersion in the radio - x - ray relationship can reproduce the observations in the same way . this would correspond to the situation in which the radio - selected objects are the tail of a population of x - ray - dominant objects , or vice versa . we investigated this by allowing either the radio or the x - ray intrinsic prominence to have a dispersion of 0.5 dex ( @xmath192 ) above the mean values determined from the radio galaxies ; this corresponds to radio - dominance or x - ray - dominance . unsurprisingly , such models do not reproduce the difference between the luminosity functions in both bands ; for example , if the x - rays have no dispersion in intrinsic prominence , then we see little difference between the x - ray luminosity functions of radio - selected and x - ray - selected objects , although their radio luminosity functions are different . we need dispersion in both bands to reproduce the differences in their luminosity functions . it has been suggested ( section [ intro ] ) that the differences between x - ray and radio - selected objects can be attributed to different lorentz factors for the regions emitting in the two wavebands . our simulations show that the observed correlation between x - ray and radio core luminosity for the b2 objects ( canosa et al.1999 ) is not consistent with isotropic x - ray emission , but we can not rule out models in which the x - ray nuclei are beamed with a lorentz factor considerably lower than that of the radio , as has been proposed to account for the less extreme properties of x - ray selected objects . values of @xmath193 do not produce enough emss bl lacs in our model ; if the lorentz factors were this low , then these sources would have to have a different parent population . with @xmath194 , @xmath195 we do not see any significant difference between the slopes or normalizations of the populations in the two samples . different lorentz factors can not alone be responsible for the difference in the observed luminosity function slopes and normalizations , although , by the same token , we can not use the observed luminosity functions to rule out mildly differing lorentz factors , since intrinsic dispersion in the radio - x - ray relation would obscure their effects . a further way to try to reproduce the difference between the radio- and x - ray - selected bl lac samples would be to include some luminosity dependence into the x - ray / radio relationship , as described by fossati et al . the simplest way to parametrise this would be in terms of an @xmath196 which varies with luminosity . unfortunately , the correlations in fossati et al . are parametrised in terms of 5-ghz luminosity , which is itself a beamed quantity ; for our purposes we need some relation between the shape of the sed and the beam power as parametrised by the low - frequency unbeamed luminosity . to test whether this kind of model can reproduce the observed difference in the luminosity functions of the different samples , we made the intrinsic x - ray flux fraction a function of intrinsic 408-mhz luminosity , while keeping the intrinsic radio flux fraction constant . we chose a dependence on luminosity such that @xmath8 increases linearly with luminosity , going from 0.6 in the least luminous sources to 1.0 in the most luminous sources we observed ; this approximately matches the magnitude of the range in @xmath8 seen by fossati et al . ( see , e.g. , their figure 8) , though over a considerably smaller range in radio luminosity . this has some unrealistic effects for example , because the majority of sources are low - luminosity and flat - spectrum , the model vastly overpredicts the number of x - ray - selected sources ( table [ numbers ] ) and also predicts an inconsistent @xmath8 distribution . however , it does give rise to clear differences between the luminosity functions of the two simulated bl lac samples , in the same sense as , though of smaller magnitude than , the differences seen in the data . we conclude that luminosity dependence of the x - ray / radio relationship is probably not in itself enough to explain the differences in the luminosity functions of the two populations . in this paper we have been trying to determine to what extent and under what assumptions b2 radio galaxies ( or , more accurately , the radio galaxy population represented by the b2 radio galaxies ) can be the parent population of two well - studied samples of bl lac objects ( the emss and 1-jy samples ) . following earlier work , we have characterized the properties of the populations using their luminosity functions : luminosity functions for the b2 and bl lac samples were constructed in sections [ b2lum ] and [ bllum ] . in section [ usmodl ] we discussed the analytical results of urry & schafer ( 1984 ) ; we argued that their models can not properly be applied to luminosity functions of samples with different selection criteria . instead , in sections [ mc ] and [ sim ] , we used monte carlo simulations to set up a parent population of beamed radio galaxies with b2-like properties and drew objects from them with selection criteria which as closely as possible matched those actually used in generating the emss and 1-jy samples . by adjusting the unknown parameters of the simulation ( section [ matching ] ) we were able to show that our simulated observations of a beamed population of radio galaxies matched to the b2 sample were able to reproduce reasonably well the observed numbers of bl lac objects and the slopes of the radio- and x - ray - selected bl lac luminosity functions , if low values ( @xmath197 ) of the beaming lorentz factor @xmath42 are adopted . in section [ dispersion ] we found that the introduction of a moderate dispersion in the rest - frame ratio of core x - ray to core radio luminosity ( for which there is some direct evidence in the observations of the b2 radio galaxies ) can help to explain the observed differences in the luminosity functions and radio - to - x - ray spectral indices of radio - selected and x - ray - selected bl lacs . this approach gives better results than other proposed ways of generating the differences between bl lac populations ( sections [ asymm ] [ ldep ] ) . these results therefore support a model in which the two apparently different bl lac populations are simply extreme objects drawn from a single parent population ( cf . laurent - muehleisen et al.1999 ) , although we have not attempted to reproduce all of the observed differences between radio - selected and x - ray - selected bl lacs , such as the apparent differences in cosmological evolution ( section [ intro ] ) . the beaming lorentz factors used here are much smaller than those inferred from superluminal motion or @xmath42-ray transparency in some bl lac objects . these typically require @xmath198 , which would grossly overpredict the number of bl lac objects that should be observed above the flux limit ( cf . table [ numbers ] ) . chiaberge et al . ( 2000 ) have previously pointed out that the nuclear luminosities of fri radio galaxies are a factor 1010@xmath199 too bright to be consistent with the ` de - beaming ' of bl lacs using these high lorentz factors . the solution they adopt , velocity structure in the nuclear jet , seems plausible in view of what we know about the existence of velocity - structures in the _ kiloparsec_-scale jets ( e.g. , laing 1996 ) . velocity structure in the jets should not significantly affect our general conclusions here . if jets have velocity structure , our values of @xmath42 parametrize the relationship between the observed luminosity and the angle to the line of sight , rather than describing real physical bulk velocities . however , the details of this parameterization make a difference to the predicted numbers of bl lac objects and the range in their luminosities . although attempting to determine a typical jet emissivity / velocity profile is beyond the scope of this paper , it is certainly the case that luminosity functions can be used to help to constrain it . the simulated observations fail to reproduce the real data in two important ways . firstly , although the simulated emss sample is a good match in redshift distribution to the observed objects ( fig.[redshifts ] , the simulated 1-jy sample contains far too few high - redshift objects . as discussed in section [ intro ] , the emss objects seem all to be fri - like in their radio structure and luminosity , while it is known that some 1-jy objects are frii - like ; since friis have higher luminosities than the parent population we use , it is not surprising that some of the 1-jy sample appear at higher redshifts . our models predict that essentially all the 1-jy objects whose parent population are fris should have redshifts less than @xmath200 ( fig . [ redshifts ] ) , which agrees well with the observations of rector & stocke ( 2001 ) . the fact that some of the 1-jy objects may have a different parent population may also help to explain the anomalously steep 1-jy x - ray luminosity function . secondly , the simulations predict a considerably larger number of bl lac objects than is observed ( by factors @xmath201 ) for @xmath202 or if a significant dispersion in intrinsic @xmath8 is introduced ; even the models which best reproduce other features of the population , such as the differences between the slopes of the x - ray and radio luminosity functions , overpredict the numbers of 1-jy bl lacs by a factor @xmath35 . the mismatch in numbers becomes even greater if some of the observed high - redshift 1-jy objects are not drawn from an fri parent population , as discussed above . we can attribute at least some of this effect to the optical selection criteria applied when defining bl lac samples . as pointed out by march et al . ( 1996 ) , a definition in terms of the strength of the 4000- break measures the strength of the optical non - thermal emission in terms of the starlight in the host galaxy , which may have very little to do with the agn , while a condition on the equivalent width of the strongest emission lines has little physical justification ( and may exclude objects which are in all other ways bl lac objects ; cf . vermeulen et al . 1995 ) . so some of the predicted objects may be present in the surveys , identified as something other than bl lacs . however , a search in the emss survey for objects intermediate between fris and traditional emss bl lacs ( rector et al . 1999 ) found relatively few candidates , and only a handful of emss sources are directly identified with radio galaxies . it remains possible that some of our simulated sources are identified as groups or clusters in the emss survey . if , as new _ chandra _ results ( section [ chandra - caveat ] ) suggest , we are overestimating the radio galaxies nuclear x - ray fluxes ( section [ chandra - caveat ] ) then we will also have overestimated the intrinsic x - ray core prominence ; correcting for this would result in a ( probably small ) reduction in the predicted total number of emss sources . in the 1-jy sample , there are additional optical magnitude and radio spectral selection criteria which are not modelled in our simulations , and so there is more scope for ` hiding ' the excess sources . on the other hand , we are extrapolating our b2 luminosity function to high redshifts to generate bl lac objects without including the effects of cosmological evolution of the fri population , now reasonably well - established ( e.g. , waddington et al . 2001 ) . because there is little evidence for evolution below @xmath203 , where most of our bl lac candidates are generated ( fig . [ redshifts ] ) , this does not have a strong effect on our models , but there may be up to an order of magnitude increase in the numbers of the most luminous sources at @xmath204 . taking this into account would lead our models to produce @xmath205 additional high - redshift 1-jy objects . without a detailed model for source evolution , we can not make this more quantitative . combining these factors with the large statistical uncertainties on the predicted and actual numbers of objects , we regard the degree of agreement between the simulations and observations as encouraging . it supports a model in which fri radio galaxies in the luminosity range of the b2 bright sample are the parent population of all the emss x - ray - selected bl lac objects , and of @xmath206 per cent of the radio - selected 1-jy objects . a higher - luminosity population , probably beamed friis , must be responsible for the remaining , higher - redshift , 1-jy objects . we are grateful for support from nasa grant nag 5 - 1882 and from pparc . cmc thanks the university of bristol for a research studentship . abraham , r.g . , mchardy , i.m . , crawford , c.s . , 1991 , mnras , 252 , 482 antonucci , r.r.j . , ulvestad , j.s . , 1985 , apj , 294 , 158 baars , j.w.m . , genzel , r. , pauliny - toth , i.i.k . , witzel , a. , 1977 , a&a , 61 , 99 browne , i.w.a . , 1983 , mnras , 204 , 23p canosa , c.m . , worrall , d.m . , hardcastle , m.j . , birkinshaw , m. , 1999 , mnras , 310 , 30 celotti , a. , maraschi , l. , ghisellini , g. , caccianiga , a. , maccacaro , t. , 1993 , apj , 416 , 118 chiaberge , m. , celotti , a. , capetti , a. , ghisellini , g. , 2000 , a&a , 358 , 104 colla , g. , et al . , 1970 , a&as , 1 , 281 colla , g. , fanti , c. , fanti , r. , gioia , i. , lari , c. , lequeux , j. , lucas , r. , ulrich , m.h . , 1975 , a&as , 20 , 1 condon , j.j . , 1989 , apj , 338 , 13 cotton , w.d . , ferretti , l. , giovannini , g. , venturi , t. , lara , l. , marcaide , j. , wehrle , a.e . , 1995 , apj , 452 , 605 fabbiano , g. , miller , l. , trinchieri , g. , longair , m. , elvis , m. , 1984 , apj , 277 , 115 falomo , r. , 1996 , mnras , 283 , 241 fanti , c. , fanti , r. , de ruiter , h.r . , parma , p. , 1987 , a&as , 69 , 57 fossati , g. , celotti , a. , ghisellini , g. , maraschi , l. , 1997 , mnras , 289 , 136 fossati , g. , maraschi , l. , celotti , a. , comastri , a. , ghisellini , g. , 1998 , mnras , 299 , 433 ghisellini , g. , maraschi , l. , 1989 , apj , 340 , 181 giommi , p. , padovani , p. , 1994 , mnras , 268 , 51 giovannini , g. , feretti , l. , gregorini , l. , parma , p. , 1988 , a&a , 199 , 73 giovannini , g. , taylor , g.b . , arbizzani , e. , bondi , m. , cotton , w.d . , feretti , l. , lara , l. , venturi , t. , 1999 , apj , 522 , 101 hardcastle , m.j . , alexander , p. , pooley , g.g . , riley , j.m . , 1999 , mnras , 304 , 135 hardcastle , m.j . , worrall , d.m . , 1999 , mnras , 309 , 969 hardcastle , m.j . , worrall , d.m . , 2000 , mnras , 314 , 359 hardcastle , m.j . , worrall , d.m . , birkinshaw , m. , laing , r.a . , bridle , a.h . , 2002 , mnras , 334 , 182 jannuzi , b.t . , smith , p.s . , elston , r. , 1994 , apj , 428 , 130 kellermann , k.i . , 1964 , aj , 69 , 205 kollgaard , r.i . , wardle , j.f.c . , roberts , d.h . , gabuzda , d.c . , 1992 , aj , 104 , 1687 laing , r.a . , 1996 , in hardee p.e . , bridle a.h . , zensus j.a . , eds , energy transport in radio galaxies and quasars , asp conference series vol . 100 , san francisco , p. 241 laing , r.a . , parma , p. , de ruiter h.r . , fanti , r. , 1999 , mnras , 306 , 513 laurent - muehleisen , s.a . , kollgaard , r.i . , moellenbrock , g.a . , feigelson , e.d . , 1993 , aj , 106 , 875 laurent - muehleisen , s.a . , kollgaard , r.i . , feigelson , e.d . , brinkmann , w. , siebert , j. , 1999 , apj , 525 , 127 machalski , j. , godlowski , w. , 2000 , a&a , 360 , 463 maraschi , l. , ghisellini , g. , tanzi , e.g. , treves , a. , 1986 , apj , 310 , 325 march , m.j.m . , browne , i.w.a . , impey , c.d . , smith , p.s . , 1996 , mnras , 281 , 425 morganti , r. , oosterloo , t.a . , fosbury , r.a.e . , tadhunter , c.n . , 1995 , mnras , 274 , 393 morris , s.l . , stocke , j.t . , gioia , i.m . , schild , r.e . , wolter , a. , maccacaro , t. , della ceca r. , 1991 , apj , 380 , 49 padovani , p. , giommi , p. , 1995 , apj , 444 , 567 padovani , p. , urry , c.m . , 1990 , apj , 356 , 75 padovani , p. , urry , c.m . , 1991 , apj , 368 , 373 perlman , e.s . , stocke , j.t . , 1993 , apj , 406 , 430 perlman , e.s . , stocke , j.t . , 1994 , aj , 106 , 56 pesce , j.e . , falomo , r. , treves , a. , 1995 , aj , 110 , 1554 rector , t.a . , stocke , j.t . , perlman , e.s . , morris , s.l . , gioia , i.m . , 2000 , aj , 120 , 1626 rector , t.a . , stocke , j.t . , 2001 , aj in press [ astro - ph/0105100 ] sadler , e.m . , et al . , 2001 , mnras submitted , astro - ph/0106173 sambruna , r.m . , maraschi , l. , urry , c.m . , 1996 , apj , 463 , 444 smith , e.p . , odea , c.p . , baum , s.a . , 1995 , apj , 441 , 113 stickel , m. , padovani , p. , urry , c.m . , fried , j.w . , khr , h. , 1991 , apj , 374 , 431 stocke , j.t . , liebert , j. , schmidt , g. , gioia , i.m . , maccacaro , t. , schild , r.e . , maccagni , d. , arp , h.c . , 1985 , apj , 298 , 619 trussoni , e. , massaglia , s. , ferrari , r. , fanti , r. , feretti , l. , parma , p. , brinkmann , w. , 1997 , a&a , 327 , 27 ulrich , m .- h . , 1989 , in maraschi l. , maccacaro t. , ulrich m .- h . , eds , bl lac objects , springer - verlag , heidelberg , p. 47 urry , c.m . , padovani , p. , 1991 , apj , 371 , 60 urry , c.m . , padovani , p. , 1995 , pasp , 107 , 803 urry , c.m . , padovani , p. , stickel , m. , 1991 , apj , 382 , 501 urry , c.m . , schafer , r.a . , 1984 , apj , 280 , 569 urry , c.m . , sambruna , r.m . , worrall , d.m . , kollgaard , r.i . , feigelson , e.d . , perlman , e.s . , stocke , j.t . , 1996 , apj , 463 , 424 venturi , t. , feretti , l. , giovannini , g. , comoretto , g. , wehrle , a.e . , 1993 , apj , 408 , 81 vermeulen , r.c . , ogle , p.m. , tran , h.d . , browne , i.w.a . , cohen , m.h . , readhead , a.c.s . , taylor , g.b . , goodrich , r.w . , 1995 , apj , 452 , l5 waddington , i. , dunlop , j.s . , peacock , j.a . , windhorst , r.a . , 2001 , mnras submitted ( astro - ph/0108048 ) wall , j.v . , peacock , j.a . , 1985 , mnras , 216 , 173 wolter , a. , gioia , i.m . , maccacaro , t. , morris , s.l . , stocke , j.t . , 1991 , apj , 369 , 314 wolter , a. , caccaniga , a. , della ceca r. , maccacaro , t. , 1994 , apj , 433 , 29 worrall , d.m . , birkinshaw , m. , 2000 , apj , 530 , 719 worrall , d.m . , birkinshaw , m. , hardcastle , m.j . , 2001 , mnras , 326 , l7 wurtz , r. , stocke , j.t . , yee , m.k.c . , 1996 , apjs , 103 , 109 wurtz , r. , stocke , j.t . , ellingson , e. , 1997 , apj , 480 , 547

in a recent work , r. pandharipande , j. p. solomon and the second author have initiated a study of the intersection theory on the moduli space of riemann surfaces with boundary . they conjectured that the generating series of the intersection numbers satisfies the open kdv equations . in this paper we prove this conjecture . our proof goes through a matrix model and is based on a kontsevich type combinatorial formula for the intersection numbers that was found by the second author .

1501.07888

the study of the intersection theory on the moduli space of riemann surfaces with boundary ( often viewed , with the boundary removed , as open riemann surfaces ) was recently initiated in @xcite . the authors constructed a descendent theory in genus @xmath0 and obtained a complete description of it . in all genera , they conjectured that the generating series of the descendent integrals satisfies the open kdv equations . this conjecture can be considered as an open analog of the famous witten s conjecture @xcite . a higher genus construction of the open descendent theory will appear in @xcite . a combinatorial formula for the open intersection numbers in all genera was found in @xcite . in this paper , using the combinatorial formula from @xcite , we present a matrix integral for the generating series of the open intersection numbers . then applying some analytical tools to this matrix integral , we prove the main conjecture from @xcite . the introduction is organized as follows . in section [ wittc ] we briefly recall the original conjecture of e. witten from @xcite . in section [ konts ] we recall kontsevich s combinatorial formula and kontsevich s proof ( @xcite ) of witten s conjecture . section [ okdv ] contains a short account of the main constructions and conjectures in the open intersection theory from @xcite . section [ ocomb ] describes the combinatorial formula of @xcite for the open intersection numbers of @xcite . throughout this text @xmath1 $ ] will denote the set @xmath2 a compact riemann surface is a compact connected smooth complex curve . given a fixed genus @xmath3 and a non - negative integer @xmath4 the moduli space of all compact riemann surfaces of genus @xmath3 with @xmath5 marked points is denoted by @xmath6 p. deligne and d. mumford defined a natural compactification of it via stable curves in @xcite in 1969 . given @xmath7 as above , a stable curve is a compact connected complex curve with @xmath5 marked points and finitely many singularities , all of which are simple nodes . the collection of marked points and nodes is the set of special points of the curve . we require that all the special points are distinct and that the automorphism group of the curve is finite . the moduli of stable marked curves of genus @xmath3 with @xmath5 marked points is denoted by @xmath8 and is a compactification of @xmath6 it is known that this space is a non - singular complex orbifold of complex dimension @xmath9 for the basic theory the reader is referred to @xcite . in his seminal paper @xcite , e. witten , motivated by theories of @xmath10-dimensional quantum gravity , initiated new directions in the study of @xmath8 . for each marking index @xmath11 he considered the tautological line bundles @xmath12 whose fiber over a point @xmath13\in { { \overline{\mathcal{m}}}}_{g , l}\ ] ] is the complex cotangent space @xmath14 of @xmath15 at @xmath16 . let @xmath17 denote the first chern class of @xmath18 , and write @xmath19 the integral on the right - hand side of is well - defined , when the stability condition @xmath20 is satisfied , all the @xmath21 s are non - negative integers , and the dimension constraint @xmath22 holds . in all other cases @xmath23 is defined to be zero . the intersection products are often called _ descendent integrals _ or _ intersection numbers_. note that the genus is uniquely determined by the exponents @xmath24 . let @xmath25 ( for @xmath26 ) and @xmath27 be formal variables , and put @xmath28 let @xmath29 be the generating function of the genus @xmath3 descendent integrals . the bracket @xmath30 is defined by the monomial expansion and the multilinearity in the variables @xmath25 . concretely , @xmath31 where the sum is over all sequences of non - negative integers @xmath32 with finitely many non - zero terms . the generating series @xmath33 is called the _ ( closed ) free energy_. the exponent @xmath34 is called the _ ( closed ) partition function_. put @xmath35 . witten s conjecture ( @xcite ) says that the closed partition function @xmath36 becomes a tau - function of the kdv hierarchy after the change of variables @xmath37 . in particular , it implies that the closed free energy @xmath38 satisfies the following system of partial differential equations : @xmath39 these equations are known in mathematical physics as the kdv equations . e. witten ( @xcite ) proved that the intersection numbers satisfy the string equation @xmath40 for @xmath41 . this equation can be rewritten as the following differential equation : @xmath42 e. witten also showed that the kdv equations together with the string equation actually determine the closed free energy @xmath38 completely . there was a later reformulation of witten s conjecture due to r. dijkgraaf , e. verlinde and h. verlinde ( @xcite ) in terms of the virasoro algebra . define differential operators @xmath43 , @xmath44 , by @xmath45 while for @xmath46 @xmath47 the virasoro equations say that the operators @xmath43 , @xmath48 , annihilate the closed partition function @xmath36 : @xmath49 it is easy to see that the virasoro equations completely determine all intersection numbers . r. dijkgraaf , e. verlinde and h. verlinde ( @xcite ) proved that this description is equivalent to the one given by the kdv equations and the string equation . witten s conjecture was proven by m. kontsevich @xcite . see @xcite for other proofs . kontsevich s proof @xcite of witten s conjecture consisted of two parts . the first part was to prove a combinatorial formula for the gravitational descendents . let @xmath50 be the set of isomorphism classes of trivalent ribbon graphs of genus @xmath3 with @xmath51 faces and together with a numbering @xmath52 $ ] . denote by @xmath53 the set of vertices of a graph @xmath54 . let us introduce formal variables @xmath55 , @xmath56 $ ] . for an edge @xmath57 let @xmath58 where @xmath11 and @xmath59 are the numbers of faces adjacent to @xmath60 . then we have @xmath61 the second step of kontsevich s proof was to translate the combinatorial formula into a matrix integral . then , by using non - trivial analytical tools and the theory of tau - functions of the kdv hierarchy , he was able to prove that @xmath36 is a tau - function of the kdv hierarchy and , hence , the free energy @xmath38 satisfies the kdv equations . in @xcite r. pandharipande , j. solomon and r.t . constructed an intersection theory on the moduli space of stable marked disks . let @xmath62 be the moduli space of stable marked disks with @xmath63 boundary marked points and @xmath5 internal marked points . this space carries a natural structure of a compact smooth oriented manifold with corners . one can easily define the tautological line bundles @xmath64 for @xmath65 $ ] , as in the closed case . in order to define gravitational descendents , as in , we must specify boundary conditions . indeed , given a smooth compact connected oriented orbifold with boundary , @xmath66 of dimension @xmath51 , the poincar - lefschetz duality shows that @xmath67 thus , given a vector bundle on a manifold with boundary , only _ relative euler class _ , relative to nowhere vanishing boundary conditions , can be integrated to give a number . the main construction in @xcite is a construction of boundary conditions for @xmath68 in @xcite , vector spaces @xmath69 of _ multisections _ of @xmath70 which satisfy the following requirements , were defined . suppose @xmath71 are non - negative integers with @xmath72 then 1 . [ it : canonic1 ] for any generic choice of multisections @xmath73 for @xmath74 the multisection @xmath75\\1\le j\le a_i } } s_{ij}\ ] ] vanishes nowhere on @xmath76 2 . [ it : canonic2 ] for any two such choices @xmath77 and @xmath78 we have @xmath79 where @xmath80 and @xmath81 is the relative euler class . the multisections @xmath82 , as above , are called _ canonical_. with this construction the open gravitational descendents in genus @xmath0 are defined by @xmath83 where @xmath84 is as above and @xmath77 is canonical . in a forthcoming paper @xcite , j. solomon and r.t . define a generalization for all genera . in @xcite a moduli space @xmath85 which classifies real stable curves with some extra structure is constructed . the moduli space @xmath85 is a smooth oriented compact orbifold with corners , of real dimension @xmath86 the stability condition is @xmath87 note that naively , without adding an extra structure , the moduli of real stable curves of positive genus is non - orientable . on @xmath85 one defines vector spaces @xmath88 , for @xmath65,$ ] for which the genus @xmath3 analogs of requirements [ it : canonic1],[it : canonic2 ] from above hold . write @xmath89 for the corresponding higher genus descendents . introduce one more formal variable @xmath77 . the _ open free energy _ is the generating function @xmath90 where @xmath91 , @xmath92 , and again we use the monomial expansion and the multilinearity in the variables @xmath93 the following initial condition follows easily from the definitions ( @xcite ) : @xmath94 in @xcite the authors conjectured the following equations : @xmath95 they were called the open string and the open dilaton equation correspondingly . put @xmath96 the main conjectures in @xcite are [ conjecture : open kdv ] the following system of equations is satisfied : @xmath97 in @xcite equations were called the open kdv equations . it is easy to see that @xmath98 is fully determined by the open kdv equations , the initial condition and the closed free energy @xmath38 . let @xmath99 be the _ open partition function_. in @xcite the authors introduced the following modified operators : @xmath100 where the operators @xmath43 were defined in section [ subsubsection : closed virasoro ] . [ conjecture : open virasoro ] the operators @xmath101 , @xmath44 , annihilate the open partition function : @xmath102 in @xcite equations were called the open virasoro equations . again it is easy to see that the open free energy @xmath98 is fully determined by the open virasoro equations , the initial condition @xmath103 and the closed free energy @xmath38 . from the closed string equation it immediately follows that the open string equation is equivalent to , for @xmath104 . moreover , from the equation @xmath105 it follows that the open dilaton equation @xmath106 is equivalent to , for @xmath107 . more precisely , in @xcite it was conjectured that there exists a definition of open intersection numbers for @xmath108 , for which the open kdv and open virasoro equations hold . the definition was later given in @xcite . although it was not clear at all that the open kdv and the open virasoro equations are compatible , in @xcite it was proved that they indeed have a common solution . for @xmath109 the conjectures were proved in @xcite . in @xcite the conjectures are proved for @xmath110 and the open string and the open dilaton equations are proved for all @xmath111 the main result of this paper is the following theorem . [ theorem : main ] conjectures [ conjecture : open kdv ] and [ conjecture : open virasoro ] are true . let @xmath112 be a power series in the variables @xmath113 with the coefficients from @xmath114 $ ] . in @xcite the following system of equations was introduced : @xmath115 it was called the half of the burgers - kdv hierarchy . this system is obviously stronger than the system of the open kdv equations . in @xcite it was actually shown that the half of the burgers - kdv hierarchy is equivalent to the open kdv equations together with equation . denote by @xmath116 a unique solution of system - specified by the initial condition @xmath117 in @xcite it was shown that @xmath116 satisfies the open kdv equations , the initial condition and the open virasoro equations . this proved the equivalence of the open analog of witten s conjecture and the open virasoro conjecture . this also shows that theorem [ theorem : main ] immediately implies the following corollary . the open free energy @xmath98 satisfies the half of the burgers - kdv hierarchy . consider more variables @xmath118 and let @xmath119 . let @xmath112 be a power series in the variables @xmath120 with the coefficients from @xmath114 $ ] . let us extend the half of the burgers - kdv hierarchy by the following equations : @xmath121 in @xcite the extended system - was called the ( full ) burgers - kdv hierarchy . let @xmath122 be a unique solution of it specified by the initial condition @xmath123 we obviously have @xmath124 . in @xcite it was proved that the function @xmath125 satisfies the following extended virasoro equations : @xmath126 where @xmath127 here we , by definition , put @xmath128 . in @xcite it was conjectured that , by adding descendents for boundary marked points , one can geometrically define intersection numbers which will be the coefficients of @xmath122 . in @xcite j. solomon an r.t . give a complete proposal for the construction in all genera , and it is proved to produce the correct intersection numbers in genus @xmath129 moreover , with this proposal the extended open free energy @xmath130 is defined , as well as the extended open partition function @xmath131 it is proved in @xcite that the following equations hold @xmath132 in @xcite , section 5.2 , it is shown that @xmath133 satisfies these equations as well . thus , theorem [ theorem : main ] implies that @xmath134 and we immediately obtain the following generalization of theorem [ theorem : main ] . [ theorem : extended main ] the extended open free energy @xmath130 is a solution of the full burgers - kdv hierarchy . from the recent result of a. alexandrov @xcite it also follows that the extended open partition function @xmath135 becomes a tau - function of the kp hierarchy after the change of variables @xmath37 and @xmath136 . in @xcite r.t . proved a combinatorial formula for the geometric models which were defined in @xcite . he showed that all these intersection numbers can be calculated as sums of amplitudes of diagrams which will be described below . in this paper a matrix model is constructed out of this combinatorial formula . using this matrix model we prove our main theorem [ theorem : main ] . a topological @xmath137-surface with boundary @xmath15 , is a topological connected oriented surface with non - empty boundary , @xmath63 boundary marked points @xmath138},$ ] and @xmath5 internal marked points @xmath139},$ ] and genus @xmath111 by genus we mean , as usual in the open theory , the doubled genus , that is , the genus of the doubled surface obtained by gluing two copies of @xmath15 along @xmath140 we require the stability condition @xmath141 let @xmath142 be non - negative integers such that @xmath143 be a finite set and @xmath144\to a$ ] a map . @xmath145 will be implicit in the definition . a _ @xmath137-ribbon graph with boundary _ is an embedding @xmath146 of a connected graph @xmath147 into a @xmath137-surface with boundary @xmath15 such that * @xmath148}\subseteq \iota(v(g))$ ] , where @xmath53 is the set of vertices of @xmath147 . we henceforth consider @xmath149 as vertices . * the degree of any vertex @xmath150 is at least @xmath151 . * @xmath152 . * if @xmath153 then @xmath154 } d_i,\ ] ] where each @xmath155 is a topological open disk , with @xmath156 . we call the disks @xmath155 faces . * if @xmath157 , then @xmath158 . the genus @xmath159 of the graph @xmath147 is the genus of @xmath15 . the number of the boundary components of @xmath147 or @xmath15 is denoted by @xmath160 and @xmath161 stands for the number of the internal vertices . we denote by @xmath162 the set of faces of the graph @xmath163 and we consider @xmath164 as a map @xmath165 by defining for @xmath166 where @xmath16 is the unique internal marked point in @xmath167 the map @xmath164 is called the labeling of @xmath168 denote by @xmath169 the set of boundary marked points @xmath148}.$ ] two ribbon graphs with boundary @xmath170 are isomorphic , if there is an orientation preserving homeomorphism @xmath171 and an isomorphism of graphs @xmath172 , such that 1 . @xmath173 2 . @xmath174 for all @xmath175.$ ] 3 . @xmath176 where @xmath177 are the labelings of @xmath178 respectively and @xmath179 is any face of the graph @xmath168 note that in this definition we do not require the map @xmath180 to preserve the numbering of the internal marked points . a ribbon graph is _ critical _ , if * boundary marked points have degree @xmath10 . * all other vertices have degree @xmath151 . * if @xmath181 then @xmath109 and @xmath182 such a component is called _ a ghost_. a _ nodal ribbon graph with boundary _ is @xmath183 , where * @xmath184 are ribbon graphs with boundary . * @xmath185 is a set of _ ordered _ pairs of boundary marked points @xmath186 of the @xmath187 s which we identify . after the identification of the vertices @xmath188 and @xmath189 the corresponding point in the graph is called a node . the vertex @xmath188 is called the legal side of the node and the vertex @xmath189 is called the illegal side of the node . * we also require the graph @xmath147 to be connected . we require that elements of @xmath190 are disjoint as sets ( without ordering ) . the set of edges @xmath191 is composed of the internal edges of the @xmath187 s and of the boundary edges . the boundary edges are the boundary segments between successive vertices which are not the illegal sides of nodes . for any boundary edge @xmath60 we denote by @xmath192 the number of the illegal sides of nodes lying on it . the boundary marked points of @xmath147 are the boundary marked points of @xmath187 s , which are not nodes . the set of boundary marked points of @xmath147 will be denoted by @xmath169 also in the nodal case . a nodal graph @xmath183 is _ critical _ , if * all of its components @xmath187 are critical . * any boundary component of @xmath187 has an odd number of points that are the boundary marked points or the legal sides of nodes . * ghost components do not contain the illegal sides of nodes . a nodal ribbon graph with boundary is naturally embedded into the nodal surface @xmath193 . the ( doubled ) genus of @xmath15 is called the genus of the graph . the notion of an isomorphism is also as in the non - nodal case . in figure [ nodal ] there is a nodal graph of genus @xmath0 , with @xmath194 boundary marked points , @xmath195 internal marked points , three components , one of them is a ghost , two nodes , where a plus sign is drawn next to the legal side of a node and a minus sign is drawn next to the illegal side . denote by @xmath196 the set of isomorphism classes of critical nodal ribbon graphs with boundary of genus @xmath3 , with @xmath63 boundary marked points , @xmath5 faces and together with a bijective labeling @xmath197,$ ] the combinatorial formula in @xcite is fix @xmath198 such that @xmath199 . let @xmath200 be formal variables . then we have @xmath201 where @xmath202 we thank a. alexandrov , l. chekhov , d. kazhdan , o. g. louidor , a. okounkov , r. pandharipande , p. rossi , s. shadrin , j. p. solomon and d. zvonkine for discussions related to the work presented here . a. b. was supported by grant erc-2012-adg-320368-mcsk in the group of r. pandharipande at eth zurich , the grants rffi 13 - 01 - 00755 and nsh-4850.2012.1 . r.t . was supported by isf grant 1747/13 and erc starting grant 337560 in the group of j. p. solomon at the hebrew university of jerusalem . part of the work was completed during the visit of a.b to the einstein institute of mathematics of the hebrew university of jerusalem in 2014 and during the visits of r.t . to the forschungsinstitut fr mathematik at eth zrich in 2013 and 2014 . in this section we present a matrix integral that is a starting point in our proof of theorem [ theorem : main ] . instead of deriving a matrix integral for the open partition function @xmath203 directly from the combinatorial formula , we write a matrix model for an auxiliary function @xmath204 that is a sum over non - nodal ribbon graphs with boundary with some additional structure . we then relate the function @xmath204 to the open partition function @xmath203 by an action of the exponent of some quadratic differential operator . the section is organized as follows . in section [ subsection : reformulation ] we give a slight reformulation of the combinatorial formula . in section [ subsection : sum over non - nodal graphs ] we introduce an auxiliary function @xmath204 and relate it to the open partition function @xmath203 . this relation is given by lemma [ lemma : relation ] . section [ subsection : brief recall ] contains a brief review of basic facts about the integration over the space of hermitian matrices . in section [ subsection : matrix integral ] we give a matrix integral for the function @xmath204 . this is the subject of proposition [ proposition : matrix integral ] . the matrix integral in this proposition is understood in the sense of formal matrix integration . in section [ subsection : convergent matrix integral ] we discuss how to make sense of it as a convergent integral . we fix an integer @xmath205 throughout this section and we set the genus parameter @xmath27 to be equal to @xmath206 here we reformulate the combinatorial formula . this step is completely analogous to what m. kontsevich did in @xcite ( see the proof of theorem 1.1 there ) . denote by @xmath207 the set of isomorphism classes of critical nodal ribbon graphs with boundary together with a labeling @xmath208 $ ] ( a coloring of faces in @xmath190 colors ) . for a graph @xmath209 and an edge @xmath210 , let @xmath211 introduce formal variables @xmath212 and consider the diagonal @xmath213 matrix @xmath214 from the combinatorial formula it follows that @xmath215 where @xmath216 and @xmath217 here @xmath218 stands for the number of the internal edges of the graph @xmath187 . here we introduce an auxiliary function @xmath204 and relate it to the open partition function @xmath203 . we denote by @xmath219 the ring of formal series of the form @xmath220 where @xmath221 is a rational function in @xmath212 homogeneous of degree @xmath222 . we denote by @xmath223 the subspace that consists of series of the form @xmath224 , where @xmath221 is a rational function in @xmath212 homogeneous of degree @xmath222 . note that the ring @xmath225 $ ] can be naturally considered as a subring of @xmath219 . let us introduce the following auxiliary set of graphs . denote by @xmath226 the set of isomorphism classes of critical non - nodal ribbon graphs with boundary , that are not ghosts , together with * a labeling @xmath227 $ ] , * a map @xmath228 $ ] ; such that on each boundary component of @xmath147 the number of the boundary marked points @xmath229 with @xmath230 is odd . vertices @xmath229 with @xmath230 will be called legal and vertices @xmath229 with @xmath231 will be called illegal . the boundary edges of @xmath147 are , by definition , the boundary segments between successive vertices of @xmath147 which are not the illegal boundary marked points . we will use the following notations : @xmath232 let @xmath233 introduce an auxiliary formal variable @xmath234 and define the following series of rational functions : @xmath235 where @xmath236 by definition , @xmath204 is an element of @xmath237 $ ] . we do not know if @xmath204 belongs to @xmath225[[s , s_-]]$ ] or not . several computations in low degrees motivate us to conjecture that @xmath238[[s , s_-]]$ ] . moreover , we conjecture that there exists a power series @xmath239[[s , s_-]]$ ] such that for any @xmath205 we have @xmath240 however , we do not need this statement in the paper . let @xmath241 [ lemma : relation ] we have @xmath242 the lemma easily follows from equation , the definition of @xmath204 and the standard result about the graphical representation of the action of the exponent of a quadratic differential operator . we recommend the book ( * ? ? ? * sections 3 , 4 ) as a good introduction to this subject . denote by @xmath243 the space of hermitian @xmath213 matrices . for @xmath244 let @xmath245 and @xmath246 . introduce a volume form @xmath247 by @xmath248 consider positive real numbers @xmath249 . recall that @xmath250 . introduce a gaussian measure @xmath251 on the space @xmath243 by @xmath252 where the normalization @xmath253 is determined by the constraint @xmath254 for any polynomial @xmath255_{1\le i , j\le n}$ ] let @xmath256 the integrals @xmath257 are described by the following result ( see e.g. ( * ? ? ? * section 3.2.3 ) ) . [ lemma : wick formula ] _ 1_. if @xmath258 is homogeneous of odd degree , then @xmath259 . + _ 2_. for any indices @xmath260 , we have @xmath261 where the sum is taken over all permutations @xmath262 of the set of indices @xmath263 such that @xmath264 and @xmath265 . + _ 3_. we have @xmath266 in this section we present a matrix integral representation for the function @xmath204 . let @xmath267.\end{aligned}\ ] ] note that @xmath268 therefore , the exponent @xmath269 can be represented in the following way @xmath270 [ proposition : matrix integral ] we have @xmath271 here the integral is understood in the sense of formal matrix integration . it means the following . we express @xmath272 as a series of the form @xmath273 where @xmath274 is a polynomial of degree @xmath275 in @xmath276 and expressions of the form @xmath277 here the degree is introduced by putting @xmath278 from lemma [ lemma : wick formula ] it follows that @xmath279 is zero , if @xmath275 is odd , and is a rational function in @xmath212 of degree @xmath280 , if @xmath275 is even . then the integral on the right - hand side of is defined by @xmath281 we begin by recalling the derivation of the matrix integral for @xmath282 from kontsevich s combinatorial formula ( @xcite ) . consider an arbitrary trivalent ribbon graph without boundary with a coloring of faces in @xmath190 colors . it can be obtained by gluing trivalent stars ( see fig . [ fig : internal vertex ] ) in such a way that corresponding indices on glued edges coincide . ( 0.5,0 ) ( 0.5,-2.5 ) ; ( -0.5,0 ) ( -0.5,-2.5 ) ; ( 0.5,0 ) ( 2.67,1.25 ) ; ( 0,0.87 ) ( 2.17,2.12 ) ; ( -0.5,0 ) ( -2.67,1.25 ) ; ( 0,0.87 ) ( -2.17,2.12 ) ; ( 0.5,-2.5 ) ( -0.5,-2.5 ) ; ( 2.67,1.25 ) ( 2.17,2.12 ) ; ( -2.67,1.25 ) ( -2.17,2.12 ) ; ( 4.5,0 ) ( 6,0 ) ; ( 4.5,0.2 ) ( 4.5,-0.2 ) ; ( 6,0 ) ( 5.7,0.2 ) ; ( 6,0 ) ( 5.7,-0.2 ) ; \(b ) at ( 0.5,-1.25 ) ; ( b ) at ( -0.5,-1.25 ) ; ( b ) at ( 1.5,0.75 ) ; ( b ) at ( 1.30,1.45 ) ; ( b ) at ( -1.5,0.75 ) ; ( b ) at ( -1.3,1.45 ) ; ( b ) at ( 8.5,0 ) ; ( 0,0.5 ) ( 2.5,0.5 ) ; ( 0,-0.5 ) ( 2.5,-0.5 ) ; ( 0,0.5 ) ( -1.25,2.67 ) ; ( -0.87,0 ) ( -2.12,2.17 ) ; ( 0,-0.5 ) ( -1.25,-2.67 ) ; ( -0.87,0 ) ( -2.12,-2.17 ) ; ( 2.5,0.5 ) ( 2.5,-0.5 ) ; ( -1.25,2.67 ) ( -2.12,2.17 ) ; ( -1.25,-2.67 ) ( -2.12,-2.17 ) ; \(b ) at ( 1.25,0.5 ) ; ( b ) at ( 1.25,-0.5 ) ; ( 0,0.5 ) ( -2.5,0.5 ) ; ( 0,-0.5 ) ( -2.5,-0.5 ) ; ( 0,0.5 ) ( 1.25,2.67 ) ; ( 0.87,0 ) ( 2.12,2.17 ) ; ( 0,-0.5 ) ( 1.25,-2.67 ) ; ( 0.87,0 ) ( 2.12,-2.17 ) ; ( -2.5,0.5 ) ( -2.5,-0.5 ) ; ( 1.25,2.67 ) ( 2.12,2.17 ) ; ( 1.25,-2.67 ) ( 2.12,-2.17 ) ; \(b ) at ( -1.25,0.5 ) ; ( b ) at ( -1.25,-0.5 ) ; to a trivalent star we associate the polynomial @xmath283 $ ] . then the wick formula ( lemma [ lemma : wick formula ] ) and kontsevich s formula imply that @xmath284 this is the famous kontsevich s integral ( @xcite ) . we recommend the reader the book ( * ? ? ? * sections 3 and 4 ) for a more detailed explanation of this techique . consider now a non - nodal critical ribbon graph with boundary @xmath285 . each boundary component is a circle with a configuration of vertices of degrees @xmath10 and @xmath151 . vertices of degree @xmath10 are boundary marked points and they are of two types : legal and illegal . in fig . [ pic2 ] we draw an example of a boundary component . legal boundary marked points are marked by @xmath286 and illegal ones by @xmath287 . each graph from @xmath226 can be obtained by gluing trivalent stars and some number of boundary components . ( 2.5,0.5 ) arc ( 11.31:78.69:2.55 ) ; ( 2.5,0.5 ) ( 5,0.5 ) ; ( 2.5,-0.5 ) arc ( -11.31:-78.69:2.55 ) ; ( 2.5,-0.5 ) ( 5,-0.5 ) ; ( 2.5,-0.5 ) ( 2.5,0.5 ) ; ( -2.5,0.5 ) arc ( 168.69:101.31:2.55 ) ; ( -2.5,0.5 ) ( -5,0.5 ) ; ( -2.5,-0.5 ) arc ( -168.69:-101.31:2.55 ) ; ( -2.5,-0.5 ) ( -5,-0.5 ) ; ( -2.5,-0.5 ) ( -2.5,0.5 ) ; ( 0.5,2.5 ) ( 0.5,5 ) ; ( -0.5,2.5 ) ( -0.5,5 ) ; ( -0.5,2.5 ) ( 0.5,2.5 ) ; ( 0.5,-2.5 ) ( 0.5,-5 ) ; ( -0.5,-2.5 ) ( -0.5,-5 ) ; ( -0.5,-2.5 ) ( 0.5,-2.5 ) ; ( 0.5,5 ) ( -0.5,5 ) ; ( -5,0.5 ) ( -5,-0.5 ) ; ( 0.5,-5 ) ( -0.5,-5 ) ; ( 5,0.5 ) ( 5,-0.5 ) ; \(b ) at ( 2.21,1.27 ) ; ( b ) at ( 2.21,1.27 ) ; ( b ) at ( 1.27,2.21 ) ; ( b ) at ( 1.27,2.21 ) ; ( b ) at ( 1.80,-1.80 ) ; ( b ) at ( 1.80,-1.80 ) ; ( b ) at ( -1.27,-2.21 ) ; ( b ) at ( -1.27,-2.21 ) ; ( b ) at ( -2.21,-1.27 ) ; ( b ) at ( -2.21,-1.27 ) ; ( 6,0 ) ( 7.5,0 ) ; ( 6,0.2 ) ( 6,-0.2 ) ; ( 7.2,0.2)(7.5,0)(7.2,-0.2)cycle ; \(b ) at ( 8,0 ) ; consider a boundary component of some non - nodal critical ribbon graph with boundary @xmath288 . recall that , by definition , the boundary edges are the boundary segments between successive vertices which are not illegal boundary marked points . let @xmath289 be the edges of the boundary component , ordered in the clockwise direction . moreover we orient each edge in the clockwise direction , so that any boundary edge points from a source vertex to a target vertex . let @xmath290 the combinatorial formula suggests that to the boundary component we should assign the following expression : @xmath291\in{\mathbb c}[h_{i , j}],\end{gathered}\ ] ] where @xmath292 denotes the automorphism group of the boundary component . an example is shown in fig . [ pic2 ] . for a moment we ignore the combinatorial coefficient . if we sum expressions over all possible choices of a boundary component , we get @xmath293.\end{gathered}\ ] ] let us deal more carefully with the combinatorial coefficient . denote by @xmath294 the number of the boundary vertices of degree @xmath151 of the graph @xmath147 and by @xmath295 the number of the boundary edges . since @xmath296 , we have @xmath297 therefore , we have to rescale expression in the following way : @xmath298.\end{gathered}\ ] ] we immediately recognize here the function @xmath299 . again , the wick formula together with kontsevich s formula and the combinatorial formula imply that @xmath300 the proposition is proved . one can show that the integral @xmath301 is absolutely convergent and determines a well - defined function of @xmath302 . it is not hard to show that the asymptotic expansion of it , when @xmath303 , is given by @xmath282 ( @xcite ) . in our case we can also make sense of the integral in as a convergent integral . suppose that @xmath55 are positive real numbers , @xmath234 is a real number such that @xmath304 , @xmath305 , and @xmath77 is a purely imaginary complex number . then the integral @xmath306 is absolutely convergent and determines a well - defined function of @xmath307 and @xmath234 . in this section we introduce a certain version of the fourier transform . it happens that after this transformation formula becomes much simpler . the main result of this section is proposition [ proposition : convergent fourier transform ] . we again fix an integer @xmath205 throughout this section and set the genus parameter @xmath27 to be equal to @xmath308 . here we define our version of the fourier transform and describe its main properties . section [ subsubsection : preliminary ] is preliminary . the main definition is contained in section [ subsubsection : first fourier transform ] . in section [ subsubsection : second fourier transform ] there is a slightly different version of it that will also be useful . section [ subsubsection : properties ] is devoted to the properties of our fourier transform . for an arbitrary positive real @xmath310 and a non - negative integer @xmath311 , we have the following classical formula : @xmath312 a power series @xmath313 $ ] is called admissible if it has the form @xmath314 , where @xmath315 $ ] . the space of admissible power series is denoted by @xmath309^{adm}\subset{\mathbb c}[[s]]$ ] . for an admissible power series @xmath313^{adm}$ ] , @xmath316 , the formal fourier transform @xmath317(z)\in{\mathbb c}[[z^{-1},z]$ ] is defined by @xmath318(z):=\sqrt\frac{z}{2\pi}\int^{form}_{{\mathbb r}}{\widetilde f}(-is+z)e^{\frac{i}{6}s^3-{\frac{1}{2}}s^2 z}ds:=\\ = \sum_{k\ge 0}\sqrt\frac{z}{2\pi}\int_{{\mathbb r}}\frac{{\widetilde f}(-is+z)}{k!}\left(\frac{i}{6}s^3\right)^ke^{-{\frac{1}{2}}s^2 z}ds\in{\mathbb c}[[z^{-1},z].\end{gathered}\ ] ] the reader can see that the formal fourier transform of @xmath319 is the asymptotic expansion of the integral @xmath320 when @xmath321 goes to @xmath322 . an element @xmath324 $ ] will be called admissible , if the series @xmath325 , @xmath326 , satisfies the following property . there exists a sequence @xmath327 , @xmath328 , of positive integers such that @xmath329 , when @xmath330 , and @xmath331 . the space of all admissible elements will be denoted by @xmath323^{adm}$ ] . for an admissible element @xmath332^{adm}$ ] , it can be seen that the element @xmath333 can be expressed in the following way : @xmath334 where @xmath335 are homogeneous linearly independent rational functions and @xmath336 $ ] are polynomials in @xmath77 . the formal fourier transform @xmath317(\lambda , z)\in{\mathcal r}_n[[z^{-1},z]]$ ] is defined by @xmath337(\lambda , z):=\sum_m g_m(\lambda)\sqrt{\frac{z}{2\pi}}\int^{form}_{\mathbb r}e^{\frac{is^3}{6}-\frac{1}{2}s^2z}p_m(-is+z)ds\in{\mathcal r}_n[[z^{-1},z]].\end{gathered}\ ] ] we again see that the formal fourier transform can be considered as the asymptotic expansion of the integral @xmath338 when @xmath321 goes to @xmath322 . a formal fourier transform for power series from @xmath340[[s]]$ ] is introduced completely analogously . we introduce a grading in the ring @xmath340 $ ] assigning to @xmath25 the degree @xmath341 . for @xmath342 let @xmath340^d$ ] be the subspace of @xmath340 $ ] that consists of power series of the form @xmath343 , where @xmath344 $ ] is a homogeneous polynomial of degree @xmath275 . the subspace @xmath340[[s]]^{adm}\subset { \mathbb c}[[t_0,t_1,\ldots]][[s]]$ ] is defined similarly to the previous section , using the filtration @xmath345={\mathbb c}[[t_0,t_1,\ldots]]^0\supset{\mathbb c}[[t_0,t_1,\ldots]]^1\supset{\mathbb c}[[t_0,t_1,\ldots]]^2\supset\ldots.\ ] ] it is easy to see that if @xmath346[[s]]$ ] is admissible , then @xmath347\ ] ] is also admissible . for an admissible element @xmath346[[s]]^{adm}$ ] the formal fourier transform is defined by @xmath337(t_*,z):=\sqrt{\frac{z}{2\pi}}\int^{form}_{\mathbb r}e^{\frac{is^3}{6}-\frac{1}{2}s^2z}{\widetilde f}(t_*,-is+z)ds\in{\mathbb c}[[t_0,t_1,\ldots]][[z^{-1},z]],\end{gathered}\ ] ] where @xmath348 . the relation to the fourier transform from the previous section is given by @xmath349\right|_{t_i=-(2i-1)!!{\mathop{\mathrm{tr}}\nolimits}\lambda^{-2i-1}}=\phi^{form}_s\left[\left.f\right|_{t_i=-(2i-1)!!{\mathop{\mathrm{tr}}\nolimits}\lambda^{-2i-1}}\right].\end{gathered}\ ] ] let us describe three basic properties of our formal fourier transform . the symbol @xmath350 will stand for @xmath219 or @xmath340 $ ] . [ lemma : injectivity ] the map @xmath351^{adm}\to{\mathcal a}[[z^{-1},z]]$ ] is injective . suppose that @xmath352 . the proof for @xmath353 $ ] is the same . suppose we have @xmath317=0 $ ] for some admissible @xmath332^{adm}$ ] . let @xmath333 . we have @xmath354 where @xmath355 are linearly independent homogeneous rational functions and @xmath336 $ ] are polynomials in @xmath77 . since @xmath317=0 $ ] , we get @xmath356 for all @xmath275 . let us choose a non - zero @xmath357 and let @xmath358 , where @xmath359 . a direct calculation shows that @xmath360 therefore this integral is non - zero . this contradiction proves the lemma . [ lemma : fourier of derivative ] for any admissible element @xmath361^{adm}$ ] , the derivative @xmath362 is also admissible and @xmath363=\frac{z^2}{2}\phi^{form}_s[f]$ ] . the proof is again presented in the case @xmath352 . if @xmath316 , then @xmath364 . therefore , @xmath362 is also admissible . now we compute @xmath365(\lambda , z)=\sqrt\frac{z}{2\pi}\int^{form}_{{\mathbb r}}e^{\frac{is^3}{6}-\frac{1}{2}s^2z}\left(\frac{(-is+z)^2}{2}{\widetilde f}(\lambda ,- is+z)+i\frac{{\partial}}{{\partial}s}{\widetilde f}(\lambda ,- is+z)\right)ds=\\ = \frac{z^2}{2}\phi^{form}_s[f](\lambda , z).\end{gathered}\ ] ] the last equality follows from integration by parts . the lemma is proved . [ lemma : fourier and multiplication ] for any admissible element @xmath361^{adm}$ ] , the product @xmath366 is also admissible and @xmath367=-\left(\frac{1}{z}\frac{{\partial}}{{\partial}z}-\frac{1}{2z^2}-z\right)\phi^{form}_s[f].\ ] ] the admissibility of the product @xmath368 is obvious . we compute @xmath369=\sqrt\frac{z}{2\pi}\frac{1}{z}\int^{form}_{\mathbb r}\left[\left(i\frac{{\partial}}{{\partial}s}-\frac{s^2}{2}\right){\widetilde f}(-is+z)\right]e^{\frac{i}{6}s^3-\frac{1}{2}s^2 z}ds=\\ = \sqrt\frac{z}{2\pi}\int^{form}_{\mathbb r}is{\widetilde f}(-is+z)e^{\frac{i}{6}s^3-\frac{1}{2}s^2 z}ds.\end{gathered}\ ] ] therefore , @xmath370=\sqrt\frac{z}{2\pi}\int^{form}_{\mathbb r}(is - z){\widetilde f}(-is+z)e^{\frac{i}{6}s^3-\frac{1}{2}s^2 z}ds=\phi_s^{form}[sf].\ ] ] the lemma is proved . recall that @xmath371[[s]].\ ] ] from formula it is easy to see that the series @xmath372 is admissible . it is also easy to see that @xmath373 , where @xmath374 . thus , for any @xmath375 , the series @xmath376 is admissible . we conclude that the fourier transform @xmath377(\lambda , z , s_-)$ ] with respect to the variable @xmath77 is well - defined . [ proposition : formal fourier transform ] we have @xmath378(\lambda , z)=\phi^{form}_s\left[e^{\frac{s^3}{6}}f^o_n\right]\left(\lambda , z,\frac{z^2}{4}\right).\ ] ] by lemma [ lemma : relation ] , we have @xmath379 from lemma [ lemma : fourier of derivative ] it follows that @xmath380(\lambda , z , s_-)=e^{\frac{z^2}{4}\frac{{\partial}}{{\partial}s_-}}\left(\phi^{form}_s\left[f^o_n e^{\frac{s^3}{6}}\right](\lambda , z , s_-)\right)=\\ = \phi^{form}_s\left[f^o_n e^{\frac{s^3}{6}}\right]\left(\lambda , z , s_-+\frac{z^2}{4}\right).\end{gathered}\ ] ] setting @xmath381 , we get the statement of the proposition . propositions [ proposition : formal fourier transform ] and [ proposition : matrix integral ] together with section [ subsection : convergent matrix integral ] imply the following proposition . [ proposition : convergent fourier transform ] the series @xmath382(\lambda , z)$ ] is the asymptotic expansion of the integral @xmath383 when @xmath384 and @xmath385 . let @xmath190 be an arbitrary positive integer . recall that the series @xmath386 $ ] is defined by @xmath233 let @xmath387 . we set the genus parameter @xmath27 to be equal to @xmath206 the following proposition is the key step in the proof of theorem [ theorem : main ] . [ proposition : main ] we have @xmath388\left(\lambda , z\right)=\tau^c_{n+1}(\lambda_z)\sqrt{\det\left(\frac{1-z\lambda^{-1}}{1+z\lambda^{-1}}\right)}.\end{gathered}\ ] ] before proving the proposition let us make a remark about the right - hand side of equation . we see that @xmath389[[z^{-1}]]$ ] and @xmath390[[z]]$ ] . so , on the right - hand side of we multiply a power series in @xmath391 and a power series in @xmath321 . in general , the multiplication of two such series may not be well - defined . in our case , the issue is resolved as follows . we have @xmath392 therefore , @xmath393 has the form @xmath394 , where @xmath395 $ ] is homogeneous of degree @xmath222 . thus , the product on the right - hand side of is well - defined . our proof of proposition [ proposition : main ] uses a famous technique , that is sometimes called the averaging procedure over the unitary group . we recall it in section [ subsection : averaging ] . after that , in section [ subsection : proof of the main proposition ] , we prove proposition [ proposition : main ] . it is well - known that an arbitrary hermitian matrix @xmath396 admits a polar decomposition , @xmath397 , where @xmath398 is a unitary matrix and @xmath399 is a diagonal matrix with real entries . denote by @xmath400 the group of unitary @xmath213 matrices . given a vector @xmath401 , define a map @xmath402 by @xmath403 for a hermitian matrix @xmath404 let @xmath405 be the vector of its eigenvalues . it is defined up to a permutation of the coordinates . for a subset @xmath406 , that is invariant under permutations of the coordinates , let @xmath407 suppose moreover that @xmath406 is compact and measurable . let @xmath408 be an arbitrary smooth function . we denote by @xmath409 the haar probability measure on @xmath400 . we have the following result ( see e.g. ( * ? ? ? * sections 3 and 4 ) ) . [ lemma : polar ] we have @xmath410 let @xmath411 be arbitrary real numbers such that @xmath412 and @xmath413 , for @xmath414 . let @xmath415 and @xmath416 . let @xmath417 be an arbitrary non - zero complex parameter . the harish - chandra - itzykson - zuber ( hciz ) formula says that @xmath418 the right - hand has no poles on the diagonals @xmath419 or @xmath420 , and so it defines a smooth function of @xmath421 real parameters @xmath411 . formula was originally found in @xcite and then was rediscovered in @xcite . we will apply the hciz formula in the following way . again , let @xmath422 be a compact measurable subset of @xmath423 invariant under permutations of the coordinates . let @xmath408 be a smooth unitary invariant function : @xmath424 , for any @xmath425 . finally , consider pairwise distinct real numbers @xmath426 and set @xmath415 . [ lemma : averaging ] we have @xmath427 where @xmath428 and @xmath429 . the lemma follows from combining lemma [ lemma : polar ] and the hciz formula . suppose that @xmath212 and @xmath321 are positive real numbers . consider kontsevich s integral @xmath430 as we already recalled in section [ subsection : convergent matrix integral ] , the integral is absolutely convergent and its asymptotic expansion , when @xmath431 , is given by the series @xmath432 ( see @xcite ) . suppose moreover that @xmath433 . using proposition [ proposition : convergent fourier transform ] we see that equation is a consequence of the following equation : @xmath434 the integral is still absolutely convergent , when @xmath55 s and @xmath321 are complex numbers with positive real parts . moreover , the integral is semi - convergent , if @xmath435 are purely imaginary non - zero complex numbers , and the asymptotic expansion of it , when @xmath436 is still given by the power series @xmath432 . see the discussion of these subtle questions in @xcite . let @xmath55 and @xmath321 be purely imaginary complex numbers with positive imaginary parts . assume moreover that @xmath437 where we choose a particular value of the square root , such that all entries of the matrix @xmath438 have positive imaginary parts . let @xmath426 be the diagonal elements of @xmath438 . let @xmath439 let us perform the change of variables @xmath440 on the left - hand side of equation . we get @xmath441 where @xmath442 . define @xmath443 to be equal to @xmath0 . by lemma [ lemma : averaging ] , the last integral is equal to @xmath444 where @xmath445 and @xmath446 . note that @xmath447 where @xmath448 . therefore , we can rewrite as follows : @xmath449,\end{gathered}\ ] ] where @xmath450 and @xmath451 . by lemma [ lemma : averaging ] , the expression in the square brackets is equal to @xmath452 redenoting @xmath453 by @xmath77 , expression is equal to @xmath454.\end{gathered}\ ] ] now we make the shift @xmath455 in the integral in the square brackets . using also that @xmath456 , we get that is equal to @xmath457.\end{gathered}\ ] ] finally , making the shift @xmath458 , we come to @xmath459 equation is proved , and , thus , also proposition [ proposition : main ] . from the dimension constraint it follows that , if an open intersection number is non - zero , then the genus @xmath3 is uniquely determined by @xmath21 s and @xmath63 . using this observation , it is easy to show that , if the open kdv equations hold for @xmath460 , then they are also true for an arbitrary @xmath27 . the same is true for the open virasoro equations . therefore , without loss of generality , we can assume that @xmath460 . let @xmath461 be the shift operator which acts on a series @xmath462 $ ] by @xmath463(t_0,t_1,\ldots):=f\left(t_0-\frac{k_0}{z},t_1-\frac{k_1}{z^3},t_2-\frac{k_2}{z^5},\ldots\right),\ ] ] where @xmath464 and , by definition , @xmath465 . let @xmath466 . proposition [ proposition : main ] implies the following corollary . [ corollary : main ] we have @xmath467=g_z[\tau^c]e^{\xi^c}.\end{gathered}\ ] ] it is easy to check that for an arbitrary @xmath205 we have @xmath468\right|_{t_i=-(2i-1)!!{\mathop{\mathrm{tr}}\nolimits}\lambda^{-2i-1}}=\tau^c_{n+1}(\lambda_z)\quad\text{and}\quad \left.e^{\xi^c}\right|_{t_i=-(2i-1)!!{\mathop{\mathrm{tr}}\nolimits}\lambda^{-2i-1}}=\sqrt{\det\left(\frac{1-z\lambda^{-1}}{1+z\lambda^{-1}}.\right)}.\ ] ] therefore , by proposition [ proposition : main ] and equation , we have @xmath469\right)\right|_{t_i=-(2i-1)!!{\mathop{\mathrm{tr}}\nolimits}\lambda^{-2i-1}}=\left.\left(g_z[\tau^c]e^{\xi^c}\right)\right|_{t_i=-(2i-1)!!{\mathop{\mathrm{tr}}\nolimits}\lambda^{-2i-1}}.\end{gathered}\ ] ] for an arbitrary @xmath470 the functions @xmath471 become algebraically independent , when @xmath190 is sufficiently large . therefore , equation implies that @xmath472=g_z[\tau^c]e^{\xi^c}$ ] . the corollary is proved . let us prove that the open free energy @xmath98 satisfies the half of the burgers - kdv hierarchy - . in @xcite it was obtained that the half of the burgers - kdv hierarchy can be written in a very convenient form using the lax formalism . let us briefly recall this result . a pseudo - differential operator @xmath438 is a laurent series @xmath473 where @xmath275 is an arbitrary integer and @xmath474 are formal power series in @xmath475 . we will always identify @xmath476 with @xmath477 . let @xmath478 the product of pseudo - differential operators is defined by the following commutation rule : @xmath479 where @xmath480 and @xmath346 $ ] . for any @xmath481 and a pseudo - differential operator @xmath438 of the form @xmath482 there exists a unique pseudo - differential operator @xmath483 of the form @xmath484 such that @xmath485 . consider the pseudo - differential operator @xmath486 . in @xcite it was proved that the half of the burgers - kdv system - is equivalent to the following system : @xmath487 let us prove that the open free energy @xmath98 satisfies equations and . let @xmath488 from the fact that the closed partition function @xmath36 is a tau - function of the kdv hierarchy it follows that the series @xmath489 is the wave function of the kdv hierarchy ( see e.g. @xcite ) , that is , it satisfies the following equations : @xmath490 from corollary [ corollary : main ] it follows that @xmath491=\psi|_{s_*=0}.\end{gathered}\ ] ] let us prove . let @xmath492 . we have @xmath493=o_{t_n}\phi^{form}_s\left[e^{f^o}\right]\stackrel{\text{by~\eqref{eq : eq2}}}{=}o_{t_n}\left(\psi|_{s_*=0}\right)=\left(o_{t_n}\psi\right)|_{s_*=0}\stackrel{\text{by~\eqref{eq : first for the wave}}}{=}0.\ ] ] by lemma [ lemma : injectivity ] , we get @xmath494 and equation is proved . let us prove . let @xmath495 . we have @xmath496\stackrel{\text{by lemma~\ref{lemma : fourier of derivative}}}{=}\left(\frac{z^2}{2}-\frac{1}{2}l\right)\phi^{form}_s\left[e^{f^o}\right]\stackrel{\text{by~\eqref{eq : eq2}}}{=}\left(\frac{z^2}{2}-\frac{1}{2}l\right)\left(\psi|_{s_*=0}\right)=\left(o_s\psi\right)|_{s_*=0}\stackrel{\text{by~\eqref{eq : second for the wave}}}{=}0.\ ] ] by lemma [ lemma : injectivity ] , we get @xmath497 and equation is proved . in @xcite it was shown that the open kdv equations follow from the half of the burgers - kdv hierarchy . therefore , conjecture [ conjecture : open kdv ] is proved . we also see that the open free energy @xmath98 satisfies the initial condition @xmath498 . in @xcite it was proved that such a solution of the half of the burgers - kdv hierarchy satisfies the open virasoro equations . thus , conjecture [ conjecture : open virasoro ] is also proved . the derivation of the open virasoro equations from the open kdv equations was given in @xcite . another derivation was obtained in @xcite . in this section we want to show that the open virasoro equations can be very easily obtained from corollary [ corollary : main ] . again we can assume that @xmath460 . by corollary [ corollary : main ] , @xmath499=g_z[\tau^c]e^{\xi^c}.\ ] ] from the closed virasoro equations it is easy to derive that ( see e.g. @xcite ) @xmath500e^{\xi^c}\right)=\left(\frac{z^{2n+2}}{2^{n+1}}\left(\frac{1}{z}\frac{{\partial}}{{\partial}z}-\frac{1}{2z^2}-z\right)+\frac{n+1}{2^{n+2}}z^{2n}\right)\left(g_z[\tau^c]e^{\xi^c}\right).\ ] ] by lemmas [ lemma : fourier of derivative ] and [ lemma : fourier and multiplication ] , we have @xmath501&=\phi_s^{form}\left[\left(-\frac{{\partial}^{n+1}}{{\partial}s^{n+1}}\circ s+\frac{n+1}{4}\frac{{\partial}^n}{{\partial}s^n}\right)\tau^o\right]=\\ & \quad\quad=-\phi_s^{form}\left[\left(s\frac{{\partial}^{n+1}}{{\partial}s^{n+1}}+\frac{3n+3}{4}\frac{{\partial}^n}{{\partial}s^n}\right)\tau^o\right].\end{aligned}\ ] ] here , as above , the circle @xmath502 means the composition of operators . in the last operator inside the fourier transform we immediately recognize the difference between the open virasoro operator @xmath503 and the closed virasoro operator @xmath43 ( see ) . therefore , we get @xmath504=0.\ ] ] lemma [ lemma : injectivity ] implies that @xmath505 . the open virasoro equations are proved . s. k. lando , a. k. zvonkin . _ graphs on surfaces and their applications_. with an appendix by don b. zagier . encyclopaedia of mathematical sciences , 141 . low - dimensional topology , ii . springer - verlag , berlin , 2004 . xvi+455 pp . a. okounkov , r. pandharipande . _ gromov - witten theory , hurwitz numbers , and matrix models_. algebraic geometry - seattle 2005 . part 1 , 325 - 414 , proc . pure math . , 80 , part 1 , amer . soc . , providence , ri , 2009 .

a complete construction of the aflt states is proposed . with this construction and for all the cases we have checked , the agt conjecture on the equivalence of nekrasov instanton counting ( nic ) to the conformal block has been verified to be true . bao shou , jian - feng wu , and ming yu .5 cm _ institute of theoretical physics , + chinese academy of sciences , beijing , 100190 , china _

1107.4784

conformal blocks , which are defined on the ( punctured ) riemann surfaces , holomorphic in each @xmath1 coordinate except when they meet each other , play an essential role in building correlation functions in two dimensional ( euclidean ) conformal field theories@xcite . they can be best understood as sewing together chiral vertex operators@xcite , which by definition , are not local objects , but the correlation functions are . the later combine both holomorphic and anti - holomorphic conformal blocks in a consistent way to make modular covariant objects . on the sphere , the @xmath2-point conformal block is represented graphically as in fig.[confblock ] , where @xmath3 is the conformal dimension of the primary field inserted at coordinate @xmath1 , and @xmath4 labels the contribution arising from the conformal family descending from a primary field with the conformal dimension @xmath4 . the global conformal invariance is @xmath5 , which may be used to fix three coordinates @xmath6 , @xmath7 and @xmath8 . so the independent variables are @xmath1 , @xmath9 , with the degrees of freedom @xmath10 for the @xmath2-point conformal blocks on the sphere . the calculation of conformal blocks is based on the conformal ward - identities , @xmath11=(z^{n+1}\partial_z + ( n+1)h z^n ) v_h(z).\ ] ] and carried out perturbatively level by level @xcite . in some special cases , the decoupling of the virasoro null vectors can be implemented as differential equations for the conformal blocks . for the general case , recursion relations have been proposed by zamolodchikov@xcite on the meromorphic structures of the conformal blocks either in complex @xmath12-plane or @xmath13-plane . however , in general , the global perspective of the sewing procedure for the conformal blocks was still not fully understood until recently when the agt duality @xcite had been proposed . agt conjecture relates 2d liouville conformal field theories to 4d @xmath14 supersymmetric gauge theories of the @xmath15 type . the main idea is coupling to the liouville field a @xmath16 field field is a gauge symmetry and can be fixed to any desired value . ] , then this system is dual to a @xmath17 superconformal 4d theory . in this case , the partition function by nekrasov instanton counting(nic)@xcite of the 4d @xmath18 theory is to be identified with the conformal blocks of the @xmath16 coupled liouville type . the liouville cft is characterized by a 2d one boson theory with center charge @xmath19 . finally , one can decouple the @xmath20 factor and obtain the instanton partition function of the @xmath21 theory which duals to liouville conformal blocks . liouville interaction breaks down the charge conservation explicitly and leads to the introduction of the screening charges . because of the existence of the screening charges , the conformal blocks of the liouville type is much more complicated than its counterpart of the @xmath16 free boson theory . however , the agt conjecture , if proven true , means that there exists an orthogonal basis upon which the @xmath22 conformal blocks are built . from the above reasoning , there exists a tree - like structure which describes the duality in coupling space of the @xmath14 4d superconformal linear quiver gauge theory . the primary objects for this tree - like structure is the `` bifundmental '' matter coupling , which , if translated correctly , should be represented by the inner products of the bra and ket descendant fields in 2d conformal families sandwiched by a `` primary '' vertex operator at position , say , @xmath23 . such kind of pants - like diagram can be sewed together to form a linear quiver diagram , which , on the 2d cft side , is just the @xmath2-point functions on the sphere for our consideration . of course , in the present context , we mean the @xmath0 2d cft . at first sight , it seems that such duality does not bring in any conveniences . however , the nekrasov instanton counting on the 4d field theory shows a rather compact form for the summands which are completely factorized in `` momentum '' @xmath24 . and the summation is well organized into the combinatorial enumeration of the young tableuax . this simple structure implies liouville theory , in particular , the evaluation of the liouville conformal blocks , could be resolved by embedding it into a bigger system . so one may expect a new construction for the liouville conformal blocks from the corresponding nic . -point conformal block on @xmath25,height=144 ] as pointed out by nakajima@xcite , the instanton counting for @xmath14 gauge theory is equivalent to the hilbert scheme of points on the corresponding seiberg - witten curve ( blow - up riemann surface)@xcite . this can be translated into a topological string description from physicists point of view . by invoking the d4-d0 brane setup@xcite for adhm construction@xcite of the instanton moduli space and the resolving process for ale singularities@xcite , these indicate that the instanton counting is a counting for d0 branes in a toric calabi - yau 3-fold . actually , there are two kinds of d0 branes in the calabi - yau 3-fold , one is the regular d0 brane , which is in regular representation of @xmath26 , the center of the corresponding ade group . it carries no flux and can move freely on the riemann surface . the other is the fractional d0 brane , which is a d2 brane wrapping on a zero - sized two sphere . it is always attached to the ale singularity since it has a nontrivial monodromy while moving around the singularity . it is these fractional d0 branes that resolve the ale singularity , and leave fluxes on the blow - up riemann surface . this property ensures that one can identify these fractional instantons as `` anyons '' on the riemann surface . on the other hand , the regular ones are `` electric charged '' particles on the riemann surface . so the total counting is equvalent to solving the problem of `` electron gas '' system with insertions of anyons at the blow - up singularity on the riemann surface . this point of view is partialy included in dijkgraaf and vafa s article@xcite . for each pants of the pants decomposition for the ( punctured ) riemann surface , one can guess that the instanton partition function can be rewritten as summation over all the intermediate states passing through the sewn holes@xcite . for the interests of the present paper , we concern ourselves only with the special pants diagram that one of the tubes is replaced by the blow - up singularity . then the summand in the instanton partition function represents itself as an inner product of the bra and ket states , sandwiched by the anyonic vertex operator . these bra and ket states should come from the interacting `` electronic '' particles . a candidate description of the `` electronic gas '' system is the integrable system of multiple calogero - sutherland model , each living on a cycle . the whole ( punctured ) riemann surface , can be obtained by sewing together these pants on nonintersecting cycles . there are many efforts on relating the conformal blocks to the nic@xcite from various points of views , and these works confirm the validity of the agt duality . however , the explicit construction for the liouville conformal blocks has remained largely unclear until the recent work @xcite by alba , fateev , litvinov and tarnopolsky . in @xcite , they have put forward the agt duality in a more explicit form [ aflt ] & = & z_bif(|p, ; p , ) , here specifically for a free field realization , @xmath27 with @xmath28 , and @xmath29 is the screening charge in the virasoro sector . the l.h.s . of eq([aflt ] ) is the pants - like ( with one of the tubes labeled by @xmath30 shrinks to a line ) conformal block . the r.h.s . of ( [ aflt ] ) reproduces @xmath31 for the instanton counting , which is given by [ zbif]z_bif(|p, ; p , ) = _ i , j=1 ^ 2 _ sy_i(q - e_y_i , y_j(p_i - p_j|s ) - ) _ ty_j ( e_y_j , y_i(p_j - p_i|t)-),where @xmath32 , @xmath33 and e_y , y(p|s ) p+b^-1(a_y(s)+1)-bl_y(s ) . here @xmath34 and @xmath35 resp . are the arm length and the leg length resp . of the box @xmath36 in the young tableau @xmath37 , defined as @xmath38 @xmath39 and @xmath40 resp . are the @xmath41-th part of the partition @xmath42 and the @xmath43-th part of the transpose partition @xmath44 respectively . ( [ aflt ] ) means that the matrix elements of a special `` chiral vertex operator '' @xmath45 in a suitably chosen basis , can be translated into a 4d theory as an instanton contribution for a special bifundamental contribution of the nic . by sewing together pants - like diagrams one gets any desired duality diagrams in the coupling space of the linear quiver gauge theory . so , the checking of the agt duality reduces to the construction of the states @xmath46 , which we shall call the aflt states@xcite , with @xmath47 the young tableaux . here the @xmath37 s , the partitions of natural numbers , or equivalently represented by young tableaux , are labels for the orthogonal basis for the descendant fields ( verma modules ) in a @xmath0 conformal family from the 2d cft point of view . by definition , the aflt states form a complete set of states for the family members in a given @xmath0 conformal family and the inner products between them , sandwiched by a vertex operator of the particular form , @xmath48 , at position , say , @xmath49 , is factorized exactly as the nic @xmath31 presented on the r.h.s . of ( [ aflt ] ) . the explicit formula , ( [ aflt ] ) , puts strong constraints on the possible forms of the aflt states and make a systematic construction of them unaccessible at first glance . in @xcite , only the explicit form of the state @xmath50 has been found , @xmath51 with @xmath52 the creator @xmath53 s valued jack symmetric function , and @xmath54 the normalization constant . in our opinion , the agt conjecture , written in the form of ( [ aflt ] ) , strongly suggests that the @xmath0 conformal family is a hamiltonian system with @xmath46 the hamiltonian eigenstates . so the construction of the aflt states becomes a quantum mechanical problem of solving the schrodinger equation . put things in this way , we propose a possible form of the hamiltonian @xmath55 and construct its eigenstates explicitly . we shall identify those eigenstates as the aflt states desired . for in all the cases we have checked , ( [ aflt ] ) is verified to be true , using the aflt states we have constructed . we shall present now as the main results of our present paper the explicit form of the hamiltonian @xmath55 along with the complete construction of the aflt states , @xmath46 . more elaborated exposition will come in the subsequent sections . [ main1 ] h&=&h_0+h_i + _ & = & j_-|p_(p).here , @xmath56 are the jack states constructed in terms of the oscillators @xmath57 s or @xmath58 s ( @xmath59 ) solely , @xmath60 the corresponding hamiltonian for the jack symmetric functions , @xmath61 . thus the eigenstate of @xmath62 is just @xmath63 with the eigenvalue @xmath64 . @xmath60 in our formalism is defined to include zero modes @xmath65 also , @xmath66 . it is important that @xmath67 is strictly triangular with respect to the basis vectors of the @xmath62 eigenstates . by `` strictly triangular '' we mean the ( upper or lower ) triangular matrix with zero diagonal entries . it is easy to see that if the interaction term @xmath67 is strictly triangular , then the eigenvalue spectrum of @xmath62 remains unperturbed and @xmath46 in ( [ main1 ] ) well defined for non - degenerate @xmath62 spectrum descending from a mother state @xmath68 for generic values of @xmath24 s . putting things all together , we have [ main2 ] & & h = h_0+h_i , h_0= h^+ + h^- , h_i=_n=1 ^ 2qn a_-n^+a_n^- , + & & h^=(z_z^)^3 + _ n=1^q n a_-n^a_n^ , + & & e_(p)=e_y_1+e_y_2 + 2p(|y_1|-|y_2| ) , e_y=_i ( y^2_ib^-1+(2i-1)y_i b ) , + & & _ ( p)= ( -)^|y_1|b^|y_1|+|y_2|_y_1(2p+(a_y_1 + 1)b^-1-l_y_2b ) _ y_2(2p - a_y_2b^-1 + ( l_y_1 + 1)b ) , + & & _ = j_-|p_(p ) , + & & h_0j_-=e_(p)j_- , h_=e_(p ) _ , -i a^_0=p notice that + 1 ) @xmath50 constructed in @xcite are included in our construction as subcases . + 2 ) the hamiltonian @xmath55 constructed by us , albeit in a disguised form , turns out to coincide up to some trivial factor with @xmath69 , one of the integrals of motion found in a different context in appendix c of @xcite . @xmath69 in @xcite , written in the form of @xmath0 , makes the virasoro symmetry manifest , but is not suitable for solving a perturbation theory with perturbation parameter @xmath70 . the hamiltonian @xmath55 written in terms of the interacting bi - jack polynomial system as in ( [ main2 ] ) , shows virasoro symmetry only implicitly , but makes the perturbation theory exactly solvable as we shall see soon after . the procedure is outlined as follows . on the 2d cft side , the @xmath71 theory can be represented as a theory of two independent scalars @xmath72 and @xmath73 . @xmath72 part is essentially a free theory of timelike oscillators , while the scalar field @xmath74 is spacelike but engaged in a liouville type interaction . the two scalars can be linearly combined to form the `` light - cone '' scalars . @xmath75 and @xmath76 . the labeling @xmath77 of the basis vectors strongly suggests that there exist a bi - jack polynomial structure , plus possibly some interactions between these two sectors . that is , the `` free '' @xmath60 spectrums should be described by @xmath78 and @xmath79 respectively , here @xmath80 denotes jack states related to young tableau @xmath37 . first we construct the `` unperturbed '' energy operator @xmath62 which just sums up the `` energies '' in @xmath81 sectors , @xmath82 . the next thing is to specify the interaction between these two sectors . strictly speaking , @xmath62 does not describe a free theory , since it also contains the interaction terms proportional to @xmath83 . but the new interaction term @xmath67 further mixes the @xmath84 s and the coupling is also a first order in @xmath83 . it is good to see that @xmath67 is strictly triangular with respect to the basis vectors of @xmath62 eigenstates . our method can be easily generalized to wider classes of integrable models , in which the interacting hamiltonian splits into two parts , @xmath85 and @xmath86 , representing respectively the shift of energies and the rotations ( mixings ) of states . the later keeps the eigenvalue spectrum untouched@xcite . besides being triangular , the form of the interaction term is however much restricted , also by the virasoro symmetry . since the total hamiltonian is of the form @xmath0 , an `` interaction energy operator '' @xmath67 is needed to make the `` full energy operator '' @xmath87 the combination of @xmath88 s and @xmath89 s only . once the hamiltonian structure is determined , then the construction of the hamiltonian eigenstate @xmath46 is just a quantum mechanical problem . @xmath62 and @xmath55 share the same eigenvalue spectrum , but only the eigenstates of @xmath55 , represented by @xmath46 s , form a complete set of basis vectors for the @xmath0 conformal family . we have checked by examples the corresponding agt duality formula , ( [ aflt ] ) up to level 4 , and have found that indeed nekrasov instanton counting can be reproduced with this construction , ( [ main2 ] ) . in fact , we have also checked more general cases and all get positive answers . but those more general results will appear elsewhere due to lacking of space to include them in this paper . the insertions of the screening charges play an important role in checking the agt duality . however , in the present work we concern ourselves only with the cases in which the screening charges can be detached away from the vertex operator @xmath45 and moved on to act on the aflt states ( similar to the felder s brst operators)@xcite . the more general cases in which screening charges can not be moved away from @xmath45 will be under our future studies . it is well known that it is possible to map the liouville theory to the analytic continuation of the calogero - sutherland(cs ) model , which was originally and in most cases considered to be a theory with the parameter @xmath90 , while in the liouville case @xmath91 is required . some explanation is given in @xcite . the physical space of the cs model are created by jack polynomials , which are symmetric functions studied in great detail in mathematics and physics literatures@xcite . the integrability of the cs model may be derived in different ways , e.g. , from the knowledge of the hidden @xmath92 symmetry of the model . a recursion relation related to the virasoro singular vectors and an integral representation based on it has appeared recently in @xcite , in which more references can be found on the subjects of the cs model and the jack symmetric functions . it should be stressed again that for @xmath90 , there is no null vectors in the cs model . so the `` null '' vectors are not the true null vectors of the cs model , since the virasoro algebra based on which the null vectors are constructed is not the true conformal algebra of the cs model in that case . but for @xmath91 , yes , there are null vectors in the cs model . it is possible to describe the @xmath93 theory in terms of the jack polynomials considered as analytic continuation from @xmath94 to @xmath95 . there is another hint that the @xmath93 theory has something to do with @xmath91 cs model . it can be found from the nekrasov partition function , in which each term in the summation can be written in the form of the carlsson - okounkov formula@xcite , for the special cases when no screening charges are inserted . carlsson - okounkov formula is a formula for the inner products between the bra jack states and the ket jack states sandwiched with a modified vertex operator . this extraordinary formula is of great help in checking the agt duality with our construction for the orthogonal basis vectors @xmath46 s defined in ( [ aflt ] ) . we notice that the construction we found shares many similarities with the construction of the jack functions themselves . namely , we take the state @xmath96 as the mother state and its descendants are constructed in such a manner that two partitions are `` squeezed '' into other pairs . the squeezing does not change the total level of the two partitions , but does make the inner products of the descendants a triangular form . although the 4d to 2d duality has just begun to be understood , it has been known for sometime that 2d conformal blocks can be equivalently described as insertions of wilson lines in 3d pure chern - simons topological gauge theory . in fact , we can interpret the @xmath2-point conformal block represented by fig.1 as a wilson line insertion inside a three - ball . the path integral in chern - simons - witten gauge theory thus creates a state living on the boundary of the three ball , which is punctured @xmath25 . so it should not be a too big surprise that 2d conformal field theory has something to do with higher dimensional quantum field theories . taking into account that jack symmetric polynomials can be taken as some special limit of the two parameter macdonald symmetric polynomials , one natural guess is that our construction can be generalized to the case of macdonald symmetric polynomials . in that case there should be a 5d to 3d duality . this paper is organized in the following way . our general formalism on the construction of the aflt states is presented in the introduction . in section 2 , we explore the general structure of the @xmath0 conformal family . we found in some cases it is more convenient to work with the bi - jack function basis . section 3 contains the major derivation of our construction . section 4 is the conclusion . and in appendix a the explicit construction of the aflt states up to level 3 is presented . we are dealing with a 4d n=2 @xmath18 linear quiver gauge theory coupled to special bi - fundmental matter in a superconformal way . according to the standard agt duality dictionary , the corresponding 2d conformal block is of the @xmath0 type , which reproduces the instanton part of the nekrasov partition function for the @xmath18 theory . there are two sets of young diagrams which measure the partitions of the instantons . if one wants to extract the virasoro basis of the conformal blocks , one need to factor out the @xmath16 factor . in this section we shall mainly explore the hilbert space for the @xmath0 theory and find the requirements that the energy operator @xmath55 should meet . our procedure depends heavily on the nekrasov instanton counting formula written more suitably for the construction of the conformal blocks , ( [ aflt ] ) . first , the 2d @xmath16 conformal block , realized in terms of the oscillators of the scaler field @xmath97 , is essentially of free theory with center charge @xmath98 . the zero modes can be integrated out trivially and does not play any significant role here . the vertex operators for @xmath97 , take a peculiar form @xmath99 here , @xmath100 means the positive ( negative ) mode part of the scalar field @xmath97 . although the above vertex operator is not the standard one in 2d cft , its contribution to the conformal block can be easily read off and factored out . second , the @xmath101 part is a liouville conformal field theory of the @xmath102 scalar field and is more complicated because of the existence of the screening charges . we have the following mode expansion for the scalar fields @xmath102 and @xmath72 , [ scalars ] ( z)&=&q+c_0 ( z)+_nz , n0z^n , + ( z)&=&+a_0 ( z)+_nz , n0 z^n , + & = & _ n+m,0 , ( c_-n)^ = c_n , = + & = & _ n+m,0 , ( a_-n)^ = -a_n , = . + virasoro generators in the @xmath101 part , @xmath89 , thus reads [ oringinalln ] l_n & = & _ kc_kc_n - k - inq c_n= _ k0,nc_kc_n - k+i(2-nq)c_n , + [ l0]l_0 & = & -^2 + 2_k>0c_-kc_k , here , @xmath103 , @xmath104 by this construction , @xmath89 defined in ( [ oringinalln]-[l0 ] ) is obviously unitary , @xmath105 in 2d cft , one frequently meets another ( more conventional ) definition of the virasoro generators , l^0_n = _ kc_kc_n - k- i q(n+1)c_n.if ( [ oringinalln ] ) and ( [ l0 ] ) are combined in this way , [ tz1 ] t(z ) = + i q ^2 + i q z^-1- = _ n l_n z^-n-2 , then @xmath106 differs from the more conventional one @xmath107 by a similarity transformation [ tz2 ] t(z ) & = & e^-iq q(+iq^2)e^iqq=_n _ n z^-n-2 + _ n & = & _ k_k_n - k- i q(n+1)_n . comparing ( [ tz1 ] ) and ( [ tz2 ] ) , we have = \ { lc_n , n0 + c_0+q , n=0 .. viewing the @xmath0 model as a 2d sigma model , since under conjugation @xmath108 and @xmath88 transform differently we recognize that @xmath109 is spacelike and @xmath110 timelike when they are considered as coordinates in target space . so the target space of the sigma model under consideration is curved in space and flat in time direction . the two scalars can also be linearly combined to form the `` light - cone '' scalars @xmath111 , [ pm ] & & ^(z)= ( z)(z ) , + & & ^(z ) ^(z)=(z - z ) , + & & ^(z ) ^(z)=0 , + & & ^(z)=^(z ) . the descendant states in the conformal family split into sub - spaces of different levels , which are measured by @xmath112 . within the sub - space of given level @xmath113 , states can be labeled either by linear conbinations of @xmath114 s or @xmath115 s , with @xmath116 , @xmath37 the young tableau , @xmath117 , @xmath118 the annihilator @xmath119 s ( or creator @xmath120 s ) valued jack symmetric functions . in either case , one can infer from agt duality that there exist a hermitian operator @xmath55 , which commutes with @xmath121 and diagonizes this subspace . hence the eigenstates of @xmath55 form an orthogonal basis . we know @xmath121 acts on this subspace trivially like an identity operator . so in order to eliminate the degeneracy , the next candidate @xmath55 we are looking for should be at least cubic in the oscillators @xmath122 s and @xmath123 s . once @xmath55 is introduced , the descendant states will organize themselves into an orthogonal basis labeled by two sets of young tableaux @xmath124 . in our opinion , it is better to start with the @xmath84 system , since there is already a hamiltonian structure @xmath60 acting separately on them . but @xmath125 does not commute with the screening charges @xmath126 pertaining to @xmath89 , @xmath127 here @xmath128 . we then add a new term @xmath67 to @xmath62 , @xmath87 and require that @xmath129=0 $ ] . if @xmath67 are chosen correctly , the eigenstates of @xmath55 will coincide with the unique orthogonal basis @xmath46 , which we call aflt states and are defined to satisfy ( [ aflt ] ) , in which the matrix elements @xmath130 is factorized in a consistent way . on the 4d theory side , one can decouple a single massless bifundamental matter . ] we shall show that under this condition the contributions can be written as the orthogonality condition for the @xmath46 s , provided ( [ aflt ] ) is satisfied . if agt conjecture is true , then the aflt states , @xmath46 s defined in ( [ aflt ] ) , form an orthogonal basis . [ bifund ] _ y_1,y_2p_y_1,y_2&= & z_bifund^u(2 ) inst(,,,;0 ) _ , . * proof * : we proceed , from ( [ aflt ] ) , [ agtbifund ] & & _ y_1,y_2p_y_1,y_2= _ y_1,y_2v_=0_y_1,y_2 + + & = & _ y_1\{q- [ ( a_y_1 + 1)b^-1- l_y_1 b]}_y_1\{(a_y_1 + 1)b^-1- l_y_1b } + & & _ y_1\{q- [ 2p+(a_y_1 + 1)b^-1- l_y_2 b]}_y_2\{-2p+(a_y_2 + 1)b^-1- l_y_1b } + & & _ y_2\{q- [ -2p+(a_y_2 + 1)b^-1- l_y_1 b]}_y_1\{2p+(a_y_1 + 1)b^-1- l_y_2b } + & & _ y_2\{q- [ ( a_y_2 + 1)b^-1- l_y_2 b]}_y_2\{(a_y_2 + 1)b^-1- l_y_2b}.we shall prove now that under this situation , if the result is non - zero , one can conclude@xmath132 if @xmath133 one gets @xmath134 to label the @xmath135-th part of the partition @xmath136 . ] since otherwise there must exist a box in the tableau @xmath136 satisfying @xmath137 this will lead to q - [ ( a_y_1 + 1)b^-1- l_y_1b ] = 0,this argument cycles and one finally conclude : @xmath138 for @xmath139 , similarly , the argument follows , and gives : @xmath140 however , the original condition @xmath141 then forces @xmath142 . q.e.d . the orthogonality condition , ( [ bifund ] ) , strongly suggests the existence of mutually commuting hermitian operators , whose common eigenstates form a complete orthogonal basis of the hilbert space . one of the operators , called the energy operator , probably cubic in @xmath88 s and @xmath108 s ( since this is most likely the case beyond @xmath121 ) , is the first object we are going to construct . however , hermiticity alone is not enough to constrain the possible forms of the construction . for example , @xmath125 is hermitian , but does not belong to @xmath0 . in fact , we shall show that altogether there should be at least 3 conditions @xmath55 are to meet in order the orthogonality of the @xmath62 eigenstates play an important role here . i)&&hermiticity + ii)&&triangularity + iii)&&reflection - invariance . now we explain what the other two conditions means . condition ii ) , triangularity , means that @xmath55 , in its matrix form , @xmath143 , where @xmath144 s are the eigenstates of @xmath145(the `` unperturbed '' energy operator ) eigenvalue equals @xmath24 . ] , is lower(or upper)-triangular with @xmath146 strictly triangular ( with zero diagonal entries ) . under such circumstances , the spectrum of @xmath55 coincides with that of @xmath62 , and the eigenstates of @xmath55 can be expressed as @xmath147 , here the normalization constant @xmath148 will be specified later on . @xmath149 , a unitriangular matrix , is again triangular with identity diagonal entries following the triangularity of @xmath55 . @xmath60 is the collective mode hamiltonian for the calogero - sutherland model in terms of the oscillators @xmath150 s . thus the eigenstate of @xmath62 is just @xmath151 . @xmath60 in our formalism ( including the zero modes @xmath65 ) is defined as @xmath152 since @xmath153 , we have @xmath154 . there is a natural question on how to define the inner products between @xmath84 s . the answer is that we need the condition iii ) reflection - invariance . notice that @xmath155 not zero means @xmath156 . in order to get a non - vanishing result , we need to shift @xmath157 to @xmath158 . we thus expect that there exists an operation which changes @xmath159 to @xmath160 . we call this operation reflection following the terminology in a similar situation in @xcite . actually , by looking closer to the nic formula i.e. the r.h.s of ( [ aflt ] ) , one can find that there exist an apparent symmetry _ y_1,y_2_y_2,y_1 . if we change either bra state @xmath161 to @xmath160 , or ket state @xmath162 to @xmath163 , on the l.h.s . of ( [ aflt ] ) , the factors on the r.h.s . of eq([aflt ] ) get reshuffling but the final result keep invariant . we may name this symmetry `` reflection '' or `` flipping '' symmetry . on the 2d cft side , from general reasoning that such an operation should be conformally invariant , it is natural to identify the insertions of the screening charges as this `` reflection '' operation . for liouville theory ( or coulomb gas model ) , we can attach to @xmath164 some screening charges for simplicity . ] , such that _ y_1,y_2v_0s^n_y_1,y_20 + s = e^2ib(z)z .now the neutrality condition forces @xmath165 if this is satisfied , then felder s contour for the integration of the screening charges actually closes and @xmath166 becomes a floating charge@xcite . now @xmath166 can move away from @xmath45 and communing through @xmath89 s and finally acts on the vacuum state @xmath157 . since @xmath166 acts by not changing the conformal weight , we deduce @xmath167 for a suitable normalization of @xmath166 . similar arguments apply to the case of @xmath168 , and one can always move a subset of screening charges , @xmath169 away from @xmath48 . since agt duality formula is valid for any @xmath24 , we may assume that @xmath2 can take arbitrary real value , as analytical continuation away from integer @xmath2 . this flipping is due to the fact that @xmath166 can be detached from @xmath170 , and act on the vacuum directly . similar operation exists in felder s brst cohomology @xcite . we are going to identify the reflection symmetry in nic as the hamiltonian symmetry in 2d cft for the insertions of the screening charges @xmath166 with @xmath171 . since hamiltonian @xmath172 , satisfies @xmath173=0 $ ] it has the property of double degeneracy . so @xmath166 with @xmath171 should map one aflt state to its partner state . if we require _ y_1,y_2h & = & _ y_1,y_2e_y_1,y_2 + _ y_1,y_2s^n h & = & _ y_1,y_2s^n e_y_1,y_2 , then we can identify @xmath174 since reflection symmetry means @xmath175 . notice that nothing has changed for the @xmath16 part . define @xmath176 then we have throughout this paper . ] : @xmath177 which obviously shows that @xmath178 . so reflection invariance means that we can identify the inner product @xmath179 with either @xmath180 or @xmath181 by the incertions of screening charges satisfying @xmath182 . having determined that @xmath46 form a normalizable orthogonal basis , the next step is the determination of their normalization . before doing this , let s review the so - called carlsson - okounkov formula@xcite which is useful for our formulation . first , define e= 1+e_1 + e_2+ = e^-_np_n = e^-_(-)(-1 ) s , @xmath183 , one may identify @xmath184 with @xmath185 , @xmath186 and @xmath187 with @xmath188 . here our convention is that @xmath189 creates a state @xmath190 . as a consequence , @xmath191 is to be identified with @xmath192 . such kind of identification is justified because they share the same values of their inner products . for more details see @xcite.]which is a vertex operator , and also a generating function for @xmath193@xmath194 here @xmath195 are elementary symmetric functions , @xmath196 is the power sum symmetric function . then e^-_(-)(-1 ) = _ n _ n p_-1^n . the conjugation of @xmath197 reads e^ = e^_(+)(-1),and we have @xmath198 now the carlsson - okounkov formula reads & & e^m ( e^)^-m-1 j_-y_1 , j_-y_2 + & = & ( -)^|y_1|^-|y_1|-|y_2|_y_1(m+(a_y_1 + 1)+l_y_2)_y_2(m - a_y_2-(l_y_1 + 1 ) ) + & = & j_y_1 e^-m-1(e^)^m j_-y_2 + & = & j_y_1e^(-k+k^-1+)_(-)(-1)e^_(+)(-1)j_-y_2 . for liouville theory , if we set @xmath200 , then the carlsson - okounkov formula reads & & j_y_1e^i(q-2)_(-)(-1)e^-2i_(+)(-1)j_-y_2 + & = & ( -)^|y_2|b^-|y_1|-|y_2|_y_1(-2+(a_y_1 + 1)b^-1-l_-y_2b ) _ y_2(-2-a_y_2b^-1+(l_y_1 + 1)b ) . the normalization of the aflt states is inherited from aflt s version of the agt duality formula , ( [ aflt ] ) and the orthogonality condition , ( [ agtbifund ] ) , [ norm1 ] & & _ y_1,y_2p_y_1,y_2= _ y_2,y_1-p_y_1,y_2 + & = & _ y_1\{-a_y_1b^-1+(l_y_1 + 1)b}\{(a_y_1 + 1)b^-1-l_y_1b } + & & _ y_2\{-a_y_2b^-1+(l_y_2 + 1)b}\{(a_y_2 + 1)b^-1-l_y_2b } + & & _ y_1\{-2p - a_y_1b^-1+(l_y_2 + 1)b)}_y_2\{-2p+(a_y_2 + 1)b^-1-l_y_1b } + & & _ y_2\{2p - a_y_2b^-1+(l_y_1 + 1)b)}_y_1\{2p+(a_y_1 + 1)b^-1-l_y_2b } + & = & ( -)^|y_1|+|y_2|j_y_1j_y_2 + & & _ y_1\{-2pb- a_y_1+(l_y_2 + 1)b^2}_y_2\{-2pb+(a_y_2 + 1)-l_y_1b^2 } + & & _ y_2\{2pb - a_y_2+(l_y_1 + 1)b^2}_y_1 \{2pb+(a_y_1 + 1)-l_y_2b^2 } + & & j_y_1j_y_2_y_2 , y_1(-p ) _ y_1 , y_2(p ) + & = & j_y_1j_y_2 j_y_2 e^i(q-2p)_(-)(-1)e^-i2p_(+)(-1 ) j_-y_1 + & & ( b^4)^|y_1|+|y_2|j_y_1 e^i(q+2p)_(-)(-1)e^i2p_(+)(-1 ) j_-y_2 . in reaching the last line in the above eqation , carlsson - okounkov formula has been applied , and we have defined [ norm2 ] _ y_1 , y_2(p ) & = & ( -)^|y_1|b^|y_1|+|y_2|_y_1(2p+(a_y_1 + 1)b^-1-l_y_2b ) _ y_2(2p - a_y_2b^-1 + ( l_y_1 + 1)b ) + & = & ( -b^2)^(|y_1|+|y_2|)j_y_1e^i(q+2p)_(-)(-1)e^i2p_(+)(-1 ) j_-y_2 + & = & b^2(|y_1|+|y_2|)j_y_1e^i(q+2p)_(-)(1)e^i2p_(+)(1 ) j_-y_2 + & & _ ( p ) . notice that @xmath201 is just a generalization of @xmath202 defined in @xcite . now we come to our main problem of the construction of the hamiltonian @xmath55 with the requirement that its eigenstates be identified with alft states satisfying ( [ aflt ] ) . we prefer to work first on the basis of jack symmetric functions @xmath203 , which already form an orthogonal basis . we found that if @xmath67 matrix elements are strictly triangular on this basis , then the orthogonality of the @xmath87 eigenstates follows immediately from the orthogonality of the @xmath62 eigenstates . this is just the simplest way to go from one orthogonal basis to another one . to see this , let s introduce an operator @xmath204 which map @xmath62 eigenstates to @xmath55 eigenstates , with @xmath148 the normalization constant [ roperator ] _ y_1,y_2 & = & r(e ) j_-y_1^+ j_-y_2 ^ -_y_1,y_2(p ) + _ y_2,y_1 & = & j_y_2^- j_y_1^+r(e)^ _ y_2y_1(-p ) + r(e)&=&1+= 1+(e ) , where the reflection symmetry has been applied to the aflt states @xmath205 and @xmath206 is strictly lower(or upper)-triangular @xmath207 @xmath208 . the way @xmath204 is expanded in ( [ roperator ] ) follows from the normalization condition , ( [ norm1]-[norm2 ] ) as we shall see in ( [ norm3 ] ) . the hermitian operator @xmath55 should satisfy : h_y_1,y_2 & = & e_y_1,y_2(p)_y_1,y_2 + _ y_2,y_1h & = & _ y_2,y_1e_y_2 , y_1(-p ) , where the energy eigenvalue has the double degeneracy : e_y_2 , y_1(-p ) = e_y_1,y_2(p ) , due to the orthogonality condition , [ norm3 ] _ y_1,y_2p_y_1 , y_2 = _ y_2 , y_1p_y_1 , y_2_,.then the next step is to determine if we get the right normalization for @xmath162 [ norm3 ] _ y_2,y_1_y_1,y_2 & = & j_y_2 ^ -j_y_2^+ r(e)^ r(e ) j_-y_1^+ j_-y_2 ^ - _ y_1 , y_2(p)_y_2 , y_1(-p ) + & = & j_y_2 ^ -j_y_1^+(1+(e)^)(1+(e))j_-y_1^+j_-y_2 ^ -_y_1 , y_2(p)_y_2 , y_1(-p ) + & = & _ y_1 , y_2(p)_y_2 , y_1(-p ) + & & + & = & _ y_1 , y_2(p)_y_2 , y_1(-p)j_y_1j_y_2.it is in agreement with ( [ norm1 ] ) . in deriving this we have used the fact that if @xmath204 is a unitriangular matrix , [ raction ] r(e)j_-y_1^+j_-y_2 ^ - = j_-y_1^+j_-y_2 ^ -+ _ c|y_1|>|y_1| + r_y_1 , y_2^y_1 y_2(e)j_-y_1^+j_-y_2^-,then it is easy to check that & & j_y_2 ^ -j_y_1^+(e)^j_-y_1^+j_-y_2 ^ - + & & = j_y_2 ^ -j_y_1^+(e)j_-y_1^+j_-y_2 ^ - + & & = j_y_2 ^ -j_y_1^+(e)^(e)j_-y_1^+j_-y_2 ^ - + & & = 0 . now we summarize the requirements for @xmath204 @xmath209 = 0 , \ \ \ \ e_{y_1 , y_2}(p)= e_{y_2 , y_1}(-p)\end{aligned}\ ] ] this means that the hamiltonian @xmath55 should also be triangular , and @xmath67 strictly triangular , h & = & h^+ + h^- + h_i + h^&=&(z_z^)^3 + _ n>0q n a_-n^a_n^,here @xmath67 is to be determined later on . + h^-&= & + + & & _ n>0q n a_-n^+a_n^++_n>0q n a_-n^-a_n^- + & = & + & + & _ n>02q n(a_-na_n+ c_-nc_n ) + & = & ( z_z)^3 + _ n>02q n a_-na_n - 2i ( z_z)(z_z)^2 + 2_n>0 q n c_-nc_n . now the requirement that @xmath55 commute with @xmath166 is equivalent to say that @xmath55 can be written in terms of @xmath89 s and @xmath88 s . to make @xmath55 triangular , we may try h_i & & _ n q n a_-n^+ a_n^- + & = & _ n q n ( a_-na_n - c_-nc_n - a_-nc_n + c_-na_n ) + & = & _ n q n ( a_-na_n - c_-nc_n)+qz_z(z_z)^2 . if we now make use of ( [ tz1 ] ) and choose @xmath210 then we get h&= & -(z_z)^3 + 4q _ n^+n a_-n a_n - 2i(z_z ) z^2 t(z ) + 2ia_0 + & = & -(z_z)^3 + 4q _ n^+n a_-n a_n - 2i_na_-nl_n+2ia_0 + & = & -i\{_n , m^+(a_-n - m^+ a_n^+ a_m^+ + a_-n - m^-a_n^-a_m^- ) . + & + & ._n , m^+(a_-n^+ a_-m^+ a_n+m^+ + a_-n^-a_-n^-a_n+m^- ) } + & + & _ n^+ q n ( a_-n^+a_n^++a_-n^-a_n^- + 2a_-n^+a_n^- ) + & + & _ n^+-2i a_0^+a_-n^+a_n^+ - 2i a_0 ^ - a_-n^- a_n^- -((a_0^+)^3+(a_0 ^ -)^3 ) .clearly , @xmath55 indeed satisfies the three requirements proposed in the previous section . i)&&hermitian + ii)&&triangular + iii)&&reflection - invarint . besides , we found that @xmath211 , where @xmath69 is defined in appendix c of @xcite as one of the infinitely many commuting operators which may makes the system integrable . the authors of @xcite have checked for the first a few levels that the aflt states , @xmath162 , which satisfies agt duality formula , ( [ aflt ] ) , are also the eigenstates of @xmath69 . but a general formula for the @xmath69 eigenstates is missing in @xcite . now the next question is : how to find all the eigenstates of @xmath55 ? first , let s consider @xmath212 h^+ & = & -i _ n , m^+ \{a_-n - m^+ a_n^+ a_m^+ + a_-n^+a_-m^+ a_n+m^+ } + & + & _ n^+ \ { n q a_-n^+ a_n^+ + 2a^+_0(-i ) a_-n^+a_n^+ } - , its eigenvalue @xmath213 @xmath214 here we have assumed the zero modes take the following eigenvalues , a_0=ip^a , c_0=ip^c , a^_0=ip^=i(p^ap^c ) for the bi - jack system , we have the following eigenequation , @xmath215 triangularity means @xmath216 since @xmath55 can be constructed in terms of @xmath89 s and @xmath88 s , so @xmath217 dose not change the eigenvalue . but @xmath166 changes @xmath218 and @xmath219 and from @xmath220 . we conclude [ sn ] s^n|p^+,p^-_y_1,y_2|p^-,p^+_y_2,y_1 . next , since @xmath221 does not play any important role , we may consider it as a gauge symmetry and can be fixed to any desired value . for convenience , we fix @xmath222 , hence , @xmath223 , @xmath224 and @xmath225 here @xmath226 if we define @xmath227 then we infer from from ( [ sn ] ) @xmath228 with the proper normalization for @xmath166 . now we are going to determine @xmath204 which satisfies @xmath229 @xmath230 here @xmath231 . @xmath204 defined in such a way should be understood as @xmath232 * proof * : first , we rewrite @xmath55 as @xmath233 then from @xmath234 one gets @xmath235 since @xmath236 is an eigenstate of @xmath62 with eigenvalue @xmath237 , we have @xmath238 hence , we conclude that @xmath239 is an eigenstate of @xmath55 with eigenvalue @xmath197 , @xmath240 q.e.d . + now we shall address the question raised in @xcite on the possible degeneracy of @xmath55 . the authors of @xcite , argued that @xmath69 has some degeneracy at level 4 and higher . we have analyzed what causes such kind of degeneracy . after analyzing the spectrum of @xmath55 , we believe that such degeneracy happens when @xmath241 , and we have @xmath242 , @xmath243 this can happen , for @xmath244 , first at level 4 , @xmath245 , and @xmath246 , @xmath247 . such degeneracy can happen at any even level higher or equal to 4 . for example at @xmath248 : @xmath249 or simply , we have @xmath250 pair @xmath251 such degeneracy does not cause any problem in constructing the eigenstate of @xmath55 for the following reasons . \i ) the mother state @xmath252 is uniquely determined by the young diagram , even for the degenerate @xmath197 . \ii ) consider power expansion @xmath253 for an intermediate state . @xmath254 since @xmath255 , @xmath256 and @xmath257 because of strictly triangularity of @xmath67 , so for a general value of @xmath24 , @xmath258 is not singular , and @xmath259 is well defined . \iii ) the construction given above leads to the orthogonality of the state @xmath260 for distinct @xmath261 even for the degenerate values of @xmath197 , cf . eqs.([norm1],[norm2],[norm3 ] ) . \iv ) it can be proven that the eigenstate of @xmath55 , constructed as in proposition 2 , is actually the common eigenstate for all the conseved charges which commute with @xmath55 , with the mild assumption that all the conserved charges are triangular in a similar way as @xmath55 is . due to lack of space for the present paper , we shall give a proof on this statement elsewhere . finally , we shall make a comment on the possible poles of @xmath204 in the complex @xmath262plane . the @xmath204 matrix elements is calculated based on the following formula , @xmath263 which always ends up with finite order perturbation because @xmath67 is strictly triangular . we found , by the explicit calculations carried out so far , that there is no pole in the finite @xmath24 complex plane . the poles in @xmath204 either cancels the zeros in @xmath264 or simply cancels by summing over all the relevant terms . of course , this property is also the necessary condition if @xmath162 s satisfy ( [ aflt ] ) . now the general aflt state can be written as [ jackconstruction]_y_1 , y_2 & = & \{_y_1,y_2(p)j_-y_1^+j_-y_2 ^ - + _ c_y_1,y_2^y_1,y_2 j_-y_1^+j_-y_2^- . + & + & _ c_y_1,y_2^y_1,y_2 j_-y_1^+j_-y_2^- + & + & .+_|y|=|y_1|+|y_2|c_y_1,y_2^y , j_-y^+},here @xmath265 is the transition coefficient which measures the changing from the young tableau vector @xmath250 to @xmath266 . we have calculated those coefficients up to level 4 , the explicit results(up to level 3 ) are included in appdix a. with the coefficients we calculated , one can check that : [ checkagt ] & & z_bif(|p , ; p , ) = + & & _ ( x_1,x_2),(y_1,y_2 ) j^-_x_1j^+_x_2 c_x_1,x_2^x_1,x_2v_c_y_1,y_2^y_1,y_2 j_-y_1j_-y_2,holds true , thus ( [ aflt ] ) is verified . here for simplicity , we have only verified the cases without the incertions of the screening charges , i.e. @xmath267 . the present work can be generalized in different ways . first , since the one parameter jack symmetric function is a special limit of the two parameter macdonald symmetric function , we expect that much of our work can be generalized to the cases where macdonald symmetric function plays a role . in that case , we expect a similar relation to the nic for 5d theory . second , the calogero - sutherland model is an integrable system . and consequently , jack symmetric function is the common eigenstate of the infinitely many commuting charges which are deformed @xmath268 charges . and for the construction of the aflt states , the conserved charges are further deformed from those for the jack symmetric functions . the final construction should give the same results as @xmath269 proposed in @xcite , which are constructed from integrable kdv equations . we find in this case , the aflt states remain to be the eigenstates for all the conserved charges . however , it is desirable to have infinitely many conserved charges constructed explicitly . third , the reflection symmetry studied in this paper is actually powerful enough to give a closed form for the construction of the aflt states . we shall present this result in our future work . another interesting idea related to our work is to consider the jack function as a perturbation away from the schur function , we have found that similar formalism applies @xcite . finally , it is very interesting to see how we present the full pants diagram for the conformal blocks , comparing to the one we have considered with one external leg . we are grateful to yi - hong gao for detailed explanations on the the agt conjecture and the related topics . + this work is part of the cas program `` frontier topics in mathematical physics '' ( kjcx3-syw - s03 ) and is supported partially by a national grant nsfc(11035008 ) . now we give the explicit construction of the aflt states up to level 3 . the transition coefficients @xmath270 are defined as , @xmath271 a. a. belavin , a. m. polyakov and a. b. zamolodchikov , `` infinite conformal symmetry in two - dimensional quantum field theory , '' nucl . b * 241 * , 333 ( 1984 ) . h. sonoda , `` sewing conformal field theories , '' nucl . * b311 * , 401 ( 1988 ) . v. a. alba , v. a. fateev , a. v. litvinov and g. m. tarnopolsky , `` on combinatorial expansion of the conformal blocks arising from agt conjecture , '' arxiv:1012.1312 [ hep - th ] . g. felder , `` brst approach to minimal methods , '' nucl . * b317 * , 215 ( 1989 ) . l. f. alday , d. gaiotto and y. tachikawa , `` liouville correlation functions from four - dimensional gauge theories , '' lett . math . phys . * 91 * , 167 ( 2010 ) [ arxiv:0906.3219 [ hep - th ] ] . n. a. nekrasov and s. l. shatashvili , `` quantization of integrable systems and four dimensional gauge theories , '' arxiv:0908.4052 [ hep - th ] . n. a. nekrasov , `` seiberg - witten prepotential from instanton counting , '' adv . theor . math . * 7 * , 831 ( 2004 ) [ arxiv : hep - th/0206161 ] . n. nekrasov and a. okounkov , `` seiberg - witten theory and random partitions , '' arxiv : hep - th/0306238 . r. dijkgraaf and c. vafa , `` toda theories , matrix models , topological strings , and n=2 gauge systems , '' arxiv:0909.2453 [ hep - th ] . m. f. atiyah , n. j. hitchin , v. g. drinfeld , y. .i . manin , `` construction of instantons , '' phys . * a65 * , 185 - 187 ( 1978 ) . l. f. alday and y. tachikawa , `` affine sl(2 ) conformal blocks from 4d gauge theories , '' lett . * 94 * , 87 ( 2010 ) [ arxiv:1005.4469 [ hep - th ] ] . m. c. n. cheng , r. dijkgraaf and c. vafa , `` non - perturbative topological strings and conformal blocks , '' arxiv:1010.4573 [ hep - th ] . h. itoyama and t. oota , `` method of generating q - expansion coefficients for conformal block and n=2 nekrasov function by beta - deformed matrix model , '' nucl . b * 838 * , 298 ( 2010 ) [ arxiv:1003.2929 [ hep - th ] ] . a. belavin and v. belavin , `` agt conjecture and integrable structure of conformal field theory for c=1 , '' nucl . b * 850 * , 199 ( 2011 ) [ arxiv:1102.0343 [ hep - th ] ] . v. belavin and b. feigin , `` super liouville conformal blocks from n=2 su(2 ) quiver gauge theories , '' arxiv:1105.5800 [ hep - th ] . n. dorey , s. lee and t. j. hollowood , `` quantization of integrable systems and a 2d/4d duality , '' arxiv:1103.5726 [ hep - th ] . h. awata , h. fuji , h. kanno , m. manabe and y. yamada , `` localization with a surface operator , irregular conformal blocks and open topological string , '' arxiv:1008.0574 [ hep - th ] . a. mironov , a. morozov and s. shakirov , `` on dotsenko - fateev representation of the toric conformal blocks , '' j. phys . a * 44 * , 085401 ( 2011 ) [ arxiv:1010.1734 [ hep - th ] ] . a. mironov , a. morozov and s. shakirov , `` conformal blocks as dotsenko - fateev integral discriminants , '' int . j. mod . a * 25 * , 3173 ( 2010 ) [ arxiv:1001.0563 [ hep - th ] ] . j. f. wu , `` note on refined topological vertex , jack polynomials and instanton counting , '' arxiv:1012.2147 [ hep - th ] . j. f. wu , y. y. xu and m. yu , `` recursions in calogero - sutherland model based on virasoro singular vectors , '' arxiv:1107.4234 [ hep - th ] . j. f. wu and m. yu , work in progress .

we give a possible generalization of a lutz twist to all dimensions . this reproves the fact that every contact manifold can be given a non - fillable contact structure and also shows great flexibility in the manifolds that can be realized as cores of overtwisted families . we moreover show that has at least three distinct contact structures .

0903.0295

lutz twists have been a fundamental tool in studying contact structures in dimension 3 . they have been used to construct contact structures on all closed oriented 3-manifolds and to manipulate the homotopy class of plane field of a given contact structure . in particular , lutz @xcite used this construction to give the first proof that all homotopy classes of plane fields can be realized by contact structures . we recall that a lutz twist alters a contact structure on a solid torus neighborhood of a transverse curve by introducing an @xmath1-family of overtwisted disks ( see below for a precise definition ) . an overtwisted disk in a contact 3-manifold @xmath2 is an embedded disk in @xmath3 that is tangent to @xmath4 only along its boundary and at one interior point . a contact manifold is called overtwisted if it contains an overtwisted disk and otherwise it is called tight . starting with eliashberg s fundamental paper @xcite defining the tight versus overtwisted dichotomy , these notions have taken a central role in 3 dimensional contact geometry . overtwisted contact structures are completely classified @xcite and exhibit a great deal of flexibility , appearing to be fairly topological in nature . much of the insight into such structures has come from careful analysis of the lutz twist construction and natural questions that arose from it . there is not a great deal known about contact structures in dimensions above 3 . specifically , there are few constructions of contact structures and few tools to manipulate a given contact structure . we introduce one such tool by giving a possible generalization of a lutz twist to all odd dimensions . as a consequence we reprove and slightly strengthen a result proved by niederkrger and van koert in @xcite that every @xmath5dimensional manifold that has a contact structure can be given a non - fillable contact structure . the proof in fact produces an embedded overtwisted family ( that is a plastikstufe in the language of @xcite ) by changing the given contact structure in a small neighborhood of any @xmath6dimensional isotropic submanifold @xmath7 ( with trivial conformal symplectic normal bundle ) . the overtwisted family is modeled on @xmath7 ( that is @xmath7 is the elliptic singular locus of the family , see subsection [ sec : families ] below ) . this construction is analogous to creating an overtwisted disk in dimension three by performing a lutz twist along a knot in the neighborhood of a point ( an overtwisted disk is a 3dimensional overtwisted family ) . overtwisted families , in dimensions above three , were first considered in @xcite as an obstruction to symplectic fillability of a contact structure , though precursors of them go back to gromov s work @xcite . in @xcite presas gave the first examples of overtwisted families in a closed contact manifold of dimension greater than three . our main result is the following . [ main ] let @xmath2 be a contact manifold of dimension @xmath8 and let @xmath7 be an @xmath6dimensional isotropic submanifold with trivial conformal symplectic normal bundle . then we may alter @xmath4 in any neighborhood of @xmath7 to a contact structure @xmath9 that contains an overtwisted family modeled on @xmath10 moreover , we may assume that @xmath9 is homotopic to @xmath4 through almost contact structures . a corollary is the following result originally proven , modulo the statement about the homotopy class of almost contact structures and core of the overtwisted family , via a delicate surgery construction in @xcite based on subtle constructions in @xcite . we call a contact structure if it contains an overtwisted family . ( we use the adjective `` weakly '' since it is not clear that this is the most appropriate generalization of overtwisted to dimensions above 3 , though it does share some properties with 3 dimensional overtwisted contact manifolds . ) [ maincor ] every odd dimensional manifold that supports a contact structure also supports a weakly overtwisted , and hence non ( semi - positive ) symplectically fillable , contact structure in the same homotopy class of almost contact structure . moreover , we can assume the overtwisted family is modeled on any @xmath6dimensional isotropic submanifold with trivial conformal symplectic normal bundle . we also observe the following non uniqueness result which can also be found in @xcite . [ nonunique ] there are at least three distinct contact structures on @xmath11 . we remark that our proof relies on cut - and - paste techniques and branch cover techniques that seem to be new to the literature . these techniques should be useful in constructing contact structures on higher dimensional manifolds and will be more fully explored and systematized in a future paper . _ acknowledgements . _ the first author was partially supported by nsf grant dms-0804820 . we are also grateful for valuable comments klaus niederkrger made on a first draft of this note . we also thank the original referees of the paper who pointed out a gap in the original proof of our main result and made many other valuable suggestions concerning the manuscript and exposition . in this section we recall some well known results and establish notation necessary in the rest of the paper . specifically in subsection [ sec : nbhds ] we prove various darboux type theorems about contact structures that agree on compact subsets . in subsection [ sec : families ] we define overtwisted families . this definition involves the `` characteristic distribution '' of a submanifold of a contact manifold and has various equivalent formulations , just as there are several equivalent definitions of overtwisted disks in a contact 3-manifold . to clarify these equivalent formulations we discuss characteristic distributions on subsection [ sec : distributions ] . in subsection [ sec : almost ] we recall a few basic facts about almost contact structures . finally in subsection [ sec:3lutz ] we recall the notion of lutz twist in dimension 3 and set up notation that will be used in the following sections . a simple application of a moser type argument yields the following result . [ thm : generalnbhd ] let @xmath12 be a compact submanifold of @xmath3 and let @xmath13 and @xmath14 be two oriented contact structures on @xmath3 such that @xmath15 moreover , assume we have contact forms @xmath16 for @xmath17 such that @xmath18 then there are open neighborhoods @xmath19 and @xmath20 of @xmath12 and a contactomorphism @xmath21 that is fixed on @xmath22 set @xmath23 noting that @xmath24 along @xmath12 and @xmath25 on @xmath12 we see that @xmath26 is contact in some neighborhood of @xmath22 thus @xmath27 now gives a family of contact structures in a neighborhood of @xmath12 that agree along @xmath22 a standard application of moser s argument @xcite now give a family of open neighborhoods @xmath28 of @xmath12 and maps @xmath29 fixed along @xmath12 such that @xmath30 for some positive functions @xmath31 setting @xmath32 and @xmath33 the map @xmath34 is the desired contactomorphism . recall if @xmath2 is a contact manifold with contact structure @xmath4 and @xmath35 is a ( locally defined ) contact form for @xmath4 then for all @xmath36 the 2form @xmath37 is a symplectic form on @xmath38 since any other contact form defining @xmath4 differs from @xmath35 by multiplication by a positive function ( we always assume a contact form for @xmath4 evaluates positively on a vector positively transverse to @xmath4 ) , we see there is a well - defined conformal symplectic structure on @xmath39 a submanifold @xmath40 is called if @xmath41 for all @xmath42 if @xmath3 has dimension @xmath8 then the dimension of an isotropic @xmath43 must be less than or equal to @xmath44 since @xmath45 is an isotropic subspace of the symplectic space @xmath46 if the dimension of @xmath43 is @xmath44 then @xmath43 is called . given an isotropic @xmath43 its is the quotient bundle with fiber @xmath47 where @xmath48 is the @xmath49-orthogonal subspace of @xmath45 in @xmath38 one may easily check that @xmath50 has dimension @xmath51 where @xmath52 is the dimension of @xmath43 and as bundles @xmath53 where the @xmath54 factor can be taken to be spanned by any reeb field for @xmath39 ( all bundle isomorphisms preserve conformal symplectic structures where they are defined . ) one may easily check that the bundle @xmath55 is isomorphic to @xmath56 so the only term on the left hand side that is not determined by the topology of @xmath43 is @xmath57 which depends on the isotropic embedding of @xmath43 in @xmath58 we now have the following result that easily follows from the above discussion and theorem [ thm : generalnbhd ] ( once one notices that the conformal symplectic normal bundles can be identified such that the symplectic structures induced by given contact forms agree ) . [ prop : csndet ] let @xmath59 and @xmath60 be two contact manifolds of the same dimension and let @xmath61 be an isotropic submanifold of @xmath62 if there is a diffeomorphism @xmath63 that is covered by a conformal symplectic bundle isomorphism @xmath64 then there are open sets @xmath65 of @xmath61 in @xmath66 and a contactomorphism @xmath67 that extends @xmath68 let @xmath69 be a @xmath70dimensional submanifold of the @xmath5dimensional contact manifold @xmath58 the singular distribution @xmath71 is called the . where the intersection is transverse the distribution has dimension @xmath72 we say @xmath69 is a if the non - singular ( that is transverse ) part of @xmath73 is integrable . we say @xmath69 is a if it is a foliated submanifold and the dimension of @xmath69 is @xmath74 ( so all the leaves of @xmath73 are locally legendrian submanifolds of @xmath2 ) . the characteristic distribution can be quite complicated as can be its singularities . here we clarify a few points that will show up in the definition of overtwisted families in the next subsection . this allows for more flexibility in the definition of overtwisted families which , in turn , makes working with overtwisted families easier . in particular we consider codimension 1 and 2 submanifolds of a maximally foliated submanifold @xmath69 that are tangent to @xmath39 by way of motivation we recall the 3-dimensional situation . in particular an overtwisted disk is usually defined to be a disk @xmath75 with characteristic foliation @xmath76 having @xmath77 as a leaf and a single elliptic singularity . alternately one could ask that there is a single elliptic singularity and @xmath77 is an isolated singular set . in particular the exact form of the foliation near @xmath77 or whether @xmath77 is a leaf or a singular set is irrelevant in the sense that given any overtwisted disk of a particular form near @xmath77 we can arrange any other suitable form . moreover , on the level of foliations there are many types of elliptic singularities , but again the exact form is irrelevant for the definition of an overtwisted disk . we will establish similar results for the characteristic distribution . suppose @xmath43 is a compact subset of the @xmath78-dimensional maximally foliated submanifold @xmath69 of the contact manifold @xmath58 further suppose @xmath43 is tangent to @xmath4 and has dimension @xmath79 thus @xmath43 is a legendrian submanifold and hence has a neighborhood contactomorphic to a neighborhood of the zero section in the jet space @xmath80 and thus by proposition [ prop : csndet ] studying the characteristic distribution on @xmath69 near @xmath43 can be done by studying embeddings of @xmath81 into @xmath82 now given any closed manifold @xmath7 suppose that @xmath83 and that @xmath43 has a neighborhood @xmath84 $ ] ( or if @xmath43 is the boundary of @xmath69 then @xmath85 $ ] ) in @xmath69 with @xmath86 and such that @xmath87 is transverse to the foliation @xmath73 and @xmath88 is ( non - singularly ) foliated by leaves of the form @xmath89 there are two cases we wish to consider . the first is if @xmath43 is also a non - singular leaf of @xmath90 in this case @xmath91 is determined by its holonomy which is non - trivial only in the @xmath1 factor . it is not hard to check that any such holonomy can be realized by the foliation induced on the image of a map from @xmath92 $ ] to @xmath93 of the form @xmath94 where @xmath95 is the zero section of @xmath96 @xmath97 is a smooth 1-parameter family of sections of @xmath98 with @xmath99 being the zero section and @xmath100\to { \mathbb{r}}.$ ] notice that arguing as in the proof of proposition [ prop : csndet ] we can see that a neighborhood of @xmath43 in @xmath3 can be made isomorphic to a neighborhood of the zero section in @xmath101 so that @xmath12 goes to the image of the above map . in the second case suppose @xmath43 consists entirely of singular points of @xmath90 here we need to assume in addition that each leaf in @xmath88 is asymptotic to @xmath102 for some @xmath103 and distinct leaves have distinct asymptotic limits . in this case @xmath12 can be identified with the image of the map @xmath104 where @xmath105 is the section of @xmath106 that sends @xmath103 to @xmath107 one may easily see that , since the two different situations discussed above only differ by the mapping into @xmath108 we may deform one of these into the other through maps whose images have integrable characteristic distributions and thus given an @xmath109 as in either case we can replace it with the other keeping @xmath69 a maximally foliated submanifold . anytime we see an @xmath43 as in one of these situations we say the . this is analogous to the situation in dimension 3 where the boundary of a seifert surface for a legendrian knot with thurston - bennequin invariant 0 can be taken to be a leaf of the characteristic foliation or a circle of singularities . we now consider the case of a submanifold @xmath110 of @xmath69 of dimension @xmath111 that consists entirely of singularities of @xmath90 so @xmath110 is an isotropic submanifold of @xmath58 we also assume that @xmath110 is an , that is there are no other singularities of @xmath73 in some neighborhood of @xmath112 we call @xmath110 if the conformal symplectic normal bundle is trivial and @xmath113 for all @xmath114 thus we may find a product neighborhood @xmath115 of @xmath110 in @xmath69 such that @xmath116 is tangent to the conformal symplectic normal bundle along @xmath112 in this situation @xmath110 has a neighborhood in @xmath3 that is contactomorphic to a neighborhood of the zero section in @xmath117 in the contact manifold @xmath118 where @xmath119 is the coordinate on @xmath54 and @xmath120 is the unit disk in the plane with polar coordinate @xmath121 ( if this is not clear see the proof of lemma [ lem : nbhdoffamily ] below . ) moreover this contactomorphism takes @xmath69 to a submanifold of @xmath122 that is tangent to the zero section times @xmath120 along the zero section times @xmath123 we say @xmath110 is if @xmath69 in @xmath124 can be parameterized by a map of the following form @xmath125 where @xmath95 is the zero section of @xmath117 and @xmath126 is some function such that @xmath127 has graph tangent to the 0 map at @xmath128 for all @xmath129 since @xmath110 is an isolated singular set it is easy to see that the foliation induced on each @xmath130 has a non - degenerate singularity at the origin ( since the disk is tangent to the conformal symplectic normal bundle there ) and moreover the type , elliptic or hyperbolic , of the singularity can not change for different @xmath114 we call @xmath110 a provided the singularity on @xmath116 is elliptic . similarly , we call @xmath110 a provided the singularity on @xmath131 is hyperbolic . one may now easily check that if @xmath110 is an elliptic singular set then we may isotope @xmath69 near @xmath110 such that @xmath69 is still a maximally foliated submanifold , the topology of the leaves in @xmath132 has not changed and @xmath133 in @xmath134 let @xmath2 be a contact manifold of dimension @xmath135 an , a closed @xmath6dimensional manifold , ( originally called in @xcite ) is an embedding @xmath136 in @xmath3 , where @xmath120 is the unit disk in @xmath137 such that 1 . the characteristic distribution @xmath138 is integrable , 2 . @xmath139 is an isotropic submanifold and the singular set of @xmath140 , 3 . @xmath7 is a normally elliptic singular set of @xmath141 4 . @xmath142 is a leaf of @xmath140 , 5 . all other leaves of @xmath140 are diffeomorphic to @xmath143 and approach @xmath144 nicely near one end and approach the normally elliptic singularity @xmath7 at the other end . we sometimes call @xmath7 the _ core _ of the overtwisted family . it is easy to see from the discussion in the last section that we may assume that @xmath144 is also an isolated singular set of @xmath145 with leaves nicely approaching it , since given this we can slightly perturb @xmath145 near @xmath144 such that @xmath144 is a non - singular leaf of @xmath140 as in the definition above . a contact manifold @xmath2 of dimension @xmath8 is called if it contains an overtwisted family modeled on any @xmath6dimensional manifold . it is not clear if this is the correct generalization of overtwisted to higher dimensional manifolds , though it does have some of the properties of 3 dimensional overtwisted contact manifolds ; hence the adjective `` weak '' . in some papers this notion is referred to as though we prefer to use the more descriptive term `` weakly overtwisted '' . currently the main evidence that this is a good generalization of 3-dimensional overtwisted contact structures is the following theorem . if @xmath2 is a weakly overtwisted contact manifold then it can not be symplectically filled by a semi - positive symplectic manifold . if the dimension of @xmath3 is less than 7 then it can not be filled by any symplectic manifold . recall that a @xmath146dimensional symplectic manifold @xmath147 is called semi - positive if every element @xmath148 with @xmath149 and @xmath150 satisfies @xmath151 note all symplectic 4 and 6 manifolds are semi - positive as are stein and exact symplectic manifolds . it seems likely that the semi - positivity condition can be removed , but we do not address that issue here . recall that an ( oriented ) almost contact structure is a reduction of the structure group of a @xmath5-dimensional manifold @xmath3 to @xmath152 that is a splitting of the tangent bundle @xmath153 where @xmath154 is a @xmath155 bundle and @xmath54 is the trivial bundle . clearly a co - oriented contact structure induces an almost contact structure as it splits the tangent bundle into @xmath156 in dimension 3 any oriented manifold @xmath3 has an almost contact structure since the tangent bundle is trivial and the homotopy classes of almost contact structures are in one to one correspondence with homotopy classes of oriented plane fields . in higher dimensions the situation is more difficult . it is known , for example , that in dimensions 5 and 7 a manifold @xmath3 has an almost contact structure if and only if its third integral stiefel - whitney class vanishes : @xmath157 see @xcite . of course this condition is equivalent to the second stiefel - whitney class @xmath158 having an integral lift . in dimension 5 the homotopy classes of almost contact structures on a simply connected manifold are in one to one correspondence with integral lifts of @xmath159 the correspondence is achieved by sending an almost contact structure to its first chern class ( recall any @xmath155-bundle has chern classes ) . as we wish to generalize lutz twists from the 3dimensional setting we digress for a moment to recall this construction . consider the contact structures @xmath160 and @xmath161 on @xmath162 given , respectively , by @xmath163 and @xmath164 where @xmath165 is the coordinate on @xmath1 and @xmath166 are polar coordinates on @xmath167 let @xmath168 be the torus @xmath169 in @xmath162 with the contact structure @xmath160 and @xmath170 the same torus in @xmath171 together with the contact structure @xmath172 furthermore set @xmath173 ) to be the solid torus in @xmath174 bounded by @xmath168 with the contact structure @xmath160 and @xmath175 the same torus with contact structure @xmath172 finally set @xmath176 and similarly for @xmath177 if we are only concerned with the solid torus or thickened torus and not the contact structure on it we will drop the subscript from the notation . that is for example @xmath178 is the solid torus @xmath179 where @xmath180 is a disk or radius @xmath181 given any @xmath182 one can use the fact that @xmath183 takes on all positive values on @xmath184 and @xmath185 to see there is a unique @xmath186 $ ] and @xmath187 $ ] such that the characteristic foliation on @xmath188 is the same as the characteristic foliation on @xmath189 and @xmath190 since the characteristic foliation determines a contact structure in the neighborhood of a surface , one can find some @xmath191 with @xmath192 sufficiently small and an @xmath193 and @xmath194 such that there is a contactomorphism @xmath195 respectively @xmath196 from @xmath197 to @xmath198 respectively @xmath199 moreover , one may explicitly construct @xmath200 and @xmath201 in such a way that @xmath201 preserves the @xmath165 and @xmath103 coordinates and @xmath200 sends them to their negatives . now given a transverse curve @xmath202 in a contact 3manifold @xmath2 there is a neighborhood @xmath12 of @xmath202 in @xmath3 that is contactomorphic to @xmath203 in @xmath204 a on @xmath202 is the process of changing the contact structure @xmath4 by removing @xmath205 from @xmath3 and gluing in @xmath206 using @xmath201 to glue @xmath207 to @xmath208 similarly a ( or sometimes called ) is performed by gluing @xmath209 in place of @xmath12 using @xmath210 the subset @xmath211 is called a . we now review a similar construction . consider the manifold @xmath212 $ ] with coordinates @xmath213 a @xmath214form of the type @xmath215 will be contact if @xmath216 moreover the contact structure is completely determined , see @xcite , by the slope of the characteristic foliation @xmath217 on @xmath218 , the slope of the characteristic foliation @xmath219 on @xmath220 and the number , @xmath221 of times that @xmath222 for @xmath223 ( notice that the contact condition implies the curve @xmath224 is monotonically winding around the origin in @xmath225 and thus that @xmath226 is monotonically decreasing with @xmath227 , here of course slope @xmath228 is allowed . ) we say that the contact structure has giroux torsion @xmath229 we denote the corresponding contact structure by @xmath230 and any contact form for this contact structure of the form discussed above by @xmath231 to connect this new notation to our notation above we notice that @xmath232 above is contactomorphic to @xmath233,\xi_0^{(-a^2 , -b^2)})$ ] and @xmath234 is contactomorphic to @xmath235 , \xi_0^{(-a\tan a , -b\tan b)})$ ] if @xmath236 or , more generally , @xmath233 , \xi_n^{(-a\tan a , -b\tan b)})$ ] if @xmath237 notice that given a transverse curve @xmath202 as above , we can find in a neighborhood @xmath203 of @xmath202 a thickened torus @xmath212 $ ] with a contact structure @xmath238 let @xmath239 be the solid torus with contact structure obtained from the one on @xmath203 by replacing @xmath240 by @xmath241 it is easy to show that @xmath239 is isotopic to @xmath209 by an isotopy fixed on the boundary . so a full lutz twist can be achieved by replacing @xmath203 with @xmath242 ( one may similarly describe a half lutz twist . ) thus we may think of performing a lutz twist as adding giroux torsion along a compressible torus . we end this section by recalling that , up to contactomorphism , the tight contact structures on @xmath243 are @xmath244 where @xmath245 are the coordinates on @xmath243 and @xmath44 is positive , @xcite . notice that @xmath246 is obtained from @xmath247 by adding giroux torsion . an in a @xmath5-dimensional contact manifold @xmath2 is a smooth map @xmath248 such that @xmath249 is a curve transverse to @xmath4 for all @xmath250 and @xmath251 is an isotropic submanifold of @xmath2 for all @xmath252 we say the family is if @xmath200 is an embedding . theorem [ thm : generalnbhd ] easily yields the following result . [ lem : nbhdoffamily ] let @xmath2 be a contact manifold of dimension @xmath135 suppose we have an embedded isotropically parameterized family of transverse curves @xmath253 in @xmath254 where the dimension of @xmath7 is @xmath255 moreover assume that the isotropic submanifold @xmath256 has trivial conformal symplectic normal bundle . then @xmath253 has a neighborhood @xmath12 in @xmath254 contactomorphic to a neighborhood of the zero section in @xmath257 in the contact manifold @xmath258 where @xmath165 is the angular coordinate on @xmath259 @xmath120 is the unit disk in the plane with polar coordinate @xmath166 and @xmath260 is the canonical 1-form on @xmath261 choose a diffeomorphism @xmath262 from @xmath253 in @xmath3 to @xmath253 in @xmath263 that respects the product structure . we can choose the normal bundle @xmath264 to @xmath253 in both manifolds to be contained in the contact hyperplanes . as the conformal symplectic normal bundle to @xmath7 is trivial we have @xmath265 and @xmath266 thus extending our diffeomorphism @xmath262 to a neighborhood of @xmath253 we can assume that it takes the contact hyperplanes along @xmath253 in @xmath3 to the contact hyperplanes along @xmath253 in @xmath267 in addition , we can scale our bundle map along the conformal symplectic normal direction such that it actually preserves the symplectic structure induced by the contact forms . thus our extension of @xmath262 can be assumed to preserve the exterior derivative of our contact forms along @xmath268 now theorem [ thm : generalnbhd ] gives the desired contactomorphic neighborhoods . a neighborhood of @xmath253 as given in lemma [ lem : nbhdoffamily ] is contactomorphic to @xmath269 where @xmath270 is a neighborhood of the zero section in @xmath271 and @xmath272 is a disk of radius @xmath273 using the notation from subsection [ sec:3lutz ] , this is contactomorphic to @xmath274 denote by @xmath145 the smooth manifold @xmath275 with no particular contact structure on it . [ lem : main ] there is a contact structure on @xmath145 that agrees with the contact structure @xmath276 near the boundary and agrees with the one on @xmath277 on @xmath278 we define the of @xmath2 along @xmath279 to be the result of removing @xmath280 from @xmath3 and replacing it with the contact structure constructed in the lemma . we call the contact manifold @xmath281 with the contact structure defined in lemma [ lem : main ] , a . given that there is a lutz tube with core @xmath7 as claimed in lemma [ lem : main ] the following theorem is almost immediate . [ thm : genlutz ] let @xmath2 be a contact manifold of dimension @xmath135 suppose we have an embedded isotropically parameterized family of transverse curves @xmath253 in @xmath254 where the dimension of @xmath7 is @xmath255 moreover assume that the isotropic submanifold @xmath256 has trivial conformal symplectic normal bundle . then we may alter @xmath4 in any neighborhood of @xmath253 to a contact structure @xmath9 that is weakly overtwisted . moreover there is an @xmath1-family of overtwisted families modeled on @xmath7 in @xmath282 to see how this theorem generalizes a lutz twist in dimension 3 , notice that the only possibility for a connected @xmath7 in a contact 3-manifold @xmath2 is @xmath283 so the embedded isotropically parameterized family of transverse curves in this case is simply a transverse knot @xmath284 clearly the lutz tube @xmath285 is an @xmath1-family of overtwisted disks . the modification mentioned in the theorem is , of course , a ( full ) lutz twist . it is clear that this can be performed in any arbitrarily small neighborhood of @xmath268 we are left to check that we have an @xmath1-family of the embedded overtwisted families modeled on @xmath286 but this is obvious as one easily checks that @xmath287 where @xmath7 is thought of as the zero section of @xmath96 is an overtwisted family modeled on @xmath7 for each @xmath288 contained in @xmath145 . the purpose of this section is to motivate as well as set up the preliminaries for the proof of lemma [ lem : main ] which will establish the existence of the lutz tube with core @xmath10 we start by setting up some preliminary notation ( and will use notation established in subsection [ sec:3lutz ] ) . suppose we are given the standard contact structure @xmath289 where @xmath290 on @xmath291 we can choose @xmath292 and set @xmath293 where @xmath70 and @xmath52 are chosen such that @xmath294 @xmath295 and @xmath296 for @xmath297\cup [ \delta',\delta]$ ] and the curve @xmath224 winds around the origin once as @xmath227 runs from @xmath298 to @xmath299 let @xmath300 be a neighborhood of the zero section @xmath301 and denote the liouville form by @xmath302 choose some @xmath303 recall we want to replace @xmath304 on @xmath305 with @xmath306 we begin to define a contact structure @xmath4 on @xmath307 as follows . @xmath308 the left hand side of figure [ fig-1 ] shows the region where the contact structure is already defined . notice that @xmath304 and @xmath309 agree on @xmath310 and on @xmath311 and thus @xmath4 is well defined where it is defined . notice that in the case when @xmath7 is a point we have already defined the contact structure on all of @xmath145 and it clearly corresponds to the ( full ) lutz twist . when @xmath7 has positive dimension we claim that @xmath4 may be extended over the rest of @xmath312 that is we need to extend @xmath4 over @xmath313 to this end we notice that if we denote by @xmath314 the 1form @xmath260 restricted to the unit cotangent bundle @xmath315 of @xmath271 then there is a diffeomorphism from @xmath316 to @xmath317\times w$ ] that takes the 1form @xmath260 to @xmath318 where @xmath319 is the coordinate on @xmath317.$ ] moreover , @xmath320 can be written @xmath321\times s^1,$ ] with coordinates @xmath322 setting @xmath323 and @xmath324 we can write @xmath325 as @xmath326\times [ \delta'',\delta']\times s^1\times s^1.$ ] this identification is orientation preserving where we have sent @xmath319 to @xmath327 and the last three coordinates are @xmath328 near @xmath329 and @xmath330 we have ( the germ of ) a contact form defined as the kernel of @xmath331 where @xmath35 is either @xmath332 or @xmath333 of course this contact structure is also defined by @xmath334 so we see that we need to construct a contact structure on @xmath325 that is equal to the kernel of @xmath314 plus the symplectization of @xmath332 near @xmath335 and @xmath336 and equal to the kernel of @xmath314 plus the symplectization of @xmath337 near @xmath338 see the right hand side of figure [ fig-1 ] . more specifically we can assume that the neighborhoods where @xmath337 and @xmath332 agree and the coordinates on @xmath339 are chosen in such a way that near @xmath340\times s^1\times s^1 $ ] the 1form is diffeomorphic to @xmath341 ( see section [ sec:3lutz ] for the notation being used here . ) similarly near @xmath342\times s^1\times s^1 $ ] the 1form is diffeomorphic to @xmath343 now to motivate a possible approach for extending the contact structure @xmath4 over all of @xmath344 we notice that _ if _ we could construct an exact symplectic structure @xmath345 on @xmath346\times [ 0,1]\times t^2 $ ] such that near @xmath347\times t^2 $ ] the @xmath214form @xmath348 near @xmath349\times t^2 $ ] we have @xmath350 and near @xmath346\times \{0,1\}\times t^2 $ ] the vector field @xmath351 is an expanding vector field for @xmath345 then we could extend @xmath345 to an exact symplectic structure on @xmath352\times t^2 $ ] that looks like the symplectization of @xmath353 for negative @xmath319 and like the symplectization of @xmath354 for @xmath319 larger than @xmath214 . by rescaling the exact symplectic form if necessary and choosing @xmath191 and @xmath355 sufficiently far apart ( notice we can clearly do this as @xmath270 minus the zero section is exact symplectomorphic to @xmath356 for some @xmath357 ) there will be some subset of @xmath358\times t^2 $ ] with contact form @xmath359 that has a neighborhood of its boundary contactomorphic to a neighborhood of the boundary of @xmath360\times [ \delta'',\delta']\times s^1\times s^1.$ ] of course it is well known that such an exact symplectic cobordism ( even non - exact ) can not exist as it would allow one to construct symplectic fillings of overtwisted contact structures . but we will show below that there is a contact structure on @xmath361\times[0,1]\times t^2 $ ] that looks like this one near the boundary and hence we can finish the argument as above . from the discussion at the end of the last section it is immediate that lemma [ lem : main ] , and hence the existence of a lutz tube with core @xmath286 will be established once we demonstrate the following lemma . [ rmn ] in the notation of section [ sec:3lutz ] , there is a contact structure on @xmath362 \times [ 0,1 ] ) \times t^2 $ ] such that the following properties are satisfied : 1 . near @xmath363\times t^2 $ ] and @xmath361\times \{0,1\}\times t^2 $ ] the contact structure is contactomorphic to @xmath364 and 2 . near @xmath365\times t^2 $ ] the contact structure is contactomorphic to @xmath366 we establish lemma [ rmn ] by first considering a similar lemma for the manifold @xmath367 \times t^3 $ ] . here the new manifold can be thought of as being obtained from the old one by identifying the boundary of @xmath346 \times t^2 $ ] by the identity . once we prove the lemma stated below , lemma [ rmn ] will follow by removing a suitable portion of the manifold @xmath367 \times t^3.$ ] [ lem:3.5 ] there exists a contact structure on @xmath368\times s^1)\times t^2 $ ] such that near one boundary component @xmath369 the contact structure is contactomorphic to @xmath370 and near the boundary component @xmath371 it is contactomorphic to @xmath372 here @xmath373 and @xmath374 are the contact forms on @xmath243 defined in section [ sec:3lutz ] . consider @xmath375 with coordinates @xmath376 the 1form @xmath377 is the primitive of the symplectic form @xmath345 on @xmath378 moreover given any point @xmath379 in @xmath225 the lift of the radial vector field @xmath380 centered at @xmath379 in @xmath225 to @xmath375 is an expanding vector field for @xmath381 let @xmath382 be this disk of radius , say , 10 in @xmath383 and let @xmath384 the expanding vector field @xmath380 is transverse to @xmath385 so @xmath386 is a hypersurface of contact type . note @xmath243 in this context is naturally thought of as @xmath387 with coordinates @xmath388 the contact structure induced on @xmath385 is easily seen to be @xmath389 where @xmath390 parameterize an ellipse about the origin in @xmath167 thus the contact structure on @xmath243 is the unique strongly fillable contact structure @xmath14 on @xmath243 ( of course this is also obvious since @xmath382 is a strong filling of the contact structure ) . similarly if @xmath391 is a disk of radius one about @xmath379 times @xmath392 then it is also a strong symplectic filling of @xmath393 moreover , @xmath394 is an exact symplectic cobordism from the symplectization of @xmath395 ( this is the boundary component of @xmath394 coming from the boundary of @xmath391 ) to the symplectization of @xmath395 ( this is the boundary component of @xmath394 that is also the boundary of @xmath382 ) . see the left hand side of figure [ fig-2 ] . let @xmath396 and @xmath397 clearly @xmath154 is a contact form on @xmath398 we now need an auxiliary observation whose proof we give below . [ lem : aux ] in any open neighborhood of @xmath399 in @xmath400 @xmath110 may be isotoped such that it is a contact submanifold of @xmath398 it is well known , and easy to prove @xcite , that if @xmath401 is the two - fold cover branched over a contact submanifold @xmath110 then @xmath402 has a contact structure that agrees with the pullback contact structure away from a neighborhood of the branch locus . we briefly recall this construction as the details will be needed below . if @xmath35 is a contact form on @xmath403 then @xmath404 is a contact form on @xmath402 away from the branch locus . near the branch locus let @xmath405 be a connection @xmath214form on the normal circle bundle and let @xmath406 be a function that is 1 near the branch locus , zero outside a slightly larger neighborhood and radially decreasing in between . for small @xmath407 the form @xmath408 where @xmath409 is the distance from @xmath410 to the branch locus , can easily be shown to be contact . set @xmath411 where @xmath412 is one of the two connected components of @xmath413 see the right hand side of figure [ fig-2 ] . notice that @xmath414 has two components @xmath415 a neighborhood of one of them , @xmath416 say , is also the boundary of @xmath417 and it is clearly contactomorphic to @xmath418\times t^3 $ ] with the contact form @xmath419 and @xmath420 the other boundary component @xmath421 is also the boundary of @xmath402 and has a neighborhood contactomorphic to @xmath361\times t^3 $ ] with contact form @xmath422 and @xmath423 thus @xmath424 is the desired contact manifold . we now establish our auxiliary lemma . we adapt a construction from @xcite . in particular see that paper for more on open book decompositions , but briefly an open book decomposition of a manifold @xmath3 is a codimension two submanifold @xmath12 with trivial normal bundle together with a locally trivial fibration @xmath425 such that the closure of the fibers are submanifolds of @xmath3 whose boundaries are @xmath22 a contact structure @xmath4 is supported by the open book decomposition if there is a contact 1form @xmath35 such that @xmath35 is contact when restricted to @xmath12 and @xmath49 is a symplectic form on the fibers of @xmath426 in addition , we also need the orientation induced on @xmath12 by @xmath35 and on @xmath12 by the fibers , which are in turn oriented by @xmath427 to agree . as every contact structure @xmath4 on @xmath3 has an open book decomposition supporting it , @xcite , it was shown in @xcite that one can use this structure to construct a contact structure on @xmath428 we show that the construction in that paper can be used to give the desired embedding of @xmath429 into @xmath398 let @xmath430 be an open book decomposition of @xmath315 that supports the contact structure @xmath431 so @xmath432 is a fiber bundle . choose a function @xmath433 that is equal to the identity near @xmath434 increasing near @xmath298 and then constant . fixing a metric on @xmath435 let @xmath436 be the distance function from @xmath7 and set @xmath437 to be @xmath438 denote the coordinate functions of @xmath439 by @xmath440 and @xmath441 ( even though @xmath442 is not defined for @xmath443 we can still define @xmath444 since @xmath445 there . ) we define the embedding of @xmath429 into @xmath446 by , @xmath447 notice that @xmath448 one may easily check that this is a contact form , or see @xcite . we finish the proof of lemma [ lem : main ] by establishing lemma [ rmn ] . set @xmath449 and @xmath450 notice that @xmath451 is a flow line of @xmath452 we consider the lift of @xmath453 to @xmath402 that intersects @xmath454 let @xmath69 be the closure of the component of @xmath455 containing the branch locus . notice that topologically @xmath69 is @xmath456\times [ 0,1]\times t^2.$ ] one easily sees that a neighborhood of @xmath457\times t^2 $ ] is contactomorphic to the kernel of @xmath314 plus the symplectization of the 1form @xmath353 on @xmath346\times t^2.$ ] similarly a neighborhood of @xmath365\times t^2 $ ] is contactomorphic to the kernel of @xmath314 plus the symplectization of the 1form @xmath458 moreover near @xmath361\times \{0,1\}\times t^2 $ ] the contact form is @xmath314 plus an exact symplectic form and we can assume the lift of @xmath380 to @xmath402 gives the @xmath319 coordinate ( that is the coordinate on the first interval factor ) . thus it is an expanding vector field for the symplectic structure . this completes the proof . in this section we prove our main theorem concerning lutz twists by finding an embedded isotropically parameterized family of transverse curves given an isotropic submanifold of dimension @xmath111 in a @xmath5dimensional contact manifold . [ lem : nbhd ] let @xmath2 be a contact manifold of dimension @xmath135 if @xmath7 is an @xmath6dimensional isotropic submanifold of a contact manifold @xmath2 with trivial conformal symplectic normal bundle then @xmath7 has a neighborhood contactomorphic to a neighborhood of the zero section of @xmath271 in the contact manifold @xmath459 where @xmath460 are cylindrical coordinates on @xmath461 and @xmath260 is the canonical 1form on @xmath261 by proposition [ prop : csndet ] we know that the conformal symplectic normal bundle of an isotropic submanifold determines the contact structure in a neighborhood of the submanifold . the lemma follows as the zero section of @xmath271 sitting inside @xmath462 clearly has trivial conformal symplectic normal bundle ( in fact it can be readily identified with @xmath463 in @xmath464 which along the zero section is just the @xmath465plane in @xmath461 ) . suppose @xmath7 is an @xmath6-dimensional isotropic submanifold of a contact manifold @xmath2 with trivial conformal symplectic normal bundle . let @xmath12 be the neighborhood of @xmath7 in @xmath3 that is contactomorphic to a neighborhood of the zero section in @xmath466 given in lemma [ lem : nbhd ] . if @xmath202 is any transverse curve in @xmath467 then consider @xmath468 in @xmath469 where we are thinking of @xmath7 as the zero section of @xmath261 clearly @xmath470 is an embedded isotropically parameterized family of transverse curves . we are now ready to prove our main result . given an @xmath6dimensional isotropic submanifold @xmath7 of @xmath2 with trivial conformal symplectic normal bundle , lemma [ lem : nbhd ] and the discussion above yield the desired embedded isotropically parameterized family of transverse curves @xmath471 in any arbitrarily small neighborhood of @xmath10 theorem [ thm : genlutz ] now allows us to perform a generalized lutz twist on @xmath472 in order to show that the homotopy class of almost contact structure is unchanged during this operation we will construct a 1parameter family of almost contact structures that starts with the contact structure obtained by performing a generalized lutz twist and ends with the original contact structure . the homotopy will be through confoliations . in the construction we use the notation from subsection [ sec : setup ] and the proof of lemma [ lem : aux ] . we will break the argument into three steps . in the first step we isotope the lutz twisted contact structure ( all isotopies will be through almost contact structures ) to a confoliation on @xmath326\times[\delta'',\delta']\times t^2 $ ] given as the kernel of a particularly simple 1form . this isotopy is fixed near the boundary so extends over all of @xmath473 in the second step we further isotope the confoliation to agree with @xmath4 on @xmath326\times[\delta'',\delta']\times t^2 $ ] in such a way that the isotopy clearly extends over all of @xmath3 except @xmath474 in the last step we complete the isotopy to @xmath4 by extending the isotopy from step two to @xmath474 * step 1 : * in this step we isotope our contact structure to a confoliation on @xmath326\times[\delta'',\delta']\times t^2 $ ] that is given as the kernel of a particularly simple 1form . to this end we set @xmath475 to be the embedding of @xmath429 into @xmath403 given by @xmath476 where we use the notation from lemma [ lem : aux ] . using a riemannian metric on @xmath403 we can extend this to a @xmath214parameter family of embeddings of the normal disk - bundle @xmath477 into @xmath398 ( notice that we can assume each of the disk - bundles have the same radius @xmath478 ) let @xmath479 be the two fold branched cover of @xmath403 over @xmath480 we have the 1parameter family of 1-forms @xmath481 on @xmath482 fix a function @xmath483\to [ 0,1]$ ] that is @xmath214 near @xmath298 , @xmath298 past @xmath227 and decreasing elsewhere . let @xmath484 where @xmath485 is the distance from the branch locus of @xmath486 and @xmath487 is the connection @xmath214form on the normal disk bundle to the branch locus . we can extend @xmath488 to all of @xmath402 and they will be a smooth family of @xmath214forms . as discussed above , or see @xcite , for small enough @xmath489 the 1form @xmath490 will be contact for @xmath491 we can choose the @xmath489 smoothly so that they are decreasing with @xmath492 and @xmath493 thus @xmath494 is a 1parameter family of hyperplane fields on @xmath402 that are contact for @xmath495 and give a confoliation for @xmath496 we claim that @xmath13 has an almost complex structure that makes it into an almost contact structure that is homotopic through almost contact structures to the almost contact structures on @xmath497 for @xmath491 to see this fix a metric on @xmath402 and let @xmath498 be the ( oriented ) unit normal vector to @xmath499 since @xmath497 is a smooth family of hyperplane fields there is some small @xmath492 such that @xmath497 is also transverse to @xmath500 we can now project @xmath497 along @xmath498 onto @xmath499 this projection will be a bundle isomorphism @xmath501 thus we can use it to define an almost complex structure on @xmath499 similarly if we take @xmath502 for @xmath503,$ ] then we can use this projection to define an almost complex structure on @xmath504 for all @xmath505 that is @xmath13 with this almost complex structure is homotopic through almost contact structures to @xmath506 we notice that @xmath507 is @xmath359 where @xmath508 is a 1form on @xmath509 ( recall @xmath510 ) so @xmath13 can be decomposed as @xmath511 where @xmath75 is a 4dimensional distribution . the projection map @xmath512 maps @xmath75 isomorphically onto the tangent space of @xmath513 we can use this isomorphism to put an almost complex structure on @xmath514 which will be used below . we have homotoped our contact structure to an almost contact structure on @xmath326\times[\delta'',\delta']\times t^2 $ ] that is given as the kernel of @xmath515 but notice that the homotopy is fixed near the boundary of our manifold so this is a homotopy of almost contact structures on our entire manifold . * step 2 : * we will further homotope the almost contact structure on @xmath326\times[\delta'',\delta']\times t^2 $ ] to be @xmath516 for this consider the @xmath214parameter family of @xmath214forms @xmath517 notice that @xmath518 is always a hyperplane field and @xmath519 where @xmath520 is the contact structure on @xmath315 and @xmath521 is a 4dimensional distribution . moreover if @xmath522\times[\delta'',\delta']\times t^2\to [ a , b]\times[\delta'',\delta']\times t^2 $ ] is the projection , then @xmath523 is an isomorphism from @xmath521 to the tangent space of @xmath524\times[\delta'',\delta']\times t^2.$ ] we can use this isomorphism to induce an almost complex structure on @xmath521 for all @xmath492 and thus @xmath497 is an almost contact structure ( as @xmath9 clearly has an almost complex structure since it is contact ) . * step 3 : * we are left to extend the homotopy above over the region @xmath474 notice that we can assume that there is some @xmath154 such that @xmath525\times[\delta'',\delta']\times t^2= \left(w\times[a , b]\times[\delta'',\delta']\times t^2\right)\cap \left(n_{\epsilon''}\times s(\delta)\right).\ ] ] thus we already have our homotopy defined on part of @xmath526 if we consider consider the @xmath214forms @xmath527 then @xmath518 is always a hyperplane field that extends the above homotopy . moreover @xmath528 where @xmath529 and @xmath521 is a 3dimensional distribution . notice that since @xmath337 and @xmath332 agree near the center of the disk and near the boundary of the disk the form @xmath530 is fixed there . moreover @xmath531 is clearly contact for @xmath492 near @xmath298 and @xmath532 we notice that it is easy to choose a nonzero vector @xmath533 in the tangent space to @xmath534 that agrees with the reeb field of @xmath531 whereever the form @xmath535 is contact . ( indeed , consider @xmath536,$ ] the last interval factor is the parameter space . the vector field @xmath533 is defined near the boundary of this space and near @xmath537 so we need to extend it over a space diffeomorphic to @xmath538 when the field defined near the boundary . moreover notice that the vector field defined so far is invariant under the rotations in the first @xmath539 it is now clear @xmath533 can be extended . ) we can use @xmath533 to induce a non - zero vector field in @xmath521 and thus we can write @xmath521 as @xmath540 where @xmath54 is a trivial line bundle and @xmath541 is an oriented plane field . notice also that @xmath9 can be split as @xmath542 where @xmath54 is a trivial bundle . this is because @xmath543 on @xmath544 defines a @xmath545bundle structure on the @xmath546 dimensional bundle @xmath282 we now have an almost complex structure on @xmath497 given by the one on @xmath547 induced by @xmath548 the unique one on @xmath541 coming from the orientation on the plane field and the unique one on the trivial plane field @xmath549 it is clear that where @xmath530 is contact this almost complex structure agrees with the one induced by @xmath550 thus we see this is a homotopy of almost contact structures and our proof is complete . using theorem [ main ] we may now easily show all manifolds admitting contact structures admit weakly overtwisted ones . let @xmath4 be a contact structure on @xmath473 in a darboux ball inside of @xmath3 with coordinates @xmath551 and contact structure @xmath552 take a unit sphere @xmath7 in the @xmath553subspace . it is clear that @xmath7 is an isotropic submanifold of @xmath3 with trivial conformal symplectic normal bundle . thus we may apply theorem [ main ] to alter @xmath4 to a contact structure containing an overtwisted family parameterized by @xmath10 the statement about finding an overtwisted family with any core is proven in theorem [ anycore ] below . let @xmath554 be the standard contact structure on @xmath0 where we are using cartesian coordinates @xmath555 let @xmath9 be the result of performing a lutz twist along some embedded isotropically parameterized family of transverse curves contained in some compact ball in @xmath556 let @xmath557 be a ball of radius @xmath558 about the integral points on the @xmath119-axis and let @xmath547 be the result of performing a lutz twist along some embedded isotropically parameterized family of transverse curves contained in each of the @xmath559 clearly @xmath4 can be contact embedded in any contact @xmath8 manifold ( by darboux s theorem ) , but neither @xmath9 nor @xmath547 can be embedded in a stein fillable contact structure ( like the standard contact structure on @xmath560 ) , thus they are not contactomorphic to @xmath39 finally notice that @xmath9 has the property that any compact set in @xmath0 is contained in another compact set whose complement can be embedded in any contact manifold , whereas @xmath9 does not have this property . thus @xmath9 is not contactomorphic to @xmath561 _ from lemma [ lem : nbhd ] it is almost immediate that we can construct a contact structure on @xmath562 which has an embedded overtwisted family modeled on any core which is parameterized by a closed embedding of @xmath563 . this is analogous to the unique `` overtwisted at infinity '' contact structure @xmath564 on @xmath565 obtained by performing a `` lutz twist along the @xmath119-axis '' in three dimension . in @xcite a contact structure on @xmath562 is constructed which contains a generalized overtwisted family at infinity ( termed a generalized plastikstufe in that paper ) . with this in mind , it would be interesting to know the answer to the following question : _ is there a unique contact structure on @xmath562 which contains an embedded overtwisted family parameterized by @xmath563 in the complement of any compact subset of @xmath562 . in this section we discuss how lutz twists affect the homotopy class of almost contact structures and which submanifolds of a contact manifold can be the core of an overtwisted family ( that is , which submanifolds can be the elliptic singularity of an overtwisted family ) . it is interesting to observe that the branched cover construction in lemma [ lem : main ] can not be used to perform a half lutz twist . however one can modify that construction to perform a half lutz twist at the expense of changing the topology of the ambient manifold . let @xmath2 be a contact manifold of dimension @xmath8 and let @xmath7 be an @xmath6dimensional isotropic submanifold with trivial conformal symplectic normal bundle . according to the discussion in section [ sec : getfamily ] we can , in any neighborhood of @xmath286 find a contact embedding of a neighborhood @xmath566 in @xmath3 where @xmath270 is a neighborhood of the zero section in @xmath271 and @xmath203 is the solid torus with contact structure from section [ sec:3lutz ] . let @xmath315 be the unit conormal bundle for @xmath7 in @xmath271 which we may think of as a submanifold of the neighborhood above . taking an @xmath1 from the @xmath203 factor we see an embedding of @xmath567 in the above neighborhood . the submanifold @xmath567 has a neighborhood @xmath568 in @xmath473 let @xmath569 be the manifold obtained from @xmath3 by removing @xmath570 and gluing in its place @xmath571 let @xmath2 be a contact manifold of dimension @xmath8 and let @xmath7 be an @xmath6dimensional isotropic submanifold with trivial conformal symplectic normal bundle . with the notation above we may extend @xmath572 over @xmath569 to obtain a contact structure @xmath9 such that for some @xmath573 the contact structure @xmath9 on @xmath574 is contactomorphic to @xmath575 following the outline in subsection [ sec : setup ] we can define the desired contact structure on @xmath576 everywhere except on @xmath577\times[\delta'',\delta']\times s^1\times s^1.$ ] in the paper @xcite , gay and kirby construct an exact near symplectic structure on @xmath578\times[\delta'',\delta']\times s^1\times s^1 $ ] that can be used as described at the end of subsection [ sec : setup ] to try to extend the contact structure on all of @xmath579 more precisely , there is a @xmath214form @xmath508 such that @xmath345 is symplectic on @xmath524\times[\delta''.\delta']\times s^1\times s^1 $ ] away from a curve @xmath580 where @xmath581 and @xmath582 thus @xmath359 is a contact form on @xmath583\times[\delta'',\delta']\times s^1\times s^1 $ ] away from @xmath584 and has the necessary boundary conditions to glue to the desired contact structure . let @xmath585 be a neighborhood of @xmath586 it is shown in @xcite that a neighborhood of the boundary of @xmath585 is contactomorphic to @xmath587\times s^1\times s^2 $ ] with the contact form @xmath588 where @xmath589 is a contact form on @xmath590 giving the minimally overtwisted contact structure ( that is the one in the same homotopy class of plane fields as the foliation of @xmath590 by @xmath591 s ) . consider @xmath585 as @xmath592\times w\times s^1\times s^2 $ ] where we use the identity diffeomorphism on most factors and @xmath593 on the interval factor . notice this is an orientation preserving diffeomorphism and the contact form in these coordinates can be taken to be @xmath594 now we can glue in a neighborhood of the zero section in @xmath271 times @xmath590 and extend the contact structure over this ( using @xmath595 ) . as discussed in subsection [ sec : almost ] an obstruction to two almost contact structures being homotopic is the chern classes of the almost contact structure . in dimension 3 it is well known that lutz twisting affects the first chern class of the contact structure . in higher dimensions this is not the case . in the proof of theorem [ main ] above we showed that the homotopy class of almost contact structure is unchanged by a lutz twist , but one might ask if the homotopy class can be affected with a generalized half lutz twist ( should it ever be defined ) . [ chern ] let @xmath2 be a closed contact @xmath5-manifold for @xmath596 suppose @xmath597 is an embedded isotropically parameterized family of transverse curves in @xmath2 with @xmath7 of dimension @xmath255 if @xmath9 is obtained from @xmath4 by altering the contact structure in a neighborhood of @xmath253 then the chern classes of @xmath598 and @xmath599 are equal for @xmath600 one can easily construct a handle decomposition of @xmath3 in which a neighborhood of @xmath253 can be taken to be a union of handles of index larger than or equal to @xmath601 moreover the contact structures @xmath4 and @xmath9 are the same outside a neighborhood of @xmath602 that is away from the handles that make up the neighborhood . as @xmath603 is the primary obstruction to the existence of a @xmath604-frame over the @xmath605 skeleton of @xmath3 we see that @xmath603 of @xmath4 and @xmath9 must be the same for @xmath606 in dimension 3 one can use lutz twists to produce contact structures in any homotopy class of almost contact structure . one might hope to do this in higher dimensions as well , but clearly proposition [ chern ] shows our notion of lutz twist ( even a more general one than defined here ) can not achieve this . so we ask the following question . is there some notion of lutz twisting , or some other modification of a contact structure , that will guarantee that any manifold @xmath3 admitting a contact structure admits one in every homotopy class of almost contact structure ? in @xcite the weakly overtwisted contact structures came from weakly overtwisted contact structures of lower dimension . more precisely , the core of the overtwisted families constructed in dimension @xmath8 were constructed as a product of @xmath1 and an overtwisted family in dimension @xmath607 starting in dimension 3 where the core is just a point , one sees that all the cores of overtwisted families constructed in @xcite are tori of the appropriate dimension . the weakly overtwisted contact structures constructed in @xcite were constructed by taking the previous examples and performing surgery on the ambient manifold without affecting the overtwisted family . thus , once again , we see that all the overtwisted families are modeled on tori . from our construction we can show the following result . [ anycore ] let @xmath2 be a contact manifold of dimension @xmath8 . given any @xmath6dimensional isotropic submanifold @xmath7 in @xmath2 with trivial conformal symplectic normal bundle there is a contact structure @xmath9 on @xmath3 that contains an overtwisted family modeled on @xmath10 moreover , if @xmath7 is any abstract @xmath6dimensional manifold ( that is not necessarily already embedded in @xmath3 ) with trivial complexified tangent bundle , then there is a contact structure on @xmath3 with overtwisted family modeled on @xmath10 the first statement is clear as we can find an embedded isotropically parameterized family of transverse curves @xmath253 as in section [ sec : getfamily ] and then use theorem [ thm : genlutz ] to perform a lutz twist to produce a contact structure @xmath9 with an overtwisted family modeled on @xmath10 for the second statement we need to see that given a @xmath7 with the required properties we can embed it in @xmath2 as an isotropic submanifold with trivial conformal symplectic normal bundle . it is clear , due to the dimensions involved , that @xmath7 can be embedded in a ball in @xmath473 it is well known that isotropic submanifolds of dimension less than @xmath44 satisfy an h - principle @xcite . this @xmath608-principle states that if an embedding @xmath609 is covered by a bundle map @xmath610 to @xmath4 sending the tangent planes of @xmath7 to isotropic spaces in @xmath4 then the embedding can be isotoped to an isotropic embedding . thus we need to construct a bundle map @xmath610 to @xmath611 sending @xmath612 to an isotropic subspace of @xmath613 in the end we will also want the conformal symplectic normal bundle to be trivial . this implies that we need to see a bundle isomorphism from @xmath614 to @xmath615 since @xmath200 can be taken to have its image in a darboux ball of @xmath3 we can assume that @xmath611 is the trivial bundle @xmath616 now it is clear that if @xmath617 is trivial then we have such an isomorphism . from this theorem we see that it is easy to produce overtwisted families modeled on many manifolds . in particular , any oriented 2manifold , respectively 3manifold , can be realized as the core of an overtwisted family in a contact 7 , respectively 9 , manifold . moreover , the vanishing of the first pontryagin class of the tangent bundle of a 4manifold is sufficient to guarantee it can be made the core of an overtwisted family in a contact 11manifold . it would be very interesting to know the answer to the following question . if @xmath618 contains an overtwisted family modeled on @xmath7 does it also contain an overtwisted family modeled on any , or even some , other @xmath6manifold @xmath470 ( satisfying suitable tangential conditions ) ? in dimension 3 , overtwisted contact structures are very flexible and various questions about them usually have a topological flavor . that is , if something is true topologically then it is frequently true for overtwisted contact structures . for example , if two overtwisted contact structures are homotopic as plane fields in dimension 3 then they are isotopic , @xcite . we also know that any overtwisted contact structure is supported by a planar open book ( just like any 3manifold ) , @xcite . thus if overtwisted families are the `` right '' generalization of overtwisted disks to higher dimensional manifolds then we would expect to have similar results . an affirmative answer to the question above would essentially say you have a lot of flexibility in the cores of overtwisted families . theorem [ anycore ] is a step in that direction . emmanuel giroux . gomtrie de contact : de la dimension trois vers les dimensions suprieures . in _ proceedings of the international congress of mathematicians , vol . ii ( beijing , 2002 ) _ , pages 405414 , beijing , 2002 . higher ed . press .

chord diagrams on circles and their intersection graphs ( also known as _ circle graphs _ ) have been intensively studied , and have many applications to the study of knots and knot invariants , among others . however , chord diagrams on more general graphs have not been studied , and are potentially equally valuable in the study of spatial graphs . we will define chord diagrams for planar embeddings of planar graphs and their intersection graphs , and prove some basic results . then , as an application , we will introduce gauss codes for immersions of graphs in the plane and give algorithms to determine whether a particular crossing sequence is realizable as the gauss code of an immersed graph .

math0508269

classically , a _ chord diagram _ is a collection of chords of a circle ; the intersection graph for these chords is called a _ circle graph_. circle graphs have been intensively studied in graph theory , with classifications given by even and itai @xcite and bouchet @xcite , among others . in recent years , interest in chord diagrams has spread to topologists as part of the theory of finite type knot and link invariants @xcite , as well as the theory of virtual knots @xcite . a natural extension of knot theory is to look at spatial embeddings of more complex graphs , and it is natural to ask whether some analogue of chord diagrams could be equally useful in this context . the goal of this paper is to define a reasonable notion of chord diagram for general graphs , and to construct some tools to study them . in particular , we will look at a variety of intersection graphs for these chord diagrams . we will use these intersection graphs to determine when an embedding of a planar graph can be extended to an embedding of a chord diagram on that graph . as an application , we will define _ gauss codes _ for immersions of arbitrary graphs and give algorithms for determining whether a crossing sequence is realizable as a gauss code . acknowledgement : the authors would like to acknowledge the hospitality of waseda university , tokyo , and professor kouki taniyama during the international workshop on knots and links in a spatial graph in july , 2004 , where the idea for this project was conceived . the second author was supported by an lmu faculty research grant . a chord diagram for a circle may be characterized as simply a circle with a set of labeled points indicating the endpoints of the chords . this is the idea we use to define a chord diagram for a general graph . [ d : chord ] let g be a graph with edge set e and vertex set v. a * chord diagram on g of degree n * is a collection of 2n points in @xmath0 ( so each point is in the interior of an edge ) , each labeled from a set @xmath1 such that each label is used exactly twice . the two points labeled @xmath2 are called the * endpoints * of chord @xmath2 . often , we will want to look at _ oriented _ chord diagrams of directed graphs . this allows us to record whether the chord is on the same or opposite sides of the edges at each endpoint . [ d : orientchord ] let g be a * directed * graph with edge set e and vertex set v. an * oriented chord diagram on g of degree n * is a collection of 2n points in @xmath0 , each labeled from a set @xmath3 such that exactly two points are labeled from each set @xmath4 . these two points are called the * endpoints * of chord @xmath2 ; if they have the same label , the endpoints are said to have the same orientation . if one is labeled @xmath2 and the other @xmath5 , they are said to have opposite orientations . generally , we will represent a chord diagram visually by drawing an arc between the endpoints . in an oriented chord diagram , the arc will be on the _ same _ side of the edges at the endpoints when the endpoints have _ opposite _ orientation , and on opposite sides when the endpoints have the same orientation . this terminology is geometrically motivated - an oriented line segment in the plane naturally has left and right sides , and when the arc is on the same side of the edges at each endpoint , the bases for @xmath6 given by the ordered pair of the tangent vectors to the chord and the edge have opposite orientations at the two endpoints , as shown in figure [ f : orient ] . [ h ] @xmath7 two examples of ( oriented ) chord diagrams are shown in figure [ f : chordexamples ] . [ h ] @xmath8 much of the work done on chord diagrams for circles has been in terms of their intersection graphs , also known as _ circle graphs_. these graphs have also been useful in the recent applications of chord diagrams to knot and link invariants @xcite . so it is natural to look at some kind of intersection graph for chord diagrams on more general graphs . for chord diagrams on the circle , two chords are said to intersect if their endpoints alternate around the circle . in a more general graph , there may not be a way to move around " the graph , or there may be many , so we need a different criterion . another point of view of circle graphs is as proper circular - arc graphs " - in other words , two chords intersect if the arcs on the circle between their endpoints overlap , but neither is a proper subset of the other @xcite . by replacing arcs on the circle " with paths in the graph , " we can generalize this notion to other graphs . [ d : igchord ] given a chord diagram c on a ( possibly directed ) graph g , the * intersection graph for c * is the graph @xmath9 whose vertices are the chords of c , and vertices @xmath2 and @xmath10 ( corresponding to chords @xmath2 and @xmath10 ) are adjacent if given any ( undirected ) path @xmath11 in g between the endpoints of @xmath2 and any ( undirected ) path @xmath12 in g between the endpoints of @xmath10 , @xmath11 and @xmath12 properly intersect ( meaning that they overlap - perhaps only in one vertex - but neither is a subset of the other ) . so for two chords to intersect , _ all _ paths between their endpoints must intersect ; for them to be disjoint , there need only be one pair of paths which are disjoint ( or in which one properly contains the other ) . this is clearly a fairly restrictive notion of intersection , but it seems to be the one which best matches our visual intuition . figure [ f : igchord ] shows the intersection graphs for the chord diagrams in figure [ f : chordexamples ] . note that , for chord diagrams on circles , this definition agrees with the usual definition of the intersection graph . [ h ] @xmath13 a common situation is to study chord diagrams when the chords are real " chords - in other words , they are chords of regions in some planar embedding of the graph . in this section , we will explore chord diagrams and their intersection graphs when we are given a planar graph together with a particular embedding in the plane . our main result is to give conditions when the embedding of the graph can be extended to an embedding of the entire chord diagram . [ d : respect ] let @xmath14 be a directed planar graph , @xmath15 be an oriented chord diagram on @xmath14 , and @xmath16 be a particular embedding of @xmath14 . consider a chord @xmath2 of @xmath15 , with endpoints on edges @xmath17 and @xmath18 in @xmath14 . we say @xmath2 * respects * @xmath19 if : 1 . both endpoints of @xmath2 lie on the boundary of a single region r of @xmath20 ( i.e. a single connected component of @xmath21 ) . if r is on the same side ( left or right ) of @xmath17 and @xmath18 , then the endpoints of @xmath2 have opposite orientations ; if r is on the opposite sides of @xmath17 and @xmath18 , then the endpoints of @xmath2 have the same orientation ( assuming @xmath17 and @xmath18 are not cut edges ) . 3 . if either @xmath17 or @xmath18 is a cut edge , then the endpoints of @xmath2 may have either orientation . if @xmath14 is not directed , then only the first condition is required . we say that @xmath15 * respects * @xmath19 if every chord of @xmath15 respects @xmath19 . we should observe that if @xmath17 and @xmath18 are both on the boundaries of two regions @xmath22 and @xmath23 , then the definition does not depend on which region is used ; if @xmath22 is on the same side of both edges , then so is @xmath23 , and vice - versa . we define the intersection graph of @xmath15 with respect to @xmath19 by restricting our attention to paths in @xmath14 which lie on the boundary of a single region of @xmath20 . [ d : igplanar ] given a chord diagram c on a ( possibly directed ) planar graph g , and an embedding @xmath16 , the * intersection graph for c with respect to f * is the graph @xmath24 whose vertices are the chords of c , and vertices @xmath2 and @xmath10 ( corresponding to chords @xmath2 and @xmath10 ) are adjacent if given any ( undirected ) path @xmath11 on the boundary of a single region of @xmath20 between the endpoints of @xmath2 and any ( undirected ) path @xmath12 on the boundary of a single region of @xmath20 between the endpoints of @xmath10 , @xmath11 and @xmath12 properly intersect . remarks : 1 . if @xmath2 does not have both endpoints on the boundary of a single region of @xmath20 , then there is no path between the endpoints which lies on the boundary of a single region . in this case , the condition for adjacency is vacuously true , so @xmath2 is adjacent to every other chord in the chord diagram . 2 . the intersection graphs shown in figure [ f : igchord ] are also the intersection graphs with respect to the embeddings shown . since the number of paths in a graph may be large , it is not clear whether @xmath9 can be constructed in polynomial time . however , @xmath24 certainly can , since an endpoint for a chord can lie on the boundary of at most two regions , there are at most 4 possible paths between the endpoints for each chord . so the number of pairs of paths which need to be compared is at most @xmath25 , where @xmath26 is the number of chords . if two chords intersect by definition [ d : igchord ] , then they certainly intersect by definition [ d : igplanar ] , but the converse is not true . so @xmath27 for every embedding @xmath19 . the last remark gives rise to the following open question : given a planar graph g and a chord diagram c on g , is @xmath28 ? our main focus in this section is to study when a planar embedding of a graph @xmath14 can be extended to give a planar embedding of a chord diagram @xmath15 on @xmath14 . to be precise , by an embedding of the chord diagram " we mean an embedding of the graph @xmath29 constructed from @xmath14 by adding vertices at the endpoints of all the chords in @xmath15 , and edges between the endpoints of each chord in @xmath15 . we will also look at embeddings of _ oriented _ chord diagrams - in this case , we will insist that the embedded chords be on the same or opposite sides of the edges at each endpoint , as dictated by the orientations ( see section [ ss : defchord ] ) . our first result is well - known ( an equivalent form can be extracted from @xcite , for example ) , but is given here for completeness and as a warm - up for our later results . [ p : circle ] let g be a directed graph consisting of a single loop , and let @xmath16 be an embedding of g ( so @xmath20 is a circle ) . let c be an oriented chord diagram on g. then f extends to an embedding of c if and only the following conditions are met : 1 . @xmath9 is bipartite . the endpoints of every chord in c have opposite orientation . if c is not oriented , then the first condition is sufficient . proof : we first note that , since every embedding of the circle in the plane is isotopic , the particular embedding @xmath19 is irrelevant . and , since @xmath20 has exactly two regions , both bounded by the circle , every path on the circle is also a path in the boundary of a region , so @xmath30 . we first show the sufficiency of the conditions of the proposition . since the endpoints of the chords have opposite orientations , the chord is on the same side of the circle at each endpoint , so each chord can be drawn entirely within one region of @xmath20 ( if @xmath15 is not oriented , we can simply choose to draw the chords on the same side of the circle at each endpoint ) . since @xmath9 is bipartite , we can divide the chords into two sets , denoted the blue " chords and the red " chords , so that no two blue chords intersect , and no two red chords intersect . we will draw the blue chords in one region of @xmath20 ( inside the circle ) and the red chords in the other region ( outside the circle ) . choose one of the blue chords , @xmath31 , and connect the endpoints by an arc inside the circle . this gives an embedding of @xmath32 , which divides the inside of the circle into two regions . since no other blue chord intersects @xmath31 , any other chord @xmath33 must have both endpoints on the boundary of one of these regions , and so there is an arc connecting these endpoints in @xmath34 . continuing in this way , we can embed all of the blue arcs inside the circle ; by the same argument , we can embed all the red arcs outside the circle . this gives us an extension of the embedding to @xmath15 . conversely , if we have an embedding of @xmath15 , then each chord must lie in a single region of @xmath20 , and so must be on the same side of the circle at each endpoint , so the endpoints must have opposite orientations . in addition , the chords inside the circle must all be disjoint , as are the chords outside the circle . so the vertices of the intersection graph can be divided into two independent sets , and so @xmath9 is bipartite . @xmath35 + + remarks : 1 . since there are polynomial - time algorithms to determine whether a graph is bipartite , proposition [ p : circle ] shows that we can determine whether the embedding extends to the chord diagram in polynomial time . proposition [ p : circle ] easily extends to general cycles ( in which the edges may not all be oriented the same way ) by replacing the condition that the endpoints of each chord have opposite orientation with the condition that every chord must respect the embedding , as in definition [ d : respect ] . now we want to consider more complicated graphs . topologically , the simplest graphs after the circle are the @xmath36-graphs , which are the graphs consisting of two vertices and @xmath26 edges between these vertices . figure [ f : theta ] shows several examples . [ h ] @xmath37 the spatial embeddings of these graphs have been studied extensively @xcite , so they are a natural set of graphs for us to consider . [ p : theta ] let g be a directed @xmath36-graph with vertices v and w , and with all edges directed from v to w. let @xmath16 be a planar embedding of g , and let c be an oriented chord diagram on g. label the edges of g by @xmath38 according to their ( counterclockwise ) cyclic order around v in f(g ) , and label the region of f(g ) enclosed by @xmath39 and @xmath40 by @xmath41 ( @xmath42 is the region enclosed by @xmath43 and @xmath17 ) . then f extends to an embedding of c if and only if the following conditions are met : 1 . the vertices of @xmath24 can be labeled by the regions @xmath41 so that : * if a vertex is labeled @xmath41 , then the corresponding chord in c has both endpoints on the boundary of region @xmath41 in f(g ) . * for each @xmath41 , the set of vertices labeled @xmath41 is independent . if both endpoints of a chord are on the same edge of g , then they have opposite orientations ; if the endpoints of a chord are on different edges of g , they have the same orientation ( i.e. @xmath15 respects @xmath19 , as in definition [ d : respect ] ) . if c is not oriented , then we can ignore the second condition . proof : we will first show that the conditions are sufficient . the idea is simply to draw the chord @xmath31 corresponding to a vertex @xmath44 labeled @xmath41 in the region @xmath41 . since both endpoints of @xmath31 lie on the boundary of @xmath41 , they can be connected by an arc in @xmath41 . if both endpoints of @xmath31 are on one edge ( @xmath39 or @xmath40 ) , then they have opposite orientations ( since @xmath15 respects @xmath19 ) , and the chord is on the same side of the edge at both endpoints ; therefore , the arc is an embedding of the oriented chord . on the other hand , if @xmath31 has one endpoint on @xmath39 and the other on @xmath40 , then its endpoints have the same orientation , and @xmath31 should be on opposite sides of the edges . but the region @xmath41 is on the left side of @xmath39 and the right side of @xmath40 , so again the arc in @xmath41 is an embedding of the oriented chord . so the embedding of @xmath14 can be individually extended to each chord in @xmath15 . we still need to show it can be extended to all the chords at once . the obstruction to this extension is when chords cross each other - i.e. when their endpoints alternate around the boundary of a region . we will first consider what happens if there _ is _ such an obstruction . [ l : thetaalternate ] if two chords @xmath31 and @xmath33 both have their endpoints on the boundary of region @xmath41 , and their endpoints alternate along the ( undirected ) cycle around the boundary of @xmath41 , then the corresponding vertices @xmath44 and @xmath45 are adjacent in @xmath24 . proof of lemma : if @xmath31 has both endpoints on the same edge , say @xmath39 , then @xmath33 has exactly one endpoint on the segment @xmath39 between the endpoints of @xmath31 , and any path between the endpoints of @xmath33 must contain one of the endpoints of @xmath31 , but not both . so any path between the endpoints of @xmath33 will intersect , but not contain , any path between the endpoints of @xmath31 . on the other hand , if neither @xmath31 nor @xmath33 have both endpoints on the same edge , then they each have one endpoint on @xmath39 and one on @xmath40 . then any path between the endpoints of the chords with lies on the boundary of a region of @xmath20 must run along the boundary of @xmath41 , since it is the only region whose boundary contains both endpoints . there are two such paths between the endpoints of @xmath31 , each of which contains one endpoint of @xmath33 . so we conclude that if the endpoints of @xmath31 and @xmath33 alternate around the boundary of @xmath41 , then @xmath44 and @xmath45 are adjacent . this proves our lemma . equivalently , if @xmath44 and @xmath45 are _ not _ adjacent , then the endpoints of @xmath31 and @xmath33 must _ not _ alternate around the boundary of @xmath41 . since the set of vertices labeled @xmath41 is independent , none of the corresponding chords have alternating endpoints around the boundary of @xmath41 . so these chords can be embedded in @xmath41 as disjoint arcs , as in proposition [ p : circle ] . this can be done for each label @xmath41 , resulting in an embedding of @xmath15 which extends @xmath19 . it remains to show that the conditions are necessary . assume that @xmath19 can be extended to @xmath15 . then each chord is embedded as an arc in some region @xmath41 . if both endpoints of the chord are on the same edge , the chord is on the same side at both endpoints , and the endpoints must have opposite orientations . on the other hand , if they are on distinct edges @xmath39 and @xmath40 , then the chord is on opposite sides of the edges at the two endpoints ( the left side of @xmath39 and the right side of @xmath40 , so the endpoints must have the same orientation . so the second condition is necessary . now label the vertex @xmath44 in @xmath24 corresponding to chord @xmath31 by the region in which @xmath31 is embedded in @xmath46 . obviously , both endpoints of @xmath31 must be on the boundary of this region . since we have embedded our chord diagram , chords which are in the same region @xmath41 do not cross , so their endpoints do not alternate around the boundary of @xmath41 . then there are clearly non - overlapping paths between the endpoints of the chords , so the corresponding vertices are not adjacent in @xmath24 . so vertices with the same label will not be adjacent . this shows that the first condition is necessary , and completes the proof . @xmath35 + + remark : it is easy to check that these conditions can be checked in polynomial time . the coloring of most chords is forced by the location of their endpoints . those that are left have both endpoints on the same edge , and it is enough to show that the subgraph in @xmath24 induced by the remaining chords on each edge is bipartite . the arguments of section [ ss : theta ] can easily be generalized to any planar graph without _ cut edges _ ( so that each edge in a planar embedding of the graph bounds a different region on either side ) . recall that @xmath47 is a _ cut edge _ of a connected graph @xmath14 if @xmath48 is not connected . [ p : nocut ] let g be a directed planar graph with no cut edges . let @xmath16 be a planar embedding of g , and let c be an oriented chord diagram on g. label the regions of f(g ) by @xmath41 ( @xmath49 number of regions ) . then f extends to an embedding of c if and only if the following conditions are met : 1 . c respects f , as in definition [ d : respect ] ( so every chord in c has both endpoints on the boundary of some region @xmath41 ) . the vertices of @xmath24 can be labeled by the regions @xmath41 so that : * if a vertex is labeled @xmath41 , then the corresponding chord in c has both endpoints on the boundary of region @xmath41 in f(g ) . * for each @xmath41 , the set of vertices labeled @xmath41 is independent . before we begin the proof of proposition [ p : nocut ] , we need a technical lemma , which generalizes lemma [ l : thetaalternate ] in the last section . [ l : alternate ] consider g , f and c as in the statement of proposition [ p : nocut ] . if c and d are chords in c whose endpoints alternate around the boundary of region @xmath50 in f(g ) , then the corresponding vertices @xmath44 and @xmath45 are adjacent in @xmath24 . proof : since @xmath14 does not have cut edges , the boundary of @xmath50 is a _ cycle _ in @xmath14 . assume that @xmath44 and @xmath45 are _ not _ adjacent . then there are paths @xmath51 between the endpoints of @xmath31 and @xmath52 between the endpoints of @xmath33 , each along the boundary of a region in @xmath20 , such that either @xmath53 or one path is contained in the other - without loss of generality , @xmath54 . so neither of the endpoints of chord @xmath33 lie on @xmath51 . since the endpoints of @xmath31 and @xmath33 alternate around the boundary of @xmath50 , @xmath51 can not be contained in the boundary of @xmath50 , or it would contain one of the endpoints of @xmath33 . so @xmath51 is in the boundary of some other region @xmath55 , meaning that the endpoints of @xmath31 lie on the boundary between regions @xmath50 and @xmath55 . first consider the case when @xmath51 and @xmath52 are disjoint . then the union of @xmath51 with one of the paths between the endpoints of @xmath31 along the boundary of region @xmath50 gives a cycle @xmath56 ( or at least a circuit which contains a cycle @xmath56 ) which contains one endpoint of @xmath33 . so @xmath52 must intersect @xmath56 . if @xmath52 lies along region @xmath50 , then it will contain one of the endpoints of @xmath31 , and so wo nt be disjoint from @xmath51 . but since @xmath52 ca nt pass _ through _ region @xmath50 ( since this is a region of the planar embedding of @xmath14 ) , it must ( by the jordan curve theorem ) pass from one side of @xmath56 to the other , and so must intersect @xmath51 , as shown in figure [ f : alternate ] . [ h ] @xmath57 this contradicts the assumption that @xmath51 and @xmath52 are disjoint . therefore , @xmath54 , so the endpoints of @xmath31 are in @xmath52 . since the endpoints of @xmath31 are on the boundary between regions @xmath50 and @xmath55 , and we know that @xmath52 can not be on the boundary of region @xmath50 , it must lie on the boundary of region @xmath55 . so the endpoints of @xmath33 are also on the boundary between regions @xmath50 and @xmath55 . since @xmath54 , the endpoints of @xmath31 and @xmath33 do _ not _ alternate around the boundary of region @xmath55 , which is a cycle . so there are paths @xmath58 and @xmath59 between the endpoints of chords @xmath31 and @xmath33 on the boundary of region @xmath55 , with @xmath60 . but this is impossible , by the argument in the last paragraph . therefore , @xmath51 and @xmath52 must properly intersect , so @xmath44 and @xmath45 are adjacent . @xmath35 + + proof of proposition [ p : nocut ] : our proof is very similar to the proof of proposition [ p : theta ] . we will first show that the conditions are sufficient . once again , the idea is simply to draw the chord @xmath31 corresponding to a vertex @xmath44 labeled @xmath50 in the region @xmath50 . since both endpoints of @xmath31 lie on the boundary of @xmath50 , they can be connected by an arc in @xmath50 . the first condition ensures that this arc will be an embedding of the oriented chord @xmath31 . so the embedding of @xmath14 can be individually extended to each chord in @xmath15 . we still need to show it can be extended to all the chords at once . let @xmath33 be another chord on region @xmath50 , with corresponding vertex @xmath45 . by lemma [ l : alternate ] , if @xmath44 and @xmath45 are _ not _ adjacent , then the endpoints of @xmath31 and @xmath33 must _ not _ alternate around the boundary of @xmath50 . since the set of vertices labeled @xmath50 is independent , none of the corresponding chords have alternating endpoints around the boundary of @xmath50 . so these chords can be embedded in @xmath50 as disjoint arcs , as in proposition [ p : circle ] . this can be done for each label @xmath50 , resulting in an embedding of @xmath15 which extends @xmath19 . it remains to show that the conditions are necessary . assume that @xmath19 can be extended to @xmath15 . then each chord is embedded as an arc in some region @xmath50 , with endpoints on edges @xmath17 and @xmath18 . if @xmath50 is on the same side of both edges , then so is the embedded chord , and the endpoints must have opposite orientations ; if it is on opposite sides , the endpoints must have the same orientation . so the first condition is necessary . now label the vertex @xmath44 in @xmath24 corresponding to chord @xmath31 by the region in which @xmath31 is embedded in @xmath46 . obviously , both endpoints of @xmath31 must be on the boundary of this region . chords which are in the same region @xmath50 do not cross , so their endpoints do not alternate around the boundary of @xmath50 ( since the boundary of @xmath50 is a cycle ) . then there are clearly non - overlapping paths between the endpoints of the chords , so the corresponding vertices are not adjacent in @xmath24 . so vertices with the same label will not be adjacent . this shows that the second condition is necessary , and completes the proof . [ c : nocut ] given a planar graph @xmath14 with no cut edges , a chord diagram @xmath15 on @xmath14 , and a planar embedding @xmath16 , there is a polynomial time algorithm to determine whether @xmath19 extends to an embedding of @xmath15 . proof : the conditions of proposition [ p : nocut ] can be checked in polynomial time . the coloring of most chords is forced by the location of their endpoints . those that are left have both endpoints on boundaries of the same two regions , and it is enough to show that the subgraph in @xmath24 induced by the remaining chords lying on each pair of regions is bipartite . @xmath35 now we want to consider planar graphs which _ do _ have cut edges . as an introduction , we will consider the simplest such graph , the complete graph on two vertices @xmath61 ( i.e. the line segment consisting of two endpoints and the edge between them ) . why do cut edges pose difficulties ? one difficulty is that the boundaries of the regions in a planar embedding of the graph are no longer cycles in the graph . for example , we might expect a line segment to be no more complicated than a circle . but consider the example in figure [ f : lineexample ] . [ h ] @xmath62 the sequence of endpoints along the graph is @xmath63 . according to definition [ d : igplanar ] , all three of these chords intersect . so the intersection graph has a cycle of length 3 , and so is not bipartite . and yet we _ can _ extend the embedding of the line segment to an embedding of the chord diagram , so proposition [ p : circle ] does not extend to @xmath61 . the fact that chords can go around the end " of the graph increases the difficulty of the problem . however , we would still like to exploit our results from section [ ss : nocut ] . so our plan is to transform a graph _ with _ cut edges into a graph _ without _ cut edges , and hope solving the problem in this new graph will give us the solution for the original graph . to this end , we introduce an operation called blowing up " an edge in a planar embedding of a graph or chord diagram . [ d : blowup ] let g be a planar directed graph with an edge e ( which is not a loop ) , and let @xmath16 be a planar embedding of g. say that the endpoints of e are vertices v and w , with e oriented from v to w. then the * blowup of @xmath47 in @xmath14 with respect to @xmath19 * is the graph @xmath64 formed by removing edge @xmath47 and replacing it with two edges e and @xmath65 , both directed from @xmath66 to @xmath67 . figure [ f : blowup ] illustrates this operation . [ h ] @xmath68 remark : figure [ f : blowup ] also shows that the embedding @xmath19 of @xmath14 is easily modified to give an embedding @xmath69 of @xmath64 . in particular , this means that @xmath64 is also a planar graph . what is the effect of blowing up an edge @xmath47 on a chord diagram @xmath15 on @xmath14 ? if @xmath70 is a chord of @xmath15 , label its endpoints by @xmath71 and @xmath72 . if @xmath73 is on edge @xmath47 in @xmath14 , it will be doubled in @xmath64 , appearing on both @xmath47 and @xmath65 . in other words , the sequence of endpoints along @xmath47 and @xmath65 will both be the same as the original sequence along @xmath47 . the chord " @xmath70 now consists of all pairings of a point labeled @xmath71 and a point labeled @xmath72 ; there are @xmath74 such pairings , where @xmath75 is the number of endpoints of the original chord @xmath70 which were on edge @xmath47 . we give each pairing its own name , @xmath76 , yielding a new chord diagram @xmath77 on @xmath64 . if the chord diagram @xmath15 is oriented , then the orientations of @xmath71 and @xmath72 are also replicated in @xmath64 . if the chord @xmath70 in @xmath15 respected the embedding @xmath19 , and @xmath70 had an endpoint on @xmath47 , then only half of the set of chords labeled @xmath70 in @xmath64 will respect the embedding @xmath69 . so we obtain the _ reduced _ oriented chord diagram @xmath78 on @xmath64 by discarding those chords @xmath70 which do not respect @xmath69 . the intersection graph @xmath79 is defined as before . we will add to this graph edges between all the chords which came from the chord @xmath70 , to get a new intersection graph @xmath80 . figure [ f : chordblowup ] illustrates this process for an example with @xmath81 . [ h ] @xmath82 [ p : line ] let @xmath83 be a directed line segment , and let @xmath16 be an embedding . let @xmath15 be an oriented chord diagram on @xmath14 . let @xmath64 be the result of blowing up the single edge @xmath47 of @xmath14 , @xmath69 be the resulting embedding of @xmath64 , and @xmath78 be the reduced oriented chord diagram on @xmath64 obtained from @xmath15 . then @xmath19 can be extended to an embedding of @xmath15 if and only if @xmath80 is bipartite . proof : in this case , every chord in @xmath15 has both endpoints on @xmath47 , so each chord of @xmath15 corresponds to 4 chords in @xmath77 , and hence 2 chords in @xmath78 . we first show the sufficiency . if @xmath80 is bipartite , then so is its subgraph @xmath79 . since every chord in @xmath15 respects @xmath19 ( since the only edge is a cut edge ) , @xmath15 respects @xmath19 ; so by definition @xmath78 respects @xmath69 and contains a chord corresponding to every chord in @xmath15 . so by proposition [ p : circle ] , @xmath69 extends to an embedding of @xmath78 . moreover , by using the partition of vertices from @xmath80 , we can extend @xmath69 so that exactly one of each pair of chords @xmath84 is inside the cycle @xmath64 and one is outside . the chords which are outside give an extension of @xmath19 to @xmath15 . conversely , if @xmath19 extends to @xmath15 , then we can draw the chords of @xmath15 as disjoint arcs in the single region of @xmath20 . these arcs can then be drawn as arcs outside the cycle @xmath64 in @xmath85 . but there is another extension of @xmath19 to @xmath15 obtained by taking the reflection of the first embedding across the axis defined by the image of @xmath14 . these arcs can be drawn _ inside _ the circle @xmath85 . together , these give an extension of @xmath69 to an embedding of @xmath78 , so @xmath79 is bipartite . since the arcs for the chords @xmath70 and @xmath86 in this extension are on opposite sides of the circle , adding edges between them in the intersection graph will not affect the partition into two independent sets , so @xmath80 is also bipartite . @xmath35 notice that , in figure [ f : chordblowup ] , the intersection graph @xmath80 is bipartite , and the chord diagram @xmath78 can be embedded as shown . we can extend the ideas of the last section to prove the analogue of proposition [ p : nocut ] for chord diagrams on graphs with cut edges . our plan is simply to blow up every cut edge , as in section [ ss : chordline ] , and consider the resulting graph , which no longer has cut edges . consider a graph @xmath14 , and let @xmath87 be the set of cut edges . let @xmath88 be the result of blowing up every edge in @xmath87 . if @xmath15 is an oriented chord diagram on @xmath14 , then @xmath89 is the corresponding chord diagram on @xmath88 . if @xmath16 is an embedding of @xmath14 in the plane , then we have a corresponding embedding @xmath90 of @xmath88 . the regions of @xmath91 are essentially the same as the regions of @xmath20 , together with the new regions created by the blown up edges ( one new region for each cut edge in @xmath14 ) . if we label the regions of @xmath20 by @xmath92 , then we can give the corresponding regions of @xmath91 the same labels ; in addition , we will give _ all _ the regions created by blowing up cut edges on the boundary of @xmath41 the same label @xmath93 . the region " @xmath93 is the union of all these regions , and its boundary is the union of the images @xmath94 of the cut edges of @xmath41 in @xmath15 , in the same order that they are found in the boundary of @xmath41 in @xmath89 ( so the boundary of @xmath93 is a subset of the boundary of @xmath41 ) . this is illustrated in figure [ f : boundaries ] . [ h ] @xmath95 as usual , a chord of @xmath89 _ respects _ @xmath90 if it satisfies definition [ d : respect ] , treating @xmath93 as a single region . so we can consider the reduced chord diagram @xmath96 which results from removing all the chords in @xmath89 which do not respect @xmath90 . the intersection graph @xmath97 is defined as usual , with the exception that a path " on the boundary of a region @xmath93 is taken to be a sequence of consecutive edges in the boundary of @xmath93 ( as defined in the last paragraph ) , even though these edges may not be adjacent in @xmath88 . as in section [ ss : chordline ] , @xmath98 is defined by adding to @xmath97 edges between vertices corresponding to chords in @xmath96 which came from the same chord in @xmath15 . [ p : cut ] let g be a directed planar graph with a set of cut edges e , and let @xmath88 be the result of blowing up the edges in e , as described above . let @xmath16 be a planar embedding of g , and let c be an oriented chord diagram on g. label the regions of f(g ) by @xmath41 ( @xmath49 number of regions ) . we define @xmath90 and @xmath89 as described above . then f extends to an embedding of c if and only if the following conditions are met : 1 . c respects f. 2 . the vertices of @xmath98 can be labeled by the regions @xmath41 and @xmath93 so that : * if a vertex is labeled @xmath50 , then the corresponding chord in @xmath96 has both endpoints on the boundary of region @xmath50 in @xmath91 . * for each region @xmath50 ( @xmath41 or @xmath93 ) , the set of vertices labeled @xmath50 is independent . * if @xmath70 was a chord of @xmath15 with both endpoints on a cut edge in the boundary of @xmath41 , then exactly one of the chords @xmath84 in @xmath96 is labeled @xmath41 , and the other is labeled @xmath93 . proof : we first show the sufficiency . from proposition [ p : nocut ] , the conditions imply that @xmath90 extends to an embedding of @xmath96 . moreover , for every chord in @xmath15 , there is exactly one corresponding chord of @xmath96 whose image lies in the union of the regions @xmath41 ( as opposed to the regions @xmath93 ) . so we can then collapse " the regions @xmath93 ( reversing the blowup operations ) to obtain an embedding of @xmath15 which extends @xmath19 . for the necessity , assume that @xmath19 does extend to @xmath15 . then , for any chord @xmath70 of @xmath15 such that @xmath99 is an arc in the region @xmath41 , there is a corresponding arc in the region @xmath41 in the complement of @xmath91 . these arcs are all still disjoint . this provides an embedding of almost all chords of @xmath96 ; label the vertices in @xmath98 corresponding these chords by the region in which each is embedded . the vertices labeled @xmath41 will still be independent , as they were in @xmath24 . the only remaining chords are doubles " of chords which have already been embedded in some @xmath41 , and which correspond to chords in @xmath15 which had both endpoints on cut edges on the boundary of @xmath41 . so these chords have both endpoints on the boundary of @xmath93 , and the corresponding vertices can be labeled @xmath93 . it only remains to check that the vertices labeled @xmath93 will be independent . however , the order of the endpoints of the chords labeled @xmath93 around the boundary " of @xmath93 is exactly the reverse of the order of the endpoints of their doubles " around the boundary of @xmath41 . so the endpoints of two chords alternate around the boundary of @xmath93 only if the endpoints of their doubles alternate around the boundary of @xmath41 . since the vertices labeled @xmath41 are all independent , so are the vertices labeled @xmath93 . so the conditions of the proposition are satisfied , completing the proof . [ c : cut ] given a planar graph @xmath14 , a chord diagram @xmath15 on @xmath14 , and a planar embedding @xmath16 , there is a polynomial time algorithm to determine whether @xmath19 extends to an embedding of @xmath15 . proof : from corollary [ c : nocut ] , we can check whether @xmath90 extends to @xmath96 in polynomial time . the only condition that is left to check is the last one , that one of each pair @xmath84 can be labeled @xmath41 and the other @xmath93 . since , for a connected graph , there is at most one division into two independent sets ( vertices with paths which are odd or even length to a given vertex ) , this can be done very quickly after determining that the set of chords with endpoints on @xmath41 and @xmath93 is bipartite . @xmath35 as an application , we will use chord diagrams and intersection graphs for graphs to study gauss codes for graphs . the gauss code was first developed by c. f. gauss to study closed curves in the plane which intersect themselves only in transverse double points ( called crossings " ) . label the crossings by some set of symbols ( such as integers or the letters of the alphabet ) , and give the curve an orientation . beginning at an arbitrary crossing , we obtain a gauss code for the curve by writing down the sequence of labels of the crossings passed as we traverse the curve , following its orientation . the gauss code provides a convenient combinatorial representation of the curve . however , not all sequences which look like " gauss codes actually represent closed curves in the plane . the problem is to determine ( in polynomial time ) which sequences of symbols can be obtained from a closed curve in this way - i.e. , which sequences are _ realizable _ as the gauss code for a closed curve in the plane . there are several different solutions to this problem @xcite , we will describe one in section [ ss : circle ] which uses proposition [ p : circle ] . kauffman s recent development of the theory of virtual knots @xcite was motivated in part by a desire to realize the unrealizable " gauss codes . this paper was largely motivated by a desire to extend kauffman s work to spatial graphs , and study virtual " spatial graphs . as a preliminary , we need to extend the notion of a gauss code to general graphs , and consider when an abstract code is realizable by an immersion of a graph in a plane . in future work , this will lead to the introduction of virtual spatial graphs @xcite . our goal is to generalize to the situation where we have an immersion of an arbitrary graph in the plane , rather than a circle . in this case , as described in section [ ss : graph ] , the gauss code " is a set of sequences associated with the edges of the graph . we will give a algorithm , based on proposition [ p : cut ] , for determining whether an abstract gauss code can be realized as the gauss code for an immersed graph and , if so , constructing the desired immersion . we will first describe a method for determining whether a classical gauss code is realizable by an immersed closed curve in the plane . this problem was first solved by dehn @xcite , and a clear presentation of his solution is given by read and rosenstiehl @xcite . our method builds on dehn s , and is equivalent to that given by de fraysseix and ossona de mendez @xcite , but is presented in the language of chord diagrams developed in section [ s : chord ] . an oriented closed curve in the plane gives rise to a crossing sequence , the _ gauss code _ for the curve , by labeling the @xmath26 self - crossings of the curve , and then writing down the sequence of @xmath100 labels encountered as we trace the curve following its orientation ( obviously , the code is only unique modulo cyclic permutations ) . an example is shown in figure [ f : gaussloop ] . the problem , initially posed by gauss , is to determine whether an arbitrary sequence of length @xmath100 , containing two occurrences each of @xmath26 symbols , is realizable as the gauss code of a closed plane curve . we will call such a sequence a _ crossing sequence _ , and refer to the symbols as _ crossings_. we follow read and rosenstiehl by defining a _ splitting _ of a crossing sequence at each crossing , and then studying the resulting _ split sequence _ or _ split code_. the notion of splitting a crossing is motivated by the idea of smoothing a self - intersection of a curve in the plane into two non - intersecting arcs . we consider a small neighborhood of a self - crossing @xmath102 . inside this neighborhood , we replace the two intersecting arcs with two disjoint arcs with the same endpoints , as shown in figure [ f : split ] . this replacement is done so as to preserve the number of components of the curve , which requires reversing the orientation of one of the arcs from @xmath102 to itself , so that the resulting curve is still oriented . aside from the change of orientation , the operation is entirely local , so if we label points @xmath104 and @xmath105 on the two new arcs , we can draw a chord @xmath106 between them which does not intersect the rest of the curve and is contained within the original neighborhood of the crossing , as shown in figure [ f : split ] . notice that , if the chord @xmath106 is oriented , its endpoints at @xmath104 and @xmath105 will have opposite orientations . [ d : splitting ] given a sequence @xmath107 , * splitting s at p * means to replace s by @xmath108 , where we say that @xmath104 and @xmath105 have opposite orientations , and @xmath109 is the result of writing the subsequence @xmath110 in reverse order and reversing the orientation of any previously split symbols in @xmath110 . given a crossing sequence @xmath55 , the result of splitting @xmath55 at _ every _ crossing is called a _ split sequence _ or _ split code _ for @xmath55 , and is denoted @xmath111 . the split code is _ not _ unique - it depends upon the order in which the crossings are split . figure [ f : splitcode ] illustrates this procedure with an example of smoothing all the crossings in a closed plane curve , and the corresponding operations on the gauss code in order to obtain the split code . the result of performing the splitting process on a closed plane curve is a _ simple _ closed curve ( i.e. an embedded circle ) labeled by the split code . we will call this the _ split curve_. clearly , the split curve is an oriented chord diagram on the circle , in the sense of definition [ d : orientchord ] . to reconstruct the original curve from the split curve , we simply need to connect each pair of points @xmath104 and @xmath105 by the chord @xmath106 shown in figure [ f : split ] , and contract the chords to bring the points back together , reversing the splitting process . so reconstructing the original curve is done by extending the embedding of the underlying circle of the split curve to an embedding of the oriented chord diagram represented by the split curve . this leads directly to the following result . [ t : closedcurve ] @xcite a crossing sequence s is realizable as the gauss code of a closed curve in the plane if and only if the the oriented chord diagram on the circle represented by the split sequence s * can be embedded in the plane . proof : from the discussion above , if the crossing sequence @xmath55 is realizable , then there is an embedding of the chord diagram @xmath111 . conversely , an embedding of @xmath111 can be turned into a collection of intersecting closed loops by expanding each chord into a crossing , reversing the splitting operation in figure [ f : split ] . it only remains to check that this process will yield a single closed curve with gauss code @xmath55 ( this is the _ complete traceability _ of @xcite ) . say that @xmath111 was produced from @xmath55 via a series of sequences @xmath113 , where @xmath114 is the result of splitting crossing @xmath11 in @xmath93 . we will expand the chords between pairs of symbols in @xmath111 in the reverse order from how they were split - i.e. the last symbol to be split is the first chord added . so we begin with the chord @xmath115 between @xmath116 and @xmath117 in @xmath118 . expanding this chord into a crossing , and reversing the orientation on the arc of @xmath119 originally directed from @xmath116 to @xmath117 , exactly reverses the splitting process of definition [ d : splitting ] . a key observation is that , since @xmath116 and @xmath117 have opposite orientations , the result of expanding the chord is a closed curve with one component . since we have exactly reversed the final splitting operation , the result is an oriented closed curve @xmath120 whose gauss code is @xmath121 . we can continue this process , always choosing to expand the chord corresponding to the last crossing which was split , so that its endpoints have opposite orientation and the resulting curve still has a single component . in this way , we can successively construct an oriented closed curve @xmath122 for each sequence @xmath93 . ultimately , we will have a curve @xmath15 whose gauss code is the original crossing sequence @xmath55 , as desired . @xmath35 + + remarks : there are several remarks that should be made at this point . 1 . to check whether a particular split code is embeddable , we apply the criteria of proposition [ p : circle ] . so we check that the two occurrences of each symbol have opposite orientation and that the intersection graph is bipartite . the importance of theorem [ t : closedcurve ] is that this can be done in polynomial time . 2 . comparing with the _ d - switch _ operation of de fraysseix and ossona de mendez @xcite , the condition that the two occurrences of each symbol have opposite orientation is equivalent to saying that a symbol and its twin are not adjacent in the interlacement graph . so , the result of theorem [ t : closedcurve ] then also follows from the proof of theorem 6 in @xcite . kauffman @xcite provides another characterization , in which the condition that each pair of symbols have opposite orientations is replaced by the condition that the crossing sequence @xmath55 is _ evenly intersticed _ , meaning that between two occurrences of any symbol is a word of even length . however , kauffman does not prove that this requirement is sufficient . it is not hard to show that , if @xmath55 is evenly intersticed , then the number of unsplit symbols between a pair of split symbols ( at any stage in the process of splitting @xmath55 ) is even exactly when the split symbols have opposite orientations . since , in @xmath111 , the number of unsplit symbols is 0 , and hence even , all pairs of split symbols have opposite orientations , so by proposition [ p : circle ] and theorem [ t : closedcurve ] the crossing sequence is realizable . this shows that kauffman s characterization is also valid . not every sequence of symbols @xmath111 which satisfies the conditions of theorem [ t : closedcurve ] is the split code of a crossing sequence . for example , if the reconstruction algorithm is applied to the sequence @xmath123 , the result will be a curve with two components ; however , this is not a contradiction because this sequence _ can not _ be obtained as the split code of a crossing sequence . to see this , observe that if we unsplit " symbol @xmath124 we are left with @xmath125 - since the two occurrences of @xmath126 have the same orientation , this could not be the result of a splitting operation ; attempting to unsplit @xmath126 leads to a similar situation . we would like to extend the ideas of the last section to general directed graphs . in this section , our graphs will be _ vertex - oriented _ , meaning that we have a cyclic order on the edges incident to each vertex . given a graph @xmath14 , a _ gauss code for g _ consists of a set of symbols @xmath126 and a set of sequences in these symbols , one for each edge of the graph . each sequence will begin with the vertex at the start of the edge , and the position of the edge in the cyclic ordering around that vertex , and end with the vertex at the end of the edge , with the position of the edge in the cyclic ordering around the end vertex . in between are symbols from @xmath126 . notice that the abstract graph @xmath14 can be reconstructed from the set of gauss codes . while an abstract graph has many possible gauss codes , we can assign a unique code ( up to cyclic permutation of the vertex orderings ) to any immersion of the graph in the plane , just as we do for immersions of circles . in this case , the sequence of symbols just records the order of the crossings along each edge . an example is shown in figure [ f : example ] . our first observation is that every symbol in the set of sequences ( other than the vertices ) must appear exactly twice . this allows us to create a _ split code _ from the original gauss code . as in the previous sections , the motivation for the split code is to split the crossings of an immersion of a graph in the plane . there are two ways to split a crossing ( one is shown in figure [ f : split ] ) - we want to perform the split so that the graph remains connected . if both splits preserve the connectedness of the graph , we can choose one arbitrarily . as before , we also want to keep track of the relative local orientations of the smoothed arcs . these operations can be defined abstractly in terms of the gauss code . if @xmath102 is a symbol in the alphabet @xmath126 , then splitting the code at @xmath102 means one of the following operations : 1 . if there is a sequence @xmath128 , then the new sequence is @xmath129 . so the subsequence @xmath110 has been reversed . if there are two sequences @xmath130 and @xmath131 , then the new sequences are either @xmath132 and @xmath133 or @xmath134 and @xmath135 . these are illustrated in figure [ f : split2 ] . notice that , in the second case , the underlying graph is changed in either choice , but at least one of these graphs is connected . we will always make a choice which keeps the underlying graph connected . also notice that the first choice preserves orientations along each edge segment , while the second choice reverses the orientations along segments coming from one of the two edges . the result of performing this splitting operation on every symbol in the set of sequences gives a split code for the original gauss code . the underlying abstract graph for the split code is called the _ split graph_. the splitting process is not unique - the final set of sequences after all crossings have been split will depend on the order in which they were split , and how they were split . any of these possibilities is called a split code for the original gauss code . a split code can be viewed as an oriented chord diagram on its split graph . [ l : planar ] if a gauss code is derived from an immersion of a connected graph , then the underlying split graph for any split code is planar and connected ; moreover , the split code determines a unique ( up to isotopy ) spherical embedding of the split graph . proof : if we begin with an immersion of a graph and split every crossing , we are left with a graph embedded in the plane ; this implies that the graph underlying the split code is planar , and the split code is realized by some planar embedding . the embedding is connected since we choose our splittings to keep the graph connected . it remains to show that this embedding is unique ( up to isotopy on the sphere ) . we will describe an algorithm for constructing a planar embedding from the split code . we will show that ( up to isotopy on the sphere ) the algorithm produces at most one embedding , and that it fails to produce an embedding respecting the split code only if one does not exist . in fact , we will not use all the information of the split code , only the underlying graph and the order of the edges around each vertex . the idea of the algorithm is to trace out faces of the planar embedding . begin with an arbitrary vertex @xmath66 , and an edge @xmath47 incident to @xmath66 . embed @xmath66 and @xmath47 . say that @xmath67 is the other endpoint of @xmath47 . let @xmath19 be the next edge adjacent to @xmath67 after @xmath66 ( using the cyclic ordering of the edges adjacent to @xmath67 ) , and add @xmath19 to the embedded graph . continue this process , moving one edge around each vertex . the first time we return to a vertex we have visited before , we will have two choices of how to draw the edge - clockwise or counterclockwise around the previously embedded parts of the graph . but these choices are isotopic on the sphere . once this first choice is made , all future edges will connect two points on the boundary of a ( topological ) disk , so all possible ways to draw them will be isotopic . as a result , there are no more choices to be made , and the resulting embedding is unique . when we come to an edge we have reached before , skip it and move to the next unused edge on that vertex . if there are no more unused edges on the vertex , then backtrack to the last vertex with an unused edge . continue in this way until all the edges have been drawn ( the procedure will terminate , since the graph is connected ) . if we are able to do this without edges crossing , we will have a planar embedding of the split graph , uniquely determined by the underlying graph and the vertex orientations . @xmath35 lemma [ l : planar ] allows us to embed the split graph in the plane . as in earlier sections , if the original gauss code is realizable then we can recover an immersion of a graph from this embedding code by connecting the two occurrences of each symbol in the split graph by disjoint chords which respect the orientations of the symbols , and then expanding the chords as in figure [ f : expand2 ] . proof : if @xmath55 is realizable , then the split code @xmath111 can be embedded in the plane as an oriented chord diagram . so the underlying split graph can be embedded ( and is therefore planar ) , and this embedding extended to the chord diagram . conversely , if the chord diagram can be embedded , we can expand each chord into a crossing as in figure [ f : expand2 ] . it only remains to check that the result is an immersion of @xmath14 with gauss code @xmath55 . as in theorem [ t : closedcurve ] , we expand the chords between symbols in the reverse order that the symbols were split . say that @xmath111 was produced from @xmath55 via a series of sequences @xmath113 , where @xmath114 is the result of splitting crossing @xmath11 in @xmath93 . since each splitting was chosen to keep the graph connected , expanding the splits in the reverse order will also keep the graph connected . our goal is for these expansions to exactly reverse the effects of the splitting process on the original gauss code . there is one possible ambiguity : the expansion process shown in figure [ f : expand2 ] assumes that crossings between different edges were always split so as to preserve the orientations of the edges ; but , in order to keep the graphs connected , it may have been necessary to reverse the orientations of one of the two edges , as shown in figure [ f : split2 ] . so the expansion may leave one of the edges with the wrong " orientation . however , since we have recorded all the sequences @xmath93 , this is easily fixed - simply compare the code for our current graph with the corresponding @xmath93 . they will differ in at most the orientation of one edge , so we can simply reverse the orientation of this edge in our graph , and continue the process . the final result will be an immersion of @xmath14 with gauss code @xmath55 . @xmath35 proof : producing the split code can be done in linear time , and there are linear - time algorithms for determining whether the split graph is planar @xcite . finally , by corollary [ c : cut ] , we can determine whether the embedding of @xmath138 extends to @xmath111 in polynomial time as well . @xmath35 + remark : in the theory of spatial graphs , a spatial embedding of a graph can be described by the _ gauss code _ for a particular diagram of the knot ( a projection of the knot to the plane ) . in this context , the gauss code comes with additional information at each crossing , indicating which edge crosses over the other and the sign of the crossing . corollary [ c : gauss ] implies that there is a polynomial time algorithm to determine whether a gauss code is realizable in this context as well . once we have determined whether the code can be realized by an immersion of the graph in the plane , and constructed such an immersion , using theorem [ t : graph ] , we simply need to check whether the sign and over / under information at each crossing are compatible . 10 d. bar - natan : on the vassiliev knot invariants , _ topology _ 34 , no . 423 - 472 , 1995 d. bar - natan and s. garoufalidis : on the melvin - morton - rozansky conjecture , _ invent . _ 125 , pp . 103 - 133 , 1996 j. birman and x. s. lin : knot polynomials and vassiliev s invariants , _ invent . _ 111 , pp . 225 - 270 , 1993 a. bouchet : circle graph obstructions , _ j. comb . theory , series b _ 60 , p.107 - 144 , 1994 j. boyer and w. myrvold : on the cutting edge : simplified @xmath139 planarity by edge addition , _ j. graph algorithms and applications _ , vol . 8 , no . 241 - 273 , 2004 s. v. chmutov , s. v. duzhin , and s. k. lando : vassiliev knot invariants i , ii , iii , _ advances in soviet mathematics _ 21 , pp . 117 - 145 , 1994 m. dehn : ber kombinatorische topologie , _ acta math . _ 67 , 1936 , pp . 123 - 168 s. even and a. itai : queues , stacks and graphs , in _ theory of machines and computations _ ( z. kohavi and a. paz , eds ) , academic press , new york , 1971 , pp . 71 - 86 t. fleming and b. mellor : virtual spatial graphs , preprint , 2005 , arxiv : math.gt/0510158 h. de fraysseix and p. ossona de mendez : on a characterization of gauss codes , _ discrete comput . _ , v. 22 , 1999 , pp . 287 - 295 h. goda : bridge index for theta curves in the 3-sphere . _ topology appl . _ , v. 79 , 1997 , pp . 177 - 196 . l. kauffman : virtual knot theory , _ europ . j. combinatorics _ , v. 20 , 1999 , pp . 663 - 691 s. kinoshita : on @xmath140-curves in @xmath141 and their constituent knots , _ topology and computer science ( atami , 1986 ) _ , kinokuniya , tokyo , 1987 , pp . 211 - 216 s. lando : on a hopf algebra in graph theory , _ j. comb . theory , series b _ 80 , p.104 - 121 , 2000 t. a. mckee and f. r. mcmorris : _ topics in intersection graph theory _ , siam , philadelphia , 1999 b. mellor : the intersection graph conjecture for loop diagrams , _ j. knot theory ramif . _ 9 , no . 2 , 2000 b. mellor : a few weight systems arising from intersection graphs , _ michigan math . j. _ , vol . 509 - 536 , 2003 r. nikkuni : delta edge - homotopy on theta curves , _ math . cambridge philos . _ , v. 138 , 2005 , pp . 401 - 420 . r. c. read and p. rosenstiehl : on the gauss crossing problem , _ colloq . janos bolyai _ , v. 18 , _ combinatorics _ , keszthely , hungary , 1976 , pp . 843 - 876

the galaxy appears to be richer in young , massive stellar clusters than previously known , due to advances in infrared surveys which have uncovered deeply embedded regions of star formation . young , massive clusters can significantly impact the surrounding interstellar medium ( ism ) and hence radio observations can also be an important tracer of their activity . several hundred cluster candidates are now known by examining survey data . here we report on multiwavelength observations of six of these candidates in the galaxy . we carried out 4.9 and 8.5 ghz vla observations of the radio emission associated with these clusters to obtain the physical characteristics of the surrounding gas , including the lyman continuum photon flux and ionized gas mass . _ spitzer _ infrared array camera observations were also made of these regions , and provide details on the stellar population as well as the dust continuum and polycyclic aromatic hydrocarbon emission . when compared to the known young , massive clusters in the galaxy , the six cluster candidates have less powerful lyman ionizing fluxes and ionize less of the h ii mass in the surrounding ism . therefore , these cluster candidates appear to be more consistent with intermediate - mass clusters ( 10 - 10 m ) .

1208.2003

massive stars return a significant fraction of their mass to the interstellar medium ( ism ) by means of stellar winds and supernovae . massive stars are rare , have short lifetimes , and are predominantly observed in young , massive ( @xmath3 20 myr , @xmath4 10@xmath1 m@xmath2 ) clusters . to date , few such clusters are known and well - studied in the galaxy because many are located in the highly obscured galactic disk . a complete census of young , massive stellar clusters in the galaxy is needed to determine their total number and distribution . estimates have shown that the galaxy should host close to 100 of these clusters ( @xcite ) . however , only a few such clusters are well known and studied . for example , the arches , quintuplet , and central clusters in the galactic center were uncovered with infrared spectroscopic observations , and are responsible for heating and shaping much of the surrounding ism ( @xcite ; @xciteb ; @xcite , @xcite , @xcite ) . in addition , several other young , massive clusters in the galaxy have been recently studied : rsgc 1 & 2 ( @xcite ) , w49a ( @xcite ) , ngc 3603 ( @xcite ) , and westerlund 1 ( @xcite ) . [ [ section ] ] over the last decade , surveys in the near - infrared and mid - infrared ( 2mass and _ spitzer_/glimpse in particular ) have revealed as many as 1000 candidate clusters ( e.g. @xcite ) . a sample of 40 possible young , massive clusters were selected from these candidates using multiwavelength data , including radio and optical observations . among these 40 sources , six were identified as having bright associated radio emission from the nrao vla sky survey ( vla - nvss ; @xcite ) . here , we present a multiwavelength study of these six cluster candidates in the galaxy . the radio observations in this study have higher sensitivity and resolution than the vla - nvss data , and provide greater insight into the properties of the gas and dust surrounding these clusters . the primary goal of the radio data is to determine the properties of the ism surrounding the candidate massive clusters , and to compare them to the known young , massive clusters in the galaxy . estimates of the ionized mass and ionizing flux can be obtained with the radio data , from which the candidate sources can be categorized as supermassive or intermediate - mass clusters . the primary goal of the infrared data is to characterize the infrared emission surrounding the clusters as well as the infrared emission arising from cluster members . comparison between the radio and infrared observations provides additional insight about the interstellar environment of the stellar clusters and the impact of the stars on the surrounding medium . presented here is a summary of the background information for the six candidate massive stellar clusters . table [ tbl-1 ] lists the six candidate sources , including the most recent distance estimates and alternative identifiers . note that the names used here have been shortened from names that originate in the bica & dutra catalogs of infrared star clusters ( @xcite ) . sharpless ( sh2 ) numbers are given for the clusters that have been observed in the sharpless catalog of h ii regions ( @xcite ) . bd52 is a stellar cluster located in the h ii region sh2 - 187 . @xcite used the spectrophotometric distance of a single star ( b2.5v ) to determine a distance of 1.44 @xmath5 0.26 kpc to the stellar cluster . according to a multifrequency study by @xcite , the h ii region is fairly young with a kinematical age of @xmath62@xmath710@xmath8 yr . optical and 1.4 ghz radio emission observed in this study suggest bd52 is enshrouded in its parental cloud , and that a b0 zero - age - main - sequence star is the exciting source . the presence of a water maser demonstrated by @xcite provides evidence that this h ii region could be undergoing star formation . bd65 is a stellar cluster located in the h ii region sh2 - 209 . @xcite have studied this complex at millimeter wavelengths to look for evidence of embedded star formation . their observations show that the region around bd65 can be described as a ring of cloud cores around a star cluster , and the complex may be a site of triggered star formation . @xcite adopted a distance of 8.2 kpc , yet there is some discrepancy in the distance to this source ( estimates range from 4.9 to 9.8 kpc ; @xcite ) . a distance to bd65 of 9.8 kpc is adopted in this work based on the kinematic and photometric distance provided by @xcite . this appears to be one of the first times bd84 has been thoroughly studied . however , it has been observed in several h ii region surveys and has been associated with the h ii region sh2 - 283 . its distance has been determined spectrophotometrically to be 8.06 @xmath5 0.30 kpc by @xcite . the galactic h ii region sh2 - 294 , or bd95 , has been studied in several papers , mostly h ii region surveys , including a multiwavelength study of the region ( @xcite ) . the distance to bd95 has been estimated several times ; most recently it has been determined spectrophotometrically to be 3.24 @xmath5 0.56 kpc ( @xcite ) . the ionizing star is believed to be a star of spectral type @xmath6b0v ( @xcite ) . @xcite confirm the ionizing source , and quote a spectral type of b0.5v . their high resolution near infrared images reveal a double cluster morphology with one cluster fully embedded and optically invisible . db7 is associated with an h ii region known as sh2 - 307 . to the sw of db7 is an ultracompact h ii region and photodissociation region ( pdr ) associated with the db8 stellar cluster ( @xcite ) . db8 s distance has been determined spectrophotometrically by @xcite to be 2.65 @xmath5 0.4 kpc . we adopt this same distance for db7 . if the two clusters are associated with the same h ii complex , db7 may have formed first and triggered the collapse of db8 ( @xcite ) . the most recent detailed study of dbcl23 was done by @xcite , who estimated an upper limit of 3.5 kpc for its distance . the spectrum of the nebula around this source indicates a temperature of 40,000 - 45,000 k ( corresponding to an o5 - o7 star ) for the ionizing stellar spectrum . however , @xcite discovered a significant discrepancy between the ionizing flux supplied by a mid o - type star and that inferred from the optical and radio parameters . it is possible that dbcl23 is ionized by an external o - type star , although the ionizing source was not identified . their results confirm that dbcl23 is not a massive star forming site . radio observations were made at 8.5 and 4.9 ghz with the very large array ( vla ) telescope of the national radio astronomy observatory ( nrao ) in august 2008 through observing program al723 . the vla was in its d array configuration allowing for resolutions of 5 - 20 . at both frequencies , j1331 + 305 was used as a flux density calibrator . standard procedures for calibration , editing and imaging were carried out using the astronomical image processing software ( aips ) of the nrao . table 2 summarizes the image parameters for the different sources . all images were made with robust = 0 weighting . images of all six targets were taken in 3.6 , 4.5 , 5.8 , and 8.0 @xmath9 m bands using the infrared array camera ( irac ) onboard the _ spitzer space telescope _ ( program i d 30734 , pi figer ) . the 5 position gaussian dithering pattern was used , and the resulting images in each wavelength cover a 5 by 5 area , with a pixel size of 1.2 . the total integration time was 471 s in each of the four bands . the _ spitzer _ mopex tool was applied to the basic calibrated data to create mosaicked images of each band . an idl adapted version of daophot was used to perform photometry on each band , and the results were cross - correlated . if a source was not present in one or more of the 4.5 , 5.8 , or 8.0 @xmath9 m bands , we reported a null result for the specific band(s ) , rather than remove the source entirely . in the following sections the morphology of the clusters and their surroundings in both radio and infrared images is discussed . the correlation between the observed radio and infrared emission in figures [ fig1]-[fig7 ] is also examined . the wavelengths at which the observations were taken provide information about the type of emission produced by the stellar clusters . ionized hydrogen in the clusters emits thermal free - free emission , which can be traced at radio frequencies . emission from the thermal dust continuum is seen at infrared frequencies . the irac bands at 3.6 , 5.8 , and 8 @xmath9 m contain polycyclic aromatic hydrocarbon ( pah ) features . pah emission typically arises in pdrs and can be used to delineate the edge of the pdr . the 3.6 @xmath9 m band is dominated by stellar emission , and the 8 @xmath9 m band contains primarily diffuse continuum and pah emission . these two bands were used for the colorscale seen in figures [ fig1]-[fig7 ] with contours overlaid representing the radio emission at 8.5 and 4.9 ghz . comparing the distribution of radio and infrared emission provides us with knowledge about the morphology and emission mechanisms of the stellar clusters and their h ii regions . figure [ fig1 ] shows bd52 with contours of radio emission ( at 8.5 ghz ( left ) and 4.9 ghz ( right ) ) overlaid on 3.6 and 8 @xmath9 m images . the radio emission associated with bd52 has an extended morphology . the 8.5 ghz radio contours in the left image show a curved , partial - shell structure of approximately 1 ( thickness ) x 3 ( diameter ) ( 0.4 pc x 1.3 pc at d = 1.44 kpc ) in size , whereas the 4.9 ghz radio emission shows a complete shell - like structure as seen in the image on the right . given the distribution of radio emission at 4.9 ghz , it appears that there must be missing flux at 8.5 ghz . indeed , the largest angular size of the vla interferometer at 8.5 ghz in d - array is @xmath63 . this means that extended structures larger than this angular size ( 3 ) will not be detectable at this frequency . the 4.9 ghz emission associated with bd52 shows that the source has a diameter of about 4 ( 1.7 pc at d = 1.44 kpc ) , therefore , flux is missing in the 8.5 ghz image . two components are visible in the 3.6 and 8 @xmath9 m emission : the stellar sources and the diffuse emission that appears to be part of a larger infrared shell extending beyond the 4.9 ghz radio shell . the emission is more apparent at 8 @xmath9 m shown in the right image , and to the north looks more complete than at 3.6 @xmath9 m . the stellar cluster is visible in the 3.6 @xmath9 m image on the left . an investigation of the stellar components of bd52 by trombley et al . ( in prep . ) suggests a cluster mass of 1300 m@xmath10 and an age of 1 myr . they conclude that several main sequence early b stars are producing the ionizing flux which powers the h ii region . in the sw corner of the image is a bright radio point source ( at approximately ra , dec ( j2000 ) : 01 23 00 , 61 48 15 ) . this radio point source has no counterpart in the infrared and is known to be a bright non - thermal extragalactic radio source ( @xcite ) . we calculate the spectral index of this compact source to be [email protected] , which is in agreement with the non - thermal nature of the source . figure [ fig2 ] shows a small area from the center of bd52 ( approximately 50 by 50 , corresponding to 0.35 x 0.35 pc at d = 1.44 kpc ) with 8.5 ghz radio contours overlaid on the 3.6 @xmath9 m image . in the center of the stellar cluster , there is a point - like radio source that appears to be coincident with an infrared source . the 8.5 ghz flux is [email protected] mjy . this association is suggestive and one interpretation is that the radio emission arises from a stellar wind in one or more of the members of the stellar cluster . presuming the radio emission associated with the stellar source is due to the massive stellar wind , it is possible to calculate the mass - loss rate using equation ( 1 ) of @xcite , based on @xcite . assuming a terminal wind velocity of @xmath61000 km s@xmath12 and a distance of 1.44 kpc , the mass - loss rate is approximately 0.35@xmath710@xmath13 m@xmath2 yr@xmath12 . this mass - loss rate is somewhat higher than expected for an early b - type main sequence star ( @xcite ) . this particular radio source also has counterparts in the near - ir ( hst / nic3 ) , mid - ir ( spitzer / irac ) , and x - ray ( cxo ) ( trombley et al . ( in prep ) ) . figure [ fig3 ] shows bd65 with contours of radio emission ( at 8.5 ghz ( left ) and 4.9 ghz ( right ) ) overlaid on 3.6 and 8 @xmath9 m images . the 3.6 and 8 @xmath9 m infrared emission associated with bd65 is primarily diffuse with denser cores of bright emission that form a shell - like structure around the central cluster located at ra , dec ( j2000 ) : 04 11 08 , 51 10 00 . the shell of infrared emission is outlined by the 8.5 ghz radio emission with a diameter of @xmath62 ( 5.7 pc at d = 9.8 kpc ) shown in contours in the left image . this emission also extends to the ne and to the sw of the central stellar cluster . the 4.9 ghz radio emission in the right image is more extended than the 8.5 ghz emission . the morphology defined by this radio emission differs from that seen in the 3.6 and 8 @xmath9 m emission . the two brightest cores of infrared emission ( e.g. ra , dec ( j2000 ) : 04 11 05 , 51 10 00 ; 04 11 05 , 51 10 30 ) are also noticeable in the radio emission , but are not resolved into two distinct components . there is another notable region south of the central cluster at ra , dec ( j2000 ) : 04 11 04 , 51 08 00 that is visible at both radio and infrared frequencies . this compact region of bright emission approximately 0.5 ( 1.4 pc at d = 9.8 kpc ) in size is not associated with the extended emission at 8.5 ghz , but is part of the extended emission at 4.9 ghz . figure [ fig4 ] shows bd84 with contours of radio emission ( at 8.5 ghz ( left ) and 4.9 ghz ( right ) ) overlaid on 3.6 and 8 @xmath9 m images . the radio emission at both frequencies for bd84 shows a symmetric distribution of contours around the central cluster about 1 ( 2.3 pc at d = 8.06 kpc ) in size with surrounding diffuse emission . the emission at 3.6 and 8 @xmath9 m shows a shell - like structure with bright edges , whereas the radio emission is more complete with a single central peak in emission . the 4.9 ghz radio emission in the image on the right contains a large amount of diffuse emission that extends well past the infrared emission with an approximate size of 3 x 2 ( 7.0 x 4.7 pc at d = 8.06 kpc ) . the 8.5 ghz radio emission in the left image only extends ne of the cluster . the peak for the infrared emission lies west of the central cluster at about ra , dec ( j2000 ) : 06 38 26 , 0 44 30 . the peak in radio emission , however , appears closer to the center of the cluster at ra , dec ( j2000 ) : 06 38 27 , 0 44 45 . the radio source at ra , dec ( j2000 ) : 06 38 34 , 0 45 00 appears distinct in the 8.5 ghz radio emission , but is part of the extended emission at 4.9 ghz . bd95 is shown in figure [ fig5 ] with contours of radio emission ( at 8.5 ghz ( left ) and 4.9 ghz ( right ) ) overlaid on 3.6 and 8 @xmath9 m images . the 4.9 ghz radio contours show a complete shell with a diameter of @xmath65 ( 4.7 pc at d = 3.24 kpc ) . there is corresponding 3.6 and 8 @xmath9 m infrared emission only along the eastern side where the brightest infrared emission can be seen . the 8.5 ghz contours shown in the left image appear drastically different with emission only around the brightest infrared emission . as with bd52 , it appears that flux is missing at 8.5 ghz , most likely due to the aforementioned largest angular size limitations of the vla . the structure of the 3.6 and 8 @xmath9 m emission is complex in comparison to the simple shell - like 4.9 ghz emission . the infrared images consist of a bright region of emission at ra , dec ( j2000 ) : 07 16 39 , -09 25 45 with diffuse arm - like features emanating out to the ne and sw . this peak emission is host to an embedded cluster ( see 2.4 ) . the cluster containing the ionizing source is visible at 3.6 @xmath9 m in the left image and is located nw of the cluster at ra , dec ( j2000 ) : 07 16 33 , -09 25 15 . the radio emission associated with db7 is shown in contours ( at 8.5 ghz ( left ) and 4.9 ghz ( right ) ) overlaid on 3.6 and 8 @xmath9 m images in figure [ fig6 ] . the infrared emission in db7 has a different morphology than the radio emission . the 8.5 and 4.9 ghz radio emission contains a denser circular region of contours around the cluster , and some surrounding diffuse emission . the densest radio contours which encompass the central cluster ( seen at 3.6 @xmath9 m ) have a diameter of @xmath62 ( 1.5 pc at d = 2.65 kpc ) . the 4.9 and 8.5 ghz radio emission have similar morphology , although the 4.9 ghz emission in the right image extends farther west . the 3.6 and 8 @xmath9 m infrared morphology resembles that of bd95 with areas of bright emission and diffuse arm - like structures . one of these structures which arcs out to the nw has corresponding radio emission , most noticeable at 8.5 ghz in the left image . other than this structure , the radio and infrared are not well correlated . the emission peaks are offset with the infrared peak occurring at ra , dec ( j2000 ) : 07 35 33 , -18 45 30 , and the radio peak occurring at ra , dec ( j2000 ) : 07 35 35 , -18 45 30 . the associated source , db8 , can be seen in the right image at ra , dec ( j2000 ) : 07 35 39 , -18 48 45 . figure [ fig7 ] shows dbcl23 with contours of radio emission ( at 8.5 ghz ( left ) and 4.9 ghz ( right ) ) overlaid on 3.6 and 5.8 @xmath9 m images . the radio contours at both frequencies are very similar , and both are well correlated to the infrared emission . there is only a slight difference in morphology between the infrared and radio with the radio emission being filled - in , and the infrared emission more shell - like with a central cavity . the contours of 4.9 ghz emission shown in the image on the right give a diameter of approximately 2 ( 2 pc at d = 3.5 kpc ) for dbcl23 . the diffuse 5.8 @xmath9 m emission seen in the right image extends farther than the 4.9 and 8.5 ghz radio contours . there are double emission peaks separated by about 30 visible in the radio and infrared emission . the location of these peaks in the radio emission is offset from the peaks in the infrared emission . examination of figures [ fig1]-[fig7 ] reveal some overall trends for the ism surrounding the candidate clusters : ( 1 ) in all but two sources , the infrared emission ( 3.6 and 8 @xmath9 m ) extends beyond the associated radio emission . the two exceptions to this are bd84 and db7 , for which the 4.9 ghz emission extends farther than the infrared emission ; ( 2 ) the morphology of the infrared emission for the cluster sources bd52 , bd84 , and dbcl23 is shell - like with a cavity and bright emission along the edges . the radio emission in bd52 , bd84 , and dbcl23 is also more symmetric and roughly follows the infrared emission . the shell - like morphology at both radio and infrared wavelengths allows for a more in - depth analysis to be carried out for these three sources ( 5.1 ) ; ( 3 ) on the other hand , bd65 , bd95 , and db7 are more complex and less symmetric in morphology at radio and infrared wavelengths ; and ( 4 ) the peak intensity seen in the radio emission is offset from the peak in the infrared emission . this trend is also noticeable in the slice analysis performed on the three cluster candidates bd52 , dbcl23 , and bd84 ( see 5.1 ) . we can estimate the physical properties of the radio continuum emission by examining the flux density of each source . flux densities have been measured within the contour representing 3 times the rms level at 4.9 ghz as seen in figure [ fig8 ] . approximate angular sizes have also been determined from the full width at zero intensity of slices taken across each source . the 3@xmath14 contour at 4.9 ghz was used to delineate the edge of the sources . the integrated flux density and angular size values are reported in table [ tbl-3 ] . based on the radio continuum parameters , physical properties of the ionized gas associated with the candidate clusters have been derived using the formulation of @xcite . in using these formulae , a uniform density , spherical , ionization bounded h ii region with an electron temperature of 10,000 k and an abundance of ionized helium to hydrogen of 0.1 is assumed . table [ tbl-4 ] presents the derived quantities from the radio continuum parameters shown in table [ tbl-3 ] . due to flux missing from a couple of the sources in the 8.5 ghz observations , an accurate measure of the flux density could not be obtained , thereby preventing a useful analysis at this frequency . using the formulae , we derive a linear radius of the equivalent sphere for the h ii region ( r ) , electron density ( n@xmath15 ) , emission measure ( em ) , total mass of ionized gas ( m@xmath10 ) , and the lyman continuum photon flux ( n@xmath16 ) . the lyman continuum photon flux is the rate of photons emitted capable of ionizing the surrounding hydrogen . this value provides a measure of the cluster s power by quantifying its ability to impact the surrounding gas . the authors acknowledge that the sources presented in this study are not well described by a uniform density sphere . however , past studies ( most notably the work done by @xcite ) have shown h ii region properties based on radio continuum emission to have relatively limited model dependence . @xcite compared three models to describe the radio continuum structure of 11 galactic h ii regions . physical parameters were derived from the models using different geometries ( homogeneous sphere and a spherical gaussian ) as well as by using peak properties in the continuum intensity maps . comparison between the three models showed no systematic differences in the derived electron density and emission measure , and resulted in an overall scatter of 15@xmath17 between modeled parameters . in addition , the ionization properties which are of greatest interest in this analysis are a function of frequency , electron temperature , distance , and flux density , all of which are model independent . the main challenge lies in accurately measuring the flux density . @xcite discovered there could be differences in measured flux density of up to @xmath650@xmath17 for extended sources , which would result in about a 50@xmath17 difference in derived ionizing flux . however , the motivation of this work is an order of magnitude comparison of ionizing photon fluxes between the sources in this study and known massive clusters in the galaxy . therefore , the results of this analysis are adequate for the primary goal of this study . it should be emphasized that the physical quantities reported in table [ tbl-4 ] derived from the radio continuum parameters are meant to be used for relative comparison . the last column in table [ tbl-4 ] gives an estimate of the spectral type for each cluster based on our calculated lyman continuum photon fluxes and model stellar atmospheres derived by @xcite . no correction for extinction was made for these estimates . the stellar spectral types given are the types that would be assigned to a single ionizing source which produces the observed lyman continuum flux . it is recognized that the ionizing photon rate calculated for the stellar clusters most likely arises from more than one cluster member , and not a single ionizing source of the estimated spectral type . however , the estimate allows for a comparison between the results of this analysis and those determined in previous studies . in general , there was close agreement between the estimated spectral types indicating these results are consistent with past observations ( e.g. bd95 ; @xcite ) . the physical parameters derived in table [ tbl-4 ] show variation in physical characteristics of the h ii regions associated with the cluster candidate sources . the equivalent radii for the h ii regions range between @xmath61 and 9 pc , with the majority of sources having linear sizes of @xmath61 - 3 pc . the electron densities range from 10 @xmath18 up to 140 @xmath18 with an average of @xmath648 @xmath18 . the em differs for all sources ranging between @xmath610@xmath0 pc @xmath19 and 6@xmath710@xmath1 pc @xmath19 . the total mass of ionzed gas associated with the h ii regions varies widely from 7 to 2800 m@xmath10 . the cluster source bd52 has the smallest ionized mass of 7.4 m@xmath10 . bd84 , bd95 , db7 , and dbcl23 have intermediate masses averaging around @xmath650 m@xmath10 , whereas bd65 contains a remarkable 2800 m@xmath10 of ionized gas . the lyman continuum photon fluxes spread over two orders of magnitude from @xmath610@xmath20 s@xmath12 to 10@xmath21 s@xmath12 with an average of @xmath66@xmath710@xmath22 s@xmath12 . this spread in ionizing photon rates corresponds to a spread in stellar spectral type from early b - type to mid o - type . the majority of the candidate sources have lyman continuum photon fluxes that would be produced by single b0 type star . in this candidate cluster comparison , bd65 stands out as having greater values for most derived parameters . it clearly has the largest linear radius of 9.1 pc , much greater than the other clusters average of @xmath62 pc . similarly , bd65 has an h ii mass that is nearly 70 times greater than the average of the other clusters ionized mass . bd65 has the greatest lyman continuum photon flux of 3.5@xmath710@xmath21 s@xmath12 , corresponding to an ionizing source of spectral type o3v . the physical paramters of the h ii regions associated with these clusters depend on the distance . therefore , the discrepancy in distance estimates to bd65 implies the physical properties derived would be less extreme if a smaller distance was adopted . for example , at a distance of 4.9 kpc ( @xcite ) bd65 has a linear radius of 4.5 pc , an ionized mass of 490 m@xmath10 , and a lyman continuum photon flux of 8.8@xmath710@xmath22 s@xmath12 . even at this smaller distance bd65 has the largest linear size , ionized mass , and ionizing photon flux of the candidate sources . although small in size and mass as compared to bd65 , dbcl23 has the largest electron density and emission measure . dbcl23 also has the second highest ionizing photon flux ( 2.7@xmath710@xmath22 s@xmath12 ) , which is about a magnitude less than that of bd65 . @xcite performed an analysis on infrared dust bubbles using radio and mid - infrared data for bubbles in the galactic plane which led to a generalized picture of infrared bubbles as ionized gas with hot dust surrounded by a pdr containing interstellar gas , pahs , and dust . the inner ionized gas is traced by radio free - free emission , and the thermal dust continuum can be traced with infrared emission , which often lies outside of the region of radio emission . the surrounding pdr is dominated by pah emission features which can be seen in irac bands at 3.6 , 5.8 , and 8 @xmath9 m . here , we carry out a similar analysis using our radio and infrared data for several of the sources that appear to be most symmetric : bd52 , dbcl23 , and bd84 . figure [ fig9 ] shows a slice through the bd52 shell at 4.9 ghz , 8.5 ghz ( convolved to match the 4.9 ghz ) and 8 @xmath9 m . the location of this slice is shown by the blue line in figure [ fig1 ] . the slight difference in flux at 8.5 ghz compared to 4.9 ghz is most likely due to missing flux , as previously discussed in 4.1.1 . figure [ fig9 ] shows that the radio emission peaks in two places near 200 and 300 along the slice . the 8 @xmath9 m emission exhibits the same double peaked structure , but peaks at @xmath6100 and 315 along the slice . this plot shows that the majority of the radio emission is contained within the 8 @xmath9 m emission indicating that the ionized gas lies primarily interior to the infrared shell ( thermal dust ) . figure [ fig10 ] shows the same slice across bd52 , but for the 3.6 , 4.5 , and 5.8 @xmath9 m data . the 3.6 and 4.5 @xmath9 m bands are more sensitive to stellar emission , and show only faint continuum emission compared to the 5.8 @xmath9 m emission . the 5.8 @xmath9 m emission closely follows the structure of the 8 @xmath9 m emission in figure [ fig9 ] . the lower intensity emission between the two emission peaks at 5.8 and 8 @xmath9 m can be attributed to hot dust . the difference between this emission and the emission peaks at 5.8 and 8 @xmath9 m is likely due to the addition of pah features in the peaks , which delineate the edge of the pdr around the cluster . figure [ fig11 ] shows a slice through the shell of dbcl23 at 4.9 ghz , 8.5 ghz ( convolved to 4.9 ghz ) and 8 @xmath9 m . the location of this slice is shown in figure [ fig7 ] . a pair of peaks in the radio emission occur at @xmath690 and 120 along the slice . the 8 @xmath9 m emission also shows double peaks that occur at @xmath665 and 130 along the slice . the majority of the radio emission is contained within the 8 @xmath9 m emission as it is in bd52 indicating that the ionized gas lies interior to the thermal dust . figure [ fig12 ] shows the same slice across dbcl23 , but at 3.6 , 4.5 , and 5.8 @xmath9 m . the 3.6 and 4.5 @xmath9 m emission roughly follows the 5.8 @xmath9 m emission , although the source is significantly fainter at these wavelengths . as in bd52 , the 5.8 @xmath9 m closely resembles the 8 @xmath9 m emission . the dip in intensity between the emission peaks at these wavelengths is likely due to hot dust with the omission of pahs . the pahs are traced by the 5.8 and 8 @xmath9 m emission in the two emission peaks . the spike at 120 seen in all three bands in figure [ fig12 ] represents stellar emission . figure [ fig13 ] shows a slice across the source bd84 at 4.9 ghz , 8.5 ghz ( convolved to 4.9 ghz ) and 8 @xmath9 m . the location of this slice is shown in figure [ fig4 ] . the radio emission for this source has a single peak near @xmath6115 - 120 along the slice , and a fwhm of @xmath630 . the 8 @xmath9 m emission has peaks on either side of the radio emission , @xmath690 and 130 . a possible interpretation is that the h ii emission indicated by the peak is surrounded by a shell of pahs and warm dust indicated by the two infrared emission peaks . as is evident in figure [ fig4 ] , the radio continuum emission is more extended than the infrared emission in bd84 . this is only slightly apparent in the slices in figure [ fig13 ] where the 4.9 ghz emission is visible above the 8 @xmath9 m emission just outside the two peaks . the different morphology of bd84 implies there is more ionized gas in this cluster system , and it is not all contained within the pdr . figure [ fig14 ] shows the same slice across bd84 at 3.6 , 4.5 , and 5.8 @xmath9 m . it can be seen in figure [ fig14 ] that the 5.8 @xmath9 m band follows the 8 @xmath9 m emission very closely . the 3.6 and 4.5 @xmath9 m bands exhibit the same general structure , although the second peak at @xmath6130 is much smaller , relatively . the double peaked radio emission for the cluster candidates bd52 and dbcl23 implies a shell - like morphology for these h ii regions . bd84 differs in this respect with a single peak of radio emission near the cluster surrounded by more diffuse extended emission . lower intensity emission at 5.8 and 8 @xmath9 m between two emission peaks for all three sources indicates pahs are probably being destroyed by stellar radiation near the clusters . this also provides more evidence that the infrared emission peaks are due to pah features , and represent the edges of the pdr . our simplistic analysis for these three sources is consistent with the infrared dust bubble model of @xcite . photometry from _ spitzer_/irac can be used to identify objects with infrared excess , including evolutionary classes of young stellar objects ( ysos ; @xcite ) . ysos serve as signposts for sites of ongoing star formation , and can be used to probe age differences in regions of extended star formation . in the simplest case , stars with infrared excess can be identified as candidate ysos ( cysos ) , as the infrared excess is often indicative of a disk structure . objects which are bright and extended in the 4.5 @xmath23 band , called extended green objects ( egos ) , have a high association rate with outflows from massive young stellar objects ( mysos ; @xcite , @xcite ) . sites of massive star formation regions , e.g. in and around massive stellar clusters , can be selected by identifying a population of egos . using simple assumptions , infrared photometry from _ spitzer_/irac can be used to construct color - color plots for sources in each field . figures [ fig : bd52midir]-[fig : dbcl23midir ] show these plots for all six cluster candidates . in each plot , the left panel shows the @xmath24 color - color plot . the right panel shows the distribution of different sources on the 8 @xmath9 m image of each cluster candidate . depending on the location of the sources on the color - color plot , they can be divided into three groups designed to probe different classes of objects : group i , ii and iii . following the procedure outlined in @xcite we construct these groups as follows . stars with little infrared color difference , e.g. main sequence stars , congregate around @xmath25 0,0 ( i.e. , no excess color ) . such objects are denoted as group i in this work . a reddening vector corresponding to a@xmath26 10 is overplotted on each color - color plot . fore- and background stars affected by differential reddening will follow the reddening vector , and are henceforth denoted as group iii . as stated earlier , excess emission at 5.8 @xmath23 can be indicative of strong pah emission from circumstellar emission , e.g. a disk around a yso , and are here referred to as group ii objects . in color - color space , these objects occupy the space defined by @xmath27 @xmath28 0.9 mag and -0.25 mag @xmath29 0.25 mag . stars with infrared colors in between groups ii and iii ( see , for example , the left panel of figure [ fig : bd65midir ] ) are also likely cysos . active galactic nuclei can contaminate infrared color - color diagrams , though the contamination rate in this sample is likely low given the low latitude of these stellar clusters . here we discuss the infrared excesses seen toward each of the target sources . only objects with valid photometry in all four bands are considered . examination of the distribution of group ii sources indicates that all six clusters are sites of ongoing star formation . younger , more deeply embedded protostars are undetectable at these wavelengths . the irac color - color diagram of bd52 is shown in the left panel of figure [ fig : bd52midir ] . the spatial distribution of group ii and group iii stars is displayed in the right panel . group iii stars are fairly evenly distributed across the face of the cluster , though a dearth of these objects are seen in the direction of bright 8.0 @xmath23 emission peaks . as these sources are distributed along the reddening vector , it is likely that their colors are affected by varying extinction . the 8.0 @xmath23 emission indicates regions of high local extinction , making it difficult to observe background stars in these areas . only one object appears in the group ii range of colors . this object ( shown as a diamond in the right panel of figure [ fig : bd52midir ] ) is located at the edge of the shell - like 8 @xmath9 m emission . figure [ fig : bd65midir ] displays the irac color - color diagram of bd65 , left , along with the spatial distribution of objects with infrared excess , right . group ii objects , indicated by a diamond in the right panel of figure [ fig : bd65midir ] appear to be distributed preferentially along ridges of the bright , extended 8.0 @xmath9 m emission surrounding bd65 . the few group iii objects in this region are associated with point sources in the 8.0 @xmath9 m image , suggesting that the majority are foreground stars distributed along the line of sight . stars farther from the reddening vector , falling mostly into group ii , are indistinguishable from cysos . the left panel of figure [ fig : bd84midir ] shows the color - color diagram for the sources detected , while the right panel shows the distribution of objects overlaid on the 8 @xmath9 m image . the number of infrared sources near bd84 is much lower than for the other target regions . there are only @xmath635 sources with valid photometry in all four bands ; @xmath6100 with valid photometry in 3 bands . the majority of sources in the field of view are likely to be fore- and background stars . group iii objects follow the reddening vector closely ; about half of these sources are located preferentially along a ridge of infrared emission to the west of bd84 and at further distances from the bright infrared emission . group ii sources are more tightly clustered around bd84 and the surrounding bright infrared emission . it is likely that many sources associated with the cluster are undetected . the left panel of figure [ fig : bd95midir ] shows the color - color diagram for the sources near bd95 and the right panel shows the distribution of the group ii and iii sources overlaid on the 8 @xmath9 m emission . both group ii and group iii stars are preferentially located close to the stellar cluster bd95 and along ridges of infrared emission . the majority of these objects are likely cysos , highlighting sites of ongoing star formation in and around bd95 . group ii sources in db7 are located along filaments and knots of bright 8.0 @xmath23 emission , which can be seen in the right panel of figure [ fig : db7midir ] . these stars are also located to the se , in the direction of the stellar cluster db8 . the majority of group iii sources are located fairly far from the reddening vector , and seem to be concentrated in and around the two clusters , making it likely these are cysos . inspection of the right panel of figure [ fig : dbcl23midir ] shows that group ii sources are located preferentially towards the bright , diffuse 8.0 @xmath23 emission surrounding dbcl23 . the group iii objects located outside of dbcl23 are likely reddened background sources , while those within the cluster are associated with bright point sources , making these sources likely cluster members . inspection of the spatial distribution of objects with infrared excess in each of the six clusters presented in this study yield similar results . each cluster contains candidate ysos in regions of extended 8.0 @xmath23 emission . it is possible that the distribution of infrared excess sources results from star formation triggered by the expansion of bubble - like h ii regions , but it can also be argued that already formed protostellar objects are being exposed by the evolution of the central systems . a more robust analysis of the yso population of each cluster would be possible at wavelengths between those investigated in this study , e.g. herschel and/or alma observations . these observations would also prove useful as probes of triggered star formation ( see @xcite for one such study ) . maser activity is often associated with shocked emission such as outflows from young stellar objects ( e.g. @xcite ) ; in the case of bd52 the possible ego is cut off by the edge of the irac frame , so a firm association can not be made . the maser - yso association is also pointed out by @xcite who focus on the association between egos and masers . the _ spitzer_/irac observations of each cluster were inspected by eye for egos , yielding no obvious sources . a lack of detection of mysos ( bright , extended sources in the 4.5 @xmath9 m discussed above ) is consistent with the general picture for all six clusters : young , partially embedded , intermediate mass stellar clusters . the lack of myso detection also implies the candidate clusters are not sites of ongoing massive star formation . for dbcl23 , this result is consistent with @xcite s confirmation that it is not a massive star forming site ( 2.6 ) . table [ tbl-5 ] lists physical properties of 14 young , massive clusters in the galaxy that have been well - studied with which the properties of the candidate clusters can be compared . the properties most relevant to this study are the h ii mass and the lyman continuum photon flux , because they quantify the impact of these stellar clusters on the surrounding ism . it is apparent that in general the average mass of ionized gas implied by the radio observations for the candidate sources is below the well - known massive clusters . with the exception of bd65 , all other cluster candidate sources have ionized masses of 7 - 70 m@xmath10 . bd84 and dbcl23 are at the upper end of this range with ionized masses of 70 and 56 m@xmath10 , which are comparable to the ionized mass associated with westerlund 1 ( @xcite ) . bd65 stands out as having a very large ionized mass of 2800 m@xmath10 . this mass is comparable to that of the w49a and quintuplet h ii regions ( @xcite , @xcite ) . the cluster candidate sources have @xmath62 - 3 orders of magnitude smaller lyman continuum photon fluxes when compared to these other well - studied massive galactic clusters , which have a typical value @xmath4 10@xmath30 s@xmath12 . our largest n@xmath16 value is 3.5@xmath710@xmath21 s@xmath12 for bd65 , and even this value is almost an order of magnitude less than the next closest n@xmath16 value of @xmath610@xmath30 s@xmath12 for ngc 3603 ( @xcite ) . the candidate cluster dbcl23 is also on the higher end of our n@xmath16 range with a value of 2.7@xmath710@xmath22 s@xmath12 . this is comparable to dbs2003 45 which had a lyman continuum flux of 8.2@xmath710@xmath22 s@xmath12 ( @xcite ) . the impact on the ism from massive clusters such as arches and quintuplet results in sizable structures 2 pc and 10 pc in size ( @xcite,@xcite ) . the size of the h ii regions around the cluster sources bd65 and bd84 fall into this range . however , the linear radii of the h ii regions around the rest of the candidates are more analogous to the sizes of the actual stellar clusters for some of the comparison clusters ( e.g. rsgc 1 & 2 ; @xcite ) . past studies have predicted the galaxy to harbor about 100 massive ( @xmath610@xmath31 m@xmath10 ) stellar clusters similar to the comparison clusters ( @xcite ) . however , using the evidence stated above , we conclude that the candidate clusters do not represent the most massive stellar clusters in the galaxy , but are rather intermediate mass clusters . @xcite extrapolated the number of known clusters with a mass larger than 10@xmath31 m@xmath10 to conclude that there are about 200 clusters with masses @xmath28 10@xmath32 m@xmath10 . the stellar clusters presented in this study may belong to this group of intermediate mass clusters in the galaxy . observations have been carried out using _ spitzer_/irac at 3.6 , 4.5 , 5.8 and 8.0 @xmath9 m and vla radio observations at 8.5 and 4.9 ghz to determine the properties of six candidate stellar clusters and their surrounding h ii regions . the following conclusions have been drawn from this study : 1 . image overlays of all the sources were created to compare the distribution of infrared and radio emission . the 3.6 and 8 @xmath9 m emission was found to be more diffuse than the radio emission in the majority of the cluster sources . in addition , the peak intensity in the radio emission is often offset from the peak intensity in the infrared emission . radio continuum properties of the h ii regions led to an estimate of the sources ionized mass and lyman continuum photon flux . bd65 is the most massive of the candidate clusters having an ionized mass of @xmath62800 m@xmath2 and n@xmath16 @xmath610@xmath21 . dbcl23 was also found to have a fairly large ionizing flux of @xmath610@xmath22 . the physical parameters of the sources from the radio continuum were compared to h ii parameters of other well - studied young , massive galactic clusters . a more complete multiwavelength analysis of the morphology in bd52 , dbcl23 , and bd84 was carried out . the emission mechanisms in the clusters and surrounding ism could be explored by investigating the intensity of emission at different wavelengths across a slice through the sources . a general picture emerged from this analysis : ionized gas surrounded by bright warm dust and pah emission . this is consistent with other studies of infrared dust bubbles ( @xcite ) . 4 . infrared photometry of stellar members of the candidate clusters was used to search for candidate ysos . the existence of ysos provides evidence of ongoing star formation . candidate ysos were found in all the cluster candidates using color - color plots . a lack of detection of mysos is consistent with the general picture for all six clusters : young , partially embedded , intermediate mass stellar clusters . comparison with other young , massive clusters ( e.g. arches , quintuplet , westerlund 1 ) leads to the conclusion that the stellar cluster candidates presented in this study are not massive ( @xmath28 10@xmath1 m@xmath2 ) clusters . more likely , they represent examples of galactic intermediate mass ( @xmath28 10@xmath0 m@xmath2 ) stellar clusters . the authors would like to thank maria messineo , qingfeng zhu , ben davies and thomas zimmermann who were involved in the initial stages of this project . ccl would like to acknowledge the university of iowa math and physical sciences funding program for supporting students emily richards and thomas zimmermann . this material is based upon work supported by the national science foundation under grant no . 0907934 . allen , l. e. , et al . 2004 , , 154 , 363 anglada , g. , estalella , r. , pastor , j. , rodriguez , l. f. , & haschick , a. d. 1996 , , 463 , 205 ascenso , j. , alves , j. , beletsky , y. , & lago , m. t. v. t. 2007 , , 466 , 137 balser , d. s. , bania , t. m. , rood , r. t. , & wilson , t. l. 1995 , , 100 , 371 bibby , j. l. , crowther , p. a. , furness , j. p. , & clark , j. s. 2008 , , 386 , l23 bica , e. , dutra , c. m. , soares , j. , & barbuy , b. 2003 , , 404 , 223 borissova , j. , ivanov , v. d. , hanson , m. m. , georgiev , l. , minniti , d. , kurtev , r. , & geisler , d. 2008 , , 488 , 151 brandner , w. , clark , j. s. , stolte , a. , waters , r. , negueruela , i. , & goodwin , s. p. 2008 , , 478 , 137 caplan , j. , deharveng , l. , pea , m. , costero , r. , & blondel , c. 2000 , , 311 , 317 comern , f. , pasquali , a. , & torra , j. 2005 , , 440 , 163 condon , j. j. , cotton , w. d. , greisen , e. w. , yin , q. f. , perley , r. a. , taylor , g. b. , & broderick , j. j. 1998 , , 115 , 1693 cotera , a. s. , erickson , e. f. , colgan , s. w. j. , simpson , j. p. , allen , d. a. , & burton , m. g. 1996 , , 461 , 750 cyganowski , c. j. , whitney , b. a. , holden , e. , et al . 2008 , , 136 , 2391 cyganowski , c. j. , brogan , c. l. , hunter , t. r. , & churchwell , e. 2009 , , 702 , 1615 cyganowski , c. j. , brogan , c. l. , hunter , t. r. , churchwell , e. , & zhang , q. 2011 , , 729 , 124 davies , b. , figer , d. f. , law , c. j. , kudritzki , r .- p . , najarro , f. , herrero , a. , & mackenty , j. w. 2008 , , 676 , 1016 de pree , c. g.,mehringer , d. m. , & goss , w. m. 1997 , , 482 , 307 de pree , c. g. , nysewander , m. c. , & goss , w. m. 1999 , , 117 , 2902 dutra , c. m. , & bica , e. 2001 , , 376 , 434 dutra , c. m. , bica , e. , soares , j. , & barbuy , b. 2003 , , 400 , 533 figer , d. f. , mclean , i. s. , & morris , m. 1999 , , 514 , 202 figer , d. f. , kim , s. s. , morris , m. , serabyn , e. , rich , r. m. , & mclean , i. s. 1999 , , 525 , 750 figer , d. f. , et al . 2002 , , 581 , 258 figer , d. f. , najarro , f. , geballe , t. r. , blum , r. d. , & kudritzki , r. p. 2005 , , 622 , l49 hanson , m. m. , & popescu , b. 2008 , iau symposium , 250 , 307 homeier , n. l. , & alves , j. 2005 , , 430 , 481 joncas , g. , durand , d. , & roger , r. s. 1992 , , 387 , 591 klein , r. , posselt , b. , schreyer , k. , forbrich , j. , & henning , t. 2005 , , 161 , 361 kothes , r. , & dougherty , s. m. 2007 , , 468 , 993 kurtev , r. , borissova , j. , georgiev , l. , ortolani , s. , & ivanov , v. d. 2007 , , 475 , 209 lang , c. c. , goss , w. m. , & wood , d. o. s. 1997 , , 474 , 275 lang , c. c. , goss , w. m. , & morris , m. 2001 , , 121 , 2681 lang , c. c. , goss , w. m. , & rodrguez , l. f. 2001 , , 551 , l143 lang , c. c. , goss , w. m. , & morris , m. 2002 , , 124 , 2677 melena , n. w. , massey , p. , morrell , n. i. , & zangari , a. m. 2008 , , 135 , 878 mercer , e. p. , et al . 2005 , , 635 , 560 messineo , m. , petr - gotzens , m. g. , schuller , f. , menten , k. m. , habing , h. j. , kissler - patig , m. , modigliani , a. , & reunanen , j. 2007 , , 472 , 471 messineo , m. , davies , b. , ivanov , v. d. , figer , d. f. , schuller , f. , habing , h. j. , menten , k. m. , & petr - gotzens , m. g. 2009 , , 697 , 701 mezger , p. g. , & henderson , a. p. 1967 , , 147 , 471 panagia , n. , & felli , m. 1975 , , 39 , 1 rauw , g. , manfroid , j. , gosset , e. , naz , y. , sana , h. , de becker , m. , foellmi , c. , & moffat , a. f. j. 2007 , , 463 , 981 russeil , d. , adami , c. , & georgelin , y. m. 2007 , , 470 , 161 samal , m. r. , pandey , a. k. , ojha , d. k. , ghosh , s. k. , kulkarni , v. k. , & bhatt , b. c. 2007 , , 671 , 555 sharpless , s. 1959 , , 4 , 257 smith , l. j. , norris , r. p. f. , & crowther , p. a. 2002 , , 337 , 1309 snell , r. l. , & bally , j. 1986 , , 303 , 683 sternberg , a. , hoffmann , t. l. , & pauldrach , a. w. a. 2003 , , 599 , 1333 watson , c. , et al . 2008 , , 681 , 1341 yun , j. l. , djurvik , a. a. , delgado , a. j. & alfaro , e. j. 2008 , , 483 , 209 zavagno , a. , et al . 2010 , , 518 , l101 zhu , q. , davies , b. , figer , d. f. , & trombley , c. 2009 , , 702 , 929 zinchenko , i. , pirogov , l. , & toriseva , m. 1998 , , 133 , 337 lccc bd52 & 01 23 06.0 , 61 51 32.0 & [email protected] & sh2 - 187 , lbn 126.70 - 00.80 , iras 01195 + 6136 + bd65 & 04 11 09.4 , 51 10 01.0 & 9.8 & sh2 - 209 , g151.59 - 0.23 , iras 04073 + 5102 + bd84 & 06 38 27.8 , 00 44 39.0 & [email protected] & sh2 - 283 + bd95 & 07 16 33.0 , -09 25 26.0 & [email protected] & sh2 - 294 , lbn 1032 + db7 & 07 35 33.7 , -18 45 39.0 & [email protected] & sh2 - 307 , eso560 - 1 , sfo49 ( db8 ) + dbcl23 & 20 45 37.4 , 44 15 14.0 & @xmath333.5 & g84.0 + 0.8 , ecx6 - 38 , 3c 423 + lccc bd52 & 4.9 & 17.93 x 11.61 & 0.313 + bd52 & 8.5 & 8.77 x 6.86 & 0.0365 + bd65 & 4.9 & 16.60 x 11.41 & 0.552 + bd65 & 8.5 & 10.75 x 6.88 & 0.156 + bd84 & 4.9 & 29.15 x 11.78 & 0.0675 + bd84 & 8.5 & 14.68 x 7.37 & 0.0425 + bd95 & 4.9 & 19.24 x 12.76 & 0.281 + bd95 & 8.5 & 12.14 x 6.97 & 0.0591 + db7 & 4.9 & 24.27 x 11.77 & 0.194 + db7 & 8.5 & 15.66 x 6.93 & 0.106 + dbcl23 & 4.9 & 24.78 x 12.30 & 0.295 + dbcl23 & 8.5 & 15.70 x 6.98 & 0.134 + lcccccc bd52 & 503@xmath525 & 250 x 250 & [email protected] + bd65 & 4100@xmath5200 & 350 x 200 & 9.8 + bd84 & 42@xmath53 & 175 x 120 & [email protected] + bd95 & 279@xmath520 & 220 x 220 & [email protected] + db7 & 511@xmath530 & 200 x 175 & [email protected] + dbcl23 & 2440@xmath5100 & 120 x 140 & @xmath333.5 + lcccccc bd52 & 1.3 & 37 & 2.7@xmath710@xmath0 & 7 & 9.3@xmath710@xmath34 & b0.5v + bd65 & 9.1 & 37 & 2.6@xmath710@xmath31 & 2800 & 3.5@xmath710@xmath21 & o3v + bd84 & 4.1 & 10 & 8.7@xmath710@xmath35 & 70 & 2.4@xmath710@xmath20 & b0v + bd95 & 2.5 & 22 & 2.5@xmath710@xmath32 & 35 & 2.6@xmath710@xmath20 & b0v + db7 & 1.7 & 43 & 6.4@xmath710@xmath32 & 22 & 3.2@xmath710@xmath20 & b0v + dbcl23 & 1.6 & 140 & 6.4@xmath710@xmath31 & 56 & 2.7@xmath710@xmath22 & o8v + lccccccccc westerlund 1 & 3.55 & 339.55 & -0.4 & 3 - 5 & 1.0 & 45 & 47 - 53 & ... & 1,2 + westerlund 2 & 2.8 & 284.27 & -0.34 & 1 - 3 & ... & 7 & ... & 6@xmath710@xmath30 & 3,4 + rcw49 & 8 & 284.31 & -0.33 & ... & ... & ... & ... & 11.5@xmath710@xmath30 & 3 + rsgc1 & 6.6 & 25.27 & -0.16 & 10 - 14 & 1.5 & 30 & ... & ... & 5 + rsgc2 & 5.8 & 26.19 & -0.07 & 14 - 20 & 3.2 & 40 & ... & ... & 5 + w49a & 11.4 & 43.17 & 0.0018 & 0.3 - 2 & ... & 50 - 70 & @xmath36 238 & 2.4@xmath710@xmath30 & 6,7 + central & 8 & 0.00 & 0.01 & 3 - 7 & 0.23 & 20 & ... & 3.2@xmath710@xmath30 & 8,9 + arches & 8 & 0.12 & 0.02 & 2 - 3 & 0.23 & @xmath33 70 & 5480 & 4@xmath710@xmath37 & 10,11 + quintuplet & 8 & 0.16 & -0.06 & 3 - 5 & 1.0 & 20 & 240 & 7.9@xmath710@xmath30 & 9,12 + ngc 3603 & 7.6 & 291.63 & -0.53 & 2.5 & 0.7 - 1.5 & 10 - 16 & 10000 & 1@xmath710@xmath30 & 13,14 + cl 1806 - 20 & 8.7 & 10.00 & -0.24 & 3 - 4.5 & ... & 3 & ... & ... & 15,16 + glimpse 30 & 7.2 & 298.76 & -0.41 & 4 - 5 & 1.36 & 3 & ... & ... & 8 + dbs2003 179 & 7.9 & 347.58 & 0.19 & 2 - 5 & 0.2 & 7 & ... & ... & 17 + dbs2003 45 & 4.5 & 283.88 & -0.91 & 5 - 8 & ... & 1 & ... & 8.2@xmath710@xmath22 & 18 +

despite their fundamental and practical interest , the physical properties of hard ellipses remain largely unknown . in this paper , we present an event - driven molecular dynamics study for hard ellipses and assess the effects of aspect ratio and area fraction on their physical properties . for state points in the plane of aspect ratio ( ) and area fraction ( ) , we identify three different phases , including isotropic , plastic and nematic states . we analyze in detail the thermodynamic , structural and self - diffusive properties in the formed various phases of hard ellipses . the equation of state ( eos ) is shown for a wide range of aspect ratios and is compared with the scaled particle theory ( spt ) for the isotropic states . we find that spt provides a good description of the eos for the isotropic phase of hard ellipses . at large fixed , the reduced pressure increases with in both the isotropic and the plastic phases and , interestingly , its dependence on is rather weak in the nematic phase . we rationalize the thermodynamics of hard ellipses in terms of particle motions . the static structures of hard ellipses are then investigated both positionally and orientationally in the different phases . the plastic crystal is shown to form for aspect ratios up to , while appearance of the stable nematic phase starts approximately at . we quantitatively determine the locations of the isotropic - plastic ( i - p ) transition and the isotropic - nematic ( i - n ) transition by analyzing the bond - orientation correlations and the angular correlations , respectively . as expected , the i - p transition point is found to increase with , while a larger leads to a smaller area fraction where the i - n transition takes place . moreover , our simulations strongly support that the two - dimensional nematic phase in hard ellipses has only quasi - long - range orientational order . the self - diffusion of hard ellipses is further explored and connections are revealed between the structure and the self - diffusion . we discuss the relevance of our results to the glass transition in hard ellipses . finally , the results of the isodiffusivity lines are evaluated for hard ellipses and we discuss the effect of spatial dimension on the diffusive dynamics of hard ellipsoidal particles .

1212.6497

hard - particle systems in general offer a great opportunity to understand the mechanism of a variety of physical phenomena , including crystal nucleation @xcite , glass transition @xcite and jamming @xcite . the spherical particles , such as hard spheres and hard disks , have been studied intensively during the past decades . in particular , for a system of hard spheres , an entropy - driven liquid - crystal transition occurs at sufficiently high density @xcite and a hard - sphere glass can also be formed above the glass transition point if crystallization is avoided @xcite . on the other hand , non - spherical particles play a fundamental role in the physics of molecular liquids and attract growing interest in recent years because of their usefulness in forming functional structures . from the theoretical point of view , it is interesting to include the rotational degrees of freedom for the particles and investigate the physical behavior of a system composed of anisotropic particles . such study uncovers new types of behavior and holds potential for providing new insights that can not be obtained from studies of spherical objects @xcite . from the practical point of view , the constituent particles often have non - spherical shapes in real materials such as laponite @xcite , and the building blocks with anisotropic shape can provide a powerful candidate for the assembly of particular targeted structures @xcite . thus , exploration of systems composed of anisotropic particles is also valuable for materials design . in this paper , we present an event - driven molecular dynamics ( edmd ) study of hard ellipses . a system of hard ellipses can be regraded as one of the simplest models of two - dimensional ( 2d ) anisotropic particles and its phase behavior is very rich . yet , the thermodynamic , structural and dynamical properties of hard ellipses remain largely unknown . the phase transitions , in particular , the isotropic - nematic ( i - n ) transition , of hard ellipses have been studied by computer simulations and various theoretical approaches . in vieillard - baron s pioneering monte carlo ( mc ) simulations in @xmath7 ensemble ( i.e. , the particle number @xmath8 , the volume @xmath9 and the termperature @xmath10 are constant ) @xcite , three different phases , including isotropic liquid , nematic liquid crystal and solid , have been identified in a hard - ellipse system with an aspect ratio of @xmath11 , where @xmath12 and @xmath13 denote semi - major and semi - minor axes . vieillard - baron observed that both i - n and nematic - solid ( n - s ) transitions are of first order , as evidenced by a small discontinuity shown in the equation of state ( eos , i.e. , density - dependence of the reduced pressure @xmath3 ) . however , computer simulations show that the i - n transition in a 2d fluid of hard needles @xcite and long hard rods @xcite , which in principle share the similar physics with elongated hard ellipses , is continuous rather than first - order . moreover , the @xmath14 ( i.e. , the particle number @xmath8 , the pressure @xmath15 and the termperature @xmath10 are constant ) mc simulations of cuesta and frenkel @xcite indicate that while there is no stable nematic phase in hard ellipses for @xmath16 , the i - n transition is of first order for @xmath17 and continuous via disclination unbinding for @xmath18 . hence , it appears that the nature of the i - n transition depends on the aspect ratio and that the phase diagram possesses a tricritical point at which the transition changes from first - order to continuous . thus , on the simulation side , a definite conclusion has not been reached on the phase transitions of hard ellipses . this issue is also contentious on the theoretical side . ward and lado @xcite found based on the percus - yevick ( py ) integral equation theory that the i - n transition does not take place in a fluid of hard ellipses although orientational ordering is allowed . while , the density - functional theory ( dft ) @xcite and the effective - liquid approach @xcite indicate that the i - n transition is continuous for hard ellipses with arbitrary aspect ratios . the other forms of dft @xcite , the onsager theory - based approach @xcite , and the scaled particle theory ( spt ) @xcite predict the existence of the i - n transition in hard ellipses , but its location differs from these theoretical predictions . therefore , a clear understanding of the phase transitions in hard ellipses is still lacking . here , it should also be mentioned that the phase behavior of binary hard ellipses was recently investigated by a scaled particle dft @xcite and a weighted dft @xcite . another important issue concerns the orientational order in the 2d nematic liquid crystals . straley has shown @xcite that true long - range order ( lro ) can not exist in a 2d nematic phase if the particles interact via a separable potential , but the existence of true lro can not be excluded if the potential is not separable into positional and orientational parts . in fact , it has been demonstrated by computer simulations @xcite that a general property of a 2d nematic phase is the lack of true lro , although earlier simulations @xcite also show that the system with the nonseparable potential exhibits true lro . for this reason , the systems that lack true lro are usually referred to as having quasi - lro . in the elastic continuum theory @xcite , quasi - lro is expected to appear in a 2d nematic phase if the free energy associated with collective fluctuations in the particle orientations can be expressed as @xmath19 where @xmath20 characterizes the orientation at position @xmath21 with respect to a fixed axis and @xmath22 is the 2d frank s elastic constant @xcite . based on eq . ( 1 ) , one can derive that the amplitude of the orientational fluctuations diverges with the system size in a logarithmic manner , @xmath23 where @xmath24 denotes the ensemble average , @xmath25 is boltzmann s constant , @xmath10 is the absolute temperature and @xmath8 is the particle number . as a consequence , both the 2d nematic order parameter @xmath26 and the angular correlation function @xmath27 decay algebraically , i.e. , @xmath28 @xmath29)>\sim r^{-2l^{2}k_{b}t/\pi k},\ ] ] where @xmath30 is a positive integer . equations ( 3 ) and ( 4 ) imply that a 2d nematic phase with quasi - lro can be characterized by a vanishing orientational order parameter in the thermodynamic limit and a power - law decay of the angular correlation function . if the transition from a 2d nematic phase with quasi - lro to an isotropic phase proceeds via the kosterlitz - thouless ( kt ) disclination unbinding mechanism @xcite , it should occur at a universal value of the renormalized frank s constant , @xmath31 @xcite . as pointed out in ref . @xcite , a different mechanism is also possible , but there is no stable nematic phase at values of the frank s constant below @xmath32 . thus , one can quantitatively identify the location where the stable nematic phase starts to form by monitoring the system size dependence of the nematic order parameter or from the spatial correlation of the particle orientations . we note that the algebraic decay of the angular correlation function has been evidenced by simulations of 2d hard needles @xcite , 2d hard rods @xcite , hard ellipses @xcite , and recently by experiments of quasi-2d suspensions of hard ellipsoids @xcite . in contrast to the efforts made on understanding the phase behavior and the structure of hard ellipses , little attention has been paid on their dynamics despite its fundamental and technological importance . the dynamical properties of hard needles have been studied by molecular dynamics ( md ) simulations about three decades ago @xcite . the diffusive process of a single hard ellipsoid confined in two dimensions was measured just several years ago in experiments @xcite . in particular , the diffusion of quasi-2d suspensions of hard ellipsoids with @xmath33 has been explored very recently by video - microscopy experiments @xcite and the enhancement of the translational diffusion with respect to the rotational diffusion has been rationalized in terms of the formation of unstable nematic - like regions with an average lifetime that exceeds the characteristic time of diffusion in a recent mc study @xcite . in spite of the aforementioned progress , our knowledge of the dynamics of hard ellipses is far from complete . to our knowledge , there are no available md results concerning the self - diffusion of hard ellipses so far . in this work , we analyze in detail the thermodynamic , structural and self - diffusive properties in the various phases of hard ellipses , with the aim of examining the effects of aspect ratio and area fraction and exploring connections between different properties . to this end , we perform edmd simulations for hard ellipses over a wide range of aspect ratios and area fractions , covering state points including the isotropic , plastic and nematic phases . we present the eos for a series of @xmath4 values and compare the simulation results with a recent prediction of the spt in the isotropic phase . we find that spt well describes the isotropic branch of the eos in hard ellipses . dependence of @xmath4 on the reduced pressure @xmath3 is explored in the various phases . at large fixed area fractions , @xmath3 increases with @xmath4 in both the isotropic and the plastic phases , and interestingly , its dependence on @xmath4 is rather weak in the nematic phase . we rationalize the thermodynamics of hard ellipses in terms of particle motions , which show different features in different phases . the static structures , both positionally and orientationally , are then investigated in the various phases . our simulations show that the plastic crystal forms for aspect ratios up to @xmath5 , while appearance of the stable nematic phase starts approximately at @xmath6 for the studied density range . the locations of the isotropic - plastic ( i - p ) transition and the i - n transition are determined by analyzing the bond - orientation correlation functions and the angular correlation functions , respectively . as expected , the i - p transition point is found to increase with @xmath4 , while a larger @xmath4 leads to a smaller area fraction where the i - n transition takes place . in addition , our results strongly support that the 2d nematic phase in hard ellipses has only quasi - lro . we further explore the self - diffusion of hard ellipses . we find that a phase transition in the translational degrees of freedom is indeed reflected by an enhancement in the rotational diffusion and vice versa . hence , a clear connection is revealed between the structure and the self - diffusion . we discuss the relevance of our results to the glass transition in hard ellipses , which is a subject of recent experimental investigation @xcite . we finally provide the results of the isodiffusivity lines for hard ellipses and discuss the effect of spatial dimension on the diffusive dynamics of hard ellipsoidal particles by comparing results of hard ellipses and uniaxial hard ellipsoids @xcite . the paper is organized as follows . in sec . ii , we specify the details of the simulation and the model system used in this work . thermodynamic , structural and self - diffusive properties of hard ellipses are presented and discussed in sec . a summary of our findings is given in sec . iv . in the appendix , we add the main ingredients for implementing an edmd simulation for hard ellipses and a brief discussion on the effect of the system size . we perform extensive edmd simulations of hard ellipses in a square box under periodic boundary conditions in @xmath7 ensemble . the main ingredients for implementing an edmd simulation for hard ellipses are described in the appendix a. in a simulation , we fix the particle number @xmath8 and the area of the simulation box @xmath34 with @xmath35 the box dimension . in order to explore the properties in different phases , we investigate two sets of aspect ratios , i.e. , @xmath36 with an interval of @xmath37 and @xmath38 with an interval of @xmath39 , and a wide range of area fractions ranging from @xmath40 to @xmath41 . at sufficiently high densities , a nematic phase ( i.e. , particles have their centers of mass at random but exhibit some long - range orientational order ) is expected to form for large @xmath4 values while a plastic phasealso known as the rotator phase , in which particles are oriented at random but their centers of mass have some ordered arrangement ) can form for small @xmath4 values . all the particles have the same mass @xmath42 and the same moment of inertia @xmath43 . both @xmath42 and @xmath43 are set to be unity for convenience in our simulations . in principle , ellipses with the same mass but different aspect ratios should have different moment of inertia . in our simulations , the function of @xmath42 and @xmath43 ( together with the particle velocities ) is to determine the temperature @xmath10 of the system via equipartition theorem . as we work on the athermal hard - particle system , the temperature only affects the time scale of the system ( see the time unit below ) . since we will report all results in reduced units , we expect that the results correctly capture the trends of static and dynamic quantities in variation with area fraction and aspect ratio . this is indeed strongly supported by the fact that our results for the static properties ( such as the reduced pressure ) are consistent with previously available mc simulations @xcite , where neither @xmath42 nor @xmath43 is relevant , suggesting that the general trends in this work are essentially unaffected by the choice of @xmath42 and @xmath43 once the results are given in reduced units . the temperature @xmath10 is irrelevant for athermal systems and remains constant due to the conservation of the total kinetic energy . we set @xmath44 to be unity in this work . note that @xmath10 contains both translational and rotational parts in the case of hard ellipses and that the separate temperature in each part fluctuates during the simulation . length , time and pressure are expressed in units of @xmath45 , @xmath46 and @xmath47 . in the following , we report results for @xmath48 . however , as can be seen from eq . ( 3 ) , it is also useful to study the effect of the system size especially in the nematic phase . we thus also perform simulations of hard ellipses with @xmath49 and @xmath50 for selected state points , and a brief discussion on the effect of the system size will be shown in the appendix b. in order to generate the starting configuration of a hard - ellipse system with specific @xmath4 and @xmath2 , we used the lubachevsky - stillinger ( ls ) compression algorithm @xcite , in which the particles collide elastically and expand uniformly at a certain growth rate . ls algorithm was initially designed for hard spherical particles and recently it has been generalized to the case of hard ellipsoidal particles @xcite . in the compression process , we started with a low - density hard - ellipse fluid at @xmath51 and used a quite small growth rate of @xmath52 so that the created initial configuration already remained close to be in an equilibrium or stable state . these initial states were then used as inputs for edmd simulations . at each aspect ratio , @xmath53 time units at @xmath54 and up to @xmath55 time units at @xmath56 were used to equilibrate the system , and then another @xmath57 time units for collecting data . four independent runs were performed for each state point in order to obtain reliable results and improve the statistics.thus , the results presented in this paper have been averaged among four independent samples .. it is worth mentioning that the slowest run takes over four weeks on a @xmath58 ghz central processing unit ( cpu ) . we note that in the solid phase ( i.e. , particles show long - range order both positionally and orientationally ) , it will be useful to allow the shape of the simulation box to change during the simulation , because the box shape is necessarily compatible with the equilibrium shape of the unit cell in order to form a solid phase . this has been done in previous @xmath14 mc simulations @xcite in order to study the transition from an isotropic or a nematic phase to a solid phase in hard ellipses . the box shape is fixed to be square in our simulations since our aim here is not to determine the location where the solid phase forms , but to focus on the physical properties of the formed phases in edmd simulations . meanwhile , we do not investigate the melting process of a perfect hard - ellipse solid ( which is a stretched triangular lattice ) . for these reasons , we did not observe the solid phase for all the state points studied in this work , please see subsection iii b for further discussion . in this section , we present and discuss our results of the thermodynamic , structural and self - diffusive properties in hard ellipses . first of all , the eos is shown for a wide range of aspect ratios and the reduced pressure ( i.e. , compressibility factor ) in the isotropic phase is compared with a recent result of the spt . we also explore the dependence of the reduced pressure on the aspect ratio in the different phases and discuss the influence of the particle motions on the thermodynamics . in subsection iii b , we analyze in detail the positional and orientational order of hard ellipses and compare our results with previous simulations when possible . finally , we investigate the self - diffusion of hard ellipses and demonstrate connections between the structure and the self - diffusion . the relevance of our results to the glass transition in hard ellipses is briefly discussed . we also show the results of the isodiffusivity lines in both the translational and the rotational degrees of freedom and discuss the effect of spatial dimension on the diffusive dynamics of hard ellipsoidal particles . as a function of area fraction @xmath2 for ( a ) hard disks and hard ellipses with @xmath59 and ( b ) hard ellipses with @xmath38 . the results are shifted up by @xmath60 from the preceding small @xmath4 for clarity . the dotted lines are simulation results and the solid lines are the results of the spt [ eq . the green dashed line in ( a ) is the spt prediction for the equation of state of hard disks in the isotropic state , i.e. , eq . ( 10 ) in ref . the circles indicate the area fractions where the system starts to form stable plastic crystals in ( a ) and stable nematic crystals in ( b ) , respectively ( see the structural analysis in subsection iii b ) . the semi - log plot in the inset of fig . 1(b ) highlights the change of dependence of @xmath3 on @xmath2 with increasing @xmath4.,scaledwidth=45.0% ] as a function of @xmath61 at fixed densities . the solid lines are the results of eq . ( 5).,scaledwidth=45.0% ] -dependence of @xmath3 at the highest densities studied . the dotted lines are a guide to the eye . for these densities , the system has a plastic phase for @xmath62 , a nematic phase for @xmath63 and an isotropic phase in between.,scaledwidth=45.0% ] knowledge of an accurate eos for a model system is of fundamental importance and particularly useful for providing information on what type of an underlying phase transition is . we present the results of eos for hard ellipses for various aspect ratios in fig . 1 . the eos for a system of hard disks is also included in fig . 1(a ) . here , the reduced pressure @xmath64 is computed by the momentum exchange from particle collisions in edmd ( see appendix a ) . we checked that our results for @xmath16 , @xmath65 and @xmath66 are consistent with the compression runs in previous mc simulations @xcite . let us first focus on hard ellipses with small @xmath4 values . in this situation , the orientational order is not appreciable in the system but the positional order can develop upon compression , leading to the formation of plastic crystals at sufficiently high densities . the appearance of the positional order can significantly affect the thermodynamics of the system and it is indeed reflected by a discontinuity in the eos , as shown in fig . 1(a ) . the plastic crystal was already observed for a system of hard ellipses with @xmath67 about four decades ago by vieillard - baron @xcite . our results demonstrate that the plastic crystal can be formed in hard ellipses with aspect ratios up to @xmath5 . the observation of the discontinuity in the eos is compatible with the scenario of a first - order transition , but it should be viewed with some caution since the nature of the formation of a 2d crystalline phase is unclear even in the case of hard disks ; e.g. , although the kosterlitz , thouless , halperin , nelson and young ( kthny ) theory may not rule out other melting scenarios @xcite , a famous one is that the melting of hard disks proceeds via two continuous phase transitions with an intermediate phase called hexatic phase @xcite , while the recent large - scale computer simulations @xcite reveal that the liquid - hexatic transition is of first order . we will further discuss this issue in subsection iii b by analyzing the structure of the system . moreover , we find in fig . 1(a ) that the discontinuity shown in the eos becomes less apparent as @xmath4 gets larger and nearly undetectable within the studied density range for @xmath68 . this can be easily understood since the particle s anisotropy tends to destroy the positional order of the system , and thus will be in favor of the isotropic phase . for hard ellipses with sufficiently large elongations , the orientational order is expected to appear and the positional order will be highly suppressed , resulting in the formation of 2d nematic liquid crystals in hard ellipses . as already found in other 2d models @xcite , the thermodynamic properties of the system are indeed not very sensitive to the occurrence of the orientational order . this is also confirmed by our results of the eos for hard ellipses with @xmath63 , as shown in fig . for these @xmath4 values , the system does have a nematic phase above some area fraction , but the corresponding change in the eos is rather mild . thus , we can not unambiguously determine the nature of the i - n transition based on solely the thermodynamics . however , by a close inspection of the results in the semi - log plot [ see the inset of fig . 1(b ) ] , some difference in the eos can be seen with varying @xmath4 . it seems that some inflection points exhibit in the eos for large aspect ratios , while the slope @xmath69 monotonically increases with @xmath2 for small aspect ratios , i.e. , the @xmath2-evolution of @xmath3 differs qualitatively from relatively small to very large aspect ratio . this suggests that the nature of the i - n transition changes with aspect ratio . in fact , previous mc simulations imply the existence of a tricritical point between @xmath17 and @xmath66 in the phase diagram of hard ellipses @xcite . the i - n transition is first - order below the tricritical point and becomes continuous above it . interestingly , we find that the appearance of the inflection points in the eos starts within the same @xmath4 range . thus , our results appear to support the existence of a tricritical point in the phase diagram of hard ellipses . however , since the location of such a point is very hard to determine even if it exists , we will focus on characterizing the properties of different phases in hard ellipses . next , we compare the reduced pressure in the isotropic phase of hard ellipses with a recent result of the spt . as mentioned in section i , the eos of hard ellipses can also be obtained by various theories . hence , it is meaningful to test these theoretical predictions as our results cover a wide window of @xmath4 and @xmath2 and contain data in the various phases . spt in general provides a simple form for the eos of a model system and its usefulness has been confirmed by the study of hard disks @xcite . other existing theories also provide relevant results , but the solution of the eos needs to be solved numerically @xcite . thus , we concentrate here on a very recent result of the spt @xcite . on the basis of the spt , boulk derived an eos for the isotropic fluid of hard ellipses @xcite , which has the following form @xmath70}{(1-\phi)^{2}},\ ] ] where the non - circularity parameter @xmath61 has been introduced . @xmath61 is calculated according to the formula @xmath71 with @xmath72 the perimeter of the ellipse . the value of @xmath72 can be expressed accurately by the complete elliptic integral of the second kind , and a rather good approximation can be obtained from the following formula : @xmath73 where @xmath74 . we use eq . ( 6 ) to calculate @xmath61 in this work and the results of the eq . ( 5 ) are shown as solid lines in fig . 1 . as a reference , the area fractions where the system starts to form stable plastic crystals and stable nematic crystals are also shown in fig . 1 . as can be seen in fig . 1(a ) , the simulation results and the spt prediction are almost indistinguishable in the isotropic phase for @xmath75 , indicating a fairly good description of eq . ( 5 ) for the eos of hard ellipses with small @xmath4 values . meanwhile , the predicted eos deviates from the simulation results above some area fraction for @xmath63 , as shown in fig . 1(b ) . as explained in ref . @xcite , the system is not isotropic but exhibits orientational order at high densities for large @xmath4 values . in fig . 1(b ) , one can find that the area fraction where eq . ( 5 ) starts to deviate from the simulations is much smaller than the i - n transition point [ see circles in the main plot of fig . however , it should be emphasized that the short - range orientational order already appears in the system at smaller area fractions ( see subsection iii b ) . taking this into account , we find that eq . ( 5 ) well describes the isotropic branch of the eos for hard ellipses even with large @xmath4 values . we can also use @xmath61 as a variable to test eq . the results for representative area fractions are shown in fig . we find that the theoretical and simulation results agree well for low densities . at high densities , deviations appear for large @xmath61 values because of the occurrence of short - range or some long - range orientational order , as evidenced by the non - monotonic dependence of @xmath3 on @xmath61 . therefore , we conclude that spt provides a good description of the eos for the isotropic phase of hard ellipses . it is interesting to further explore dependence of the reduced pressure on the aspect ratio in the various phases of hard ellipses . to this end , we plot @xmath3 as a function of @xmath4 for @xmath76 , @xmath77 and @xmath41 in fig . 3 . for these densities , the system has a plastic phase for @xmath62 , a nematic phase for @xmath63 and an isotropic phase in between ( see subsection iii b ) . we find that @xmath3 monotonically increases with @xmath4 at fixed densities in both the isotropic and the plastic phases ( i.e. , @xmath75 at @xmath78 ) , suggesting that the thermodynamics of the system can be significantly influenced by the particle shape when the ellipses are oriented at random . note that this conclusion is still valid for even larger @xmath4 values ( see the results for @xmath79 in fig . 2 ) . interestingly enough , in the nematic phase ( i.e. , @xmath63 at @xmath78 ) , when the particles show some long - range orientational order but have their centers of mass at random , @xmath3 depends on @xmath4 very weakly . although not pointed out before , the same conclusion can be drawn from previous @xmath14 mc simulations @xcite , which indicate that the area fractions are almost identical at the same pressure in the nematic phase of hard ellipse at least for @xmath17 and @xmath66 . to understand the above findings , we recall that the reduced pressure is calculated from the momentum exchange during collisions of two ellipses , which indeed contains translational and rotational parts ( although they are coupled with each other ) . it is expected that the contribution from the rotational motions becomes more apparent as the ellipses get more elongated . this is the reason why @xmath3 increases with @xmath4 in the plastic and the isotropic phases , where the particles can freely move both translationally and rotationally ( see fig . 14 ) . in the nematic phase , however , particles on average line up preferentially along a common direction . in this case , the rotational dynamics becomes extremely slow , but the motions of the particles have little hindrance in the translational degrees of freedom ( see fig . 14 ) and thus the translational part becomes dominant in determining the reduced pressure of the system . we then speculate that the translational part must be very similar for all aspect ratios at the same density in the nematic phase so that @xmath3 has only weak dependence on @xmath4 . therefore , the thermodynamics of hard ellipses could be explained in terms of the particle motions . we now turn to the structural properties of hard ellipses . as a system of hard ellipses with sufficiently large aspect ratios has a transition from the isotropic liquid to the 2d liquid crystal , much attention has been paid on the structure of the nematic phase in hard ellipses , i.e. , on characterizing the orientational order of elongated hard ellipses . instead of focusing solely on the orientational correlations , we explore both positional and orientational order in the various phases of hard ellipses in this subsection . of hard ellipses at @xmath80 for @xmath68 . the results are shifted up by @xmath81 from the preceding small @xmath4 for clarity . as there are two basic length scales in a system of hard ellipses , @xmath82 also peaks at @xmath83 , as indicated by the arrows for @xmath84.,scaledwidth=45.0% ] of hard ellipses with @xmath85 in the vicinity of the isotropic - plastic transition . the solid lines are the results of the oz fittings and the dash dotted lines are the results of the power - law fittings . the green dashed line indicates @xmath86 . the results are similar for other aspect ratios with @xmath62.,scaledwidth=45.0% ] as a function of @xmath4 . the solid line is a guide to the eye.,scaledwidth=45.0% ] as a function of nematic director @xmath87 for hard ellipses with @xmath18 for various @xmath2 . note that the result is shown for a single sample because @xmath88 can not be averaged among different samples since the nematic direction differs from one sample to another.,scaledwidth=45.0% ] as a function of @xmath2 for various @xmath4 . lower : contour plot of @xmath89 in the plane of @xmath2 and @xmath4.,scaledwidth=45.0% ] for hard ellipses with @xmath75 at @xmath80 . the green dashed line indicates @xmath90.,scaledwidth=45.0% ] for a system of hard ellipses with @xmath18 in the vicinity of the isotropic - nematic transition . the solid lines are the results of the power - law fittings . the green dashed line indicates @xmath90 . the results are similar for other aspect ratios with @xmath63 . , scaledwidth=45.0% ] as a function of @xmath4 . the triangle and the squares are the mc results taken from ref . @xcite and ref . @xcite , respectively . the dashed and solid lines are the fits to the edmd data by @xmath91 with @xmath92 and @xmath93 with @xmath94 and @xmath95 , respectively.,scaledwidth=45.0% ] for hard ellipses with @xmath63 at @xmath80.,scaledwidth=45.0% ] and area fraction @xmath2 . ( b)-(e ) representative snapshots for a high - density isotropic phase with @xmath96 and @xmath80 , a low - density isotropic phase with @xmath18 and @xmath97 , a plastic phase with @xmath85 and @xmath80 , and a nematic phase with @xmath84 and @xmath80 . note that there is visually no distinction for the orientation of an ellipse with @xmath98 and @xmath99 so that particles with @xmath98 are shown as the same color as those with @xmath99 in the snapshots.,scaledwidth=45.0% ] we first consider the pair correlation function of hard ellipses , which is defined as @xmath100 where @xmath101 is the distance between the centers of mass of the particles . in general , characteristic peaks will appear in @xmath82 at large distances if the system possesses some long - range positional order . this is indeed observed in systems of hard ellipses with @xmath62 at high densities ( data not shown ) , where the centers of mass of the particles form a triangular ( or hexagonal ) lattice . in hard ellipses with larger aspect ratios , an ordered structure in principle can also form translationally although it does not need to be a perfect triangular one . we present the results of @xmath82 at @xmath80 for @xmath68 in fig . 4 . note that there are two basic length scales ( i.e. , the semi - major axis @xmath12 and the semi - minor axis @xmath13 ) in a system of hard ellipses so that @xmath82 also peaks at the distance of @xmath102 . beyond the distance of @xmath102 , we find from fig . 4 that @xmath82 rapidly decays to the value of @xmath39 , indicating the absence of any long - range translational order even at the largest density studied . the reason , why we did not observe the formation of a translationally ordered structure for large aspect ratios , has been explained in section ii . moreover , it is clear that several peaks are also shown in @xmath82 at between @xmath45 and @xmath102 , suggesting the existence of some short - range translational order . as we shall see later , this is indeed due to the appearance of the orientational order . the pair correlation function only provides a qualitative way to see whether a system possesses long - range positional order or not . meanwhile , we can study the bond - orientation correlation functions @xcite , in order to quantitatively identify the location of the transition from the isotropic liquid to the hexagonal crystal . we first define the sixfold bond - orientation order parameter @xmath103 where @xmath104 , @xmath105 is the number of the neighbors of the @xmath106th particle and @xmath107 is the angle between the vector @xmath108 and the @xmath109 axis . here , we define two ellipses to be neighbors if they overlap when uniformly expanding their sizes by @xmath110 . then , the spatial correlation of @xmath111 can be calculated by @xmath112 the quasi - long - range bond - orientation order is evidenced by an algebraic decay of @xmath113 , i.e. , @xmath114 in particular , according to the kthny theory @xcite , @xmath115 has a value of @xmath116 at the boundary between the liquid phase and the hexatic phase . we thus use the criterion @xmath86 to separate the plastic states from the isotropic states for hard ellipses with @xmath62 . in the liquid phase , the system only exhibits short - range bond - orientation order and @xmath113 should decay exponentially . however , for high densities and especially in vicinity of the formation of the hexatic phase , we find that the bond - orientation correlation function can indeed be better described by the ornstein - zernike ( oz ) function @xmath117 with @xmath118 the static correlation length . in fact , it is observed that the oz function can well characterize the behavior of @xmath113 in 2d supercooled liquids @xcite . we confirm the above claims by presenting the results of @xmath113 for a system of hard ellipses with @xmath85 in fig . as can be seen , @xmath113 decays slowly with increasing @xmath2 . below @xmath119 , the envelopes of @xmath113 can be well fitted by the oz function . with further increasing @xmath2 , a power - law decay with an exponent of @xmath120 holds for the behavior of @xmath113 , indicating the emergence of quasi - long - range bond - orientation order . we thus estimate the location of the isotropic - plastic transition to be @xmath121 for @xmath85 . note that @xmath122 obtained in this way is actually an upper bound to the exact value . the values of @xmath122 for other aspect ratios are shown in fig . and we confirmed that the obtained value for hard disks ( i.e. , @xmath123 ) is in agreement with other studies within the simulation accuracy @xcite . the above analysis indicates that hard ellipses for @xmath62 share the similar behavior with hard disks on the structure as well as the thermodynamics ( see discussion in subsection iii a ) . thus , the nature of the isotropic - plastic transition of hard ellipses should be the same as that of hard disks . turning to the orientational order of hard ellipses , we first focus on the nematic order parameter as a function of @xmath2 and @xmath4 . the 2d nematic order parameter is defined as @xmath124 where @xmath125 is the angle characterizing the orientation of the @xmath106th ellipse with respect to the @xmath109 axis and @xmath87 the orientation of the nematic director . note that this definition is equivalent to eq . ( 3 ) . as the nematic director is not known _ a priori _ , @xmath126 and @xmath87 are usually determined by finding the eigenvalues and the eigenvectors of a tensor order parameter @xcite . as pointed out and observed in previous work @xcite , the nematic order parameter obtained from the above method has a clearly nonzero value even in the isotropic phase due to the finite system size used in simulations . instead of using the tensor order parameter , we set @xmath87 as a variable and calculate @xmath126 as a function of @xmath87 in this work . we show the results of @xmath88 for various area fractions for @xmath18 in fig . 7 as an illustration . as can be seen , @xmath88 shows extrema , and the absolute values of these extrema are the same ( we denote it by @xmath89 ) because of the periodicity of the cosine function in the definition of @xmath126 . we then use @xmath89 to quantify the nematic order of hard ellipses . in fig . 8 , @xmath89 is presented as a function of @xmath2 for various aspect ratios and also shown as a contour plot in the plane of @xmath2 and @xmath4 . we find from fig . 8 that @xmath89 is nearly zero at low densities and has clearly nonzero values at high densities for @xmath63 . thus , it appears that @xmath89 is more sensitive to the onset of the orientational order than @xmath26 . for @xmath75 , a clearly nonzero but still very small value of @xmath89 sets in above @xmath127 , implying the absence of any long - range orientational order . the quasi - long - range orientational order is expected to appear for systems of hard ellipses with larger aspect ratios , as confirmed by the results of @xmath89 for @xmath63 . moreover , the area fraction where the system starts to orientate becomes small as @xmath4 increases , revealing the fact that a system of more elongated hard ellipses has a stronger tendency to form nematic liquid crystals . it is interesting to explore whether @xmath89 is dependent on the system size in the nematic phase . unlike the order parameter @xmath26 computed from the tensor order parameter ( dependence of the system size on @xmath26 has been confirmed by mc studies @xcite ) , we find that the system size does nt strongly influence the value of @xmath89 for both the isotropic and the nematic phases studied ( see appendix b ) . hence , it is not proper to estimate the point of the i - n transition from the dependence of the system size on @xmath89 . we can quantitatively determine the location of the i - n transition @xmath128 from the angular correlation function defined by eq . we mainly concentrate on the case of @xmath129 , i.e. , @xmath130)>\sim r^{-\eta_{2}},\ ] ] where @xmath131 and the average is performed over all pairs with the distance @xmath101 . as introduced in section i , the stable nematic phase is then identified by a power - law decay with @xmath132 for @xmath133 . the results of @xmath133 are shown for @xmath75 at @xmath80 in fig . consistent with the results of @xmath89 , we find that @xmath133 decays much faster than @xmath90 even at the largest studied density for these aspect ratios , indicating again there is no quasi - long - range orientational order in the system . the algebraic decay appears at high densities for @xmath63 , as evidenced in fig . 10 for a system of hard ellipses with @xmath18 . we then use the power law @xmath134 to fit @xmath133 at @xmath135 and estimate the location of the i - n transition @xmath128 to be the area fraction when the value of @xmath136 is smaller than @xmath116 . again , @xmath128 obtained in this way is actually an upper bound to the exact value . it should also be stressed that the i - n transition in a finite system tends to occur at a lower density than in an infinite system because the frank s constant obtained from simulations on small systems is larger than that in the infinite system size @xcite . we show the results of @xmath128 as a function of @xmath4 in fig . as can be expected , @xmath128 decreases with increasing @xmath4 . comparing with previous mc results , we find that @xmath128 for @xmath18 obtained in this work is larger than that in ref . @xcite but close to the result in ref . the reason , why the i - n transition was observed at a smaller area fraction by vieillard - baron , has been hinted by frenkel and eppenga @xcite . similar to the case of @xmath18 , the value of @xmath128 for @xmath17 in this work is also a little smaller than that of ref . the slight difference between the results of ref . @xcite and ours is possibly due to the different particle number and different method used ( @xmath137 and @xmath14 mc in ref . @xcite versus @xmath48 and @xmath7 md in this work ) . we also note a recent mc study @xcite , where @xmath128 for @xmath84 is roughly estimated to be @xmath138 , which is much larger than ours . our results also provide an opportunity to test theoretical predictions on the @xmath4-dependence of @xmath128 for hard ellipses . several theories @xcite have predicted the occurrence of the i - n transition in hard ellipses . in particular , the onsager theory - based approach @xcite and spt @xcite predict that the relation @xmath139 emerges for sufficiently elongated hard ellipses . since the values of @xmath4 studied in this work are not large enough and the hard - needle regime has probably not been reached yet , such a relation can not well describe our results ( see the dashed line in fig . 11 ) . instead , we find that a different form , @xmath93 with @xmath140 and @xmath141 adjustable parameters , provides a fairly good description of our data , as shown by a solid line in fig . 11 with @xmath94 and @xmath95 . as mentioned in section i , the nature of the orientational order in the 2d nematic phase is also a central theme in the study of liquid crystals . the algebraic decay of @xmath133 already indicates the lack of the true long - range orientational order in the nematic phase of hard ellipses . we further confirm that the algebraic orientational order holds even for the largest studied density , as shown in fig . 12 , although @xmath133 may decay very slowly in this case . thus , our simulations strongly support that the nematic phase in hard ellipses has quasi - lro . based on the above structural analysis , we present a phase diagram of hard ellipses and several representative snapshots in fig . 13 . our simulations show three different phases within the investigated state points . for @xmath62 , the system has a transition from a low - density isotropic liquid to a plastic crystal at high densities [ fig . 13(d ) ] . with increasing the aspect ratio up to @xmath16 , we did not observe any phase transition within the whole density range studied in this work [ fig . 13(b ) ] . for hard ellipses with @xmath63 , an isotropic phase [ fig . 13(c ) ] will transform into a nematic phase [ fig . 13(e ) ] at sufficiently high densities . again , we confirm from the snapshot in fig . 13(c ) that some short - range orientational order exists even in the isotropic phase of elongated hard ellipses . on the other hand , although the simulated phase diagram does not include any solid phase , we should stress again that hard ellipses do have the transition from an isotropic or a nematic phase to a solid phase , as revealed in previous mc simulations @xcite . we even expect that a transition from a plastic crystal to a solid phase will occur for small @xmath4 values at even larger densities than @xmath80 . the reason why such a transition was not observed in our work , as explained in section ii , is due to the fact that a solid is hard to spontaneously form from a liquid in @xmath7 md simulations . on the other hand , it will be very interesting to explore the high - density part of the phase diagram for hard ellipses , where the stable crystalline phase will exist . in fact , our knowledge is very limited so far about the equilibrium structure of the solid in hard ellipsoidal particles , although it can be readily recognized that the densest packing of hard ellipses has the same density as the densest packing of hard disks and has a stretched triangular lattice structure @xcite . however , such explorations are hindered by the present method . more elegant methods will be needed to clarify this issue , which remains a challenge from both theoretical and computational points of view . self - diffusion is of great importance in many real processes , but it is still poorly understood in hard ellipses . the diffusive property of hard ellipses with @xmath84 has been studied recently by experiments @xcite and mc simulations @xcite . yet , the effects of aspect ratio and area fraction on the self - diffusion of hard ellipses are unclear . in this subsection , we investigate in detail the self - diffusion of hard ellipses over a wide range of aspect ratios and area fractions , including the isotropic , plastic and nematic phases , and demonstrate the connections between the structure and the self - diffusion . we further discuss the relevance of our results to the glass transition in hard ellipses and provide the results of the isodiffusivity lines . for three aspect ratios . the system has a plastic phase for @xmath142 , an isotropic phase for @xmath16 and a nematic phase for @xmath18.,scaledwidth=45.0% ] the self - diffusion is measured here by the translational and rotational mean squared displacements ( msd ) @xmath143 @xmath144 and the corresponding diffusion constants obtained from the long - time data of the msd @xmath145 @xmath146 we first focus on the behavior of the msd in the different phases of hard ellipses . the results of @xmath147 and @xmath148 are shown at @xmath80 for the plastic ( @xmath142 ) , isotropic ( @xmath16 ) and nematic ( @xmath18 ) phases of hard ellipses in fig . as can be seen , both translational and rotational motions of particles are ballistic at short times , and cross over into diffusive behavior at sufficiently long times in both the plastic and the isotropic phases . moreover , an apparent subdiffusive behavior is seen in the intermediate time regime of the translational msd due to the cage effect in the plastic phase [ see the solid line in fig . by contrast , the rotational diffusion of particles becomes extremely slow in the nematic phase , while the translational msd can still retain a diffusive regime at long times within the simulation time window ( see the dotted lines in fig . we have seen in subsection a that the above features shown in the particle motions may be responsible for the thermodynamics in the different phases of hard ellipses . and ( b ) rotational diffusion constant @xmath149 as a function of area fraction @xmath2 for various @xmath4.,scaledwidth=45.0% ] as a function of translational diffusion constant @xmath150 for various @xmath4 . the green dashed line indicates @xmath151.,scaledwidth=45.0% ] to better illustrate the effects of aspect ratio and area fraction on the self - diffusion of hard ellipses , we extract the translational and rotational diffusion constants from those of the corresponding msds retaining a diffusive regime . the results of @xmath150 and @xmath149 are shown in fig . 15 . at fixed area fractions , we observe that increasing @xmath4 leads to a slowing down of the rotational diffusion , while @xmath150 increases with aspect ratio . moreover , the occurrence of the phase transition seems not to affect these trends at least within the studied density range . then , a natural question is how self - diffusion responds to the phase transition . in a recent md study @xcite , it was found for a system of hard ellipses with @xmath84 that if the diffusion constants are normalized by the values in the infinite dilution limit , then both diffusion constants still monotonically decrease with @xmath2 but the normalized @xmath150 will exceed the normalized @xmath149 at the area fraction where the i - n transition takes place , i.e. , the translational diffusion is enhanced with respect to the rotational diffusion as a result of the appearance of the quasi - long - range orientational order . we did nt observe crossing of the two normalized curves for the same aspect ratio by performing the same analysis ( data not shown ) . instead , our simulations reveal that the self - diffusion responds to the phase transition in a rather straight manner , which can be detected already in each curve ; i.e. , a sudden increase of @xmath149 is seen when a plastic crystal forms , as evidenced by a small peak in @xmath149 in fig . 15(b ) , while the emergence of a nematic phase leads to an enhancement of @xmath150 although this observation seems to become less evident as @xmath4 gets smaller . the small peaks in fig . 15 immediately indicate that the formation of the orientational or positional order will increase the translational or rotational mobilities of particles . hence , a phase transition in the translational degrees of freedom is reflected by a corresponding change in the rotational diffusion and vice versa . we thus uncover a clear connection between the structure and the self - diffusion in hard ellipses . it is interesting to further explore the effect of the particle s anisotropy on the self - diffusion in each degrees of freedom by plotting @xmath149 as a function of @xmath150 at varying @xmath4 . as shown in fig . 16 , @xmath149 is larger than @xmath150 within the full density range for @xmath75 and the particles diffuse slower in the rotational degrees of freedom than in the translational degrees of freedom for @xmath152 . in addition , the curves show an upturn at high densities for @xmath75 , while they turn down for @xmath63 . this means that the translational mobility of hard ellipses will be similar with the rotational mobility at some aspect ratio between @xmath153 and @xmath154 . if a glass is allowed to form in hard ellipses , we then expect that the rotational glass transition sets in at a lower density than the translational glass transition for @xmath155 , which was indeed observed recently in experiments of quasi-2d hard ellipsoids with @xmath156 @xcite . instead , the translational glass transition density will be smaller than the rotational one for @xmath157 . this immediately suggests that there is a point in the phase diagram of glassy hard ellipses , where the translational and the rotational glass transition lines will intersect . therefore , glasses in hard ellipses can be classified into three categories : a plastic glass in which dynamic arrest occurs in the translational degrees of freedom but the system behaviors as a liquid in the rotational degrees of freedom , a liquid glass in which the system becomes glass in the rotational degrees of freedom but the translational motions of particles remains ergodic , and a glass in which motions of the particles are dynamically arrested in both the translational and the rotational degrees of freedom . and area fraction @xmath2 . the solid lines are isodiffusivity lines from translational diffusion coefficients @xmath150 and the dashed lines are isodiffusivity lines from rotational diffusion coefficients @xmath149 . the green squares and the red circles indicate the locations of the isotropic - plastic transition and the isotropic - nematic transition , respectively.,scaledwidth=45.0% ] another relevant property is the isodiffusivity of hard ellipses . as there are related results for hard ellipsoids of revolution @xcite , we can evaluate such isodiffusivity lines for hard ellipses by proper interpolation and thus explore the qualitative effect of spatial dimension . we present the results of the isodiffusivity lines in fig . we find that an almost perpendicular crossing of the translational and the rotational isodiffusivity lines is shown in the plane of @xmath2 and @xmath4 . moreover , the rotational isodiffusivity lines of hard ellipses reproduce qualitatively the shape of the i - n transition line at least for sufficiently small @xmath149 values , which is similar as the result in hard ellipsoids of revolution @xcite . the latter observation indicates that the rotational motions of the ellipses is mostly controlled by the particle s anisotropy . on the other hand , the translational isodiffusivity lines in uniaxial hard ellipsoids mimic the swallowlike shape of the coexistence line between the isotropic liquid and the crystalline phase @xcite , suggesting a non - monotonic change of @xmath150 with @xmath4 at fixed density ( see fig . 2 in ref . note that the swallowlike shape appears because there is a distinction between prolate ( rodlike ) and oblate ( disklike ) particles for hard ellipsoids , while such a distinction does not exist in hard ellipses since there are only two symmetry axes . our results , already apparent from fig . 15 , show that @xmath150 monotonically increases with @xmath4 at fixed @xmath2 , which remains valid even when a phase transition intervenes . thus , it appears that the spatial dimension does influence the translational dynamics . however , it should be noted that our result does not rule out a non - monotonic dependence of @xmath150 on @xmath4 at even larger fixed densities than those presented in fig . in fact , our preliminary study for polydisperse hard ellipses ( which prevent the formation of crystal phases ) does show that the translational glassy dynamics depends on aspect ratio non - monotonically at large fixed densities . therefore , a common mechanism will be responsible for the translational glassy dynamics of hard ellipsoidal particles although different origins may exist at low densities . in summary , we have presented a numerical study of hard ellipses over a wide range of aspect ratios and area fractions , covering state points including the isotropic , plastic and nematic phases . we have provided the results of the eos for a series of @xmath4 values and compared the simulations with a recent prediction of the spt in the isotropic phase . we find that spt well describes the isotropic branch of the eos in hard ellipses . dependence of @xmath4 on the reduced pressure @xmath3 has been explored in the various phases . at large fixed area fractions , @xmath3 increases with @xmath4 in both the isotropic and the plastic phase , and interestingly , its dependence on @xmath4 is rather weak in the nematic phase . we rationalize the thermodynamics of hard ellipses in terms of particle motions , which show different features in different phases . for the static structures , our simulations show that the plastic crystal forms for aspect ratios up to @xmath5 , while appearance of the stable nematic phase starts approximately at @xmath6 . the locations of the i - p transition and the i - n transition have been determined respectively by analyzing the bond - orientation correlation functions and the angular correlation functions and compared with previous simulations . we demonstrate that the i - p transition point increases with @xmath4 as the particle s anisotropy tends to destroy the positional order , while a larger @xmath4 leads to a smaller area fraction where the i - n transition takes place , revealing the fact that a system of more elongated hard ellipses has a stronger tendency to form nematic liquid crystals . in addition , our results strongly support that the 2d nematic phase in hard ellipses has only quasi - lro , as evidenced by the algebraic decay of the angular correlation function in the nematic phase . we have investigated the self - diffusion of hard ellipses and found that a phase transition in the translational degrees of freedom is indeed reflected by a corresponding change in the rotational diffusion and vice versa . hence , we reveal a clear connection between the structure and the self - diffusion . we discuss the relevance of our results to the glass transition in hard ellipses and infer that there is an intersection point between the translational and the rotational glass transition lines in the phase diagram of glassy hard ellipses . finally , we have evaluated the isodiffusivity lines for hard ellipses and discussed the effect of spatial dimension on the translational dynamics by comparing the diffusive dynamics of hard ellipses with that of hard ellipsoids . our results are also valuable for understanding the structure and the diffusion of anisotropic molecules at membranes and interfaces . this work is subsidized by the national basic research program of china ( 973 program , 2012cb821500 ) , and supported by the national natural science foundation of china ( 21222407 , 21074137 , 50930001 ) programs . we describe the main ingredients for implementing an edmd simulation of hard ellipses in this section . in order to perform a md or a mc simulation for a hard - particle system , a key task is to detect overlap between two particles . this is a trivial thing for spherical particles , while it is highly non - trivial for non - spherical particles . in the case of hard ellipses , vieillard - baron @xcite first introduced an overlap criterion about four decades ago . later on , perram and wertheim @xcite derived a contact function . both criteria can be easily applied in a numerical simulation of hard ellipses . however , it appears that the perram - wertheim ( pw ) approach is more convenient both computationally and theoretically . moreover , in addition to an overlap criterion , we also need the first - order time derivative of the distance of closest approach between two moving ellipses in order to determine the contact time , which is a central quantity in an edmd simulation @xcite . indeed , donev _ et al _ @xcite have derived the time derivatives of the pw contact function and developed an edmd algorithm for hard ellipsoidal particles . we thus follow their results and adopt the pw approach in our study . however , we note that other approaches also exist for detecting overlap between two ellipses @xcite . let us first describe the orientation of a rigid body . in two dimensions , it needs only one angle to represent the orientation of the rigid body as there is only one rational degree of freedom . let @xmath98 denote the angle between the semi - major axis of the ellipse and the @xmath109 axis , then the rotation matrix reads @xmath158 taking its shape into account , an ellipse can be described by @xmath159 where @xmath160 is a diagonal matrix containing the semi - major and semi - minor axes along the diagonal , @xmath161 the inverse of @xmath160 and @xmath162 the transpose of @xmath163 . we define the pw contact function for two ellipses @xmath164 and @xmath165 as @xcite @xmath166 where @xmath167 with @xmath168 the position of ellipse @xmath164 , and @xmath169 . perram and wertheim @xcite has proven that @xmath170 is strictly concave on the interval @xmath171 $ ] and has a unique maximum @xmath172 at @xmath173 . then the numerically determined maximum @xmath172 has the following properties : @xmath174 note that @xmath170 is indeed a rational function of @xmath175 and hence its maximum can be readily found using a newton - raphson method @xcite . when @xmath172 and @xmath176 are known , we can scale the two ellipses by a common factor @xcite so that they are externally tangent . the contact point is then determined by @xmath177 where @xmath178 is the unnormalized common normal vector at the contact point . then @xmath179 and @xmath180 are used to compute the first - order time derivative of @xmath172 , which has been derived in ref . @xcite and reads @xmath181 where @xmath182 is the projection of the relative velocity at the contact point @xmath183 where @xmath184 , @xmath185 , @xmath186 and @xmath187 are the translational and angular velocities , respectively , and the cross product @xmath188 is defined as @xmath189 in principle , the contact time of two moving ellipses can be then numerically determined by using a newton - raphson method @xcite . however , some numerical problems are probably met and some tricks must be employed . the reader can find more details in refs . @xcite . when the two ellipses @xmath164 and @xmath165 become externally tangent , their new translational velocities @xmath190 , @xmath191 , and angular velocities @xmath192 and @xmath193 can be exactly found from the conservation of total energy , linear and angular momentum . specifically , let @xmath194 , @xmath195 , @xmath196 and @xmath197 denote the translational and angular velocities of ellipses @xmath164 and @xmath165 before collision , their new velocities are updated by the following formula : @xmath198 @xmath199 @xmath200 @xmath201 where @xmath42 is the mass of the particle , @xmath43 is the moment of inertia , @xmath202 is the normalized common normal vector at the contact point , and @xmath203 is the momentum exchange during the collision , which is calculated according to @xmath204 meanwhile , the reduced pressure can be computed from the momentum exchange during a time interval @xmath205 : @xmath206 to accelerate the simulation , we use a neighbor list ( nl ) method . note that the nl method for hard ellipses is more complicated than that for hard spheres @xcite , but the essence remains unchanged . and @xmath65 , effect of the system size @xmath8 on ( a ) reduced pressure , ( b ) nematic order parameter , ( c ) translational diffusion constant and ( d ) rotational diffusion constant . the results for @xmath17 in ( a ) are shifted up by @xmath60 for clarity , and the error bars in ( b ) correspond to the standard deviation over four independent samples . the system shows an isotropic phase in the whole density range in the case of @xmath16 , while the i - n transition occurs at @xmath127 for @xmath17.,scaledwidth=45.0% ] in this section , we assess the effect of the system size . as introduced in sec . i , an analysis of the system - size dependence of the orientational order parameter is very useful especially in the 2d nematic phase . however , we also want to explore the influence of the system size on other quantities in order to confirm the validity of the conclusions given in this work . to this end , we also performed simulations of hard ellipses with @xmath49 and @xmath50 for @xmath16 and @xmath65 . we have concluded in sec . ii that the system shows an isotropic phase at area fractions up to @xmath80 in the case of @xmath16 during the compression run , while the isotropic liquid transforms into the nematic liquid crystal at @xmath127 for @xmath17 . we find from fig . 18 that the system size does not affect the above conclusion . in addition , all properties , including the reduced pressure @xmath3 , the nematic order parameter @xmath89 and the diffusion constants @xmath208 , do not depend on the system size within the statistical accuracy at least for the studied particle numbers . in particular , @xmath89 is also independent of @xmath8 even in the nematic phase of hard ellipses . this immediately indicates that the computed @xmath89 here is different from the order parameter @xmath26 obtained from the tensor order parameter , since previous simulations do show that @xmath26 decreases algebraically with @xmath8 @xcite . thus , we can not determine the point of the i - n transition from the dependence of the system size on @xmath89 , although we have shown that @xmath89 is more sensitive to the onset of the orientational order than @xmath26 at fixed particle number . moreover , as shown in fig . 19 , the angular correlation function @xmath133 decays faster for larger @xmath8 in the nematic phase , but this feature seems to become less evident and even disappears as @xmath2 gets larger . however , the results confirm again that the nematic phase in hard ellipses does have only quasi - lro . note that the free energy density for a nematic liquid crystal in three dimensions is composed of splay , twist and bend energy terms and that each term is associated with a separate elastic constant , while the twist term is absent in two dimensions . since the three elastic constants are of the same order of magnitude in most cases , the free energy is often derived by assuming that the three elastic constants are equal . based on this so - called one - constant approximation , the resulting free energy has a very compact form and thus facilitates the theoretical analysis @xcite . note that even if the term `` bond - orientation '' is used , the calculation of the bond - orientation correlation function is based solely on the centers of mass of the particles and thus it indeed measures the positional order .

a new thermodynamics for liquids related to von laue s approach ( 1917 ) substitutes some particle priors of gibb s rational thermodynamics . this allows the definition of a new dynamic entity ( defect ) whose diffusion properties also claim a largest causal region ( speckle ) . in the frame of the hidden charge model it is discussed , whether this new thermodynamics can be applied to an initial liquid for cosmology , where the defects lead to the later galaxies and the speckles to a finite expanding universe of diameter . far below a hadronic compton wave length of order 1 fermi , , there is no room left for too small filter elements , that would , however , be necessary for a filter convergence to an isolated cold quantum mechanical point particle . when the expansion of the universe comes to , i.e. for , then many hadrons are created . the related negative pressure gets large amounts and leads to an intense hadronic cosmological inflation . the defects are formed by the shaping power of levy distribution ( preponderant component , hierarchy , damping factor ) . a relation between the number of galaxies , the tilt in the density fluctuation , and the temperature amplitude of the cmb is obtained . for ( much vacuum ) , the electromagnetic tangent objects are large and flat . this allows a geometric interpretation of the stony dark energy as flatness on the golden side of the einstein equation , . * keywords : * 1 . inflation 2 . power spectrum 3 . dark energy theory 4 . gravity

0901.1050

about 1960 , a change in education and evaluation of general phenomenological thermodynamics was observed . before 1960 , it was mainly based on arguments by carnot , kelvin and clausius . after 1960 , it was mainly based on the gibbs distribution of statistical mechanics . this gibbs rational thermodynamics has the following priors for a system of interacting structureless particles : 1 . the only general entities are the particles ; the rational is obtained from the thermodynamic limit : number of particles @xmath11 under the condition of finite density ( @xmath12 given ) , 2 . there is a large heat reservoir that feeds the temperature @xmath13 as an equivalence - class index into the system , 3 . the non - fluctuation of the reservoir temperature , @xmath14 , is transferred into the definition of the general temperature ( to justify the sharp @xmath15 denominator in the boltzmann factor for energy in the gibbs distribution ) . these priors are tried and tested in systems where the spatial infinities , sufficiently separated from the particles , are contained inside the system itself , quasi as an internal heat reservoir . examples are the infinite lattice in a solid or an infinite vacuum of vacuum elements in a gas with isolated particles imbedded . we ask , whether these priors can be applied to liquids , especially to liquid dynamics , where such separated infinities are not obvious . the existence of two experimentally well - tried particularities could be useful for cosmology by starting the universe from an initial liquid ( initial liquid hypothesis ) . smaller ones ( glarum levy defects for dynamic heterogeneity of the dynamic glass transition , @xmath0 defect as a new entity ) could be the seeds for the later galaxies @xmath7 we use now the symbol @xmath16 for their number in the universe @xmath7 , and a larger one ( fischer speckle ) could be the largest causal region in the initial liquid , wherefrom the increasing diameter @xmath17 of the expanding universe results . in the dispersion zone of the dynamic glass transition of a classical molecular liquid , the @xmath0 defects must dynamically be described by a levy distribution with a levy exponent @xmath18 , @xmath19 for relaxation ( frequency @xmath20rad / s ) . this follows from the representativeness theorem ( section 3.1 ) . the difference @xmath21 is called tilt and is later related to the tilt @xmath22 of the spectral index @xmath23 in cosmic density fluctuations . the preponderant component of the levy sum for @xmath24 corresponds to a levy diffusion step of a molecule through the cage door of its nearest neighbor particles . a relation between the small @xmath0 defects and the large @xmath1 speckle is obtained via an analytic continuation of the diffusion steps . if the application to the cosmological initial liquid is reliable , a relation results between the number of galaxies @xmath16 in the universe of diameter @xmath3 and the tilt @xmath22 . all this is worded by the phrase shaping power of levy distribution . the thermodynamic entity is the key concept of this second paper second part . this entity is defined dynamically by a representative containing one @xmath0 defect which is formally labeled by one boltzmann constant @xmath25 from an internal quantum mechanical experiment of nyquist ( 1928 ) , cf . the remark of section 3.1 . the internal experiment is a self - experiment in the liquid @xmath7 without any reservoir and without any conscious observer @xmath7 that is thermodynamically described by the fluctuation dissipation theorem ( fdt ) . this self - experiment may be understood as a common filter construction of the hidden charge model @xcite , common in relation to several particles , or to quantums in a nyquist transmission line . in the model , a common filter construction from several eigensolutions must be allowed for the construction of e.g. a baryon : from a 1-class lepton , a 2-class prebaryon and a 3-class confinon eigensolution . the entity is , so to speak , a common thermodynamic culminating event of the model . the term defect diffusion was introduced by glarum @xcite . an experimental retardation function that could later be identified with the characteristic function of the symmetric levy distribution was detected by kohlrausch @xcite in 1847 . many experimental hints of the connection of glass relaxation data with an underlying levy distribution where collected , e.g. by ngai @xcite . a theoretical review was e.g. by shlesinger @xcite . the fischer dispersion zone @xcite and his speckles @xcite are reviewed by himself . the levy character of the defects was further developed in the book @xcite containing also the representativeness theorem . the fdt as an equation for thermodynamic self - experiments was developed on the base of the nyquist model @xcite , first in 1982 @xcite and later sophisticated in @xcite . glass transition arguments were used in cosmology e.g. in the paper of she @xcite . the hidden charge model is discussed in part 1 @xcite . in classical liquids , local glarum - levy defects ( @xmath0 defects ) are necessary , densily neighbored dynamic phenomena . the levy property is a consequence of the representativeness theorem ( section 3.1 ) , their locality ( finite volume , equation ( [ eq.2.2 ] ) below ) is a consequence of the local high - frequency break throughs they are ( levy instability @xcite , otherwise a general negative pressure would be obtained ) , and the dense arrangement is a consequence of their mutual coexistence condition emerging from spatial aspects of the individual levy damping factors ( @xmath26 ) . the @xmath0 defects are connected with thermodynamic (= internal quantum mechanical ) self - experiments . @xmath7 on the other hand , the rational gibbs thermodynamics is not the consequence of such a self - experiment , if a large external heat reservoir @xmath7 i.e. a macroscopic apparatus or observer @xmath7 is considered to be necessary . the absence of infinitive attributes for self - experiments means that there are no further restrictions to equilibrium fluctuations . then the @xmath0-defect entity can be connected with an internal temperature fluctuation @xmath27 . the rational gibbs thermodynamics has no such a @xmath27 , because @xmath28 is supposed , and new entities ( besides the particles ) can not be described there . von laue [ @xcite ( 112 ) ] suggested a thermodynamics for two independent pairs of variables ( entropy @xmath29 and temperature @xmath13 , volume @xmath30 and pressure @xmath31 ) that allows to calculate the average temperature fluctuation of a representative subsystem : @xmath32 where @xmath33 is the extensive heat capacity , @xmath34=$ ] j / k , and defines the size of the subsystem . the smallest representative ( in the time scale of the dynamic glass transition ) contains one @xmath0 defect @xmath7 otherwise it would not be representative @xmath7 , its characteristic volume is therefore @xcite @xmath35 where the index @xmath36 means the main transition , i.e. the dispersion zone of the dynamic glass transition ; the small letter @xmath37 is the specific heat capacity , @xmath38=$ ] j / kg@xmath39k . the cubic root of @xmath40 is called the characteristic length , @xmath41 . the characteristic volume can be determined by dynamic heat capacity , e.g. @xcite , similar results are obtained from dynamic compressibility , @xmath42 . for high temperatures ( and frequencies ) a characteristic volume of order the volume one molecule is obtained , for low temperature near the conventional glass temperature @xmath43 , the order of 100 molecules is systematically reached @xcite . this behavior suggests that the characteristic length is a reasonable quantity . if a similar formula would be throwed together @xcite from gibbs , a thermodynamic compliance @xmath44 is obtained instead of a modulus ( @xmath45 in equation ( [ eq.2.2 ] ) , @xmath46 where @xmath47 is now calculated from the @xmath36 transformation interval corrected by freezing - in effects . using the same set of experimental data as for equation ( [ eq.2.2 ] ) @xcite , the @xmath48 values are much larger than @xmath40 from ( eq.2.2 ) @xmath7 reflecting the landau - lifshitz subsystem size [ @xcite , 1 , 2 , 35 ] based on statistical independence as defined by the gibbs distribution ( or static correlation functions ) @xmath7 and do not show a systematic increase with falling temperature . this behavior suggests that the construct ( [ eq.2.3 ] ) is not a reasonable quantity . today , the majority of thermodynamic colleagues does not believe in equation ( [ eq.2.2 ] ) . to prove unambiguously the existence of such an entity , we need another experiment that uses a definite , continuously varying length testing the glass transition zone ( @xmath36 ) at different temperatures @xmath13 and frequencies @xmath49 , e.g. a wave vector @xmath50/length , and which length can be compared with the characteristic length from thermodynamics . static xray or neutron scattering is not suitable , because the appropriate contrast in pure liquids is too low . more suitable is e.g. the dynamic neutron scattering ( dns ) in partly deuterated , otherwise pure liquids . but today , a direct comparison with e.g. dynamic calorimetry ( dc ) is not possible because there is a frequency gap @xcite between the former ( @xmath51 mz ) and the latter ( @xmath52 mz ) method . according to an judgement of leading people in dns , e.g. dieter richter @xcite , and in dc , e.g. christoph schick @xcite , the gap can be closed in some years , in 2012 , say . if the lengths obtained from the two methods would be consistent ( and well outside the gibbs formula ) , then this experimentum crucis would be positive for the new entity . a @xmath53 plot could be used for the comparison . the relevant ( diffusion , relaxation ) part of the dynamic intermediate scattering function gives a levy extended - exponential time @xmath54 @xcite that can be transferred ( via the levy exponent @xmath18 @xcite ) in a frequency @xmath55 . from this , a raster @xcite of dns isotherms in the above plot can be constructed . the ( @xmath56 ) points from dc can be put in this raster , and the lengths from dns-@xmath57 values on the abscissa , say , can directly compared with characteristic lengths from dc . it remains to check the minor uncertainty , how density and entropy responses can be compared . let us explain the concept of _ shaping power of levy distribution _ in more detail than above . consider a common space in two dimensions : the @xmath58 axis be the natural space across the @xmath0 defect and the @xmath49 axis be the frequency axis of the event space of probability . after a common scaling ( shorter modes are quicker , i.e. have a higher probability ) , the center @xmath59 is the middle of the defect with the larger frequency @xmath49 values ( @xmath60 ) and the periphery is the borderline region to the neighbor defect with the lower @xmath49 values . center and periphery of the @xmath0 defect substitute thermodynamically a cooperatively rearranging region @xcite , an entity with low density contrast . in the common space , the particular properties of the levy distribution are differently located . this is called shaping power . the preponderant component @xcite with an influence of order of the tilt ( @xmath61 ) on the whole levy sum of random variables , and the upper , fractal hierarchy @xcite of this ordered sum ( by @xmath62 ) are located at the center ; while the damping factor @xmath26 of the lower hierarchy ( large @xmath62 ) is arranged near the periphery in the entities where the sum breaks down by diving below the neighbor defect . in other words , irrespective of the low static density contrast , the borderline between the entities is defined by the mutual disturbation of the differently centered ( in space ) shaping powers of neighboring @xmath0 defects ; the borderlines define the characteristic volume ( [ eq.2.2 ] ) . since the gauss distribution ( @xmath63 ) has no particular properties such as hierarchy etc . , a shaping power can not be defined for gauss and gibbs : the shaping power is for the dynamic heterogeneity . the levy distribution density @xmath64 is thus realized by a spectral density @xmath65 in the measure @xmath66 with @xmath67 . for large frequency at the @xmath0 defect center we find fractality , @xmath68 the center is then connected with the local break through of the mobility ( levy instability ) . the preponderant component corresponds to the diffusion step of the molecule through a cage door of neighbor molecules in the center . as known from probability theory @xcite , for @xmath24 the levy exponent @xmath18 is often the only important parameter of the limit distribution . theoretically , the levy exponent @xmath18 must be realized during a renormalizing limit process of the sum , @xmath69 . the levy exponent @xmath18 depends on a plurality index @xmath70 that depends for a given defect on the kind of disturbance / response , on temperature , and pressure . the reason for plurality is that the numbers @xmath62 for the levy sum of a @xmath0 defect are not so large as e.g. the huge number of filter elements for the renormalization of charge and mass of elementary particles , where a pure huge - number approximation seems to be a reasonable approach ( cf . this part , sections 5.2 and 5.3 there ) . as a consequence of this picture , the local diffusion is determined by the fractality , as observed by dns @xcite . a sublinear _ levy diffusion _ is obtained as @xmath71 having the cauchy limit @xmath72 for _ normal diffusion _ , @xmath73 ( @xmath74 being the relevant length ) . remark . very general reasons are necessary to apply the above shaping - power scenario to a cosmological initial liquid . the representativeness theorem @xcite shows that the levy dynamic heterogeneity of the dynamic glass transition ( @xmath36 ) can even be derived from arguments of classical thermodynamics alone . this theorem says that the dynamic compliances have levy - distributed times with a levy exponent @xmath75 , whereas the corresponding static spatial correlations have a gauss distribution ( @xmath76 ) . roughly : the levy distribution follows from a general limit theorem of probability theory ( excluding others than gauss and levy , cf . feller in @xcite ) , since fluctuation ( variance ) and expectation is infinite due to the levy instability in the cage . this is a new entity , thermodynamically a minimal representative with one @xmath0 defect ( representative for the whole system , cf . the equation ( [ eq.2.2 ] ) arguments ) ; the infinities follow from the negative pressure inside , especially @xmath75 follows from the infinitive expectation of the diffusion step . since the probability density of classical systems is symmetric , the characteristic function is proportional to @xmath77 leading directly to the experimental kohlrausch function . the use of compliance ( instead of a modulus ) follows from the relation to the additivity of the ordered levy sum via the fdt as measuring equation . one of the greatest surprise in classical glass transition research was the discovery that the debye bche inhomogeneities in frozen glasses debye correspond to a slow dispersion zone . these fischer modes @xmath78 have relaxation times e.g. 10@xmath79 times longer and mode lengths e.g. 50 times larger than those of the dynamic glass transition ( @xmath36 ) fischer , patkowski . the corresponding fischer speckles contain , according to the length ratio @xmath80 , a large number @xmath81 of @xmath0 defects . the origin of such large numbers is the next task . we assume @xmath7 against wild structural speculations in the literature , e.g. fischer clusters @xmath7 that the fischer modes are a direct consequence of the dynamic glass transition @xmath36 without any structural input @xcite . then we have to look for a physical @xmath0 defect process that eats its way through large space @xmath74 and time @xmath82 ranges : a levy diffusion process equation ( [ eq.3.2 ] ) . we expect , therefore , that the tilt ( @xmath61 ) with the levy exponent @xmath18 defines the large number @xmath81 ( figure 1 ) . process length ( @xmath83 ) and time ( @xmath84 ) in a diffusion plot , log@xmath82 vs @xmath85 , using the relevant tilt ( @xmath61 ) for the glarum levy defects ( @xmath0 defects ) of the main transition ( @xmath36 ) . forward = levy ( @xmath86 ) , backward = normal ( @xmath87 ) , cf . figure 2 below . [ fig1 ] ] forward , how probes the fast @xmath0 defect process the slow @xmath78 process ? the @xmath0 defects have a large density , in classical liquids ( @xmath88 ) from equation ( [ eq.2.2 ] ) of order @xmath89nm@xmath90 . the molecular diffusion is therefore step by step with short times @xmath91 and short lengths @xmath92 , from one @xmath0 defect to the next neighboring one along a diffusion line ( figure 2a ) . we find therefore the levy diffusion ( [ eq.3.2 ] ) as observed @xcite , along to full distances of the @xmath1 speckles by analytic continuation with the sublinear slope . ) and backward ( @xmath93 ) probes . details see text . [ fig2 ] ] backward , how probes the slow @xmath78 process the fast @xmath0 defect process ? in the large @xmath78 time order @xmath84 , the fast dynamics of the small @xmath36 subsystems are mutually statistically independent ( for @xmath94 ) . the many @xmath36 subsystems allow a quasi thermodynamic treatment . the thermodynamic variables @xmath95 have a finite variance . the relaxation can then be described by @xmath96 const @xmath97 , with a positive constant . this gives the exponential decay leading via a cauchy law ( @xmath72 ) to normal diffusion in equation ( eq.3.2 ) . consequently , we find the space and time scales for the @xmath1 speckles ( @xmath98 ) at the upper ( @xmath78 ) corner of the triangle in figure 1 : the levy diffusion , to be effective , must be faster than the normal diffusion . in larger scales , beyond the @xmath78 corner , the construction is exhausted , which means that the @xmath1 speckle defines a largest causal region in the classical liquid . to get a formula , we start from the triangle basis @xmath99 . this is the logarithmus - of - time difference between the normal and levy diffusion at the @xmath0 defect length scale @xmath92 . the levy diffusion is at high fractal frequencies near the @xmath0 defect center at the cage door , i.e. at some @xmath100 in the short - time tail of the @xmath36 process . the normal diffusion , however , is near an average of time ( in logarithmic measure ) , @xmath101 : @xmath102 i.e. @xmath99 is of the order @xmath103 , the width of the response function on the @xmath104 axis ( figure 3.7b of @xcite ) . figure 1 shows that the length and time ratios ( @xmath80 ) are mainly determined by the levy exponent tilt of the @xmath0 defect dynamics , ( @xmath61 ) . using @xmath105 , we get the slope difference ( forward minus backward ) as ( @xmath106 and the ratios as @xmath107 both ratios tend to infinity for vanishing tilt , @xmath108 ( or @xmath109 , cauchy ) . large ratios can therefore easily be obtained for small tilts , also such ones as observed in the classical liquids @xcite . example . the number @xmath81 of @xmath0 defects as claimed by the fischer speckles as largest causal region in @xmath110 dimensions is @xmath111 for \{@xmath112 , @xmath113 , @xmath114 } we get @xmath115 ; this @xmath115 value can also be obtained e.g. for @xmath116 and @xmath117 , and so on . for cosmological applications ( section 6 ) , we should discuss the relation of the @xmath0 defect tilt ( @xmath61 ) to the scalar index tilt ( @xmath22 ) from the power law fit of the density fluctuation spectrum . it seems useful , however , first to devote the next sections 4 and 5 to gravitation and inflation of our model . our model does not have an eigensolution for gravitons . if the graviton is some type of charge ( gravitation as a charge is discussed by feynman feynman ) , and if a charge is characterized by at least one @xmath119 torus in the model , then an extra torus beyond the eigensolutions must be constructed : a circle @xmath119 from four @xmath120 segments of @xmath121 may be linked by four random connections called spots ( @xmath122 in figure 3 ) . the mass concentration into galaxies can give spots in the particle environment , if the culminating - point filter construction from large to small filter elements is considered as some map of the universe in the mach s principle layer of the tangent objects ( @xcite , table 2 of appendix c there ) . e.g. the central black holes of galaxies may give the randomly selected linking spots in the dollhouse universe around any particle ( section 7 ) . @xmath121 segments in the tangent objects of any particle . [ fig3 ] ] the number of spots is equal to the number of galaxies in the universe , @xmath123 . assuming some statistical independence in the random selection of the four spots , we get a probability for the graviton of about @xmath124 this probability may be compared with the probability of order @xmath125 to find a charge torus for the other particles participating in an interaction considered . this means that the gravitation is , for given distances , weaker by a factor of order 10@xmath126 than the electroweak and strong interaction . a review of our model constructions for particles ( eigensolutions [ this part , section 3 ] , for common eigensolutions [ @xcite , section 5 ] , for exchange boson construction ( @xcite , appendix c ) , and for the dollhouse universe ( section 7 , below ) seems to show that now the application of torus constructions for particles is exhausted ( table 1 ) . our model is now complete . .list of @xmath127 torus topologies defining charges of our model.[tab.1 ] [ cols="^,^,^ " , ] via an adapted van der waerden ( vdw ) construction @xcite , the gravitation torus of figure 3 defines a metric tensor @xmath128 with spin @xmath129 on the reality space from @xmath9 tangent objects . because the gravitation is so weak and stems from the far universe , we can speak about a _ shadow metric over the reality _ @xmath9 _ space _ , if the particles or mass concentrations are widely isolated in a vacuum ( much vacuum ) . in our model , quantum - mechanically behaving cold hadrons are not realized before the expansion of the universe reaches the length scale of order @xmath130 fermi . moreover , in this scale the first black holes are formed in the center of @xmath0 defects . the empirical mass scale @xmath131 above the neutrinos is from mev to tev , without the electrons we find a scale from gev to tev ; inclusive the stable atoms . in sufficiently large @xmath9 tangent objects , we have a compton wave - length of @xmath132 assuming that only the heavy particles are of cosmological influence for a cold expansion , then the energy scales for quantum flavor dynamics ( qfd ) and chromo dynamics ( qcd ) have length scales of @xmath133 shortly , we call this generous region hadronic : @xmath134 with @xmath135 fermi @xmath136 gev and the proton mass @xmath137 as typical parameters . for high temperatures , @xmath138 , the de broglie wave lengths are much shorter . this is called hot ; our hadronic scale ( [ eq.5.3 ] ) is then called cold . we try to describe the expansion of the universe as a cold phenomenon , because our model mass scale is confined from above by existential instability ( generalisation of figure 5 in @xcite ) . consider in advance a box with a length @xmath139 ( @xmath139 in meters ) . without @xmath140 or @xmath141 separation , the box _ must _ be filled with particles that _ must _ be hot for small boxes @xmath142 ( because of the quantum mechanical uncertainty relation ) . with @xmath143 or @xmath141 separation , however , an empty small box is possible : a culminating - point particle at @xmath58 supposes the existence of small filter elements @xmath144 . this is not possible for cold particles in case of @xmath142 , because then smaller elements ( @xmath145 ) are not possible due to the upper limit of the hadronic region ( equation ( [ eq.5.3 ] ) , see also figure 6 of @xcite ) . the uncertainty relation follows from @xmath146 assuming the possibility of existence of a particle at @xmath58 as the basic prerequisite . this _ must _ be assumed for a manifold diffeomorphism ( without @xmath143 separation ) , but this is excluded for cold particles in the too small boxes ( @xmath142 ) from our independent culminating - point particles ( with @xmath143 separation ) . let us now specify the concept of _ filter elements _ sometimes used above . since a filter is considered as some kind of a map from the hidden @xmath147 in the reality @xmath148 space , it contains things from the origin @xmath147 ( e.g. vacuum elements ) and the image @xmath148 ( metric for @xmath9 or @xmath149 tangent objects ) . our mathematical filter sets contain , therefore , a certain size variation and get some relation to a cauchy filter ( figure 1b of @xcite ) . the filter sets of vacuum elements resulting from existential instability are therefore called filter elements . the method used for our inflation scenario is the _ cutoff filter _ of figure 4 . the filter elements on the @xmath9 tangent objects for cold model hadrons ( in the vdw spinor space ) should not become smaller than the hadronic scale @xmath4 , because then the compton wave length ( as condition for their cold quantum mechanical existence ) is not available for virtual filter elements in the too small universe ( @xmath5 ) . such cold quantum mechanical particles can not exist as points there , but only as larger constructs of the model , hidden by the eigensolutions on the @xmath150 space ( @xmath151 ) , and lurking there for their convergence to particles in the reality space for @xmath152 . in other words , for @xmath5 there is no room enough to realize a cold quantum mechanical particle ( a hadron ) in the reality space ( cf . with the box discussion above ) . and @xmath149 tangent objects . * a * ( left ) . for cold particles we find a cutoff filter for @xmath153 . the cutoff at @xmath6 has a negative pressure from creation of many hadrons during expansion . * b * ( right ) . for non - isolated dark matter particles in an @xmath149 euclidean environment , the filter is not cut and we find always positive pressure , @xmath154 . if , however , because of expansion , dark matter particles become isolated , @xmath7 i.e. if they will be imbedded in minkowski space @xmath7 then they are also to be treated by the left hand side cutoff scenario ( a ) . [ fig4 ] ] this holds also for 2-class particles ( e.g. dark matter or prebaryons ) with internal @xmath149 tangent objects , if they are isolated by imbedding in an external , sufficiently large minkowski space from @xmath9 tangent objects of the other particles or other participants of common culminating points . this does not hold , however , if they are not isolated , i.e. if they are living in an @xmath149 environment from tightly neighboring particles of 2-class eigensolutions , e.g. from dark matter particles . the missing time in @xmath149 does not allow to define the planck constant @xmath155 needing the seconds . then the filter is not cut ( figure 4 ) , and cold dark matter particles ( and prebaryons ) can also exist at high density for @xmath156 . dark matter particles are preferred in the small universe . in other words , if the cosmological expansion of a cold initial dark - matter universe crosses the hadronic length scale , @xmath6 , then many hadrons ( and hadron pairs ) are formed which were absent before in the reality - space of the universe . this is a typical situation for negative pressure , @xmath157 ( figure 4 again ) . [ remember : thermodynamically , the pressure @xmath31 is one of the variables that are necessary for maintaining the considered equilibrium state . for representative subsystems with length @xmath139 , here for @xmath158 , this pressure @xmath31 would be the external pressure . there is a widespread misunderstanding , that negative pressure can be defined by radially escaping particle trajectories . such behavior would also be typical for positive external pressures @xmath159 with @xmath160 . ] during a period of forming many new particles , controlled by the lagrangian , the ( stationary ) state can only be maintained , if the pressure @xmath161 is negative , @xmath157 , and the system ( our universe ) , therefore , is expanding . this expansion stage directly driven by hadronic forces is called _ hadronic inflation_. if this is connected with a hadronic cosmological constant , suitably compared to gravitation , we would get the factor 10@xmath162 of equation ( [ eq.4.1 ] ) . the cold hadronic inflation is therefore an extremely violent stage of expansion without hot x particles from a gauge unification at 10@xmath163 gev . the resulting scenario for the hadronic inflation is characterized by the following concepts ( figure 5 ) . the infinitely large liquid system ( primeval soup ) is called original , because it exists also before the universe comes into being as consequence of a linde type of fluctuation . the developing small ( @xmath156 ) universe is called initial , because it is related to its beginning and its independence ; it is a first step in a series of processes , developing from a linde fluctuation in a maximal causal region claimed by an @xmath1 defect of the original liquid . the violent increase near @xmath6 is called hadronic inflation as defined above . immediately after this violent hadronic pulse , the universe can not be controlled by the too weak gravitation . later , however , for @xmath164 , the expansion control is captured by gravitation and is then called primordial . ] the stages of the scenario are discussed according to our model with respect to the crumbled space ( mix of @xmath165 ) , to gravitation , and to expansion . _ original state_. assuming a large volume of the original liquid , we get point particles from 1- , 2- , and 3-class eigensolutions via the non - virtuality of the small filter elements , e.g. photons , hadrons , and dark matter point particles . the high density corresponds to a crumbled space of @xmath166 and @xmath149 tangent objects . although @xmath0 defects seem to bee possible , we get no gravitation , because their number @xmath81 tends to infinity in the large volume ( @xmath167 ) and , additionally , the links for gravitons ( @xmath168 in figure 3 ) are expected to be too weak because of their weak contrast . dynamic @xmath1 speckles as a consequence of @xmath0 defects should still define maximal causal regions for linde fluctuations . _ initial liquid_. since only the dark matter particles and prebaryons remain points for @xmath5 , figure 4b ) , their fraction in the expanding linde fluctuation should be large . we expect a fluctuation ( boltzmann zacke ) with many dark matter particles in the high - density liquid , and with no isolated hadrons because of virtuality in their cut off filter . if there are cold photons ( no @xmath169 ) but too weak for an imbedding @xmath9 space , we extremely get an euclidic @xmath149 space , or at least a crumbled space with a high @xmath149 fraction . in the extremum , time seems to be standing still , and we find only a shape of time , some succession of fourdimensional euclidic @xmath149 spaces . the reason for the expansion in this stage is therefore highly speculative also in the frame of our model and rests on a certain cold photon fraction . e.g. , one possibility would be the vanishing of the original @xmath9 particles ( hadrons ) due to their transfer into the virtuality in such a boltzmann zacke ( figure 4a ) . this could mean a positive pressure @xmath159 , @xmath170 , in the universe for @xmath171 . the total topology of the whole universe from the @xmath1 speckle may be formed in this expansion stage . @xmath3 is the radius of the emerging universe in meters . the topology may be induced from the two hidden heegaard tori of the charge symmetry and from imbedding of @xmath3 in a crumbled space with dominating fourdimensional @xmath149 tangent elements . this is , i think , a situation that favors a total @xmath172 sphere instead of a @xmath173 disk , irrespective of local curvature fluctuations inside . _ hadronic inflation_. for @xmath6 the hadrons come into the play . this generates the negative pressure leading to a violent expansion transferring the dark matter particles in the isolation . they are then imbedded in the minkowski space allowing their quantum mechanical behavior for @xmath164 . if the @xmath0 defects from the zacke are intensified to become the @xmath168 links for the gravitons , then gravitation is switched on during the inflation . this intensification could be possible by their transfer to black holes ( section 5.3 ) . [ this would mean that the galaxies come finally from the shaping power of levy distributions , section 3.1 . ] _ primordial stage_. if we put on a material variant then the emerging particles drive the expansion , not a vacuum field : this stage develops from the motion of interacting hadronic particles . in the inflation stage ( @xmath6 ) and for high particle density in the cold universe , the gravitation has no or at most a minor influence ( 10@xmath174 , equation [ eq.4.1 ] ) . in the further expansion , we get increasing distances between the particles and particle congregations . for electric neutrality , the vacuum gets influence via the gravitation ( section 5.3 ) . a control by gravitation , however , is only possible for not too small @xmath3 . the expansion can then be captured by the friedmann lamaitre ( fl ) equations , containing besides the newtonian gravitational constant @xmath175 the mysterious cosmological parameter for gravitation , @xmath176 ( section 7.2 ) . this material variant has no superluminar particle velocities and is based on special relativity in the minkowski space from the @xmath9 tangent objects . we have no serious horizon exit and reentry problems during and after the violent pulse from cold inflation . our material variant is an alternative to the conventional vacuum variant : the superluminal velocities there may result from requirements to the fl equations , e.g. stationary during inflation : @xmath177 constant , density @xmath178 const , and a field @xmath179 for the vacuum @xmath30 . let us repeat : it is the field @xmath78 that is considered in the vacuum variant , not cold particles with mass in the material variant as used here . the hawking temperature @xcite for black holes , @xmath180 connects gravitation ( @xmath175 ) , hole mass from particles ( @xmath148 ) , photon ( @xmath181 ) and thermodynamics ( @xmath25 ) . equation ( [ eq.5.4 ] ) is sometimes considered as the basis for an unification of gravitation with the other interactions . using , however , conventional gauge invariance principles , the unification with gravitation is usually obtained in a hot universe that is characterized by temperatures or energies of order the planck mass ( @xmath182gev ) . we try to find a cold alternative in the frame of our model . consider the @xmath0 defect as representative thermodynamic entity ( from which the one @xmath25 for the hawking temperature comes ) . the physical process for the equilibration of temperature is now the exchange of particles ( mass @xmath183 ) at the horizon of the black hole . @xmath7 we consider first the particle concentration inside the @xmath0 defect . in cosmological approximation , the extensive entropy is equal to the number of particles in the mass concentration of the @xmath0 defect , identified with @xmath148 : @xmath184 the corresponding bekenstein hawking entropy @xcite for the black hole is @xmath185 we ask where the two formulas cross each other ( @xmath186 ) , equation ( [ eq.5.5 ] ) for the @xmath0 defect mass concentration without a black hole , and ( [ eq.5.6 ] ) for a black hole formed by this mass concentration of the @xmath0 defect ( figure 6 ) . -defect mass concentration ( @xmath187 ) with a horizon area entropy ( @xmath188 ) for a corresponding black hole . the drawn lines are from an entropy maximum principle , the broken lines are extrapolations . @xmath189 is the cross length scale related to the schwarzschild radius of the black hole , @xmath190 . below @xmath191 of order the hadron compton length ( equation ( eq.5.9 ) below ) , the particle concentration of the @xmath0-defect with no black hole is the optimum . [ fig6 ] ] using the schwarzschild radius @xmath192 we get from ( [ eq.5.5 ] ) and ( [ eq.5.6 ] ) for the crossing of figure 6 @xmath193 where @xmath131 is the mass of particles considered . then we have @xmath194 schwarzschild radius and hawking temperature at the formation of first black holes are in the hadronic region . the number of particles with mass @xmath131 is @xmath195 for protons ( @xmath196 ) we get a mass @xmath197 of the first black holes of order 10@xmath198 kg , for dark matter particles ( @xmath199 ) of order 10@xmath200 kg . this mass contains about 10@xmath201 protons or 10@xmath202 dark matter particles : gigantic numbers . the reason for such numbers is the small ( 10@xmath203 ) gravitational constant in the denominator of equation ( [ eq.5.10 ] ) . a black hole can only come if the particle numbers ( and their density ) are gigantic . we see also that any galaxy gets a central black hole during inflation . there is no danger in the lhc experiment that such cosmological black holes can artificially be generated . the corresponding @xmath204 energy is of order 10@xmath205ev ( calculated with protons ) , whereas lhc gets maximally 10@xmath206ev . higher dimensions as from string theory for small black holes are excluded in our model . our earth experiments can not generate a second central black hole in our galaxy . the crossover mass ( [ eq.5.8 ] ) for protons ( 10@xmath198 kg ) is of order 10@xmath207 ( @xmath208 is the sun mass ) and is much smaller than the mass of the today central black holes of the galaxies ( 10@xmath209 ) . there are two reasons for this smallness in our scenario . the first black holes from mass concentration are in the center of @xmath0 defects smaller than the @xmath0 defects themselves . in a liquid , the center is at the fractal tails of their levy distributions at large frequencies @xmath49 ; the preponderant component of the levy distribution is a part of the mass concentration . in the @xmath149 space , this corresponds to large wave vectors @xmath210 ( section 6 ) . this means that the first black holes are formed at a small part not only of @xmath0 defects , but also a very small part of the much larger @xmath211 speckles defining @xmath3 of the universe . the first black holes are formed at diameters @xmath3 larger than @xmath4 , i.e. at a later stage of hadronic inflation . since @xmath212 is decreased during the inflation ( figure 6 ) , the growing black hole enters at some time a nonequilibrium state , and the further growth becomes irreversible . this is a self - amplifying process , because larger black holes become stronger spots @xmath213 in figure 3 which strengthens to linking of @xmath121 graviton segments increasing the gravitational constant , @xmath214 . there is much enough material left for forming of the galaxy structure . a structure can also be formed in the early primordial state where essential ingredients such as mass concentration and central black holes are already available . in formulas , the results can bee summarized as @xmath215 if the optimum is really defined by the maximum of entropy from the boltzmann probability , @xmath216 , we conclude 1 . the first black holes of the expanding universe are formed in the spatial center of the @xmath0 defects during later stages of the hadronic inflation , ( where @xmath217 is reached ) . these black holes are formed in thermodynamic equilibrium ; later in the larger expanding universe , the black holes fall out of the equilibrium , and their growth becomes irreversible and is mechanically controlled by the newtonian gravitational constant @xmath175 . in our model scenario , gravitation is unified with the other three interactions insofar as they are relevant to our cold hadronic inflation : via the hawking temperature equilibration in the hadronic scale @xmath218 of the expansion , @xmath6 . this is called cold hawking unification . the gravitational interaction comes into being not before the inflation : gravitation is first realized by strengthening of torus links @xmath219 for the graviton torus of figure 3 . from a methodical point of view , the cold unification corresponds to some calibration of mass values for leptons and hadrons . the filter construction ( @xcite , section 5.3 there ) calculates sharp mass ratios , e.g. @xmath220 , that can be related e.g. to @xmath221 as electron or to the proton mass . absolute values can be obtained by comparison with gravitation in the primordial state of the universe . the gravitational constant @xmath175 , however , has some fundamental randomness from the number @xmath81 of @xmath0 defects inside the fischer speckle ( equation ( [ eq.4.1 ] ) and figure 2b leads to the relative order of @xmath222 for the fluctuation variance ) , whereas e.g. the parameters of the electroweak interaction are sharp from the huge numbers of the filter construction . the way from a levy tilt ( @xmath61 ) in the original liquid to the tilt ( @xmath224 ) of cosmic density fluctuation in the primordial stage is discussed in this section . first three model priors : * a. * in the classical liquid , the number of @xmath0 defects ( @xmath81 ) in the maximal causal region ( @xmath1 speckle ) can bee estimated using the relevant levy exponent tilt ( @xmath61 ) with @xmath72 for the classical cauchy diffusion ( equation ( [ eq.3.5 ] ) , figure 1 ) . small tilts ( @xmath61 ) correspond to large numbers @xmath16 . the estimation is based on an analytic continuation of the levy diffusion from the @xmath0 scale to the @xmath1 scale ( figure 2 ) . * b. * the initial liquid before the hadronic inflation is dominated by euclidic @xmath149 tangent objects ( section 5.1 ) . to get a tilt in the initial liquid , the shaping power from a classical diffusion levy instability by free volume may be substituted by the shaping power from an @xmath149-potential levy instability in the initial liquid . the elliptic type of interaction between dark matter particles favors high densities in @xmath149 instead high frequencies from large local free volume in classical liquids . * c. * the first black holes emerge from a mass concentration in the center of @xmath0 defects of the initial liquid ( section 5.2 ) . now three changes of the tilt . first change . during the way from the original classical liquid in @xmath9 to the initial liquid in @xmath149 the time is lost ( hawking s shape of time ) . instead of a frequency @xmath49 for diffusion ( besides three spatial dimensions ) we have then a fourth wave vector component for density fluctuation in four spatial dimensions . we assume that the levy exponent @xmath18 for the levy diffusion is transferred into a tilt in the new @xmath225 direction . since @xmath149 is symmetric , the transfer effect must be distributed in all the four @xmath225 directions . second tilt change . in the further way , the loss of time can not be absolute ; we need a time for the definition of expansion @xmath226 . in the speculation of section 5.1 , this time is connected with photons as massless points from the filter elements . there remain only three @xmath225 directions which means a second change of the tilt , giving @xmath227 for a density fluctuation before the inflation . third tilt change . the next stage of the way is the hadronic inflation up to the primordial stage from which the cosmic microwave background cmb is measured today . the primordial tilt ( @xmath22 ) follows after this third change ( @xmath228 ) during the inflation . assume that , via the above three changes , a small levy tilt ( @xmath61 ) in the original liquid results in another small but finite tilt in the primordial state , measured as the deviation ( @xmath22 ) from the zeldovich harrison scaling @xmath229 . then we get from the three above priors , the three changes , and equation ( [ eq.3.4 ] ) , @xmath230 where the three changes are collected by the factor @xmath231 , @xmath232 . i think that @xmath233 . equation ( [ eq.6.1 ] ) shows that a finite universe ( containing a finite number of galaxies @xmath234 ) as consequence of a linde type expansion can only exist for a finite tilt ( @xmath235 ) . for ( @xmath236 we would have @xmath237 and , because of equation ( eq.4.1 ) , the gravitational constant tends also to zero , @xmath238 . without the finite tilt , i.e. for zeldovich harrison scaling of density fluctuation , we would get an infinitely large universe ( @xmath239 ) with no gravitation . the average cmb temperature fluctuation @xmath223 can be estimated from gauss statistics of the smallest representative subsystems , i.e. from the number of galaxies , @xmath81 . the consistency of levy distribution for estimation of the size of universe and the gauss distribution for smaller ( and therefore faster ) subsystems is explained in the text around figure 2 . the size of @xmath223 can therefore estimated by @xmath240 in the model , the finiteness of the cmb temperature fluctuation components @xmath241 follows from the finiteness of the universe ( @xmath242 ) and is , from equation ( [ eq.6.1 ] ) , related to a finite tilt in the density fluctuation . the model allows to treat the cosmological constant @xmath176 exclusively on the golden side of the einstein equation ; the dark energy can be reduced to the flatness of the underlying @xmath9 tangent objects giving @xmath243 . this degrades the dark energy density parameter to @xmath244 figures of arithmetics of the golden side . the main idea is : feeding the torus construction for the graviton ( figure 3 ) into the algebraic vdw spinor construction with physical coordinates for the tangent object @xcite defines the manifold diffeomorphism with @xmath245 for the einstein hilbert action . this gravitational @xmath246 is very weak ( 10@xmath203 ) when compared with the electromagnetic nature of the flat @xmath166 tangent objects in a state with much vacuum . it is therefore the geometric golden side of the einstein equation which determines flatness , not the energetic stony side . the filter elements mediate some kind of map between the universe and the immediate environment on the @xmath9 tangent object of any isolated culminating point particle ( section 5.1 ) . because of conformal invariance of the hidden space , the large filter elements pass over the whole universe down to the small elements of size the particle ( s compton wave length ) . assume that the central black holes of galaxies are mapped into spots on the tangent objects , we get a _ dollhouse universe _ around any particle ( figure 7 ) . these spots are related to the links of gravitons ( @xmath247 in figure 3 ) and can therefore be included in the einstein equation . since the gravitation is a small interaction ( equation [ eq.4.1 ] ) , we find for isolated objects an additional shadow metric as varying part of @xmath248 over the flat @xmath9 tangent objects . this defines a new , the mach principle layer of the tangent objects ( @xcite , appendix c there ) . defects ( more precisely : the galaxies emerging from the @xmath0 defects of the original or initial state ) of the reality universe via the filter elements into the @xmath0 spots on the ( corresponding layer of the ) tangent objects around any culminating point particle : the dollhouse universe . [ fig7 ] ] if this dollhouse universe from the filter is put on the stony side of the einstein equation , we would get some field @xmath78 that is neither independent nor a quantizable higgs field . it is located wherever a particle is located , we do not find this field without a particle there . it suits , however , to the equivalence principle of the gravitation theory , since it represents the same situation around any particle . from the viewpoint of a map from a finer model ( original ) to a coarser model ( image ) , the priors of the latter are image constraints of the map ( this part , section 4 there ) . the finer model may now be characterized by \{flat @xmath9 tangent objects around isolated particles or matter congregations , dollhouse universe , shadow metric } , the coarser model by \{total manifold with @xmath248 metric , priors of cosmological standard model}. we consider the primordial stage ( and later ones ) of the universe : _ much vacuum _ , i.e. only flat and very large @xmath9 tangent objects around culminating - point constructions such as leptons , hadrons , dark matter particles , baryons etc . , and also around particle congregations such as stars and black holes . we try to reduce the dark energy problem to flatness of the underlying @xmath9 tangent objects in the vacuum of the universe . a prior of the general theory of relativity ( gravitation theory ) is the @xmath250 field that is described by the einstein hilbert action , @xmath251 where @xmath252 is then a einstein riemann metric assumed to be a manifold for all things of the cosmos . @xmath253 is the ricci scalar curvature . from equation ( [ eq.7.1 ] ) follows by a variation @xmath254 and by addition of a material term ( @xmath255 ) the einstein equation . it occurs in two forms , a geometric form with the cosmological parameter @xmath256 on the golden side , or an energetic form with @xmath176 on the stony side : @xmath257 @xmath258 flatness is a concept for @xmath176 on the golden side ( curvature is diminished by @xmath259 for given @xmath255 ) , dark energy is a concept of the stony side ( energy is raised by @xmath259 ) . [ in the literature , the golden side is sometimes called a modification of gravity . ] absolutizing the metric diffeomorphism , both sides can not be distinguished . the @xmath143 separation @xcite , however , can distinguish golden from stony , the two equations are not longer equivalent . the additional prior of the conventional standard cosmology is the copernican principle ( total homogeneity , isotropy , and cosmic time ) that usually is condensed in a robertson walker metric for the supposed manifold , @xmath260 with @xmath261 as indicator for the gauss curvature of \{@xmath172 , @xmath262 , @xmath263 } , @xmath264 lobachevsky . at the moment , @xmath225 is a symmetry parameter of a total copernican solution , not of the einstein equation . the total copernican prior for the standard model of cosmology are therefore the friedmann lamaitre ( fl ) equations , @xmath265 , \label{eq.7.5}\ ] ] @xmath266 where ( @xmath267 ) are the energy density and the pressure of matter and radiation . the expansion is accelerated by the dominance of @xmath176 , with the consequence that the hubble rate @xmath147 increases with larger @xmath259 . in the language of the stony side , an energy is connected with @xmath176 , the dark energy . for a time @xmath268 in the primordial or a later stage , e.g. the expansion stage today ( why me , why now @xcite ) , the density parameters for matter and radiation ( @xmath148 ) and for @xmath176 are defined as @xmath269 with @xmath270 and @xmath271 . then the space curvature @xmath225 is determined by the mysterious dark energy density @xmath272 . for decreasing @xmath273 and given @xmath274 , the dominance of dark energy @xmath275 starts at a certain time @xmath276 ( now ) ; @xmath277 for large dominating @xmath278 is also possible for a flat space . experimental flatness means a local ( not the total ) property @xmath279 , i.e. @xmath280 . the total @xmath172 or @xmath263 topology may be consistent with small local fluctuations near local flatness , @xmath280 , especially for sufficiently large @xmath3 and large @xmath281 ( very much vacuum ) . in our language of the golden side , much vacuum between widely isolated phenomena is necessary for a measurable influence of @xmath176 . in these large primordial scales , the vacuum is based on the @xmath9 tangent objects from the biquaternion construction @xcite . they are flat because the hidden electromagnetic structure is conformal . - photons and gravitons are the only massless ( free ) particles in the empty space . the cosmological photons in the shadow metric are not bound and change , therefore , their frequency during the primordial expansion . the flat @xmath9 tangent objects do not come from photons alone but are bound to the charges of our model ( appendix c@xcite ) . their @xmath9 properties , therefore , do not change during the expansion : they remain exactly flat . the electromagnetic structure of the tangent objects is always much stronger than the gravitons for the shadow ( equation ( [ eq.4.1 ] ) ) . this means that the empty basic space ( much vacuum ) underlying the shadow metric is always experimentally flat irrespective that the curved additional shadow metric is decisive for the gravitational aspects of the primordial universe . both the flatness and the shadow curvature can be placed in the fl equations , i.e. in the priors of conventional standard cosmology . on the golden side , the above stony alternative is then to be interpreted by the local relation @xmath282 . the stony term @xmath271 is not an energy in the golden - side interpretation ; i.e. the stony dark energy is geometrically defined by golden flatness , @xmath283 the universe is experimentally flat since the primordial stage . the @xmath284 term in equation ( [ eq.7.8 ] ) is degraded to figures of arithmetics . there is no extra dark energy in the cosmos and a search for it , e.g. in form of certain higgs particles , is not necessary from the viewpoint of our model . guth s flatness is possible without guts.the destiny of our model universe is local flatness without matter , @xmath285 , i.e. @xmath286 . if the universe has got a total @xmath172 topology ( section 5.1 ) , then the universe in the limit gets practically the state of the hidden charge , conformality without gravitation , and the play can start anew , without any trace from the foregoing universe , but with the same parameters so far as they are determined by the culminating point filters . [ on the other hand , this means , e.g. that the new gravitational constant is not exactly the same , since there is a variance of the number of @xmath0 defects inside the fischer speckles ( of order of 10@xmath287 for @xmath115 , cf . the end of section 5.3 ) ] . * equation of state parameter @xmath288 . introduce a new parameter of the stony side , @xmath289 . because of the huge number of vacuum elements in our model , we have no fluctuation of the stony @xmath78 field . from thermodynamics , especially from the fluctuation dissipation theorem fdt , we get no response from no fluctuation . from the fl equations we obtain , via @xmath290 the relevant response as @xmath291 from no response we have then @xmath288 as announced . this justifies also the only use of @xmath176 alone in section 7.2 . the photon / baryon ratio is of order 10@xmath198 . the pass - over of the large filter elements corresponds to one cosmic interaction diagram between a feynman vertex of the baryon pro one galaxy spot . the spot acts as catalyst , because in the large number approximation there is no feedback from the feynman vertex to the galaxy . the vertex forms a surviving baryon in the model lepton capture reaction with a confinon and a prebaryon ( this part , equation ( [ eq.5.10 ] ) there ) . the surviving rests on a different partial chirality of the participating prebaryon ( @xcite , appendix c ) . since the prebaryon has an electric charge , we assume that there is about one photon pro one interaction . for @xmath115 we obtain the announced ratio . the hidden charge model ( shortly : the model ) is introduced by two concepts : @xcite conceptual separation of mass and energy behind @xmath204 , and @xcite a new entity for an initial cosmological liquid . the model remains in four dimensions , has a clearly arranged particle spectrum confined by the tev scale , and is supplied with a rich thermodynamics . this allows a serious prosy representation and a heuristic approach which two are suited to the early stage of the conventional new standard cosmology . the model is much simpler and more robust than highly sophisticated approaches , e.g. string theory . a hadronic inflation emerges because cold baryons can first be realized in the length scale of @xmath292fermi . for @xmath5 , the filter elements for culminating - point baryons with @xmath9 tangent objects are too small for quantum mechanics in a cold reality . dense dark - matter liquids , however could be realized for @xmath5 because their tangent objects are @xmath149 euclidean allowing points smaller than @xmath4 . the sudden occurrence of baryons ( and hadron pairs ) at @xmath293 generates a large ( hadronic ) amount of negative pressure and therefore a violent pulse for inflationary expansion . during its cosmic time scale the first small black holes emerge that allow a cold hawking unification of the four interactions . the new entities in the liquid are called glarum levy defects ( @xmath0 defects ) . it is finally the shaping power of the levy distribution that forms the galaxies of the universe and allows a relation between their number @xmath81 and the tilt ( @xmath22 ) in the density fluctuation . @xmath239 would mean no tilt and no gravitation . the decrease of the filter element size during forming the culminating points corresponds to a map of the universe in the @xmath9 tangent objects near culminating point particles . in this dollhouse universe are , in accordance with mach s principle , the galaxies mapped to small spots that are used to construct gravitons as charges from four hidden @xmath121 segments randomly connected by the spots . this defines gravitation , in the vacuum as a shadow metric over flat @xmath9 tangent objects as basis . the underlying flatness is based on electromagnetism , much stronger than the weakness of the shadow metric from gravitation ( factor 10@xmath126 ) . this allows a purely geometric interpretation of the cosmological parameter @xmath256 ( that was called dark energy before ) as a consequence of this flatness since the primordial stage of the universe , @xmath294 . not all reality objects are quantized so as usually assumed . the electroweak interaction remains quantized in the form that follows from feynman path integrals . dark matter , if not isolated in the vacuum of minkowski space , is not conventionally quantized ( e.g. in dense liquids ) . somewhere between these limiting cases is the quantization of strong interaction for hadrons , in particular for baryons where @xmath149 tangent objects are inside from prebaryons , and also the gravitation . although gravitons are constructed as charges , the possibility of a further general quantization of the einstein riemann metric manifold must seriously be questioned . dark energy as geometrical flatness does not allow any quantization . some conventional priors important for cosmology are substituted by the model as follows : there are no higgs particles because corresponding eigensolutions are missing ; the filter elements are not independent fields and can not be quantized because they do not fluctuate due to their huge numbers . a hot big bang is not necessary , because cold dark matter particles remain points also at large densities in the initial liquid for they can not be quantized in the euclidean @xmath149 tangent objects there . the cosmological parameter @xmath176 needs not to be an dark energy , because it can be reduced to spatial flatness . @xmath7 in the frame of our hidden charge model , a search for higgs particles or independent dark energy is not useful . the above part of the paper corresponds to arxiv 0901.1050 v2 [ physics.gen-ph ] 20 jan 2009 , apart from twenty - two short adaptions and six corrections of mistakes . it seems useful to explain how this paradigmatic result was obtained . i used a neural network method without computer . the author motivated systematically , since 1964 , the neuron structure in his own head to find possibly new paradigmatic ideas in astroparticle physics . partitions into a growing number of promising `` remainders '' of the known physics in growing specific time were used as changing starting conditions . at the end there are roughly twenty remainders . the iterations of the method were organized only by internal consistencies : completely _ inside _ the paradigm , i.e. consistencies with all foregoing internal consistencies : between the changing remainders , between their binary relations , and the many new relations between foregoing consistencies . most of these internal iterations were documented by the author ( all in german ) : the first with number 1 is of 4 may 1964 ( physical point and its environment ) , the last with number 1381 is of 19 oct 2003 ( @xmath295 torus for graviton ) . several hundreds of additional iterations became later implicit parts of rejected papers ( about a dozen papers ) . two further examples . * 1*. a new spin statistic theorem ( sst ) was firstly documented with number 352 of 16 may 1970 with headline sst . its derivation from a threedimensional rotation excluded later susy ( supersymmetry ) ; see part i , sect . 4.1 , table i , and eqs . ( 7 ) , ( 8) , and ( 9 ) there . * 2*. the mass operator was indicated in number 899 of 8 dec 1986 with headline pair generation , and was documented in number 948 of 9 aug 1987 with the headline : difference between charge ( coordinates @xmath296 and @xmath297 and mass ( coordinates @xmath295 and @xmath298 ) , see again part i , section 4.1 there . the mystery of the higgs particles and their connection with the flatness puzzle compensation in the universe ( m. veltmann , facts and mysteries in elementary particle physics , world scientific , new jersey 2003 , section 10 there ) gets a paradigmatic background from our application of the 90@xmath299 polar @xmath295 coordinate for both , the mass operator and the graviton torus . it turned out that the development of conventional new physics and the paradigmatic iterations diverge , i.e. they are drifting away one another , especially if the conventional ( i.e. communicable ) was organized as alternative ( @xmath300/@xmath62 ) sequence . communicable is defined by only a few ( e.g. three or less ) new concepts for understanding . an `` off - site convergence '' of the paradigmatic iterations was reached in 2008 . this convergence was defined as the event where an important phenomenon not analyzed before could be explained without introduction of any new specified idea . this was the so - called `` dark energy '' of part ii ( in section 7.2 of our paper ) . as mentioned above , roughly two thousand internal paradigmatic iterations were necessary . nevertheless , the paradigm did not become communicable and remains woolly today . the woolliness is assumed to decrease in the future . to improve understanding , it seems useful to characterize the new paradigm by three `` basic principles a , b , c '' . this will be tried here for the example of the 1-class solution ( part i , tab . i there ) . the basic idea is a map from a hidden thing to the physical space ( a @xmath301c ) . a. : : conformal hidden charge model ( i , section 2.2 , points ( 1 ) , ( 2 ) , ( 3 ) there ) . b. : : existential instability as a consequence of the maximum for the lepton masses ( i , fig.6 there ) . this excludes relevant particles above the hadronic region ( ii , eq.(12 ) here ) . c. : : electroweak feynman fabric in the physical space ( i , fig.2 there ) as a consequence of the above a conformality . the relationship between the present new physics and an off - site converging paradigm was briefly presented in a comment ; see arxiv : 0906.2425 v2 [ physics.gen-ph ] 3 aug 2009 . this is a radical variant of a discussion in the common probability space for an alternative new physics sequence and the off - site development of a converging paradigm . sociological aspects of the hierarchies in the present large collaborations are also presented . the radical excludes any perceptible probability to find a new paradigm inside the new physics . in reality , the relationship may be gentler . from my experience with the common probability space , i think that the probability to find a new paradigm inside a sequence of communicable new physics remains rather small . this work was partly supported by the dfg sonderforschungsbereich sfb 418 and by the fonds chemische industrie , fci . the author thanks in particular steffen trimper , the speaker of the sfb , and has profited from stimulating discussions with e.w . fischer ( mainz ) , dieter richter ( jlich ) , christoph schick ( rostock ) , klaus schrter ( halle ) , and our experimental group at the universitt halle collecting dynamic calorimetry data supporting the belief to the characteristic length of glass transition .

we study the free energy of the directed polymer in random environment model in dimension and . for dimension one , we improve the statement of comets and vargas in concerning very strong disorder by giving sharp estimates on the free energy at high temperature . in dimension two , we prove that very strong disorder holds at all temperatures , thus solving a long standing conjecture in the field . + 2000 _ mathematics subject classification : 82d60 , 60k37 , 82b44 _ + + _ keywords : free energy , directed polymer , strong disorder , localization , fractional moment estimates , quenched disorder , coarse graining . _

0901.0699

we study a directed polymer model introduced by huse and henley ( in dimension @xmath0 ) @xcite with the purpose of investigating impurity - induced domain - wall roughening in the 2d - ising model . the first mathematical study of directed polymers in random environment was made by imbrie and spencer @xcite , and was followed by numerous authors @xcite ( for a review on the subject see @xcite ) . directed polymers in random environment model , in particular , polymer chains in a solution with impurities . in our set up the polymer chain is the graph @xmath2 of a nearest neighbor path in @xmath3 , @xmath4 starting from zero . the equilibrium behavior of this chain is described by a measure on the set of paths : the impurities enter the definition of the measure as _ disordered potentials _ , given by a typical realization of a field of i.i.d . random variables @xmath5 ( with associated law @xmath6 ) . the polymer chain will tend to be attracted by larger values of the environment and repelled by smaller ones . more precisely , we define the hamiltonian @xmath7 we denote by @xmath8 the law of the simple symmetric random walk on @xmath3 starting at @xmath9 ( in the sequel @xmath10 , respectively @xmath11 , will denote the expectation with respect to @xmath8 , respectively q ) . one defines the polymer measure of order @xmath12 at inverse temperature @xmath13 as @xmath14 where @xmath15 is the normalization factor which makes @xmath16 a probability measure @xmath17 we call @xmath15 the _ partition function _ of the system . in the sequel , we will consider the case of @xmath18 with zero mean and unit variance and such that there exists @xmath19 $ ] such that @xmath20 finite exponential moments are required to guarantee that @xmath21 . the model can be defined and it is of interest also with environments with heavier tails ( see e.g. @xcite ) but we will not consider these cases here . in order to understand the role of disorder in the behavior of @xmath16 , as @xmath12 becomes large , let us observe that , when @xmath22 , @xmath16 is the law of the simple random walk , so that we know that , properly rescaled , the polymer chain will look like the graph of a @xmath23-dimensional brownian motion . the main questions that arise for our model for @xmath24 are whether or not the presence of disorder breaks the diffusive behavior of the chain for large @xmath12 , and how the polymer measure looks like when diffusivity does not hold . many authors have studied diffusivity in polymer models : in @xcite , bolthausen remarked that the renormalized partition function @xmath25 has a martingale property and proved the following zero - one law : @xmath26 a series of paper @xcite lead to @xmath27 and a consensus in saying that this implication is an equivalence . for this reason , it is natural and it has become customary to say that _ weak disorder _ holds when @xmath28 converges to some non - degenerate limit and that _ strong disorder _ holds when @xmath28 tends to zero . carmona and hu @xcite and comets , shiga and yoshida @xcite proved that strong disorder holds for all @xmath13 in dimension @xmath29 and @xmath30 . the result was completed by comets and yoshida @xcite : we summarize it here [ strdis ] there exists a critical value @xmath31 $ ] ( depending of the law of the environment ) such that * weak disorder holds when @xmath32 . * strong disorder holds when @xmath33 . moreover : @xmath34 \text { for } d\ge 3 . \end{split}\ ] ] we mention also that the case @xmath35 can only occur when the random variable @xmath36 is bounded . in @xcite and @xcite a characterization of strong disorder has been obtained in term of localization of the polymer chain : we cite the following result ( * ? ? ? * theorem 2.1 ) if @xmath37 and @xmath38 are two i.i.d . polymer chains , we have @xmath39 moreover if @xmath40 there exists a constant @xmath41 ( depending on @xmath13 and the law of the environment ) such that for @xmath42 one can notice that has a very strong meaning in term of trajectory localization when @xmath28 decays exponentially : it implies that two independent polymer chains tend to share the same endpoint with positive probability . for this reason we introduce now the notion of free energy , we refer to ( * ? ? ? * proposition 2.5 ) and ( * ? ? ? * theorem 3.2 ) for the following result : the quantity @xmath43 exists @xmath6-a.s . , it is non - positive and non - random . we call it the _ free energy _ of the model , and we have @xmath44 moreover @xmath45 is non - increasing in @xmath13 . we stress that the inequality @xmath46 is the standard _ annealing _ bound . in view on , it is natural to say that _ very strong disorder _ holds whenever @xmath47 . one can moreover define @xmath48 the critical value of @xmath13 for the free energy i.e. : @xmath49 let us stress that , from the physicists viewpoint , @xmath48 is the natural critical point because it is a point of non - analyticity of the free energy ( at least if @xmath50 ) . in view of this definition , we obviously have @xmath51 . it is widely believed that @xmath52 , i.e. that there exists no intermediate phase where we have _ strong disorder _ but not _ very strong disorder_. however , this is a challenging question : comets and vargas @xcite answered it in dimension @xmath0 by proving that @xmath53 . in this paper , we make their result more precise . moreover we prove that @xmath54 . the first aim of this paper is to sharpen the result of comets and vargas on the @xmath0-dimensional case . in fact , we are going to give a precise statement on the behavior of @xmath45 for small @xmath13 . our result is the following [ pasgaussi ] when @xmath55 and the environment satisfies , there exist constants @xmath41 and @xmath56 ( depending on the distribution of the environment ) such that for all @xmath57 we have @xmath58 \le p({\beta})\le -c { \beta}^4.\ ] ] we believe that the logarithmic factor in the lower bound is an artifact of the method . in fact , by using replica - coupling , we have been able to get rid of it in the gaussian case . [ gaussi ] when @xmath55 and the environment is gaussian , there exists a constant @xmath41 such that for all @xmath59 . @xmath60 these estimates concerning the free energy give us some idea of the behavior of @xmath16 for small @xmath13 . indeed , carmona and hu in ( * ? ? ? * section 7 ) proved a relation between @xmath45 and the overlap ( although their notation differs from ours ) . this relation together with our estimates for @xmath45 suggests that , for low @xmath13 , the asymptotic contact fraction between independent polymers @xmath61 behaves like @xmath62 . the second result we present is that @xmath54 . as for the @xmath0-dimensional case , our approach yields an explicit bound on @xmath45 for @xmath13 close to zero . [ 1 + 2uplb ] when @xmath63 , there exist constants @xmath41 and @xmath64 such that for all @xmath65 , @xmath66 so that @xmath67 and @xmath9 is a point of non - analyticity for @xmath45 . after the appearance of this paper as a preprint , the proof of the above result has been adapted by bertin @xcite to prove the exponential decay of the partition function for _ linear stochastic evolution _ in dimension @xmath30 , a model that is a slight generalisation of directed polymer in random environment . unlike in the one dimensional case , the two bounds on the free energy provided by our methods do not match . we believe that the second moment method , that gives the lower bound is quite sharp and gives the right order of magnitude for @xmath68 . the method developped in @xcite to sharpen the estimate on the critical point shift for pinning models at marginality adapted to the context of directed polymer should be able to improve the result , getting @xmath69 for all @xmath59 for any @xmath70 . the various techniques we use have been inspired by ideas used successfully for another polymer model , namely the polymer pinning on a defect line ( see @xcite ) . however the ideas we use to establish lower bounds differ sensibly from the ones leading to the upper bounds . for this reason , we present first the proofs of the upper bound results in section [ rough ] , [ onedim ] and [ twodim ] . the lower bound results are proven in section [ lb11 ] , [ dim11 ] and [ dim12 ] . to prove the lower bound results , we use a technique that combines the so - called _ fractional moment method _ and change of measure . this approach has been first used for pinning model in @xcite and it has been refined since in @xcite . in section [ rough ] , we prove a non - optimal upper bound for the free energy in the case of gaussian environment in dimension @xmath0 to introduce the reader to this method . in section [ onedim ] we prove the optimal upper bound for arbitrary environment in dimension @xmath0 , and in section [ twodim ] we prove our upper bound for the free energy in dimension @xmath1 which implies that _ very strong disorder _ holds for all @xmath13 . these sections are placed in increasing order of technical complexity , and therefore , should be read in that order . concerning the lower bounds proofs : section [ lb11 ] presents a proof of the lower bound of theorem [ pasgaussi ] . the proof combines the second moment method and a directed percolation argument . in section [ dim11 ] the optimal bound is proven for gaussian environment , with a specific gaussian approach similar to what is done in @xcite . in section [ dim12 ] we prove the lower bound for arbitrary environment in dimension @xmath1 . these three parts are completely independent of each other . before going into the core of the proof , we want to present here the starting step that will be used repeatedly thourough sections [ rough ] , [ onedim ] and [ twodim ] . we want to find an upper bound for the quantity @xmath71 however , it is not easy to handle the expectation of a @xmath72 , for this reason we will use the following trick . let @xmath73 , we have ( by jensen inequality ) @xmath74 hence @xmath75 we are left with showing that the fractional moment @xmath76 decays exponentially which is a problem that is easier to handle . to introduce the reader to the general method used in this paper , combining fractional moment and change of measure , we start by proving a non optimal result for the free energy , using a finite volume criterion . as a more complete result is to be proved in the next section , we restrict to the gaussian case here . the method used here is based on the one of @xcite , marorizing the free energy of the directed polymer by the one of multiplicative cascades . let us mention that is has bee shown recently by liu and watbled @xcite that this majoration is in a sense optimal , they obtained this result by improving the concentration inequality for the free energy . the idea of combining fractional moment with change of measure and finite volume criterion has been used with success for the pinning model in @xcite . [ lb ] there exists a constant @xmath41 such that for all @xmath59 @xmath77 for @xmath78 sufficiently small , we choose @xmath79 to be equal to @xmath80 for a fixed constant @xmath81 ( here and thourough the paper for @xmath82 , @xmath83 , respectively @xmath84 will denote the upper , respectively the lower integer part of @xmath85 ) and define @xmath86 . for @xmath87 we define @xmath88\right){\mathbf{1}}_{\{s_n = x\}}.\ ] ] note that @xmath89 . we use a statement which can be found in the proof of theorem 3.3 . in @xcite : @xmath90^{\theta } \quad \forall m\in { \mathbb{n}}.\ ] ] this combined with implies that @xmath91^{\theta}.\ ] ] hence , to prove the result , it is sufficient to show that @xmath92^{\theta}\le e^{-1 } , \label{eq : aprouver}\ ] ] for our choice of @xmath93 and @xmath79 . in order to estimate @xmath94^{\theta}$ ] we use an auxiliary measure @xmath95 . the region where the walk @xmath96 is likely to go is @xmath97\times[-c_2\sqrt{n},c_2\sqrt{n}]\right)\cap { \mathbb{n}}\times { \mathbb{z}}$ ] where @xmath98 is a big constant . we define @xmath95 as the measure under which the @xmath99 are still independent gaussian variables with variance @xmath29 , but such that @xmath100 where @xmath101 . this measure is absolutely continuous with respect to @xmath6 and @xmath102\right).\ ] ] then we have for any @xmath87 , using the hlder inequality we obtain , @xmath103={\widetilde}q \left[\frac{{\,\text{\rm d}}q}{{\,\text{\rm d}}{\widetilde}q}\left(w_n(x)\right)^\theta\right]\le \left({\widetilde}q \left[\left(\frac{{\,\text{\rm d}}q}{{\,\text{\rm d}}{\widetilde}q}\right)^{\frac{1}{1-\theta}}\right]\right)^{1-\theta}\left ( { \widetilde}q w_n(x)\right)^{\theta}.\ ] ] the first term on the right - hand side can be computed explicitly and is equal to @xmath104 where the last inequality is obtained by replacing @xmath105 and @xmath93 by their values ( recall @xmath106 ) . therefore combining and we get that @xmath107 to bound the right hand side , we first get rid of the exponent @xmath93 in the following way : @xmath108 if @xmath79 is sufficiently large ( i.e. , @xmath13 sufficiently small ) the first term on the right - hand side is smaller than @xmath109 so that @xmath110 we are left with showing that the expectation of @xmath111 with respect to the measure @xmath95 is small . it follows from the definition of @xmath95 that @xmath112 and therefore @xmath113 one can choose @xmath98 such that the first term is small , and the second term is equal to @xmath114 that can be arbitrarily small by choosing @xmath81 large compared to @xmath115 . in that case is satisfied and we have @xmath116 for small enough @xmath13 . the upper bound we found in the previous section is not optimal , and can be improved by replacing the finite volume criterion by a more sophisticated coarse graining method . the technical advantage of the coarse graining we use , is that we will not have to choose the @xmath93 of the fractional moment close to @xmath29 as we did in the previous section and this is the way we get rid of the extra @xmath72 factor we had . the idea of using this type of coarse graining for the copolymer model appeared in @xcite and this has been a substantial source of inspiration for this proof . we will prove the following result first in the case of gaussian environment , and then adapt the proof to general environment . let @xmath79 be the smallest squared integer bigger than @xmath117 ( if @xmath13 is small we are sure that @xmath118 ) . the number @xmath79 will be used in the sequel of the proof as a scaling factor . let @xmath119 be fixed ( say @xmath120 ) . we consider a system of size @xmath121 ( where @xmath122 is meant to tend to infinity ) . let @xmath123 denote the interval @xmath124 . in order to estimate @xmath76 we decompose @xmath28 according to the contribution of different families path : @xmath125 where @xmath126.\ ] ] = 6.5 cm [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] then , we apply the inequality @xmath127 ( which holds for any finite or countable collection of positive real numbers ) to this decomposition and average with respect to @xmath6 to get , @xmath128 in order to estimate @xmath129 , we use an auxiliary measure as in the previous section . the additional idea is to make the measure change depend on @xmath130 . for every @xmath131 we define the set @xmath132 as @xmath133 where @xmath134 is equal to zero . note that for big values of @xmath79 and @xmath122 @xmath135 we define the measure @xmath136 the measure under which the @xmath137 are independent gaussian variables with variance @xmath29 and mean @xmath138 where @xmath139 . = 6.5 cm [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] the law @xmath136 is absolutely continuous with respect to @xmath6 and its density is equal to @xmath140\right).\ ] ] using hlder inequality with this measure as we did in the previous section , we obtain @xmath141={\widetilde}q_{y } \left[\frac{{\,\text{\rm d}}q}{{\,\text{\rm d}}{\widetilde}q_{y}}{\widecheck}w_{(y_1,y_2,\dots , y_m)}^{\theta}\right]\\ \le { \widetilde}q_{y}\left ( \left[\left(\frac{{\,\text{\rm d}}q}{{\,\text{\rm d}}{\widetilde}q_{y}}\right)^{\frac{1}{1-\theta}}\right]\right)^{1-\theta}\left ( { \widetilde}q_{y } { \widecheck}w_{(y_1,\dots , y_m)}\right)^{\theta } \label{eq : hhol}.\end{gathered}\ ] ] the value of the first term can be computed explicitly @xmath142\right)^{1-\theta}=\exp\left(\frac{\ # j_y\theta{\delta}_n^2}{2(1-\theta)}\right)\le \exp(3 m ) , \label{eq : dens}\ ] ] where the upper bound is obtained by using the definition of @xmath105 , and the fact that @xmath120 . now we compute the second term @xmath143\}}.\ ] ] we define @xmath144 equation implies that ( recall that @xmath145 is the law of the simple random walk starting from @xmath85 , and that we set @xmath146 @xmath147 combining this with , and we have @xmath148.\ ] ] if the quantity in the square brackets is smaller than @xmath149 , by equation we have @xmath150 . therefore , to complete the proof it is sufficient to show that @xmath151 is small . to reduce the problem to the study of a finite sum , we observe ( using some well known result on the asymptotic behavior of random walk ) that given @xmath152 we can find @xmath153 such that @xmath154 to estimate the remainder of the sum we use the following trivial bound @xmath155 then we get rid of the @xmath156 in the sum by observing that if a walk starting from @xmath85 makes a step in @xmath157 , the walk with the same increments starting from @xmath9 will make the same step in @xmath158 ( recall ) . @xmath159 now we are left with something similar to what we encountered in the previous section @xmath160 if @xmath161 is chosen large enough , the first term can be made arbitrarily small by choosing @xmath161 large , and the second is equal to @xmath162 and can be made also arbitrarily small if @xmath163 is chosen large enough once @xmath161 is fixed . an appropriate choice of constant and the use of and can leads then to @xmath164 this combined with completes the proof . in the case of a general environment , some modifications have to be made in the proof above , but the general idea remains the same . in the change of measure one has to change the shift of the environment in @xmath132 by an exponential tilt of the measure as follow @xmath165\right).\ ] ] the formula estimating the cost of the change of measure becomes @xmath166\right)\le \exp(2 m),\ ] ] where the last inequality is true if @xmath167 is small enough if we consider that @xmath120 and use the fact that @xmath168 ( @xmath169 has @xmath9 mean and unit variance ) . the next thing we have to do is to compute the effect of this change of measure in this general case , i.e. find an equivalent for . when computing @xmath170 , the quantity @xmath171\ ] ] appears instead of @xmath172 . using twice the mean value theorem , one gets that there exists @xmath173 and @xmath174 in @xmath175 such that @xmath176=-{\beta}{\delta}_n{\lambda}''(-h{\delta}_n+h'{\beta}).\ ] ] and as @xmath169 has unit variance @xmath177 . therefore if @xmath13 and @xmath105 are chosen small enough , the right - hand side of the above is less than @xmath178 . so that can be replaced by @xmath179\}}.\ ] ] the remaining steps follow closely the argument exposed for the gaussian case . in this section , we prove the main result of the paper : very strong disorder holds at all temperature in dimension @xmath30 . the proof is technically quite involved . it combines the tools of the two previous sections with a new idea for the change a measure : changing the covariance structure of the environment . we mention that this idea was introduced recently in @xcite to deal with the _ marginal disorder _ case in pinning model . we choose to present first a proof for the gaussian case , where the idea of the change a measure is easier to grasp . before starting , we sketch the proof and how it should be decomposed in different steps : * we reduce the problem by showing that it is sufficient to show that for some real number @xmath119 , @xmath76 decays exponentially with @xmath12 . * we use a coarse graining decomposition of the partition function by splitting it into different contributions that corresponds to trajectories that stays in a large corridor . this decomposition is similar to the one used in section [ onedim ] . * to estimate the fractional moment terms appearing in the decomposition , we change the law of the environment around the corridors corresponding to each contribution . more precisely , we introduce negative correlations into the gaussian field of the environment . we do this change of measure in such a way that the new measure is not very different from the original one . * we use some basic properties of the random walk in @xmath180 to compute the expectation under the new measure . we fix @xmath79 to be the smallest squared integer bigger than @xmath181 for some large constant @xmath182 to be defined later , for small @xmath13 we have @xmath183 . the number @xmath79 will be used in the sequel of the proof as a scaling factor . for @xmath184 we define @xmath185\times[b\sqrt{n},(b+1)\sqrt{n}-1]$ ] so that @xmath186 are disjoint and cover @xmath180 . for @xmath187 , we decompose the normalized partition function @xmath28 into different contributions , very similarly to what is done in dimension one ( i.e. decomposition ) , and we refer to the figure [ fig : wnn2 ] to illustrate how the decomposition looks like : @xmath188 where @xmath189 \right){\mathbf{1}}_{\left\{s_{in}\in i_{y_i},\forall i= 1,\dots , m \right\}}.\ ] ] we fix @xmath119 and apply the inequality @xmath190 ( which holds for any finite or countable collection of positive real numbers ) to get @xmath191 in order to estimate the different terms in the sum of the right hand side in , we define some auxiliary measures @xmath136 on the the environment for every @xmath192 with @xmath193 . we will choose the measures @xmath194 absolutely continuous with respect to @xmath6 . we use hlder inequality to get the following upper bound : @xmath195 now , we describe the change of measure we will use . recall that for the @xmath29-dimensional case we used a shift of the environment along the corridor corresponding to @xmath196 . the reader can check that this method would not give the exponential decay of @xmath28 in this case . instead we change the covariance function of the environment along the corridor on which the walk is likely to go by introducing some negative correlation . we introduce the change of measure that we use for this case . given @xmath197 we define @xmath122 blocks @xmath198}$ ] and @xmath132 their union ( here and in the sequel , @xmath199 denotes the @xmath200 norm on @xmath180 ) : @xmath201 we fix the covariance the field @xmath169 under the law @xmath202 to be equal to @xmath203 \text { such that } ( i , z ) \text { and } ( j , z')\in b_k \\ { \mathbf{1}}_{\{(i , z)=(j , z')\ } } & \text { otherwise , } \end{cases}\end{split}\ ] ] where @xmath204 we define @xmath205 one remarks that the so - defined covariance matrix @xmath206 is block diagonal with @xmath122 identical blocks which are copies of @xmath207 corresponding to the @xmath208 , @xmath209 $ ] , and just ones on the diagonal elsewhere . therefore , the change of measure we describe here exists if and only if @xmath207 is definite positive . the largest eigenvalue for @xmath210 is associated to a positive vector and therefore is smaller than @xmath211 for the sequel we choose @xmath79 such that the spectral radius of @xmath210 is less than @xmath212 so that @xmath207 is positive definite . with this setup , @xmath136 is well defined . the density of the modified measure @xmath202 with respect to @xmath6 is given by @xmath213 where @xmath214 for any matrix @xmath215 of @xmath216 with finite support . then we can compute explicitly the value of the second term in the right - hand side of @xmath217 note that the above computation is right if and only if @xmath218 is a definite positive matrix . since its eigenvalues are the same of those of @xmath219 , this holds for large @xmath79 thanks to . using again the fact that @xmath206 is composed of @xmath122 blocks identical to @xmath207 , we get from @xmath220 in order to estimate the determinant in the denominator , we compute the hilbert - schmidt norm of @xmath210 . one can check that for all @xmath79 @xmath221 we use the inequality @xmath222 for all @xmath223 and the fact that the spectral radius of @xmath224 is bounded by @xmath225 ( cf . ) to get that @xmath226&=\exp\left({\text{trace}}\left(\log\left(i-\frac{{\widehat}v}{1-\theta}\right)\right)\right)\ge \exp\left(-\frac{\| { \widehat}v\|^2}{(1-\theta)^2}\right)\\ & \ge \exp\left(-\frac{1}{(1-\theta)^2}\right ) . \end{split}\ ] ] for the numerator , @xmath227 implies that that @xmath228 . combining this with and we get @xmath229 now that we have computed the term corresponding to the change of measure , we estimate @xmath230 under the modified measure ( just by computing the variance of the gaussian variables in the exponential , using ) : @xmath231 replacing @xmath206 by its value we get that @xmath232 now we do something similar to @xmath233 : for each `` slice '' of the trajectory @xmath234}$ ] , we bound the contribution of the above expectation by maximizing over the starting point ( recall that @xmath145 denotes the probability distribution of a random walk starting at @xmath85 ) . thanks to the conditioning , the starting point has to be in @xmath235 . using the translation invariance of the random walk , this gives us the following ( @xmath236 stands for maximum ) : @xmath237.\end{gathered}\ ] ] for trajectories @xmath4 of a directed random - walk of @xmath79 steps , we define the quantity @xmath238 combining with , and , we finally get @xmath239^m.\ ] ] the exponential decay of @xmath76 ( with rate @xmath79 ) is guaranteed if we can prove that @xmath240 is small . the rest of the proof is devoted to that aim . we fix some @xmath152 . asymptotic properties of the simple random walk , guarantees that we can find @xmath241 such that @xmath242 to estimate the rest of the sum , we use the following trivial and rough bound @xmath243^{\theta } \le r^2 \left[\max_{x\in i_0 } p_x \exp\left(-\frac{{\beta}^2}{2}g(s)\right)\right]^{\theta}\ ] ] then we use the definition of @xmath244 to get rid of the @xmath156 by reducing the width of the zone where we have negative correlation : @xmath245 we define @xmath246 . we get from the above that @xmath247 one can make the first term of the right - hand side arbitrarily small by choosing @xmath248 large , in particular on can choose @xmath248 such that @xmath249 } |s_n|\ge ( c_6 - 1)\sqrt{n}\right\}\le ( { \varepsilon}/r^2)^{\frac{1}{\theta}}.\ ] ] to bound the other term , we introduce the quantity @xmath250 and the random variable @xmath251 , @xmath252 for any @xmath253 , we can find @xmath254 such that @xmath255 . we fix @xmath254 such that this holds for some good @xmath256 ( to be chosen soon ) , and by remarking that @xmath257 almost surely , we obtain ( using markov inequality ) @xmath258 moreover we can estimate @xmath259 getting that for @xmath79 large enough @xmath260 using and we get @xmath261 due to the choice of @xmath79 we have made ( recall @xmath262 ) , the second term is less than @xmath263 . we can choose @xmath256 , @xmath254 and @xmath182 such that , the right - hand side is less that @xmath264 . this combined with , , and allow us to conclude that @xmath265 so that with a right choice for @xmath70 , implies @xmath266 then allows us to conclude that @xmath150 . the case of general environment does not differ very much from the gaussian case , but one has to a different approach for the change of measure in . in this proof , we will largely refer to what has been done in the gaussian case , whose proof should be read first . let @xmath267 be a large constant . one defines the function @xmath268 on @xmath269 as to be @xmath270 recall the definitions and , and define @xmath271 function of the environment as @xmath272 multiplying by @xmath271 penalizes by a factor @xmath273 the environment for which there is to much correlation in one block . this is a way of producing negative correlation in the environment . for the rest of the proof we use the notation @xmath274 we do a computation similar to to get @xmath275\le \left(q\left [ g_y({\omega})^{-\frac{\theta}{1-\theta}}\right]\right)^{1-\theta}\left ( q\left [ g_y({\omega}){\widecheck}w_{(y_1,\dots , y_n)}\right]\right)^{\theta}.\ ] ] the block structure of @xmath271 allows to express the first term as a power of @xmath122 . @xmath276=\left(q\left[\exp\left(-\frac{\theta}{1-\theta}f_k\left(u_1\right)\right)\right]\right)^m.\ ] ] equation says that @xmath277 so that @xmath278 and hence @xmath279\\ \le 1+\exp\left(-2k^2+\frac{\theta}{1-\theta}k\right)\le 2,\end{gathered}\ ] ] if @xmath267 is large enough . we are left with estimating the second term @xmath280=p q g_y({\omega})\exp\left(\sum_{i=1}^{nm}[{\beta}{\omega}_{i , s_i}-{\lambda}({\beta } ) ] \right){\mathbf{1}}_{\{s_{kn}\in i_{y_k},\forall k=1\dots m\}}.\ ] ] for a fixed trajectory of the random walk @xmath4 , we consider @xmath281 the modified measure on the environment with density @xmath282 \right).\ ] ] under this measure @xmath283 as @xmath284 has zero - mean and unit variance under @xmath6 , implies @xmath285 around zero and that @xmath286 for all @xmath287 if @xmath13 is small enough . moreover @xmath281 is a product measure , i.e. the @xmath288 are independent variables under @xmath289 . with this notation becomes @xmath290{\mathbf{1}}_{\{s_{kn}\in i_{y_k},\forall k=1,\dots , m \}}.\ ] ] as in the gaussian case , one wants to bound this by a product using the block structure . similarly to , we use translation invariance to get the following upper bound @xmath291 using this in with the bound we get the inequality @xmath292^{\theta}\right)^m\!\!\!\!.\ ] ] therefore to prove exponential decay of @xmath76 , it is sufficient to show that @xmath293^{\theta}\ ] ] is small . as seen in the gaussian case ( cf . , ) , the contribution of @xmath294 far from zero can be controlled and therefore it is sufficient for our purpose to check @xmath295 for some small @xmath256 . similarly to , we force the walk to stay in the zone where the environment is modified by writing @xmath296}|s_i|\ge ( c_6 - 1)\sqrt{n}\}\\ + \max_{x\in i_0 } p_x { \widebar}q_s \exp\left(f_k\left(u_1\right)\right){\mathbf{1}}_{\{|s_n - s_0|\le ( c_6 - 1)\sqrt{n}\}}.\end{gathered}\ ] ] the first term is smaller than @xmath297 if @xmath248 is large enough . to control the second term , we will find an upper bound for @xmath298}|s_i - s_0|\le ( c_6 - 1)\sqrt{n}\}},\ ] ] which is uniform in @xmath299 . what we do is the following : we show that for most trajectories @xmath4 the term in @xmath268 has a large mean and a small variance with respect to @xmath289 so that @xmath300 with large @xmath281 probability . the rest will be easy to control as the term in the expectation is at most one . the expectation of @xmath301 under @xmath281 is equal to @xmath302 when the walk stays in the block @xmath303 we have ( using definition ) @xmath304 the distribution of @xmath251 under @xmath145 is the same for all @xmath305 . it has been shown earlier ( cf . and ) , that if @xmath254 is chosen large enough , @xmath306 as @xmath307 if @xmath13 is small , if @xmath182 is large enough ( recall @xmath262 ) , this together with gives . @xmath308}|s_i - s_0|\le ( c_6 - 1)\sqrt{n}\bigg\}\le \frac{{\delta}}{6}.\ ] ] to bound the variance of @xmath301 under @xmath281 , we decompose the sum @xmath309 and hence ( using the fact that @xmath310 ) . @xmath311 where we used that @xmath312 ( which is true for @xmath13 small enough ) . the last term is less than @xmath313 thanks to , so that we just have to control the first one . independently of the choice of @xmath314 we have the bound @xmath315 moreover it is also easy to check that @xmath316 ( these two bounds follow from the definition of @xmath317 : ) . therefore @xmath318\max_{(j , z)\in b_1}\sum_{1\le i\le n}v_{(i , s_i),(j , z')}\le 1.\ ] ] injecting this into guaranties that for @xmath13 small enough @xmath319 with chebyshev inequality , if @xmath267 has been chosen large enough and @xmath320 we have @xmath321 hence combining with gives @xmath322}|s_i - s_0|\le ( c_6 - 1)\sqrt{n}\right\}\le { \delta}/3.\ ] ] we use this in to get @xmath323 so that our result is proved provided that @xmath267 has been chosen large enough . in this section we prove the lower bound for the free - energy in dimension @xmath29 in arbitrary environment . to do so we apply the second moment method to some quantity related to the partition function , and combine it with a percolation argument . the idea of the proof was inspired by a study of a polymer model on hierarchical lattice @xcite where this type of coarse - graining appears naturally . [ th : prop ] there exists a constant @xmath324 such that for all @xmath59 we have @xmath325 we use two technical lemmas to prove the result . the first is just a statement about scaling of the random walk , the second is more specific to our problem . [ th : lem1 ] there exists an a constant @xmath326 such that for large even squared integers @xmath79 , @xmath327 [ th : lem2 ] for any @xmath152 we can find a constant @xmath328 and @xmath64 such that for all @xmath329 , for every even squared integer @xmath330 we have @xmath331<{\varepsilon}.\ ] ] let @xmath79 be some fixed integer and define @xmath332 which corresponds to the contribution to the partition function @xmath111 of paths with fixed end point @xmath333 staying within a cell of width @xmath333 , with the specification the environment on the last site is not taken in to account . @xmath334 depends only of the value of the environment @xmath169 in this cell ( see figure [ fig : restr ] ) . = 4.5 cm [ c]o [ c]@xmath79 [ c]@xmath333 one also defines the following quantities for @xmath335 : @xmath336}{\mathbf{1}}_{\{s_n=\sqrt{(y+1)n } , 0<s_i - y \sqrt n<\sqrt{n } \text { for } 0<i < n\}}\right],\\ { \widebar}w_i^{(y , y-1)}&:=p_{\sqrt{n}y}\left[\ e^{\sum_{k=1}^{n-1}\left[{\beta}{\omega}_{in+k , s_k}-{\lambda}({\beta})\right]}{\mathbf{1}}_{\{s_n=(y-1)\sqrt{n } , -\sqrt{n}<s_i - y\sqrt{n}<0 \text { for } 0<i < n\}}\right ] . \end{split}\ ] ] which are random variables that have the same law as @xmath334 . moreover because of independence of the environment in different cells , one can see that @xmath337 is a family of independent variables . let @xmath187 be a large integer . we define @xmath338 as the set of path @xmath339,\ |s_{in}-s_{(i-1)n}|=\sqrt{n},\ \forall j\in [ 1,n-1],\ s_{(i-1)n+j}\in \left ( s_{(i-1)n},s_{in}\right)\},\ ] ] where the interval @xmath340 is to be seen as @xmath341 if @xmath342 , and @xmath343 \right\ } .\ ] ] we use the trivial bound @xmath344,\ ] ] to get that @xmath345 ( the exponential term is due to the fact the @xmath334 does not take into account to site in the top corner of each cell ) . the idea is of the proof is to find a value of @xmath79 ( depending on @xmath13 ) such that we are sure that for any value of @xmath122 we can find a path @xmath346 such that along the path the values of @xmath347 are not to low ( i.e. close to the expectation of @xmath334 ) and to do so , it seems natural to seek for a percolation argument . let @xmath348 be the critical exponent for directed percolation in dimension @xmath0 ( for an account on directed percolation see ( * ? ? ? * section 12.8 ) and references therein ) . from lemma [ th : lem2 ] and chebyshev inequality , one can find a constant @xmath349 and @xmath64 such that for all @xmath350 and @xmath65 . @xmath351 we choose @xmath79 to be the biggest squared even integer that is less than @xmath352 . ( in particular have @xmath353 if @xmath13 small enough ) . as shown in figure [ fig : wnnn ] , we associate to our system the following directed percolation picture . for all @xmath354 is even : * if @xmath355 , we say that the edge linking the opposite corners of the corresponding cell is open . * if @xmath356 , we say that the same edge is closed . equation and the fact the considered random variables are independent assures that with positive probability there exists an infinite directed path starting from zero . = 6.5 cm [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] [ c ] when there exists an infinite open path is linking zero to infinity exists , we can define the highest open path in an obvious way . let @xmath357 denotes this highest path . if @xmath122 is large enough , by law of large numbers we have that with a probability close to one , @xmath358 \ge -2m{\lambda}({\beta}).\ ] ] using this and and the percolation argument with we finally get that with a positive probability which does not depend on @xmath122 we have @xmath359^m.\ ] ] taking the @xmath72 and making @xmath122 tend to infinity this implies that @xmath360\ge -\frac{c}{n}\log n.\ ] ] for some constant @xmath41 , if @xmath79 is large enough ( we used lemma [ th : lem1 ] to get the last inequality . the result follows by replacing @xmath79 by its value . let @xmath79 be square and even . @xmath361 , @xmath362 denote the first hitting time of @xmath363 by the random walk @xmath4 ( when @xmath364 it denotes the return time to zero ) . we have @xmath365 where the second equality is obtained with the strong markov property used for @xmath366 , and the reflexion principle for the random walk . the last line is equal to @xmath367 } s_k\in [ \sqrt{n}/2,\sqrt{n})|t_0=n\}p\{t_0=n\}.\ ] ] we use here a variant of donsker s theorem , for a proof see ( * ? ? ? * theorem 2.6 ) . the process @xmath368\ ] ] converges in distribution to the normalized brownian excursion in the space @xmath369,{\mathbb{r}})$ ] . we also know that ( see for example ( * ? ? ? * proposition a.10 ) ) for @xmath79 even @xmath370 . therefore , from we have @xmath371 } { \bf e}_t\in ( 1/2,1)\right]+o(n^{-3/2}).\end{gathered}\ ] ] where @xmath372 denotes the normalized brownian excursion , and @xmath373 its law . let @xmath13 be fixed and small enough , and @xmath79 be some squared even integer which is less than @xmath374 . we will fix the value @xmath328 independently of @xmath13 later in the proof , and always consider that @xmath13 is sufficiently small . by a direct computation the variance of @xmath375\right)\bigg| \ s_n=\sqrt{n } , 0<s_i<\sqrt{n } \text { for } 0<i < n \right]\ ] ] is equal to @xmath376 - 1.\ ] ] where @xmath377 and @xmath378 ( recall that @xmath379 ) , and @xmath380 , @xmath381 denotes two independent random walk with law denoted by @xmath382 . from this it follows that if @xmath79 is small the result is quite straight forward . we will therefore only be interested in the case of large @xmath79 ( i.e. bounded away from zero by a large fixed constant ) . we define @xmath383 the set where the walks meet ( it can be written as an increasing sequence of integers ) . by the markov property , the random variables @xmath384 are i.i.d . , we say that @xmath385 is a _ renewal sequence_. we want to bound the probability that the renewal sequence @xmath385 has too many returns before times @xmath386 , in order to estimate . to do so , we make the usual computations with laplace transform . from @xcite , we know that @xmath387 thanks to the the local central limit theorem for the simple random walk , we know that for large @xmath79 @xmath388 so that we can get from that when @xmath85 is close to zero @xmath389 we fix @xmath390 such that @xmath391 for all @xmath392 . for any @xmath393 we have @xmath394|\ge k\}=p^{\otimes 2}\{\tau_k\le n-1\}&\le \exp((n-1)x ) p^{\otimes 2 } \exp ( -\tau_k x)\\ & \le \exp\left[nx+k \log p^{\otimes 2}\exp(-x\tau_1)\right ] . \end{split}\ ] ] for any @xmath395 one can choose @xmath396 in the above and use the definition of @xmath390 to get that @xmath397|\ge k\}\le \exp\left(-k^2/(32 n)\right).\ ] ] in the case where @xmath398 we simply bound the quantity by @xmath397|\ge k\}\le \exp\left(k_0 ^ 2/(32 n)\right)\le \exp\left(-n x_0/4\right).\ ] ] by lemma , if @xmath79 is large enough @xmath399 a trivial bound on the conditioning gives us @xmath400|\ge k \ \big| \ a_n \right)&\le\min\left(1,2c_{rw}^{-2 } n^3\exp\left(-k^2/(32 n)\right)\right ) \text { if } k\le k_0,\\ p^{\otimes 2}\left(|\tau\cap [ 1,n-1]|\ge k \ \right)&\le 2c_{rw}^{-2}n^{3 } \exp\left(-n x_0/4\right ) \text { otherwise}. \end{split}\ ] ] we define @xmath401 . the above implies that for @xmath79 large enough we have @xmath402|\ge k \ \big| \ a_n \right)&\le 1 \text { if } k\le k_1 , \\ p^{\otimes 2}\left(|\tau\cap [ 1,n-1]|\ge k \ \big| \ a_n \right)&\le \exp\left(-k^2/(64 n)\right ) \text { if } k_1\le k\le k_0,\\ p^{\otimes 2}\left(|\tau\cap [ 1,n-1]|\ge k \ \big| \ a_n \right)&\le\exp\left(-n x_0/8\right ) \text { otherwise}. \end{split}\ ] ] now we are ready to bound . integration by part gives , @xmath403|\right)\ \big| \ a_n\right]-1\\ & \quad\quad\quad\quad\quad\quad\quad\quad={\gamma}({\beta})\int_{0}^{\infty}\exp({\gamma}({\beta } ) x)p^{\otimes 2}\left(|\tau\cap [ 1,n-1]|\ge x \ a_n\right){\,\text{\rm d}}x . \end{split}\ ] ] we split the right - hand side in three part corresponding to the three different bounds we have in : @xmath404 $ ] , @xmath405 $ ] and @xmath406 $ ] . it suffices to show that each part is less than @xmath407 to finish the proof . the first part is @xmath408|\ge x\ \big| \ \le { \gamma}({\beta})k_1\exp({\gamma}({\beta})k_1).\ ] ] one uses that @xmath409 and @xmath410 to get that for @xmath13 small enough and @xmath79 large enough if @xmath411 is well chosen we have @xmath412 so that @xmath413 . + we use our bound for the second part of the integral to get @xmath414|\ge x\ \big| \ d}}x\\ & \quad\le { \gamma}({\beta})\int_{0}^{\infty}\exp\left({\gamma}({\beta})x - x^2/(64 n)\right){\,\text{\rm d}}x = \int_{0}^{\infty}\exp\left(x-\frac{x^2}{64 n { \gamma}({\beta})^2}\right){\,\text{\rm d}}x . \end{split}\ ] ] replacing @xmath79 by its value , we see that the term that goes with @xmath415 in the exponential can be made arbitrarily large , provided that @xmath411 is small enough . in particular we can make the left - hand side less than @xmath407 . + finally , we estimate the last part @xmath416|\ge x \ \big| \ a_n\right){\,\text{\rm d}}x\\ & \quad\quad\quad\quad\quad\quad\quad\quad \le { \gamma}({\beta})\int_{0}^{n}\exp({\gamma}({\beta})x - n x_0/8){\,\text{\rm d}}x = n\exp(-[{\gamma}({\beta})-x_0/8]n ) . \end{split}\ ] ] this is clearly less than @xmath407 if @xmath79 is large and @xmath13 is small . in this section we use the method of replica - coupling that is used for the disordered pinning model in @xcite to derive a lower bound on the free energy . the proof here is an adaptation of the argument used there to prove disorder irrelevance . the main idea is the following : let @xmath417 denotes the renormalized partition function for inverse temperature @xmath13 . a simple gaussian computation gives @xmath418 where @xmath37 and @xmath38 are two independent random walk under the law @xmath382 . this implies that for small values of @xmath13 ( by the equality of derivative at @xmath419 ) , @xmath420 this tends to make us believe that @xmath421 however , things are not that simple because is only valid for fixed @xmath12 , and one needs some more work to get something valid when @xmath12 tends to infinity . the proofs aims to use convexity argument and simple inequalities to be able to get the inequality @xmath422 the fact that convexity is used in a crucial way make it quite hopeless to get the other inequality using this method . let use define for @xmath13 fixed and @xmath423 $ ] @xmath424\right),\ ] ] and for @xmath425 @xmath426\right).\ ] ] one can notice that @xmath427 and @xmath428 ( recall the definition of @xmath429 ) , so that @xmath430 is an interpolation function . via the gaussian integration by par formula @xmath431 valid ( if @xmath169 is a centered standard gaussian variable ) for every differentiable functions such that @xmath432 , one finds @xmath433\right){\mathbf{1}}_{\left\{s_j = z\right\}}}{p\exp\left(\sum_{i=1}^n \left[\sqrt{t}{\beta}{\omega}_{i , s_i}-\frac{t{\beta}^2}{2}\right]\right)}\right)^2 \\ & = -\frac{{\beta}^2}{2n}q \left(\mu^{(\sqrt{t}{\beta})}_n\right)^{\otimes 2 } \left[\sum_{i=1}^n { \mathbf{1}}_{\{s_i^{(1)}=s_i^{(2)}\}}\right ] . \end{split}\ ] ] this is ( up to the negative multiplicative constant @xmath434 ) the expected overlap fraction of two independent replicas of the random walk under the polymer measure for the inverse temperature @xmath435 . this result has been using it formula in ( * ? ? ? * section 7 ) . + for notational convenience , we define @xmath436.\ ] ] we use gaussian integration by part again , for @xmath437 : @xmath438 the above implies that for every @xmath423 $ ] and @xmath425 @xmath439 comparing and , and using convexity and monotonicity of @xmath440 with respect to @xmath441 , and the fact that @xmath442 one gets @xmath443 where in the last inequality we used @xmath444 and . integrating this inequality between @xmath9 and @xmath29 and recalling @xmath428 we get @xmath445 on the right - hand side of the above we recognize something related to pinning models . more precisely @xmath446 where @xmath447 is the partition function of a homogeneous pinning system of size @xmath12 and parameter @xmath448 with underlying renewal process the sets of zero of the random walk @xmath449 . this is a well known result in the study of pinning model ( we refer to ( * ? ? * section 1.2 ) for an overview and the results we cite here ) that @xmath450 where @xmath451 denotes the free energy of the pinning model . moreover , it is also stated @xmath452 then passing to the limit in ends the proof of the result for any constant strictly bigger that @xmath453 . the technique used in the two previous sections could be adapted here to prove the results but in fact it is not necessary . because of the nature of the bound we want to prove in dimension @xmath30 ( we do not really track the best possible constant in the exponential ) , it will be sufficient here to control the variance of @xmath111 up to some value , and then the concentration properties of @xmath454 to get the result . the reader can check than using the same method in dimension @xmath29 does not give the right power of @xmath13 . as the above quantity is increasing in @xmath79 , it will be enough to prove the result for @xmath79 large . for technical convenience we choose to prove the result for @xmath465 ( recall @xmath378 ) which does not change the result since @xmath410 . the result we want to prove seems natural since we know that @xmath466 converges to an exponential variable ( see e.g. @xcite ) , and @xmath467 . however , convergence of the right hand side of requires the use of the dominated convergence theorem , and the proof of the domination hypothesis is not straightforward . it could be extracted from the proof of the large deviation result in @xcite , however we include a full proof of convergence here for the sake of completeness . to prove the result , we compute bounds on the probability of having too many point before @xmath79 in the renewal @xmath385 . as in the @xmath29 dimensional case , we use laplace transform to do so . from @xcite , we know that let @xmath390 be such that for any @xmath392 , @xmath472 . for @xmath363 such that @xmath473 , we replace @xmath85 by @xmath474 in to get @xmath475|\ge k\}\le \exp \left(\frac{k}{\log(n / k)}-\frac{3 k}{\log\left[k/(n\log n / k)\right]}\right)\le \exp\left(-\frac{k}{\log(n / k)}\right),\ ] ] where the last inequality holds if @xmath476 is small enough . we fix @xmath477 for some small @xmath256 . we get that @xmath478|\ge k\}&\le \exp\left(-\frac{k}{\log(n / k)}\right ) \quad \text{if } k\le k_0\\ p^{\otimes 2}\{|\tau\cap [ 1,n]|\ge k\}&\le \exp\left(-\frac{k_0}{\log(n / k_0)}\right)=\exp\left(-\frac{{\delta}n}{\log ( 1/{\delta})}\right ) \quad \text{if } k\ge k_0 . \end{split}\ ] ] we are ready to bound . we remark that using integration by part we obtain @xmath481|\ge x){\,\text{\rm d}}x\le \int_{1}^{{\delta}n}{\gamma}({\beta})\exp\left({\gamma}({\beta})x-\frac{x}{\log(n / x)}\right){\,\text{\rm d}}x\\ \le \int_{1}^{{\delta}n}{\gamma}({\beta})\exp\left({\gamma}({\beta})x-\frac{{\gamma}({\beta}){\beta}x}{c_{{\varepsilon}}}\right)\le \frac{c_{{\varepsilon}}}{1-c_{{\varepsilon}}}.\end{gathered}\ ] ] this is less that @xmath407 if @xmath328 is chosen appropriately . the last part to bound is @xmath482|\ge x)\le n{\gamma}({\beta})\exp\left({\gamma}({\beta})n-\frac{{\delta}n}{\log 1/{\delta}}\right)\le { \varepsilon}/3,\ ] ] where the last inequality holds if @xmath79 is large enough , and @xmath13 is small enough . by a martingale method that one can find a constant @xmath483 such that @xmath484 ( see ( * ? ? ? * proposition 2.5 ) and its proof for more details ) . + therefore chebyshev inequality gives @xmath485 using lemma [ th : lemmvar ] and chebyshev inequality again , we can find a constant @xmath486 such that for small @xmath13 and @xmath487 we have @xmath488 this combined with implies that the author is very grateful to giambattista giacomin for numerous suggestions and precious help for the writing of this paper , to francesco caravenna for the proof of lemma [ th : lem1 ] and to fabio toninelli and francis comets for enlightening discussions . the author also acknowledges the support of anr , grant polintbio .

we study the distribution modulo of the values taken on the integers of linear forms in variables with random coefficients . we obtain quenched and annealed central limit theorems for the number of simultaneous hits into shrinking targets of radii . by the khintchine - groshev theorem on diophantine approximations , is the critical exponent for the infinite number of hits .

1605.00311

an important problem in diophantine approximation is the study of the speed of approach to @xmath5 of a possibly inhomogeneous linear form of several variables evaluated at integers points . such a linear form is given by @xmath6 @xmath7 more generally , one can consider @xmath1 linear forms for @xmath8 , corresponding to @xmath9 and @xmath10 , where each @xmath11 , is a vector in @xmath12 . diophantine approximation theory classifies the matrices @xmath13 and vectors @xmath14 according to how `` resonant '' they are ; i.e. , how well the vector @xmath15 approximates @xmath16 as @xmath17 varies over a large ball in @xmath18 one can then fix a sequence of targets converging to @xmath5 , say intervals of radius @xmath19 centered at @xmath5 with @xmath20 , and investigate the integers for which the target is _ hit _ , namely the integers @xmath17 such that and @xmath21 $ ] for every @xmath22 . an important class of targets is given by radii following a power law , @xmath23 for some @xmath24 ( see for example @xcite or @xcite for a nice discussion related to the diophantine properties of linear forms ) . fix a norm @xmath25 on @xmath26 , and let @xmath27 denote the euclidean norm on @xmath28 . for @xmath29 and @xmath30 or @xmath31 , we define sets @xmath32^r \subset \r^r\end{aligned}\ ] ] for @xmath30 and @xmath33 for @xmath34 . we then introduce @xmath35 a matrix @xmath36 is said to be _ badly approximable _ if for some @xmath29 , the sequence @xmath37 is bounded . by contrast , matrices @xmath13 for which @xmath38 is unbounded , where @xmath39 is defined as @xmath40 , but with radii @xmath41 instead of @xmath42 are called _ very well approximable _ or vwa . the obvious direction of the borel - cantelli lemma implies that almost every @xmath43 is not very well approximable ( cf . the celebrated khintchine - groshev theorem on diophantine approximation implies that badly approximable matrices are also of zero measure @xcite . analogous definitions apply in the inhomogeneous case of @xmath44 , and similar results hold . for targets given by a power law , the radii @xmath42 are thus the smallest ones to yield an infinite number of hits almost surely . a natural question is then to investigate statistics of these hits , which we call _ resonances_. in the present paper we address in this context the behavior of the resonances on average over @xmath13 and @xmath14 , or on average over @xmath13 while @xmath14 is fixed at @xmath45 or fixed at random . introduce the `` expected '' number of hits @xmath46 when @xmath14 and @xmath13 are uniformly distributed on corresponding tori . let @xmath47 denote the normal distribution with mean @xmath48 and variance @xmath49 . [ thclt - u ] suppose @xmath50 and let @xmath13 be uniformly distributed on @xmath51 . then , @xmath52 converges in distribution to @xmath53 as @xmath54 , where @xmath55 where @xmath56 is the unit ball in @xmath25-norm and vol denotes the euclidean volume . the restriction @xmath57 above is necessary . in fact , it is shown in @xcite using continued fractions that in that case the central limit theorem ( clt ) still holds for @xmath58 , but the correct normalization should be @xmath59 rather than @xmath60 [ thclt - v ] let @xmath13 and @xmath14 be uniformly distributed in @xmath61 . then , @xmath62 converges in distribution to @xmath63 as @xmath54 . the preceding theorems give clts in the cases of @xmath14 fixed to be @xmath5 or @xmath14 random . the clt also holds for for almost every fixed @xmath14 . [ thclt - v - as ] suppose @xmath64 for almost every @xmath14 , if @xmath13 is uniformly distributed in @xmath61 , then @xmath65 converges in distribution to a normal random variable with zero mean and variance one . using a by now standard approach of dani correspondence ( cf . @xcite ) we deduce our results about diophantine approximations from appropriate limit theorems for homogeneous flows . namely we need to prove a clt for siegel transforms of piecewise smooth functions ; these limit theorems are formulated in section [ sclat ] . the reduction of the theorems of section [ scintro ] to those o section [ sclat ] is given in section [ scda - lat ] . the clts in the space of lattices are in turn deduced from an abstract central limit theorem ( theorem [ thcltabs ] ) for weakly dependent random variables which is formulated and proven in section [ scabstract ] . in order to verify the conditions of theorem [ thcltabs ] for the problem at hand we need several results about regularity of siegel transforms which are formulated in section [ sch - sl ] and proven in the appendix . in section [ sec.proofs ] we deduce our central limit theorems for homogeneous flows from the abstract theorem [ thcltabs ] . section [ sec.variances ] contains the proof of the formula on the variances . section [ scgengroup ] discusses some applications of theorem [ thcltabs ] beyond the subject of diophantine approximation . d. d. was partially supported by the nsf . i.v . appreciates the support of fondation sciences mathmatiques de paris during his stay in paris . we let @xmath66 the multiplication rule in @xmath67 takes form @xmath68 . we regard @xmath69 as a subgroup of @xmath67 consisting of elements of the form @xmath70 . we let @xmath71 be the abelian subgroup of @xmath69 consisting of matrices @xmath72 and @xmath73 be the abelian subgroup of @xmath67 consisting of matrices @xmath74 where @xmath5 is an origin in @xmath75 @xmath76 is an @xmath1-dimensional vector , @xmath13 is a @xmath77 matrix and @xmath78 let @xmath79 be the space of @xmath80 dimensional unimodular lattices and @xmath81 be the space of @xmath80 dimensional unimodular affine lattices . we identify @xmath79 and @xmath81 respectively with @xmath82 and @xmath83 . we will need spaces @xmath84 @xmath85 , and @xmath86 of functions which can be well approximated by smooth functions , given @xmath87 . recall first that the space @xmath88 consists of functions @xmath89 whose derivatives up to order @xmath90 are bounded . to define spaces @xmath91 and @xmath92 , fix bases for @xmath93 and @xmath94 ; then , @xmath91 and @xmath92 consist of functions whose derivatives corresponding to monomials of order up to @xmath90 in the basis elements are bounded ( see appendix for formal definitions ) . now we define @xmath95-norm on a space equipped with a @xmath96-norm by @xmath97 some properties of these spaces are discussed in the appendix . given a function @xmath98 on @xmath99 we consider its siegel transforms @xmath100 and @xmath101 defined by @xmath102 we emphasize that siegel transforms of smooth functions are never bounded but the growth of their norms at infinity is well understood , see subsection [ sec.truncation ] . in this section we present general central limit theorems for siegel transforms . the reduction of theorems [ thclt - u ] , [ thclt - v ] , [ thclt - v - as ] to the results stated here will be performed in section [ scda - lat ] . let @xmath103 be a non - negative function supported on a compact set which does not contain @xmath5 . ( the assumption that @xmath98 vanishes at zero is needed to simplify the formulas for the moments of its siegel transform . see proposition [ lmrogers ] . ) denote @xmath104 we say that a subset @xmath105 is _ @xmath106-regular _ if @xmath107 is a union of codimension 1 submanifolds and there is a one - parameter subgroup @xmath108 such that @xmath109 } \cl\cap s\neq \emptyset)\leq k\eps^\alpha.\ ] ] we say that a function @xmath110 is _ @xmath106-regular _ if @xmath111 has a @xmath106-regular boundary and the restriction of @xmath112 on @xmath111 belongs to @xmath113 with @xmath114 @xmath106-regular functions on @xmath81 are defined similarly . let @xmath115 be subgroup of @xmath69 consisting of diagonal matrices . we use the notation @xmath116 for haar measure on @xmath115 . we say that @xmath112 is _ @xmath117-centrally smoothable _ if there is a positive function @xmath118 supported in a unit neighborhood of the identity in @xmath115 such that @xmath119 and @xmath120 has @xmath121 norm less than @xmath122 we say that a function @xmath112 on @xmath81 is _ @xmath117-centrally smoothable _ if @xmath123 is a @xmath117-centrally smoothable function on @xmath124 as before , we write @xmath125 for the normal distribution with @xmath48 and variance @xmath49 and `` @xmath126 '' stands for convergence in distribution . we write @xmath127 for a certain diagonal element of @xmath69 and @xmath67 formally introduced in . [ thclt - u - lat ] suppose that @xmath64 ( a ) there is a constant @xmath128 such that if @xmath129 is uniformly distributed on @xmath79 then @xmath130 as @xmath54 . \(b ) fix constants @xmath131 with @xmath132 suppose that @xmath129 is distributed according to a density @xmath133 which is @xmath134-regular and @xmath135-centrally smoothable . then @xmath136 as @xmath54 . [ thclt - v - lat ] ( a ) let @xmath137 be uniformly distributed on @xmath138 then there is a constant @xmath139 such that @xmath140 as @xmath54 . \(b ) fix constants @xmath131 with @xmath132 suppose that @xmath137 is distributed according to a density @xmath133 which is @xmath134-regular and @xmath135-centrally smoothable . then @xmath141 as @xmath54 . let @xmath142 be an unstable rectangle , that is @xmath143 where @xmath144 is a fixed affine lattice and @xmath145 and @xmath146 are rectangles in @xmath147 and @xmath28 respectively . consider a partition @xmath148 of @xmath142 into @xmath71-rectangles . thus elements of @xmath148 are of the form @xmath149 for some fixed @xmath150 [ thclt - v - lat - as ] suppose that @xmath64 then for each unstable cube @xmath142 and for almost every @xmath151 , if @xmath137 is uniformly distributed in @xmath152 , then @xmath153 as @xmath54 . the explicit calculation of @xmath128 and @xmath139 when @xmath98 is an indicator functions ( the case needed for theorems [ thclt - u]-[thclt - v - as ] ) will be given in section [ sec.variances ] . central limit theorems for partially hyperbolic translations on homogeneous spaces are proven in @xcite ( for bounded observables ) and in @xcite ( for @xmath154 observables ) . ( see also @xcite for important special cases ) . it seems possible to prove theorem [ thclt - u - lat ] for sufficiently large values of @xmath80 by verifying the conditions of @xcite . instead , we prefer to present in the next section an abstract result which will later be applied to derive theorems [ thclt - u - lat ] , [ thclt - v - lat ] , and [ thclt - v - lat - as ] . we chose this approach for three reasons . first , this will make the paper self contained . second , we replace the @xmath154 assumption of @xcite by a weaker @xmath155 assumption which is important for small @xmath156 third , our approach allows to give a unified proof for theorems [ thclt - u - lat ] , [ thclt - v - lat ] , and [ thclt - v - lat - as ] . in this section we reduce theorem [ thclt - u ] , [ thclt - v ] , and [ thclt - v - as ] to theorems [ thclt - u - lat][thclt - v - lat - as ] . to fix our notation we consider @xmath157 and @xmath158 the analyis of @xmath159 and @xmath160 being similar . we also drop the extra subscript and write @xmath157 and @xmath161 as @xmath58 and @xmath162 , respectively , until the end of this section . in subsections [ sec.dani ] and [ sec.measure ] we explain how to reduce theorem [ thclt - u ] to theorem [ thclt - u - lat ] . the reductions of theorems [ thclt - v ] and [ thclt - v - as ] to theorems [ thclt - v - lat ] and [ thclt - v - lat - as ] require only minor modifications which will be detailed in section [ ssred - v ] . in this subsection we use the dani correspondence principle to reduce the problem to a clt for the action of diagonal elements on the space of lattices of the form @xmath163 where @xmath13 is random . let @xmath13 be the matrix with rows @xmath164 , @xmath165 . for @xmath166 and @xmath167 , we denote the @xmath168 diagonal matrix @xmath169 we then consider the following matrices @xmath170 let @xmath118 be the indicator of the set @xmath171 , |x|^{d / r } y_j \in [ 0,c ] \text { for } j=1,\dots , r\right\}\ ] ] and consider its siegel transform @xmath172 . now @xmath173 with @xmath174 contributes to @xmath175 ( from ) precisely when there exists @xmath176 such that @xmath177 clearly such a vector @xmath178 is unique . it is elementary to see that holds if and only if @xmath179 for some integer @xmath180}$ ] . from we obtain [ lmred1 ] for each @xmath181 @xmath182}^{[\log_2 n ] } \phi\circ g^t(\lambda_{{\mathbf a}}\z^{d+r})+r_n,\ ] ] with @xmath183 and @xmath184 denoting the @xmath185-norm with respect to the lebesgue measure on the unit cube in @xmath186 from , it follows that @xmath187 } \leq |n|<n,\\ \text{or } |n|\leq 2^{\log_2^\eps n } \end{array } : ( l_1(n,0,\a_1 ) , \dots , l_r(n,0,\a_r ) ) \in b(n , d , r , c ) \right\ } .\ ] ] note that for a fixed @xmath188 and @xmath189 , the form @xmath190 is uniformly distributed on the circle . hence @xmath191 and so @xmath192}\le |n|\le n } \frac{1}{|n|^d}+\sum_{|n|\le 2^{\log_2^\eps n } } \frac{1}{|n|^d } \ll ( \log_2 n)^\eps.\ ] ] hence , to prove theorem [ thclt - u ] we can replace @xmath175 by @xmath193}^{[\log_2 n ] } \phi\circ g^t(\lambda_{{\mathbf a}}\z^{d+r}).\ ] ] note that the action of @xmath194 on the space of lattices @xmath79 is partially hyperbolic and its unstable manifolds are orbits of the action @xmath195 with @xmath196 ranging in the set of @xmath77 matrices . this will allow us to reduce the proof of clts to clts for the diagonal action on the space of lattices . a similar reduction is possible for the @xmath194-action on the space of affine lattices since in that case the unstable manifolds for the action of @xmath194 are given by @xmath197 , with @xmath196 , @xmath198 let @xmath199 where @xmath200 $ ] . for @xmath201 let @xmath202 be independent uniformly distributed in @xmath203 $ ] . also introduce a random matrix @xmath204 where all the entries of @xmath205 are independent uniformly distributed in @xmath206 $ ] . let @xmath207 and @xmath208 let @xmath209 . it is clear that if @xmath13 is distributed uniformly in a unit cube , then @xmath210 is distributed according to a @xmath211-regular and @xmath135-centrally smoothable density . note that @xmath212 observe also that for @xmath213 and @xmath214 , we have @xmath215 and hence for @xmath216 we have @xmath217 where @xmath218 denotes a @xmath219 neighborhood of the boundary of @xmath220 now the same argument as in lemma [ lmred1 ] gives [ lmred2 ] for each @xmath181 , @xmath193}^{[\log_2 n ] } \phi\circ g^t(\lambda_{{\mathbf a}}\z^{d+r})= \sum_{t=[(\log_2 n)^\eps]}^{[\log_2 n ] } \phi\circ g^t(\tilde{\lambda}({{\mathbf a } } , { { \mathbf b}},{{\mathbf t } } ) \z^{d+r } ) + \tilde{r}_n\ ] ] with @xmath221 now theorem [ thclt - u ] follows from theorem [ thclt - u - lat ] except for the formula for @xmath128 which is derived in section [ sec.variances ] . the reduction of theorems [ thclt - v ] and [ thclt - v - as ] to theorems [ thclt - v - lat ] and [ thclt - v - lat - as ] requires only small changes compared to the preceding section . to wit , lemmas [ lmred1 ] and [ lmred2 ] take the following form . [ lmredv ] ( a ) for each @xmath181 , @xmath222}^{[\log_2 n ] } \tcs(\one_{e_c})(g^t(\lambda_{{\mathbf a}}\z^{d+r}+(0 , { { \bf x } } ) ) ) + r_n\ ] ] where @xmath223 \(b ) let @xmath205 and @xmath224 have the same distribution as in subsection [ sec.measure ] and @xmath225 be uniformly distributed in @xmath226^d$ ] then for each @xmath181 @xmath227}^{[\log_2 n ] } \tcs(\one_{e_c})(g^t(\lambda_{{\mathbf a}}\z^{d+r}+(0 , { { \bf x } } ) ) ) \\ = \sum_{t=[(\log_2 n)^\eps]}^{[\log_2 n ] } \tcs(\one_{e_c})(g^t(\tlambda({{\mathbf a } } , { { \mathbf b } } , { { \mathbf t } } ) \z^{d+r}+({{\bf y } } , { { \bf x } } ) ) ) + \tilde{r}_n \end{gathered}\ ] ] with @xmath228 note that in part ( a ) the error has small @xmath184 norm for each fixed @xmath229 this follows from the fact that for each @xmath14 and @xmath17 , @xmath230 is uniformly distributed on @xmath231 this is useful in the proof of theorem [ thclt - v - as ] since we want to have a control for each ( or at least , most ) @xmath229 we also note that part ( b ) is only needed for theorem [ thclt - v ] since in theorem [ thclt - v - lat - as ] we start with initial conditions supported on a positive codimension submanifold of @xmath138 ( one of the steps in the proof of theorem [ thclt - v - lat - as ] consists of fattening the support of the initial measure , see subsection [ scasclt ] , however lemma [ lmredv](b ) is not needed at the reduction stage ) . in what follows @xmath232 , and @xmath233 are fixed constants . let @xmath234 be a sequence of random variables satisfying the following conditions . write @xmath235 for the corresponding centered random variable . 1 . given any @xmath117 , there is a sequences @xmath236 of random variables such that 1 . @xmath237 almost surely ; 2 . @xmath238 @xmath239 2 . there exists a filtration @xmath240 defined for @xmath241 such that for every @xmath242 there exists a variable @xmath243 that is @xmath244-measurable , @xmath245 almost surely , @xmath246 , and @xmath247 3 . for @xmath242 , there exists an event @xmath248 such that @xmath249 and for @xmath250 @xmath251 4 . there exists a numerical sequence @xmath252 for @xmath253 such that for @xmath250 and @xmath254 , @xmath255 [ thcltabs ] ( a ) under conditions ( h1)-(h4 ) if @xmath256 is distributed according to @xmath257 then @xmath258 with @xmath259 \(b ) suppose conditions ( h1)(h4 ) are satisfied . fix @xmath260 such that @xmath261 and set @xmath262 . suppose that @xmath256 is distributed according to a measure @xmath263 which has a density @xmath264 satisfying ( d1 ) @xmath265 ( d2 ) for each @xmath17 there is an @xmath266-measurable density @xmath267 such that @xmath268 ( d3 ) for each @xmath269 , @xmath270 then , @xmath271 here we show that the normalized variance converges . under conditions ( h1)(h4 ) we have that @xmath272 with @xmath128 as in . first we record a property of cross - terms in the sum that lets us pass from @xmath234 to the truncated sequence . we have @xmath273 indeed , we have @xmath274 now applying the cauchy - schwarz inequality followed by ( h1a ) and ( h1b ) , we arrive at the bound since @xmath233 . next , applying ( h3 ) and ( h4 ) with @xmath275 gives @xmath276 combing and we get @xmath277 take a small number @xmath278 and assume that @xmath279 @xmath280 taking @xmath281 we see that @xmath282 therefore for each @xmath283 , the following limits exist , @xmath284 and moreover @xmath285 next we claim that under the condition @xmath286 there exists @xmath287 such that @xmath288 indeed , @xmath289 using ( h1a ) and ( h2 ) , we get @xmath290 while ( h3 ) shows that @xmath291 proving . combining and we see that for @xmath117 satisfying we have @xmath292 take a small @xmath293 suppose first that @xmath294 then we can take @xmath295 and holds giving @xmath296 for some @xmath297 . in particular combining , and with @xmath298 @xmath299 we obtain @xmath300 also and show that under , @xmath301 choosing @xmath302 and assuming @xmath303 , @xmath304 to prove the lemma , we need to control the sum @xmath305.\ ] ] using for @xmath306 and using otherwise we see that @xmath307 combining this with we see that the limit in exists and moreover that @xmath308 for all the sequel , we fix @xmath309 and let @xmath310 . let @xmath311 $ ] , @xmath312:=m_n$ ] . denote @xmath313 so that @xmath314 we claim that @xmath315 converge to @xmath5 in probability ; it will therefore suffice to prove that @xmath316 to prove the claim , observe first that following the proof of , we have @xmath317 as for the second sum in , we note that @xmath318\ ] ] hence ( h1b ) gives @xmath319 therefore @xmath320 we turn now to the proof of . we start by defining an exceptional set @xmath321 on which we will not be able to exploit the almost independence of @xmath322 from @xmath323 . the exact reasons for the definition of each condition on @xmath324 will appear during the proof . let @xmath325 to be able to use ( h2)(h4 ) we let for @xmath326 @xmath327 next define for @xmath328 @xmath329 and @xmath330 for @xmath331 $ ] with @xmath332 define @xmath333 and let @xmath334 and @xmath335 : k'-k\geq n^{\eps^2/2 } } { \bar g}_{m , k , k'}.\ ] ] finally set @xmath336 observe that ( h2)(h4 ) show that @xmath337 the main step in the proof of the clt is the following [ lemma.clt3 ] for @xmath338 , @xmath339 where @xmath340 is uniform in @xmath341 . to prove we note that @xmath342 where the last step uses that @xmath343 . next , ( h3 ) implies that on @xmath324 @xmath344 to finish the proof of it suffices to show that for @xmath338 , @xmath345 let us estimate the individual terms in this sum . to fix our notation let us suppose that @xmath346 let @xmath347 be a large constant and consider two cases . \(a ) @xmath348 in this case ( h4 ) and give @xmath349 \(b ) @xmath350 and hence @xmath351 then @xmath352 the second term is @xmath353 since @xmath354 for the first term use that @xmath355 to obtain @xmath356 so both @xmath357 and @xmath358 are negligible . combining the estimates of cases ( a ) and ( b ) we obtain completing the proof of . to finish the proof of part ( a ) it remains to derive the central limit theorem from . for @xmath359 set @xmath360 then @xmath361 and so @xmath362 therefore , @xmath363 iterating this recurrence relation @xmath364 times we obtain @xmath365 completing the proof of part ( a ) of theorem [ thcltabs ] . part ( b ) can be established by a similar argument and we just briefly describe the necessary changes . let @xmath366 denote the expectation with respect to @xmath367 that is , @xmath368 to extend the proof of the central limit theorem to the setting of part ( b ) we need to prove and with @xmath366 instead of @xmath369 . for we need to prove the analogues of and . we claim the following . first , still holds with @xmath366 instead of @xmath370 second , @xmath371 ( note that in contrast to the sum here starts with @xmath372 not @xmath373 ) third , @xmath374 where @xmath375 to prove note that @xmath376 so follows from ( d3 ) . observe that once these three points of the claim are established , the rest of the proof of part ( b ) proceeds exactly as in part ( a ) . to obtain the the other two points of our claim we will need the following [ measurability ] there exists a set @xmath377 with @xmath378 such that for @xmath379 , for @xmath380 and for @xmath381 a bounded random variable we have that @xmath382 let @xmath381 be a bounded random variable , @xmath380 and @xmath256 be such that @xmath383 then @xmath384 we prove now that the set where fails has measure that is exponentially small in @xmath385 . the proof consists of two steps . first , it follows form ( d1 ) and ( d2 ) that @xmath386 so @xmath387 is exponentially small by markov s inequality . second , @xmath388 and so @xmath389 is exponentially small due to . lemma [ measurability ] , together with lemma [ lemma.clt3 ] , imply that holds for @xmath366 instead of @xmath369 provided that we decrease slightly the set @xmath324 to @xmath390 . it remains to prove . note that the argument used to establish and in fact gives for @xmath391 @xmath392 hence lemma [ measurability ] implies that for @xmath393 @xmath394 this estimate implies by direct summation . the proof of theorem [ thcltabs ] is thus completed . in case @xmath395 are bounded , one can take @xmath396 @xmath397 so the conditions of theorem [ thcltabs ] simplify as follows . 1 . there exists filtration @xmath398 such that for every @xmath242 there exists a bounded @xmath244-measurable random variable @xmath399 with @xmath400 such that @xmath401 2 . for @xmath242 , there exists @xmath248 such that @xmath249 and for @xmath250 @xmath402 3 . for @xmath250 and @xmath254 @xmath403 if @xmath395 is a bounded sequence satisfying @xmath404@xmath405 then @xmath406 converges as @xmath407 to a normal distribution with zero mean and variance @xmath408 in this section we use the abstract theorem [ thcltabs ] to prove theorem [ thclt - u - lat ] , [ thclt - v - lat ] , [ thclt - v - lat - as ] . for this , we just have to check ( h1)(h4 ) for the case where our probability space is @xmath79 equipped with the haar measure and @xmath409 @xmath410 , where @xmath103 is a positive function supported on a compact set which does not contain @xmath411 before we construct the filtrations and prove ( h1)(h4 ) for the sequence @xmath409 we recall and prove preliminary results about functions defined on the space of lattices , on siegel transforms , and on the action of diagonal matrices . we will cover this in sections [ sec.rogers ] , [ sec.truncation ] , and [ sec.representative ] respectively . then we will prove theorems [ thclt - u - lat ] , [ thclt - v - lat ] , [ thclt - v - lat - as ] in section [ sec.proofs ] . in section [ sec.variances ] , we compute the variances in the special case that interests us of @xmath98 being the characteristic function of @xmath412 given in . this will finish the proof of theorems [ thclt - u ] , [ thclt - v ] , [ thclt - v - as ] . [ lmrogers ] ( * ? ? ? * theorems 3.15 and 3.16 ) , ( * ? ? ? * appendix b ) suppose that @xmath413 is piecewise smooth function that is supported on a compact set which does not contain @xmath411 then , @xmath414 ^ 2 ( \cl ) \ ; d\mu(\cl)= \left[\int_{\r^{d+r } } f({{\bf x } } ) d{{\bf x}}\right]^2\\&\hspace{5 cm } + \sum_{\overset{(p , q)\in \n^2}{gcd(p , q)=1 } } \int_{\r^{d+r } } \left[f(p{{\bf x } } ) f(q{{\bf x}})+f(p{{\bf x } } ) f(-q{{\bf x}})\right ] dx;\\ \intertext{suppose that $ f\colon \tcm \to \r$ is piecewise smooth function of compact support . then , } & ( c ) \quad \int_{\tcm } \tcs(f)(\tcl ) \ ; d\mu(\tcl)=\int_{\r^{d+r } } f({{\bf x } } ) d{{\bf x } } ; \\ & ( d)\quad \int_{\tcm } \left[\tcs(f)\right]^2 ( \tcl ) \ ; d\mu(\tcl)= \left[\int_{\r^{d+r } } f({{\bf x } } ) d{{\bf x}}\right]^2+\int_{\r^{d+r } } f^2 ( { { \bf x } } ) d{{\bf x}}. \end{aligned}\ ] ] let @xmath415 be a one parameter subgroup of @xmath416 for example one can take the matrices with ones on the diagonal , an arbitrary number in the upper left corner and zeros elsewhere . note that @xmath417 the filtrations for which we will prove ( h2)(h4 ) for the sequence @xmath409 will consist of small arcs in the direction of the flow of @xmath415 . the exponential mixing of the @xmath69-action will underly the equidistribution and independence properties that are stated in ( h1)(h4 ) . we will need the notion of representative partitions that was already used in @xcite . these will be partitions of @xmath79 whose elements are segments of @xmath415 orbits , whose pushforwards by @xmath418 will become rapidly equidistributed . to guarantee the filtration property , we would ideally consider an increasing sequence of such partitions with pieces of size @xmath419 , @xmath420 . however , such partitions with fixed size pieces do not exist because @xmath415 is weak mixing . we overcome this technical difficulty due to the following observations : 1 . rudolph s theorem ( see ( * ? ? ? * section 11.4 ) ) shows that for each @xmath421 we can find a partition @xmath422 into @xmath415-orbits such that the length of each element is either @xmath423 or @xmath424 and if @xmath425 then @xmath426 : \cp_u(\cl ) \text { has length } l/\sqrt{2})\leq \breps.\ ] ] 2 . given @xmath427 , it suffices to check the properties ( h2)(h4 ) away from a set of measure less than @xmath428 . having fixed @xmath309 , we will therefore abuse notation and say that a partition is of size @xmath423 if @xmath429 in 1 ) is less than @xmath430 . in light of this , let @xmath422 be a partition of size 1 and @xmath431 be its sub - partition of size @xmath419 . due to ( 1 ) and ( 2 ) we can assume without loss of generality that for every fixed @xmath432 $ ] , the partitions @xmath433 form an increasing sequence and that as a consequence the sequence @xmath434 of @xmath128-algebras generated by @xmath435 forms a filtration . fix a small constant @xmath436 . given a collection @xmath437 a set of natural numbers @xmath438 , and a number @xmath423 , we call a partition @xmath422 of size @xmath423 is _ representative _ with respect to @xmath439 if for each @xmath440 and for each @xmath441 , @xmath442 the curve @xmath443 is of the form @xmath444 with @xmath445 $ ] , and we use the notation @xmath446 for the normalized integral @xmath447 . given a finite collection @xmath448 , @xmath438 , and @xmath449 , we let @xmath450 assume that @xmath451 . then we have as in ( * ? ? ? * proposition 7.1 ) [ prop.rep ] let @xmath452 $ ] be the set of @xmath453 such that @xmath454 is representative with respect to @xmath439 . then @xmath455 . we quickly recall how proposition [ prop.rep ] can be deduced , exactly as in ( * ? ? ? * proposition 7.1 ) , from the polynomial mixing of the unipotent flow @xmath415 . indeed , assuming that @xmath456 , polynomial mixing implies implies that @xmath457 with @xmath458 thus for curves @xmath459 of the form @xmath460 with @xmath461 $ ] we get that @xmath462 with @xmath463 this implies that if we consider a partition @xmath422 of size @xmath423 and its corresponding shifted partitions @xmath454 , @xmath464 $ ] we get for the measure @xmath465}$ ] that @xmath466 : \left| \int_{\cp_u(\cl ) } a \right| > \ck_a l^{-\kappa_0}\right ) \leq c l^{-\kappa_0}\ ] ] where @xmath467 denotes the piece of @xmath454 that goes through @xmath129 . the claim of proposition [ prop.rep ] then follows by markov s inequality . proposition [ prop.rep ] will be used in the next section to obtain a partition of @xmath79 into pieces of @xmath415 orbits satisfying the condition of theorem [ thcltabs ] . we could also use a partition into whole @xmath71-orbits . the proof of proposition [ prop.rep ] in that case would be simpler since we could use effective equidistribution of horospherical subgroups @xcite . we prefer to use @xmath415 orbits instead since it allows us to give unified proofs of theorems [ thclt - u - lat ] , [ thclt - v - lat ] , and [ thclt - v - lat - as ] ( as well as theorem [ thgenaction ] in section [ scgengroup ] ) . in this section we give some useful results on truncations of a siegel transform of a compactly supported function @xmath468 . these bounds are essential to control the truncated @xmath236 that appear in the abstract clt of section [ scabstract ] . we will leave all the proofs and constructions to appendix a. in particular , we will define @xmath469 , with the properties described in lemma [ crhs - cs ] below . we will always use the following notation for @xmath470 ( or @xmath471 ) : @xmath472 in the sequel we will consider @xmath473 . [ crhs - cs ] @xcite there exists a constant @xmath474 such that for each pair of integers @xmath475 and each @xmath347 there is a constant @xmath476 such that the following holds . let @xmath98 be supported on @xmath477 in @xmath478 \(a ) if @xmath479 then @xmath480 \(b ) if @xmath481 then @xmath482 \(c ) if @xmath481 then @xmath483 [ prop.ha1 ] for every @xmath484 there exists @xmath485 such that @xmath470 or @xmath486 satisfy @xmath487 if @xmath488 , then @xmath486 satisfies @xmath489 in addition , the same inequalities hold if the expectation is considered with respect to a measure that has a @xmath135-centrally smoothable density . recall that an @xmath71-rectangle is a set of the form @xmath490 where @xmath13 belongs to the box @xmath491 in @xmath492 also define @xmath493 by @xmath494 [ truncation.xfix ] for each @xmath495 there exists a constant @xmath485 such that for any box @xmath491 whose sides are longer than @xmath421 and which is contained in @xmath496^{dr}$ ] and for any @xmath137 with @xmath497 @xmath498 satisfies @xmath499 where @xmath500 is a restriction of the haar measure on @xmath501 to @xmath148 and @xmath502 is expectation with respect to @xmath503 we are ready now to prove theorem [ thclt - u - lat ] , [ thclt - v - lat ] , [ thclt - v - lat - as ] using theorem [ thcltabs ] . for @xmath98 as in the statement of theorem [ thclt - u - lat ] , recall that we defined @xmath410 and @xmath504 recall also the notation @xmath505 and @xmath506 . we will now prove ( h1)(h4 ) for the sequence @xmath507 . fix any @xmath508 . property ( h1 ) follows from inequalities and of lemma [ prop.ha1 ] . the fact that ( h1 ) fails to hold when @xmath509 is the reason why the clt does not hold in this case . we will use the notion of representative partitions of section [ sec.representative ] to construct the desired filtrations . first of all , note that to prove ( h4 ) we need to deal with function of the form @xmath510 . therefore we define for every @xmath511 the collection of functions @xmath512 fix constants @xmath513 and define for every @xmath514 the following collection of functions and sequences of integers @xmath515 next let @xmath422 be a partition of size 1 and @xmath431 be its subpartition of size @xmath516 by proposition [ prop.rep ] and [ crhs - cs ] there is @xmath453 such that for each @xmath517 , @xmath435 is representative with respect to the collections of integers and functions in . let @xmath434 be the filtration of @xmath128-algebras generated by @xmath435 . denote @xmath518 we claim that @xmath519 satisfies ( h1)(h4 ) with @xmath520 provided that @xmath521 is sufficiently close to 1 . since ( h1 ) has been checked above it remains to verify ( h2)(h4 ) . if @xmath522 then ( h2 ) holds if we take @xmath453 sufficiently large . by lemma [ crhs - cs ] there are functions @xmath523 such that @xmath524 then @xmath525 where @xmath526 and @xmath527 are defined analogously to @xmath528 and @xmath529 with @xmath530 replaced by @xmath531 since @xmath523 are lipschitz , we have @xmath532 so if @xmath533 and @xmath135 is sufficiently large then @xmath534 implies @xmath535 hence markov s inequality gives @xmath536 this proves ( h2 ) provided that @xmath453 is large enough and @xmath537 ( h3 ) follows from the definition of representative partition if @xmath538 while for @xmath539 , @xmath540 provided that @xmath521 is sufficiently close to 1 . likewise , if @xmath541 then ( h4 ) holds by the definition of the representative partition with @xmath542 if @xmath543 we consider two cases \(a ) @xmath544 and so @xmath545 in this case ( h4 ) trivially holds similarly to ( h3 ) . \(b ) @xmath546 and so @xmath547 accordingly to establish ( h4 ) with @xmath548 it suffices to show that there is a constant @xmath549 such that @xmath550 we are going to show that follows from already established ( h1)(h3 ) . the argument is similar to the proof of . namely , denoting by @xmath551 we get @xmath552 ( h1a ) and ( h2 ) imply that @xmath553 next , @xmath554\big|\cf_l\right)\right|\ ] ] and ( h3 ) shows that the expected value of the rhs is @xmath555 now markov s inequality shows that @xmath556 combining the estimates of @xmath357 and @xmath358 we obtain . having checked ( h1)(h4 ) , we have established theorem [ thclt - u - lat](a ) via theorem [ thcltabs](a ) . to prove theorem [ thclt - u - lat](b ) we just need to check condition ( d1)(d3 ) of theorem [ thcltabs](b ) for @xmath557 property ( d1 ) follows from @xmath134-regularity since @xmath558 to check ( d2 ) let @xmath559 then if @xmath560 } \cl\cap \partial(\supp(\rho))=\emptyset\ ] ] then @xmath133 is holder on the element of @xmath433 containing @xmath129 so @xmath561 so the exceptional set for ( d2 ) consists of points violating . this set has a small measure since @xmath562 is @xmath134-regular . finally ( d3 ) follows from inequalities and of lemma [ prop.ha1 ] applied to the centrally smoothable density @xmath133 . the proof of theorem [ thclt - u - lat ] is thus complete . for @xmath98 as in the statement of theorem [ thclt - v - lat ] we define @xmath486 and @xmath563 , with @xmath564 distributed according to haar measure . we also use @xmath505 and @xmath506 . if @xmath57 the analysis is exactly the same as in the case of linear lattices . if @xmath565 , and of lemma [ prop.ha1 ] are not sufficient anymore to prove property ( h1 ) , and we replace them by and . the proof of properties ( h2)(h4 ) proceeds exactly as in the case of linear lattices . theorem [ thclt - v - lat](a ) thus follows from theorem [ thcltabs](a ) . the changes needed for theorem [ thclt - v - lat](b ) are the same as for for theorem [ thclt - u - lat](b ) . observe that ( d3 ) and ( h1 ) hold since are valid if haar measure is replaced with measures having centrally - smoothable densities . here we deduce theorem [ thclt - v - lat - as ] from a refined version of theorem [ thclt - v - lat ] . using we conclude that it is sufficient to prove the central limit theorem for sums with a shorter range of summation , @xmath566 next , take a large constant @xmath567 and for @xmath568 let @xmath569 let @xmath570 we have a partition @xmath571 of @xmath572 into sets of the form @xmath573 [ lmfat ] if @xmath574 is sufficiently large large then there is a constant @xmath575 such that @xmath576 except possibly for a set of @xmath577 of measure @xmath578 first we use to replace @xmath471 by @xmath579 hence denoting @xmath580 we get functions @xmath581 such that @xmath582 denoting @xmath583 we claim that @xmath584 consider for example the case where @xmath585 the opposite case being similar . then @xmath586 @xmath587 the second term can be estimated by @xmath588 proving our claim . next if @xmath574 is sufficiently large then @xmath589 while @xmath590 now the lemma follows from markov s inequality . lemma [ lmfat ] allows to reduce theorem [ thclt - v - lat - as ] to the following result . [ thclt - v - lat - as - thick ] suppose that @xmath64 for each @xmath591 and each @xmath181 the following holds . if @xmath592 denotes the uniform distribution on @xmath593 then @xmath594 the proof of theorem [ thclt - v - lat - as - thick ] is also very similar to the proof of theorem [ thclt - u - lat ] . let us describe the necessary modifications . the property ( h1 ) follows from lemma [ truncation.xfix ] instead of and . to define the required filtration of ( h2)(h4 ) we need to adapt proposition [ prop.rep ] as follows . take @xmath595 going to @xmath5 sufficiently slowly , for example , @xmath596 we let @xmath422 be a partition into segments of @xmath415 orbits of size @xmath595 and @xmath431 the corresponding subpartitions of pieces with length @xmath597 we let @xmath433 be the translates by @xmath415 of these partitions and denote by @xmath598 the collection of pieces of @xmath433 which are contained in @xmath599 we say that @xmath577 is @xmath600-good if there exists @xmath601 $ ] such that for each @xmath602 , @xmath598 is representative with respect to the families .the proof of proposition [ prop.rep ] also shows the following . [ lmmostgood ] given @xmath603 , if we take @xmath604 in sufficiently large then @xmath605 on the other hand if @xmath577 is @xmath600-good then the filtration generated by the partitions @xmath598 satisfies ( h2)-(h4 ) . theorem [ thclt - v - lat - as - thick ] thus follows from theorem [ thcltabs ] . [ rkeff ] the argument given above does not tell us for which @xmath14 theorem [ thclt - v - as ] holds . of course rational @xmath14 have to be excluded due to theorem [ thclt - u ] . now a simple baire category argument shows that theorem [ thclt - v - as ] also fails for very liouvillian @xmath229 it is of interest to provide explicit diophantine conditions which are sufficient for theorem [ thclt - v - as ] . the papers @xcite provide tools which may be useful in attacking this question . the arguments of the previous section are by no means limited to @xmath606 in particular , we have the following result . [ thgenaction ] let @xmath607 be a @xmath608 diffeomorphism , @xmath609 of a manifold @xmath610 and @xmath611 be a @xmath608 flow on that space . suppose that both @xmath607 and @xmath611 preserve a probability measure @xmath612 and there exists @xmath29 such that @xmath613 assume that @xmath614 is polynomially mixing , that is , there exist positive numbers @xmath615 and @xmath117 such that if @xmath616 @xmath617 fix @xmath449 and let @xmath618 be a random variable uniformly distributed on @xmath619.$ ] let @xmath620 satisfy . then for @xmath612 almost all @xmath621 @xmath622 converges as @xmath54 to a normal random @xmath623 variable with zero mean and variance @xmath624 moreover for each @xmath625 there is a constant @xmath135 such that @xmath626 the constant @xmath135 can be chosen uniformly when @xmath423 varies over an interval @xmath627 $ ] for some @xmath628 we note that @xmath610 need not be compact , so different @xmath608 norms on @xmath610 need not be equivalent . in the theorem [ thgenaction ] above we assume that the compositions with @xmath607 preserve @xmath608 norm and moreover @xmath629 the proof of theorem [ thgenaction ] is similar to but easier than the proof of theorem [ thclt - v - lat - as ] . namely since @xmath620 is bounded we only need to check conditions @xmath404@xmath630 fix a partition @xmath148 of @xmath610 into @xmath611 orbit segments of size @xmath631 @xmath632 denote the partition of @xmath610 of the form @xmath633 which has @xmath634 as the boundary point where @xmath635 is the smallest positive number with this property . as in section [ scasclt ] we let @xmath636 be the subpartition of @xmath637 into segments of size @xmath638 consider the following collections . @xmath639\right)\ ] ] and @xmath640 we say that @xmath634 is @xmath600-good if for each @xmath641 the partition @xmath636 is representative with respect to families and . lemma [ lmmostgood ] is easily extended to show that for each @xmath1 , @xmath642 provided that @xmath643 are large enough . hence almost every @xmath634 is @xmath600-good for all sufficiently large @xmath644 next let @xmath645 be the filtration corresponding to @xmath646 if @xmath634 is @xmath600-good then @xmath647 satisfies the conditions @xmath404-@xmath405 which implies the clt in view of theorem [ thcltabs ] . we note the following consequence of theorem [ thgenaction ] . [ corgengroup ] let @xmath648 be a semisimple lie group without compact factors , and @xmath649 be an irreducible lattice . let @xmath415 be a unipotent subgroup which is expanded by an element @xmath650 in the sense that @xmath651 for some @xmath29 . fix @xmath449 and let @xmath618 be a random variable uniformly distributed on @xmath619.$ ] suppose that @xmath1 is large enough and let @xmath620 be a @xmath608 function of zero mean . then for haar almost all @xmath652 @xmath653 converges as @xmath54 to a normal random @xmath623 variable with zero mean and variance @xmath654 moreover for each @xmath625 there is a constant @xmath135 such that @xmath655 the constant @xmath135 can be chosen uniformly when @xmath423 varies over an interval @xmath627 $ ] for some @xmath628 corollary [ corgengroup ] follows from theorem [ thgenaction ] with @xmath656 and @xmath607 and @xmath614 actions of @xmath127 and @xmath657 respectively . the polynomial mixing follows from ( * ? ? ? * theorem 3.4 ) . here we establish . we prove the formula for @xmath658 the computation for @xmath659 is the same . by it suffices to compute @xmath660 - 1 } \cs(\one_{e_c } ) ( g^n \cl ) \right)\ ] ] where the variance is taken with respect to the haar measure on the space of lattices . note that @xmath661 - 1 } \cs(\one_{e_c } ) ( g^n \cl)$ ] can be replaced by @xmath662 where @xmath663 , |x|{^{d / r } } y_j\in [ 0 , c]\ } .\ ] ] now proposition [ lmrogers](b ) gives @xmath664 since @xmath665 , the last integral equals to @xmath666}(|px| ) \one_{[1 , n]}(|qx| ) \prod_j \left(\one_{[0 , c ] } ( p^{1+d / r}|x|{^{d / r } } y_j ) \one_{[0 , c ] } ( q^{1+d / r}|x|{^{d / r } } y_j ) \right ) dx dy\\ = \int_{\r^{d+r } } \one_{{[1/p , n / q]}}(|x| ) \prod_j \left(\one_{[0 , c ] } ( q^{1+d / r}|x|{^{d / r } } y_j ) \right ) dx dy={\frac{c^r}{q^{d+r}}\ ; \int_{\r^d } \one_{{[1/p , n / q]}}(|x| ) \frac{dx}{|x|^d } } \ ] ] to evaluate the last integral we pass to the polar coordinates @xmath667 where @xmath668 is a unit vector in the euclidean norm . then , @xmath669}}(|x| ) \frac{dx}{|x|^d}= \int_{\mathbb{s}^{d-1 } } ds \int_{1/p|s|}^{n / q|s| } \frac{\rho^{d-1 } d\rho}{|s|^d\rho^d}= \ln \left(\frac{np}{q}\right ) \ ; \int_{\mathbb{s}^{d-1 } } \frac{ds}{|s|^d } .\ ] ] the second factor here equals to @xmath670 therefore , @xmath671 where the last step relies on ( * ? ? ? * theorem 288 ) . here we compute the limiting variance for @xmath672 as in section [ ssvaru ] we consider the case of boxes , the computations for balls being similar . the same computation as in section [ ssvaru ] shows that we need to compute @xmath673 where the variance is taken with respect to the haar measure on the space of affine lattices . by proposition [ lmrogers](d ) this variance equals to @xmath674 ^ 2 ( x , y ) dx dy= \int_{\r^{d+r}}\one_{e_c(n ) } ( x , y ) dx dy=\widehat{v}_n.\ ] ] the fact that the variance of @xmath162 has a simpler form than the variance of @xmath58 has the following explanation . let @xmath675 so that @xmath676 then @xmath677 s are pairwise independent ( even though triples @xmath678 are strongly dependent ) and hence uncorrelated ( see e.g. @xcite ) while @xmath679 s are not pairwise independent . for a fixed dimension @xmath166 , we denote by @xmath79 the space of @xmath680 dimensional lattices . we let @xmath681 denote the space of smooth functions on @xmath79 . let @xmath682 be a basis in the space of left invariant vector fields on @xmath124 we let @xmath683 the space @xmath684 of smooth functions on the space of @xmath1-dimensional affine lattices is defined similarly . without the loss of generality we may assume that @xmath691 suppose first that @xmath692 given @xmath693 let @xmath694 be the functions such that @xmath695 without the loss of generality we may assume that @xmath696 since otherwise we can replace @xmath694 by @xmath697 where @xmath698 is an appropriate cutoff function . then @xmath699 this proves the result in case @xmath692 in general , write @xmath700 where @xmath701 @xmath702 and apply the foregoing argument to each term of the product @xmath703 [ lmst - covol ] for each sufficiently large @xmath347 there is a constant @xmath705 such that if @xmath98 is supported on the ball of radius @xmath347 centered at the origin , then @xmath706 and @xmath707 also @xmath708 and are taken from ( ( * ? ? ? * lemma 5.1 ) ) . follows from . indeed suppose that @xmath709 then there exists @xmath710 such that @xmath711 now we have @xmath712 where @xmath713 applying to @xmath714 we get . for example , one can take @xmath727 where @xmath648 is a non negative function with integral one supported on the set @xmath728 we write @xmath729 we can also regard @xmath730 as functions on @xmath81 defined by the formula @xmath731 we are ready now to give the proofs of the statements from section [ sec.truncation ] . \(c ) we already know from part ( b ) that @xmath740 a similar argument shows that @xmath741\circ g^j \| _ { c^{{{\mathbf s } } , { { \mathbf r}}}}=o(2^{j { { \mathbf s } } } k).\ ] ] now the result follows by lemma [ lmprod ] . now we deal with the case of affine lattices and @xmath565 . let @xmath129 be such that @xmath744 we claim that @xmath745 this gives the required improvement of an extra power of @xmath746 that is sufficient to verify and using lemma [ lmcovoltail ] . to show let @xmath747 be the shortest vector in @xmath748 note that @xmath749 is of order @xmath750 thus @xmath129 is contained in a union of lines going in the direction of @xmath747 so that the distance between the lines is almost @xmath751 if we shift @xmath14 in the direction perpendicular to @xmath747 then the probability that one of the shifted lines intersects the ball of a fixed radius around the origin is @xmath752 since @xmath753 due to , the estimate follows . we now show that and hold if haar measure is replaced by a measure having a @xmath135-centrally smoothable density with respect to haar measure . we just prove in the lattice case , since the proofs of as well as the proofs for affine lattices are exactly the same . @xmath754 where the inequality follows from lemma [ lmst - covol ] since @xmath410 with @xmath98 having a compact support . next , the @xmath117-central smoothability of @xmath112 and lemma [ lmcovoltail ] imply that 1 . equation still holds for a measure with density @xmath756 2 . the tail estimate of lemma [ lmcovoltail ] can be proved for measures with centrally smoothable densities following the same lines as the proof of . let @xmath501 and @xmath759 denote the haar measures on @xmath71 and @xmath69 respectively . then @xmath760 similarly for @xmath761 , @xmath762 where the last step follows from . since the integrand depends only on projection of @xmath763 to @xmath79 the integral can be estimated by @xmath764 thus and follow lemma [ lmcovoltail ] .

in this paper , we study the motion of spirals by mean curvature type motion in the ( two dimensional ) plane . our motivation comes from dislocation dynamics ; in this context , spirals appear when a screw dislocation line reaches the surface of a crystal . the first main result of this paper is a comparison principle for the corresponding parabolic quasi - linear equation . as far as motion of spirals are concerned , the novelty and originality of our setting and results come from the fact that , first , the singularity generated by the attached end point of spirals is taken into account for the first time , and second , spirals are studied in the whole space . our second main result states that the cauchy problem is well - posed in the class of sub - linear weak ( viscosity ) solutions . we also explain how to get the existence of smooth solutions when initial data satisfy an additional compatibility condition . [ [ ams - classification ] ] ams classification : + + + + + + + + + + + + + + + + + + + 35k55 , 35k65 , 35a05 , 35d40 [ [ keywords ] ] keywords : + + + + + + + + + spirals , motion of interfaces , comparison principle , quasi - linear parabolic equation , viscosity solutions , mean curvature motion .

1002.0326

in this paper we are interested in curves @xmath0 in @xmath1 which are half lines with an end point attached at the origin . these lines are assumed to move with normal velocity @xmath2 where @xmath3 is the curvature of the line and @xmath4 is a given constant . we will see that this problem reduces to the study of the following quasi - linear parabolic equation in non - divergence form @xmath5 this paper is devoted to the proof of a comparison principle in the class of sub - linear weak ( viscosity ) solutions and to the study of the associated cauchy problem . [ [ continuum - mechanics . ] ] continuum mechanics . + + + + + + + + + + + + + + + + + + + + from the viewpoint of applications , the question of defining the motion of spirals in a two dimensional space is motivated by the seminal paper of burton , cabrera and frank @xcite where the growth of crystals with the vapor is studied . when a screw dislocation line reaches the boundary of the material , atoms are adsorbed on the surface in such a way that a spiral is generated ; moreover , under appropriate physical assumptions , these authors prove that the geometric law governing the dynamics of the growth of the spiral is precisely given by where @xmath6 denotes a critical value of the curvature . we mention that there is an extensive literature in physics dealing with crystal growth in spiral patterns . [ [ different - mathematical - approaches . ] ] different mathematical approaches . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + first and foremost , defining geometric flows by studying non - linear parabolic equations is an important topic both in analysis and geometry . giving references in such a general framework is out of the scope of this paper . as far as the motion of spirals is concerned , the study of the dynamics of spirals have been attracting a lot of attention for more than ten years . different methods have been proposed and developed in order to define solutions of the geometric law . a brief list is given here . a phase - field approach was first proposed in @xcite and the reader is also referred to @xcite . other approaches have been used ; for instance , `` self - similar '' spirals are constructed in @xcite by studying an ordinary differential equation . in @xcite , spirals moving in ( compact ) annuli with homogeneous neumann boundary condition are constructed . from a technical point of view , the classical parabolic theory is used to construct smooth solutions of the associated partial differential equation ; in particular , gradient estimates are derived . we point out that in @xcite , the geometric law is anisotropic , and is thus more general than . in @xcite , the geometric flow is studied by using the level - set approach @xcite . as in @xcite , the author of @xcite considers spirals that typically move into a ( compact ) annulus and reaches the boundary perpendicularly . the starting point of this paper is the following fact : to the best of our knowledge , no geometric flows were constructed to describe the dynamics of spirals by mean curvature by taking into account both the singularity of the pinned point and the unboundedness of the domain . [ [ the - equation - of - interest . ] ] the equation of interest . + + + + + + + + + + + + + + + + + + + + + + + + + we would like next to explain with more details the main aims of our study . by parametrizing spirals , we will see ( _ cf . _ subsection [ subsec : level - set ] ) that the geometric law is translated into the quasi - linear parabolic equation . we note that the coefficients are unbounded ( they explode linearly with respect to @xmath7 ) and that the equation is singular : indeed , as @xmath8 , either @xmath9 or first order terms explode . moreover , initial data are also unbounded . in such a framework , we would like to achieve : uniqueness of weak ( viscosity ) solutions for the cauchy problem for a large class of initial data , to construct a unique steady state ( i.e. a solution of the form @xmath10 ) , and finally to show the convergence of general solutions of the cauchy problem to the steady state as time goes to infinity . this paper is mainly concerned with proving a uniqueness result and constructing a weak ( viscosity ) solution ; the study of large time asymptotic will be achieved in @xcite . [ [ classical - parabolic - theory . ] ] classical parabolic theory . + + + + + + + + + + + + + + + + + + + + + + + + + + + classical parabolic theory @xcite could help us to construct solutions but there are major difficulties to overcome . for instance , giga , ishimura and kohsaka @xcite studied a generalization of in domains of the form @xmath11 with @xmath12 and @xmath13 , with neumann boundary conditions at @xmath14 . roughly speaking , we can say that our goal is to see what happens when @xmath15 and @xmath16 . first , we mentioned above that the equation is not ( uniformly ) parabolic in the whole domain @xmath17 . second , in such analysis , the key step is to obtain gradient estimates . unfortunately , the estimates from @xcite in the case of explode as @xmath18 goes to @xmath19 . third , once a solution is constructed , it is natural to study uniqueness but _ even in the setting of classical solutions _ there are substantial difficulties . to conclude , classical parabolic theory can be useful in order to get existence results , keeping in mind that getting gradient estimates for is not at all easy , but such techniques will not help in proving uniqueness . recently , several authors studied uniqueness of quasilinear equations with unbounded coefficients ( see for instance @xcite ) by using viscosity solution techniques for instance . but unfortunately , eq . does not satisfy the assumptions of these papers . [ [ main - new - ideas . ] ] main new ideas . + + + + + + + + + + + + + + + new ideas are thus necessary to handle these difficulties , both for existence and uniqueness . as far as uniqueness is concerned , one has to figure out what is the relevant boundary condition at @xmath20 . we remark that solutions of satisfy at least formally a neumann boundary condition at the origin @xmath21 in some sense , we thus can say that the boundary condition is embedded into the equation . second , taking advantage of the fact that the neumann condition is compatible with the comparison principle , viscosity solution techniques ( also used in @xcite ) permit us to get uniqueness even if the equation is degenerate and also in a very large class of weak ( sub- and super- ) solutions . but there are remaining difficulties to be overcome . first , the boundary condition is only true asymptotically ( as @xmath22 ) and the fact that it is embedded into the equation makes it difficult to use . we will overcome this difficulty by making a proper change of variables ( namely @xmath23 , see below for further details ) and proving a comparison principle ( whose proof is rather involved ; in particular many new arguments are needed in compare with the classical case ) in this framework . second , classical viscosity solution techniques for parabolic equations do not apply directly to because of polar coordinates . more precisely , the equation do not satisfy the fundamental structure conditions as presented in ( * ? ? ? * eq . ( 3.14 ) ) when polar coordinates are used . but the mean curvature equation has been extensively studied in cartesian coordinates @xcite . hence this set of coordinates should be used , at least far from the origin . [ [ perrons - method - and - smooth - solutions . ] ] perron s method and smooth solutions . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we hope we convinced the reader that it is really useful , if not mandatory , to use viscosity solution techniques to prove uniqueness . it turns out that it can also be used to construct solutions by using perron s method @xcite . this technique requires to construct appropriate barriers and we do so for a large class of initial data . the next step is to prove that these weak solutions are smooth if additional growth assumptions on derivatives of initial data are imposed ; we get such a result by deriving non - standard gradient estimates ( with viscosity solution techniques too ) . we would like also to shed some light on the fact that this notion of solution is also very useful when studying large time asymptotic ( and more generally to pass to the limit in such non - linear equations ) . indeed , convergence can be proved by using the half - relaxed limit techniques if one can prove a comparison principle . see @xcite for more details . in this section , we make precise the way spirals are defined . we will first define them as parametrized curves . [ [ parametrization - of - spirals . ] ] parametrization of spirals . + + + + + + + + + + + + + + + + + + + + + + + + + + + we look for interfaces @xmath24 parametrized as follows : @xmath25 for some function @xmath26 . if now the spiral moves , i.e. evolves with a time variable @xmath27 , then the function @xmath28 also depends on @xmath29 . a moving _ spiral _ is a family of curves @xmath0 of the following form @xmath30 for some function @xmath31 . this curve is oriented by choosing the normal vector field equal to @xmath32 . with the previous definition in hand , the geometric law implies that @xmath28 satisfies with the initial condition @xmath33 [ [ link - with - the - level - set - approach . ] ] link with the level - set approach . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + in view of , we see that our approach is closely related to the level - set one . we recall that the level - set approach was introduced in @xcite ; in particular , it permits to construct an interface moving by mean curvature type motion , that is to say satisfying the geometric law . it consists in defining the interface @xmath34 as the @xmath19-level set of a function @xmath35 and in remarking that the geometric law is verified only if @xmath36 satisfies a non - linear evolution equation of parabolic type . in an informal way , we can say that the quasi - linear evolution equation is a `` graph '' equation associated with the classical mean curvature equation ( mce ) , but written in polar coordinates . more precisely , if @xmath37 with @xmath38 , then @xmath28 will satisfy as long as @xmath36 solves the following level - set equation @xmath39 ( where @xmath40 and @xmath41 for @xmath42 ) . notice that the angle @xmath43 is multivalued , i.e. only defined modulo @xmath44 . such an approach is for instance systematically developed in @xcite . [ [ comparison - principle . ] ] comparison principle . + + + + + + + + + + + + + + + + + + + + + our first main result is a comparison principle : it says that all sub - solutions lie below all super - solutions , provided they are ordered at initial time . [ thm comp spirales sous lin en r ] assume that @xmath45 is a globally lipschitz continuous function . consider a sub - solution @xmath28 and a super - solution @xmath46 of , ( in the sense of definition [ defi::1 ] ) such that there exist @xmath47 and for all @xmath48 and @xmath49 , @xmath50 if @xmath51 for all @xmath52 , then @xmath53 in @xmath54 . the growth of the sub - solution @xmath55 and the super - solution @xmath56 is made precise by assuming condition . such a condition is motivated by the large time asymptotic study carried out in @xcite ; indeed , we construct in @xcite a global solution of the form @xmath57 . the proof of theorem [ thm comp spirales sous lin en r ] is rather involved and we will first state and prove a comparison principle in the set of bounded functions for a larger class of equations ( see theorem [ thm comp spirales ] ) . we do so in order to exhibit the structure of the equation that makes the proof work . we then turn to the proof of theorem [ thm comp spirales sous lin en r ] . both proofs are based on the doubling of variable method , which consists in regularizing the sub- and super - solutions . obviously , this is a difficulty here because one end point of the curve is attached at the origin and the doubling of variables at the origin is not well defined . to overcome this difficulty , we work with logarithmic coordinates @xmath23 for @xmath7 close to @xmath19 . but then the equation becomes @xmath58 we then apply the doubling of variables in the @xmath59 coordinates . there is a persistence of the difficulty , because we have now to bound terms like @xmath60 that can blow up as @xmath61 . we are lucky enough to be able to show roughly speaking that @xmath62 can be controlled by the doubling of variable of the term @xmath63 which appears to be the main term ( in a certain sense ) as @xmath59 goes to @xmath64 . in view of the study from @xcite , @xmath65 has to be chosen sub - linear in cartesian coordinates and thus so are the sub- and super - solutions to be compared . the second difficulty arises when passing to logarithmic coordinates for large @xmath7 s ; indeed , the sub - solution and the super - solution then grow exponentially in @xmath66 at infinity and we did not manage to adapt the previous reasoning in this setting . there is for instance a similar difficulty when dealing with the mean curvature equatio . indeed , in this framework , for super - linear initial data , the uniqueness of the solution is not known in full generality ( see @xcite ) . in other words , the change of variables do not seem to work far from the origin . we thus have to stick to cartesian coordinates for large @xmath7 s ( using a level - set formulation ) and see the equation in different coordinates when @xmath7 is either small or large ( see section [ sec : sublinear ] ) . [ [ existence - theorem . ] ] existence theorem . + + + + + + + + + + + + + + + + + + in order to get an existence theorem , we have to restrict the growth of derivatives of the initial condition . we make the following assumptions : the initial condition is globally lipschitz continuous and its mean curvature is bounded . we recall that the mean curvature of a spiral parametrized by @xmath28 is defined by @xmath67 we can now state our second main result . [ th : cauchywc ] consider @xmath68 . assume that @xmath69 is globally lipschitz continuous and that @xmath70 . then there exists a unique solution @xmath55 of , such that for all @xmath71 , there exists @xmath72 such that for all @xmath73 and @xmath49 , latexmath:[\[\label{baru : baru0sst } @xmath28 is lipschitz continuous with respect to space and @xmath75-hlder continuous with respect to time . more precisely , there exists a constant @xmath76 depending only on @xmath77 and @xmath78 such that @xmath79 and @xmath80 allows us to consider an initial data @xmath65 which does not satisfy the compatibility condition ( [ neum ] ) , like for instance @xmath81 with @xmath82 . notice also that we do not know if the solution constructed in theorem [ th : cauchywc ] is smooth ( i.e. belongs to @xmath83 ) . to get such a result , we first construct smooth solutions requiring that the compatibility condition ( [ neum ] ) is satisfied by the initial datum , like in the following result . [ th : cauchy ] assume that @xmath84 with @xmath85 and that it satisfies the following compatibility condition for some @xmath86 : @xmath87 which is @xmath88 in @xmath89 , which satisfies , , and such that there exists @xmath90 such that @xmath91 and @xmath92 ) and to replace it by the lipschitz estimate ( [ baru : baru0 ] ) . with the help of this lipschitz estimate ( [ baru : baru0 ] ) , we can conclude that the solution constructed in theorem [ th : cauchy ] is smooth . notice also that our space - time lipschitz estimates on the solution allow us to conclude that @xmath93 satisfies ( [ cond : compatibilite ] ) with the constant @xmath76 replaced by some possible higher constant . this implies in particular that @xmath94 satisfies the compatibility condition ( [ neum ] ) for all time @xmath95 . [ [ open - questions . ] ] open questions . + + + + + + + + + + + + + + + @xmath96 + * a. weaker conditions on the initial data * + it would be interesting to investigate the existence / non - existence and uniqueness / non - uniqueness of solutions when we allow the initial data @xmath65 to be less than globally lipschitz . for instance what happens when the initial data describes an infinite spiral close to the origin @xmath20 , with either @xmath97 or @xmath98 ? on the other hand , what happens if the growth of @xmath65 is super - linear as @xmath7 goes to @xmath99 ? + * b. more general shapes than spiral * + one of our main limitation to study only the evolution of spirals in this paper is that we were not able to prove a comparison principle in the case of the general level - set equation ( [ eq::rm2 ] ) . the difficulty is the fact that the gradient of the level - set function @xmath36 may degenerate exactly at the origin where the curve is attached . the fact that a spiral - like solution is a graph @xmath100 prevents the vanishing of the gradient of @xmath36 at the origin @xmath20 . if now we consider more general curves attached at the origin , it would be interesting to study the existence and uniqueness / non - uniqueness of solutions with general initial data , like the curves on figure [ f4 ] . [ [ organization - of - the - article . ] ] organization of the article . + + + + + + + + + + + + + + + + + + + + + + + + + + + + in section [ prel ] , we recall the definition of viscosity solutions for the quasi - linear evolution equation of interest in this paper . the change of variables that will be used in the proof of the comparison principle is also introduced . in section [ sec : bounded ] , we give the proof of theorem [ thm comp spirales sous lin en r ] in the case of bounded solutions . the proof in the general case is given in section [ sec : sublinear ] . in section [ sec : classical ] , a classical solution is constructed under an additional compatibility condition on the initial datum ( see theorem [ th : cauchy ] ) . first , we construct a viscosity solution by perron s method ( subsection [ subsec : perron ] ) ; second , we derive gradient estimates ( subsection [ subsec : gradestim ] ) ; third , we explain how to prove that the viscosity solution is in fact a classical one ( subsection [ subsec : class ] ) . the construction of the solution without compatibility assumption ( theorem [ th : cauchywc ] ) is made in section [ sec : withoutcompatibility ] . finally , proofs of technical lemmas are gathered in appendix [ app : a ] . [ [ notation . ] ] notation . + + + + + + + + + if @xmath18 is a real number , @xmath101 denotes @xmath102 and @xmath103 denotes @xmath104 if @xmath105 , @xmath106 , then @xmath107 denotes @xmath108 and @xmath109 denotes @xmath110 . in view of , it is convenient to introduce the following notation @xmath111 we first recall the notion of viscosity solution for an equation such as . [ defi::1 ] + let @xmath112 $ ] . a lower semi - continuous ( resp . upper semi - continuous ) function @xmath113 is a ( viscosity ) _ super - solution _ ( resp . _ sub - solution _ ) of , if for any @xmath114 test function @xmath115 such that @xmath116 reaches a local minimum ( resp . maximum ) at @xmath117 , we have * if @xmath118 : @xmath119 * if @xmath120 : @xmath121 a continuous function @xmath122 is a ( viscosity ) _ solution _ of , if it is both a super - solution and a sub - solution . we do not impose any condition at @xmath20 ; in other words , it is not necessary to impose a condition on the whole parabolic boundary of the domain . this is due to the `` singularity '' of our equation at @xmath20 . since we only deal with this weak notion of solution , ( sub-/super-)solutions will always refer to ( sub-/super-)solutions in the viscosity sense . when constructing solutions by perron s method , it is necessary to use the following classical discontinuous stability result . the reader is referred to @xcite for a proof . [ pro : stability ] consider a family @xmath123 of sub - solutions of , which is uniformly bounded from above . then the upper semi - continuous envelope of @xmath124 is a sub - solution of , . we will make use of the following change of unknown function : @xmath125 with @xmath23 satisfies for all @xmath29 and @xmath126 @xmath127 submitted to the initial condition : for all @xmath126 , @xmath128 where @xmath129 . . can be rewritten @xmath130 with @xmath131 remark that functions @xmath132 and @xmath133 are related by the following formula @xmath134 since the function @xmath135 is increasing and maps @xmath136 onto @xmath137 , we have the following elementary lemma which will be used repeatedly throughout the paper . [ lem : equiv ] a function @xmath28 is a solution of , if and only if the corresponding function @xmath55 is a solution of - with @xmath138 . the reader is referred to @xcite ( for instance ) for a proof of such a result . when proving the comparison principle in the general case , we will also have to use cartesian coordinates . from a technical point of view , the following lemma is needed . [ lem : cartesien ] we consider a sub - solution @xmath55 ( resp . super - solution @xmath56 ) of - and we define the function @xmath139 ( resp . @xmath140 ) @xmath141 by @xmath142 where @xmath143 is defined such that @xmath144 . then @xmath139 ( resp . @xmath140 ) is sub - solution ( resp . super - solution ) of @xmath145 in lemma [ lem : cartesien ] , for @xmath146 , the angle @xmath147 is only defined modulo @xmath44 , but is locally uniquely defined by continuity . then @xmath148 are always uniquely defined . as explained in the introduction , we first prove a comparison principle for in the class of bounded weak ( viscosity ) solutions . in comparison with classical comparison results for geometric equations ( see for instance @xcite ) , the difficulty is to handle the singularity at the origin ( @xmath20 ) . in order to clarify why a comparison principle holds true for such a singular equation , we consider the following generalized case @xmath149 which can be written , with @xmath150 , @xmath151 where @xmath152 . [ [ assumption - on - mathbf - bsigma . ] ] assumption on @xmath153 . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + * @xmath154 ; * there exists @xmath155 such that * * for all @xmath156 and @xmath157 , @xmath158 * * for all @xmath159 and @xmath160 , @xmath161 @xmath162 * * for all @xmath163 @xmath164 * * we have @xmath165 . in our special case , @xmath166 and @xmath167 , and the assumption on @xmath168 is satisfied . [ thm comp spirales ] assume that @xmath169 is lipschitz continuous . consider a bounded sub - solution @xmath55 and a bounded super - solution @xmath56 of , in the sense of definition [ defi::1 ] with @xmath170 given by the right hand side of ( [ eq : gen - r ] ) . then @xmath171 in @xmath172 . for radial solutions of the heat equation @xmath173 on @xmath174 , we get @xmath175 and @xmath176 . therefore the assumption on @xmath168 is satisfied if and only if @xmath177 . notice that in particular for @xmath178 this assumption is not satisfied . we classically fix @xmath71 and argue by contradiction by assuming that @xmath179 [ lem : penal ] for @xmath180 small enough , and any @xmath181 , the supremum @xmath182 is attained at @xmath183 with @xmath29 , @xmath184 , @xmath185 and @xmath186 for some @xmath187 only depending on @xmath188 and @xmath189 . the fact that @xmath190 means that there exist @xmath191 and @xmath192 such that @xmath193 since @xmath55 and @xmath56 are bounded functions , @xmath194 is attained at a point @xmath183 . by optimality of @xmath183 , we have in particular @xmath195 for @xmath196 and @xmath197 small enough ( only depending on @xmath198 ) . in particular , @xmath199 hence , there exists a constant @xmath200 only depending on @xmath188 and @xmath189 such that latexmath:[\[\label{estim : penal } assume that @xmath120 . in this case , we use the fact that @xmath202 is lipschitz continuous and in order to get @xmath203 which is absurd for @xmath204 small enough ( depending only on @xmath205 and @xmath206 ) . hence @xmath29 and the proof of the lemma is complete . in the remaining of the proof , @xmath207 is fixed ( even if we will choose it small enough ) and @xmath196 goes to @xmath19 ( even if it is not necessary to pass to the limit ) . in view of the previous discussion , we can assume that , for @xmath207 small enough , we have @xmath29 for all @xmath208 small enough ( independent on @xmath207 ) . we thus can write two viscosity inequalities . it is then classical to use jensen - ishii s lemma and combine viscosity inequalities in order to get the following result . [ lem : ji ] @xmath209 where @xmath210 . jensen - ishii s lemma @xcite implies that for all @xmath211 , there exist four real numbers @xmath212 such that @xmath213 moreover @xmath214 satisfy the following inequality @xmath215 ( it is in fact an equality ) and for any @xmath216 small enough , there exist two real numbers @xmath217 satisfying the following matrix inequality @xmath218 \le \frac{e^{kt}}{\eps } ( 1+{\gamma_1 } ) \left[\begin{array}{cc } 1 & -1 \\ -1 & 1 \end{array } \right ] \ , .\ ] ] this matrix inequality implies @xmath219 for all @xmath220 . use this inequality with @xmath221 and @xmath222 and let @xmath223 in order to get the desired inequality . we can next make appear which error terms depend on @xmath196 and which ones depend on @xmath207 . for all @xmath224 , @xmath225 where @xmath226 and @xmath227 with @xmath228 it remains to estimate @xmath229 . [ lem : estim1 ] for all @xmath230 , @xmath231 where @xmath232 . through a taylor expansion , we obtain @xmath233 where @xmath234 for some @xmath235 $ ] , and we have used the fact that for all @xmath236 , @xmath237 we get for any @xmath230 @xmath238 where we have used that @xmath239 . now , since @xmath240 , we also have @xmath241 and we can conclude . combining and finally yields @xmath242 it suffices now to choose @xmath243 , @xmath244 , such that @xmath245 and then choose @xmath246 and @xmath207 small enough to get : @xmath247 . the following lemma permits to estimate @xmath248 . [ lem : estim2 ] for @xmath249 large enough , we have @xmath250 using assumptions on @xmath251 immediately yields @xmath252 where we used that @xmath253 . we next consider @xmath12 such that for all @xmath254 , we have @xmath255 we now distinguish cases . * case 1 : @xmath256 for some @xmath257 . * we choose @xmath258 large enough so that @xmath259 and we get @xmath260 which implies @xmath261 . contradiction . * case 2 : @xmath262 for all @xmath196 small enough . * we use and get @xmath263 and we let @xmath264 to get @xmath261 in this case too . the proof is now complete . thanks to the change of unknown function described in subsection [ subsec : change ] , we can consider the functions @xmath55 and @xmath56 defined on @xmath265 which are sub- and super - solutions of . we can either prove that @xmath53 in @xmath266 or that @xmath171 in @xmath265 . for @xmath267 , we define @xmath268 note that @xmath269 and @xmath270 are respectively sub and super - solution of @xmath271 we fix @xmath71 and we argue by contradiction by assuming that @xmath272 , x , \theta\in { { \mathbb r } } } \{u(t , x,\theta)-v(t , x,\theta)\}>0.\ ] ] in order to use the doubling variable technique , we need a smooth interpolation function @xmath273 between polar coordinates for small @xmath7 s and cartesian coordinates for large @xmath7 s . precisely , we choose @xmath273 as follows . [ lem : psi ] there exists a smooth ( @xmath88 ) function @xmath274 such that @xmath275 and such that there exists two constants @xmath276 and @xmath277 such that for @xmath278 and @xmath279 , @xmath280 and such that for all @xmath281 , if @xmath282 and @xmath283 , then @xmath284-tensor @xmath285 and a @xmath286-tensor @xmath287 by @xmath288 the proof of this lemma is given in appendix [ app : a ] . [ [ penalization . ] ] penalization . + + + + + + + + + + + + + we consider the following approximation of @xmath198 @xmath289,x,\theta , y,\sigma \in { { \mathbb r } } } \bigg\ { u(t , x,\theta)-v(t , y,\sigma ) \\ -e^{kt}\frac{|\psi(x,\theta)- \psi(y,\sigma)|^2}{2{\varepsilon}}-\frac 1 { \varepsilon}\left(|\theta-\sigma|-\frac \pi 3\right)_+^2 -\frac \a 2 \left|\psi(x,\theta)\right|^2 -\frac \eta { t - t}\bigg\}\end{gathered}\ ] ] where @xmath290 are small parameters and @xmath291 is a large constant to be fixed later . for @xmath292 and @xmath197 small enough we remark that @xmath293 . in order to prove that the maximum @xmath294 is attained , we need the following lemma whose proof is postponed until appendix [ app : a ] . [ lem : est ] there exists a constant @xmath295 such that the following estimate holds true for any @xmath296 @xmath297 using this lemma , we then deduce that @xmath298 using the @xmath44-periodicity of @xmath299 , the maximum is achieved for @xmath300 $ ] . then , using the previous estimate and the fact that @xmath301 as @xmath302 , we deduce that the maximum is reached at some point that we still denote @xmath303 . [ [ penalization - estimates . ] ] penalization estimates . + + + + + + + + + + + + + + + + + + + + + + + using estimate and the fact that @xmath304 , we deduce that there exists a constant @xmath305 such that @xmath306 on the one hand , an immediate consequence of this estimate is that @xmath307 for @xmath204 small enough . on the other hand , we deduce from and @xmath308 hence , we have @xmath309 for @xmath204 small enough so that the constraint @xmath310 is not saturated . we can also choose @xmath204 small enough so that @xmath311 in the sequel of the proof , we will also need a better estimate on the term @xmath312 ; precisely , we need to know that @xmath313 as @xmath314 . even if such a result is classical ( see @xcite ) , we give details for the reader s convenience . to prove this , we introduce @xmath315 , x,\theta , y , \sigma \in { { \mathbb r } } } \bigg\ { & u(t , x,\theta)-v(t , y,\sigma)-e^{kt}\frac{|\psi(x,\theta)- \psi(y,\sigma)|^2}{2{\varepsilon}}\\ & -\frac 1 { \varepsilon}\left(|\theta-\sigma|-\frac \pi 3\right)^2_+ -\frac \eta { t - t}\bigg\}\end{aligned}\ ] ] which is finite thanks to . we remark that @xmath316 and that @xmath294 is non - decreasing when @xmath196 decreases to zero . we then deduce that there exists @xmath317 such that @xmath318 as @xmath314 . a simple computation then gives that @xmath319 and then @xmath320 [ [ initial - condition . ] ] initial condition . + + + + + + + + + + + + + + + + + + we now prove the following lemma . [ avoiding @xmath120][lem : ci ] for @xmath204 small enough , we have @xmath29 for all @xmath322 small enough . we argue by contradiction . assume that @xmath120 . we then distinguish two cases . if the corresponding @xmath59 and @xmath323 are small ( @xmath324 and @xmath325 ) then , since @xmath202 is lipschitz continuous and holds true , there exists a constant @xmath326 such that @xmath327 which is absurd for @xmath204 small enough . the other case corresponds to large @xmath59 and @xmath323 ( @xmath328 and @xmath329 ) . in this case , since @xmath65 is lipschitz continuous , we know that there exists a constant @xmath330 such that @xmath331 using the fact @xmath332 we get a contradiction for @xmath204 small enough . thanks to lemma [ lem : ci ] , we will now write two viscosity inequalities , combine them and exhibit a contradiction . we recall that we have to distinguish cases in order to determine properly in which coordinates viscosity inequalities must be written ( see the introduction ) . [ [ case-1-there - exists - a_nto-0-such - that - xge - frac-32-and - yge - frac-32 . ] ] case 1 : there exists @xmath333 such that @xmath334 and @xmath335 . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we set @xmath336 and @xmath337 . consider @xmath139 and @xmath140 defined in lemma [ lem : cartesien ] . remark that , even if @xmath338 is defined modulo @xmath44 , the quantity @xmath339 is well defined ( for @xmath340 and @xmath341 ) and thus so is @xmath342 . recall also that @xmath343 are respectively sub and super - solutions of the following equation @xmath344 moreover , using the explicit form of @xmath299 , we get that @xmath345 , x , y \in { { \mathbb r}}^2 \backslash b_1(0 ) } \left\{\tilde u(t , x)-\tilde v(t , y ) - \frac { e^{kt}}{2{\varepsilon } } |x - y|^2 - \frac \a 2 |x|^2 - \frac \eta{t - t}\right\}.\ ] ] moreover , @xmath346 ( in the viscosity sense ) . we set @xmath347 we now use the jensen - ishii lemma @xcite in order to get four real numbers @xmath348 such that @xmath349 moreover , @xmath350 satisfies the following estimate @xmath351 satisfy the following equality @xmath352 and @xmath217 satisfy the following matrix inequality @xmath353 \le \frac { 2 e^{kt}}{{\varepsilon } } \left[\begin{array } { cc } i&-i\\ -i&i \end{array}\right].\ ] ] this matrix inequality implies @xmath354 for all @xmath355 . subtracting the two viscosity inequalities , we then get @xmath356 where we have used successively , and . recalling , by that @xmath357 , we get a contradiction for @xmath292 small enough . [ [ case-2-there - exists - a_nto-0-such - that - xle -- frac-12-and - yle -- frac-12 . ] ] case 2 : there exists @xmath333 such that @xmath358 and @xmath359 . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + using the explicit form of @xmath299 and the fact that @xmath360 and @xmath361 with @xmath55 and @xmath56 respectively sub and super - solution of , we remark that @xmath362 moreover , the maximum is reached at @xmath183 , where we recall that @xmath303 is the point of maximum in . using the jensen - ishii lemma @xcite , we then deduce the existence , for all @xmath211 , of four real numbers @xmath212 such that @xmath363 where @xmath364 these inequalities are exactly and . moreover @xmath214 satisfy the following inequality @xmath365 and we obtain . moreover , @xmath217 satisfy the following matrix inequality @xmath366 \le \frac{e^{kt}}{\eps } ( 1+{\gamma_1 } ) \left[\begin{array}{cc } 1 & -1 \\ -1 & 1 \end{array } \right]\ ] ] which implies . on one hand , from , , and , we can derive . on the other hand , , the fact that @xmath367 , @xmath368 and lemma [ lem : psi ] imply ( with a different constant ) . we thus can apply lemmas [ lem : estim1 ] , [ lem : estim2 ] and deduce the desired contradiction . [ [ case-3-there - exists - a_nto-0-such - that--1le - x - y - le-2 . ] ] case 3 : there exists @xmath333 such that @xmath369 . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + since @xmath370 , there then exists @xmath371 ( only depending on the function @xmath299 ) such that for all @xmath372 $ ] and @xmath373 $ ] , @xmath374 for simplicity of notation , we denote @xmath375 by @xmath376 and @xmath377 by @xmath378 . we next define @xmath379 we have @xmath380 and we set @xmath381 a basis of @xmath1 . we have for @xmath382 @xmath383 where @xmath384 recall that @xmath269 and @xmath270 are respectively sub and super - solution of and use the jensen - ishii lemma @xcite in order to deduce that there exist two real numbers @xmath214 and two @xmath385 real matrices @xmath217 such that @xmath386 where @xmath387 remark that , since @xmath388 , there exists @xmath389 such that @xmath390 moreover @xmath214 satisfy the following equality @xmath391 and @xmath217 satisfy the following matrix inequality @xmath392 \le \frac{2 e^{kt}}{\eps}\bigg\ { \left[\begin{array}{cc } ( \psi(\bar x)-\psi(\bar y))\odot d^2\psi(\bar x ) & 0 \\ 0 & -(\psi(\bar x)-\psi(\bar y))\odot d^2\psi(\bar y ) \end{array } \right]\\ + \left[\begin{array}{cc } d\psi(\bar x)^t \odot d\psi(\bar x ) & -d\psi(\bar y)^t\odot d\psi(\bar x ) \\ -d\psi(\bar y)^t\odot d\psi(\bar x ) & d\psi(\bar y)^t\odot d\psi(\bar y ) \end{array } \right]\bigg\ } \ , .\end{gathered}\ ] ] this implies @xmath393 for all @xmath394 . combining the two viscosity inequalities and using the fact that @xmath395 , we obtain @xmath396 where @xmath397 and @xmath398 are defined respectively as @xmath399 and @xmath400 with @xmath401 replaced by @xmath402 . remarking that there exists a constant @xmath403 such that @xmath404 and @xmath405 and sending @xmath314 ( recall that @xmath406 lie in a compact domain ) , we get . [ lem : tech1 ] there exists a constant @xmath407 such that @xmath408 thanks to and @xmath409 , we have @xmath410 where we have used the fact that @xmath411 ( see ) . this finally gives that there exists a constant @xmath407 ( depending on @xmath412 and @xmath413 ) such that holds true . using the fact that @xmath414 , we can prove in a similar way the following lemma . [ lem : tech2 ] there exist three positive constants @xmath415 and @xmath416 such that @xmath417 use now lemmas [ lem : tech1 ] and [ lem : tech2 ] in order to derive from the following inequality @xmath418 with @xmath419 . choosing @xmath420 , we get a contradiction . in this section , our main goal is to prove theorem [ th : cauchy ] which claims the existence and uniqueness of classical solutions under suitable assumptions on the initial data @xmath65 . notice that assumptions on the initial data imply in particular that @xmath421 to prove theorem [ th : cauchy ] , we first construct a unique weak ( viscosity ) solution . we then prove gradient estimates from which it is not difficult to derive that the weak ( viscosity ) solution is in fact smooth ; in particular , it thus satisfies the equation in a classical sense . before constructing solutions of submitted to the initial condition , we first construct appropriate barrier functions . [ prop : barriers ] assume that @xmath84 and @xmath422 with @xmath69 such that holds true . then there exists a constant @xmath423 such that @xmath424 are respectively a super- and a sub - solution of , . it is enough to prove that the following quantity is finite @xmath425 with @xmath426 } \frac{| \bar{f}(r,(\bar{u}_0)_r ( r ) , ( \bar{u}_0)_{rr}(r))|}r , \quad \bar{c}_2 = \sup_{r \in [ { \color{black } r_0},+\infty ) } \frac{| \bar{f}(r,(\bar{u}_0)_r ( r ) ( x ) , ( \bar{u}_0)_{rr}(r))|}r.\ ] ] on one hand , thanks to and the lipschitz regularity of @xmath65 , we have @xmath427 is finite . on the other hand , thanks to lipschitz regularity and @xmath428 or @xmath429 , @xmath430 is also finite . the proof is now complete . we now construct a viscosity solution for , ; this is very classical with the results we have in hand , namely the strong comparison principle and the existence of barriers . however , we give a precise statement and a sketch of proof for the sake of completeness . [ pro : cauchy ] assume that @xmath431 and that there exists @xmath432 for some continuous function @xmath433 satisfying @xmath434 , which are respectively a super- and a sub - solution of , . then , there exists a ( continuous ) viscosity solution @xmath435 of , such that holds true for some constant @xmath436 depending on @xmath433 . moreover @xmath28 is the unique viscosity solution of , such that holds true . in view of lemma [ lem : equiv ] , it is enough to construct a solution @xmath55 of satisfying with @xmath437 . consider the set @xmath438 remark that it is not empty since @xmath439 ( where @xmath440 with @xmath23 ) . we now consider the upper envelope @xmath55 of @xmath441 . by proposition [ pro : stability ] , it is a sub - solution of . the following lemma derives from the general theory of viscosity solutions as presented in @xcite for instance . [ lem : bump ] the lower envelope @xmath442 of @xmath55 is a super - solution of . we recall that the proof of this lemma proceeds by contradiction and consists in constructing a so - called bump function around the point the function @xmath442 is not a super - solution of the equation . the contradiction comes from the maximality of @xmath55 in @xmath443 . since for all @xmath444 , @xmath445 with @xmath434 we conclude that @xmath446 if @xmath28 satisfies , we use the comparison principle and get @xmath447 in @xmath448 for all @xmath71 . since @xmath449 by construction , we deduce that @xmath450 is a solution of . the comparison principle also ensures that the solution we constructed is unique . the proof of proposition [ pro : cauchy ] is now complete . in this subsection , we derive gradient estimates for a viscosity solution @xmath435 of satisfying . [ prop : gradestim ] consider a globally lipschitz continuous function @xmath69 . we denote by @xmath326 and @xmath451 such that for all @xmath49 , @xmath452 let @xmath28 be a viscosity solution @xmath435 of , satisfying . then @xmath453 is also lipschitz continuous in space : @xmath454 , @xmath455 , @xmath456 moreover , if @xmath84 with @xmath457 and holds true , then @xmath28 is @xmath458-lipschitz continuous with respect to @xmath459 for all @xmath460 where @xmath458 denotes the constant appearing in proposition [ prop : barriers ] . + * step 1 : gradient estimates * + proving is equivalent to prove that the solution @xmath55 of satisfies the following gradient estimate : @xmath454 , @xmath461 , @xmath462 where @xmath463 . we will prove each inequality separately . since @xmath435 is sublinear , there exists @xmath464 such that for all @xmath126 @xmath465 eq . is equivalent to prove @xmath466 we first prove that @xmath467 . we argue by contradiction by assuming that @xmath468 and we exhibit a contradiction . the following supremum @xmath469 is also positive for @xmath292 and @xmath197 small enough . using the fact that , by assumption on @xmath65 , @xmath470 and the fact that @xmath471 as @xmath472 or @xmath473 , we deduce that the supremum is achieved at a point @xmath183 such that @xmath474 and @xmath475 . moreover , we deduce using and the fact that @xmath476 , that there exists a constant @xmath477 such that @xmath59 and @xmath323 satisfy the following inequality @xmath478 thanks to jensen - ishii s lemma ( see e.g. @xcite ) , we conclude that there exist @xmath479 such that @xmath480 \le 0 .\end{aligned}\ ] ] subtracting the viscosity inequalities and using the last line yield @xmath481 using the fact that the functions @xmath482 and @xmath483 are @xmath484-lipschitz , we deduce that @xmath485 remarking that the function @xmath486 is bounded from above by a constant @xmath487 , we have @xmath488 where @xmath489 * case a : @xmath490 * + we now rewrite @xmath491 in the following way @xmath492 and use the fact that @xmath493 to deduce that @xmath491 is non - decreasing . hence , we finally get @xmath494 which is absurd for @xmath292 small enough . in order to prove that @xmath495 , we proceed as before and we obtain where @xmath496 remarking that @xmath491 is decreasing permits us to conclude in this case . + * case b : @xmath497 * + we simply notice that the equation is not changed if we change @xmath498 in @xmath499 . * step 2 : lipschitz in time estimates * + it remains to prove that @xmath435 is @xmath458-lipschitz continuous with respect to @xmath459 under the additional compatibility condition . to do so , we fix @xmath500 and we consider the following functions : @xmath501 remark that @xmath502 and @xmath503 satisfy . moreover , proposition [ prop : barriers ] implies that @xmath504 thanks to the comparison principle , we conclude that @xmath505 in @xmath506 ; since @xmath507 is arbitrary , we thus conclude that @xmath435 is @xmath458-lipschitz continuous with respect to @xmath459 . the proof of proposition [ prop : gradestim ] is now complete . it is now easy to derive theorem [ th : cauchy ] from propositions [ pro : cauchy ] and [ prop : gradestim ] . consider the viscosity solution @xmath435 given by proposition [ pro : cauchy ] with @xmath508 where the constant @xmath509 is given in the barrier presented in proposition [ prop : barriers ] . this function is continuous . moreover , thanks to proposition [ prop : gradestim ] , @xmath510 and @xmath511 are bounded in the viscosity sense ; hence @xmath55 is lipschitz continuous . in particular , there exists a set @xmath512 of null measure such that for all @xmath513 , @xmath435 is differentiable at @xmath514 . thanks to the equation @xmath515 with @xmath516 we also have that @xmath517 is locally bounded in the viscosity sense . this implies that @xmath435 is locally @xmath518 with respect to @xmath7 , and in particular , we derive from alexandrov s theorem @xcite that for all @xmath29 there exists a set @xmath519 of null measure such that for all @xmath520 , @xmath521 is twice differentiable with respect to @xmath7 , i.e. there exist @xmath522 such that for @xmath523 in a neighborhood of @xmath7 , we have @xmath524 from @xmath525 and @xmath526 , we can construct a set @xmath527 of null measure such that for all @xmath528 , @xmath435 is differentiable with respect to time and space at @xmath514 and there exists @xmath529 such that holds true . we conclude that @xmath530 in particular , holds true for @xmath528 . we deduce from the previous discussion that @xmath531 holds true almost everywhere , and thus in the sense of distributions . from the standard interior estimates for parabolic equations , we get that @xmath532 for any @xmath533 . then from the sobolev embedding ( see lemma 3.3 in @xcite ) , we get that for @xmath534 , and @xmath535 , we have @xmath536 . we now use that holds almost everywhere . therefore we can apply the standard interior schauder theory ( in hlder spaces ) for parabolic equations . this shows that @xmath537 . bootstrapping , we finally get that @xmath538 , which ends the proof of the theorem . the main goal of this section is to prove theorem [ th : cauchywc ] . we start with general barriers , hlder estimates in time and finally an approximation argument . [ prop : barriers2 ] let @xmath68 be such that there exists @xmath200 such that latexmath:[\[\label{eq : hypu0 } exists a constant @xmath423 ( depending only on @xmath200 ) such that for any function @xmath540\to { { \mathbb r}}$ ] with @xmath541 and @xmath542 , @xmath543 are respectively a super- and a sub - solution of , . we only do the proof for the super - solution since it is similar ( and even simpler ) for the sub - solution . we also do the proof only in the case @xmath490 , noticing that the equation is unchanged if we replace @xmath498 with @xmath499 . it is convenient to write @xmath62 for @xmath544 and do the computations with this function . since @xmath545 , we have @xmath546 since @xmath547 , there exists @xmath548 such that @xmath549 we then have @xmath550 using , we can write @xmath551 we get @xmath552 we now set @xmath523 such that @xmath553 and distinguish two cases : * case 1 : @xmath554 . * in this case , @xmath555 where for the second line , we have used the fact that for @xmath556 , @xmath557 . on the other hand , we have @xmath558 . choosing @xmath559 we get the desired result in this case . * case 2 : @xmath560 or @xmath561 . * in this case @xmath562 then @xmath563 we distinguish two sub - cases : subcase 2.1 : @xmath564 . in this sub - case , we get @xmath565 and we obtain the desired result taking @xmath566 . subcase 2.2 : @xmath567 . in this subcase , @xmath568 and @xmath569 and thus @xmath570 where for the last line we have used the fact that we can find @xmath571 ( only depending on @xmath572 ) such that @xmath573 for all @xmath460 . we finally get the desired result taking @xmath574 . the proof is now complete . [ pro : holder1 ] let @xmath575 satisfying . let @xmath28 be a solution of , satisfying . if @xmath28 is @xmath576-lipschitz continuous with respect to the variable @xmath7 , then there exists a constant @xmath76 , depending only on @xmath200 and @xmath576 such that @xmath577 where @xmath578 is defined in proposition [ prop : barriers2 ] . [ rem : holder ] let us note that in proposition [ prop : barriers2 ] , we can choose @xmath579 . hence , we deduce from proposition [ pro : holder1 ] that there exists @xmath403 such that for all @xmath580 $ ] , latexmath:[\[\label{eq::rm3 } let @xmath86 . using proposition [ prop : barriers2 ] and the comparison principle , we deduce that there exists a constant @xmath509 and a function @xmath578 such that @xmath582 since @xmath28 is @xmath576-lipschitz continuous in @xmath7 , we also have @xmath583 taking the minimum over @xmath584 in the right hand side , we get that @xmath585 with @xmath586 . we finally deduce that @xmath587 the desired result is obtained by remarking that , if @xmath588 , then @xmath589 , while if @xmath590 , then @xmath591 . the next proposition asserts that the previous proposition is still true if we do not assume that @xmath28 is lipschitz continuous with respect to @xmath7 . [ pro : holder2 ] let @xmath575 satisfying . then there exists a solution @xmath28 of , satisfying . moreover there exists a constant @xmath76 , depending only on @xmath200 such that @xmath577 where @xmath578 is defined in proposition [ prop : barriers2 ] , and there exists a constant @xmath576 ( only depending on @xmath200 ) such that @xmath592 the initial datum is approximated with a sequence of initial data satisfying and the compatibility condition ; passing to the limit will give the desired result . we can assume without loss of generality that @xmath593 . then we consider @xmath594 where @xmath595 is such that @xmath596 and @xmath597 where the non - increasing function @xmath598 satisfies @xmath599 [ claim ] the initial condition @xmath600 satisfies the compatibility condition and for some constant @xmath200 which does not depend on @xmath204 . let @xmath601 denote the unique solution of with initial condition @xmath600 given by proposition [ pro : cauchy ] , using the barrier ( proposition [ prop : barriers2 ] ) provided by the claim [ claim ] . in particular , @xmath601 satisfies for some constant @xmath602 depending on @xmath204 . using proposition [ prop : gradestim ] , we deduce that @xmath603 is @xmath576-lipschitz continuous with @xmath604 . then proposition [ pro : holder1 ] can be applied to obtain the existence of a constant @xmath76 ( depending only on @xmath200 , because @xmath576 now depends on @xmath200 ) such that for all @xmath204 @xmath605 taking @xmath606 and using the stability of the solution and the uniqueness of , , we finally deduce the desired result . we now prove the claim . we have @xmath607 hence , since @xmath608 and @xmath609 , we get @xmath610 which means that @xmath600 satisfies . using the fact that @xmath611 and ( [ eq::rm5 ] ) , we get . since @xmath612 and @xmath613 and @xmath614 are @xmath200-lipschitz continuous , we have @xmath615 let @xmath616 denote @xmath617 hence @xmath618 let us now obtain an estimate on @xmath619 . using the previous bound , we only have to estimate @xmath620 if @xmath621 , then @xmath622 and the estimate follows from . if @xmath623 , it is enough to estimate @xmath624 . we have @xmath625 moreover there exists a constant @xmath626 ( depending only on @xmath200 ) such that for all @xmath627 , @xmath628 . let @xmath629 denote @xmath630 . we then have @xmath631 we finally deduce that for @xmath623 , @xmath632 which proves that @xmath600 satisfies with a constant @xmath633 depending only on @xmath200 . we now turn to the proof of theorem [ th : cauchywc ] . the existence of @xmath28 and its lipschitz continuity with respect to @xmath7 follows from proposition [ pro : holder2 ] . the uniqueness ( and continuity ) of @xmath28 follows from the comparison principle ( theorem [ thm comp spirales sous lin en r ] ) . let us now prove that @xmath28 is @xmath75-hlder continuous with respect to time . by remark [ rem : holder ] , there exists a constant @xmath76 such that for @xmath634 @xmath635 with @xmath76 given in ( [ eq::rm3 ] ) . proceeding as in step 2 of the proof of proposition [ prop : gradestim ] , we get for @xmath636 : @xmath637 the reverse inequality is obtained in the same way . this implies ( [ eq::rm1 ] ) . the proof is now complete . we look for @xmath299 under the following form : for @xmath638 , @xmath639 where @xmath640 is non - decreasing , smooth ( @xmath88 ) and such that @xmath641 if @xmath642 and @xmath643 if @xmath644 . remark that and are readily satisfied . it is convenient to use the following notation : @xmath648 . we first write in terms of @xmath649 : @xmath650 ( we used a different norm in @xmath651 and @xmath412 is changed accordingly ) . it is enough to prove @xmath652 we choose @xmath653 and we remark that such an inequality is clear if @xmath654 or @xmath655 . through a taylor expansion and using the fact that @xmath656 , this reduces to check that @xmath657 which reduces to @xmath658 for @xmath59 far from @xmath19 , a simple computation shows that @xmath659 ( for @xmath660 ) and this permits us to conclude . for @xmath59 in a neighborhood of @xmath19 , @xmath661 and @xmath662 and we can conclude in this case too . in @xmath663\backslash [ 0,1]$ ] , the conclusion is straightforward . we next write in terms of @xmath649 @xmath664 where @xmath665 once again , the previous inequality is true for @xmath666 and for @xmath667 , we choose @xmath412 such that @xmath668 the same reasoning as above applies here too . it remains to prove that for @xmath645 given , we can find @xmath389 such that , as soon as @xmath670 and @xmath671 , then @xmath672 . we argue by contradiction by assuming that there exists @xmath673 and two sequences @xmath674 and @xmath675 such that @xmath676 as @xmath677 . since @xmath678 is bounded from below by @xmath484 , we deduce that @xmath679 . up to a subsequence , we can assume that @xmath680 and we thus deduce that @xmath681 . hence , @xmath682 for large @xmath258 s . thanks to a taylor expansion in @xmath683 , we can also get that @xmath684 . because @xmath685 , we then get that @xmath686 and @xmath687 remain in a bounded interval . we can thus assume that @xmath688 and @xmath689 . finally , we have @xmath690 for @xmath691 and @xmath692 which is impossible . the proof of the lemma is now complete . [ [ case-2-xle-1-and - yge-1-or - xge-1-and - yle-1 . ] ] case 2 : ( @xmath695 and @xmath329 ) or ( @xmath328 and @xmath696 ) . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the two cases can be treated similarly and we assume here that @xmath700 and @xmath329 . in that case @xmath701 with @xmath702 and @xmath703 with @xmath704 ( see ) and @xmath705 moreover , there exists a constant @xmath76 such that @xmath706 we also have @xmath707 hence , @xmath708 which gives the desired estimate . , _ linear and quasilinear equations of parabolic type _ , translated from the russian by s. smith . translations of mathematical monographs , vol . 23 , american mathematical society , providence , r.i . ,

we analyze x - ray spectra of 43 palomar - green quasars observed with _ xmm - newton _ in order to investigate their mean fe k line profile and its dependence on physical properties . the continuum spectra of 39 objects are well reproduced by a model consisting of a power law and a blackbody modified by galactic absorption . the spectra of the remaining four objects require an additional power - law component absorbed with a column density of . a feature resembling an emission line at 6.4 kev , identified with an fe k line , is detected in 33 objects . approximately half of the sample show an absorption feature around 0.650.95 kev , which is due to absorption lines and edges of and . we fit the entire sample simultaneously to derive average fe line parameters by assuming a common fe line shape . the fe line is relatively narrow (kev ) , with a center energy of 6.48kev and a mean equivalent width ( ew ) of 248ev . by combining black hole masses estimated from the virial method and bolometric luminosities derived from full spectral energy distributions , we examine the dependence of the fe k line profile on eddington ratio . as the eddington ratio increases , the line becomes systematically stronger ( ew = 130 to 280 ev ) , broader ( to 0.7kev ) , and peaks at higher energies ( 6.4 to 6.8kev ) . this result suggests that the accretion rate onto the black hole directly influences the geometrical structure and ionization state of the accretion disk . we also examine a two - component model consisting of a gaussian and a diskline to constrain the intensity of the broad line . the mean equivalent widths areev for the four eddington ratio groups , although the standard deviations in each group are very large . this suggests that the broad line is not ubiquitous .

astro-ph0703350

fluorescent fe k lines are common in the x - ray spectra of accreting black holes , ranging from x - ray binaries to active galactic nuclei ( agns ) , and provide one of the best probes for studying accretion disks . if the fe k line is emitted from the inner part of the accretion disk , it becomes broad and asymmetric due to both doppler shift and gravitational redshift ( fabian et al . 1989 ) . the best example of a broad fe k line is seen in the seyfert 1 galaxy mcg63015 ( tanaka et al . 1995 ) . broad fe k lines have been observed in many other sources and are thought to be very common in the x - ray spectra of seyfert 1 nuclei ( nandra et al . 1997a ) . recent observations of agns with _ xmm - newton _ show that the broad fe k line is not as common as previously believed . while a few seyfert 1 galaxies ( e.g. , mcg63015 , mrk 205 , mrk 509 ) indeed show an unambiguous broad line , fe lines in other galaxies are dominated by a relatively narrow feature . from an analysis of 53 type 1 agns observed with _ xmm - newton _ , page et al . ( 2004a ) showed that a broad fe k line ( @xmath4kev ) is seen in only 13 sources , while the remaining 40 sources have a narrow line ( @xmath5kev ) or no lines . some of the 13 sources show a narrow line as well as the broad line . in a similar study , yaqoob & padmanabhan ( 2004 ) reported the results of 18 observations of 15 seyfert 1 galaxies observed with the _ chandra _ high energy grating . they measured the width of the line core and obtained a weighted mean of fwhm = 2380@xmath6 760 kms@xmath7 , which is slightly larger than the instrument resolution ( fwhm @xmath8 1860 kms@xmath7 ) . evidence of underlying broad - line emission was also found in at least four sources . based on these recent studies , it is still controversial whether or not relativistically broadened fe k emission is truly common in nearby agns . streblyanska et al . ( 2005 ) derived an average rest - frame spectrum of agns detected in a 770 ksec _ xmm - newton _ observation of the lockman hole field . they used a sample of 104 x - ray sources with optical redshifts measured by lehmann et al . ( 2001 ) , analyzing separately the type 1 and type 2 subsamples defined by optical spectra ( schmidt et al . 1998 ; lehman et al . 2001 ) . from composite rest - frame spectra generated for the type 1 and type 2 sources , they found evidence for a broad line peaking at a rest - frame energy @xmath96.4kev with an equivalent width ( ew ) @xmath9560ev and @xmath9460ev , respectively . however , it should be noted that there are some systematic uncertainties in modelling the continuum , since it is virtually impossible to distinguish a very broad line clearly from the continuum for sources as faint as those analyzed by streblyanska et al . in addition , making composite spectra of faint sources may introduce spurious spectral features ( yaqoob 2005 ) . brusa et al . ( 2005 ) and comastri et al . ( 2007 ) studied average spectra of agns detected in the _ chandra _ deep field north and south . they stacked the x - ray counts in the observed frame from spectroscopically identified agns , using a large number of source spectra within sufficiently narrow redshift ranges such that the energy ( redshift ) spread is negligible . a broad fe line is seen in some of the stacked spectra and is fitted with a relativistic disk line model peaking at a rest - energy of 6.4 kev . since the fluxes of the objects in the sample are low , this study may suffer from the same uncertainties affecting the streblyanska et al . analysis of the lockman hole . what parameters of the agn affect the strength or profile of the fe k line ? an inverse correlation between the equivalent width of fe k line and x - ray luminosity , known as the x - ray baldwin effect , was first pointed out by iwasawa & taniguchi ( 1993 ) . this correlation has been confirmed with _ asca _ ( nandra et al . 1997b ) and _ xmm - newton _ data ( page et al . 2004a ; see also jimnez - bailn et al . the dependence of the iron line profile with luminosity was investigated by nandra et al . ( 1997b ) using 18 seyfert 1s and 21 quasars observed with _ asca_. they divided the sample into five luminosity bins and examined average fe k line profiles . the line shows a very similar profile , which is composed of a narrow core and a broad red wing for groups with a luminosity below @xmath10 . the intensity of the narrow core , however , becomes weak above this luminosity . the red wing becomes weak and the peak energy shifts to higher energy for the luminosity range of @xmath11 ; no evidence for line emission is observed above @xmath12 . jimnez - bailn et al . ( 2005 ) , analyzing a sample of 39 palomar - green ( pg ; schmidt & green 1983 ) quasars observed with _ xmm - newton _ , found a similar result as that reported by nandra et al . ( 1997b ) . the luminosity dependence of the fe k line may be related to the ionization level of the accretion disk . this paper examines the behavior of the fe k line as a function of accretion rate , as parameterized by the eddington ratio , taking advantage of the availability of black hole mass and bolometric luminosity estimates . the eddington ratio is a fundamental parameter that may control both the ionization state and geometric structure of the accretion disk . if so , we expect the profile of the fe k line , in particular its width and central energy , to vary systematically with eddington ratio . the goal of this study is to investigate this issue . this paper is organized as follows . section 2 describes the observations , and section 3 shows the results of the spectral analysis of average spectra and its eddington ratio dependence . we discuss the physical origin of the eddington ratio dependence in section 4 . our findings are summarized in section 5 . we selected our sample from the 87 low - redshift ( @xmath13 ) palomar - green quasars in boroson & green ( 1992 ) . x - ray data for 45 objects are available in the _ xmm - newton _ science archive ( xsa ) . only epic - pn data were used in the following analysis since the effective area of pn is much larger than that of epic - mos . pg 1001 + 045 was excluded from the present study because no epic - pn data are available . pg 1226 + 023 ( 3c 273 ) , a very bright quasar , was also excluded since it overwhelms the x - ray flux in the average spectrum examined below . basic data ( name , redshift , central black hole mass , and eddington ratio ) for the final sample of 43 objects are given in table [ tab : the sample ] . _ xmm - newton _ results of 34 objects have been published , and the references are also given in table [ tab : the sample ] . a log of the observations is given in table [ tab : the sample xray ] . the black hole masses and eddington ratios are taken from the study of l. c. ho ( 2007 , in preparation ) . ho calculates black hole masses employing the virial method of kaspi et al . ( 2000 ) , using broad h@xmath14 line widths from boroson & green ( 1992 ) and 5100 continuum luminosities from neugebauer et al . his bolometric luminosities come from direct integration of the full , broad - band ( radio to x - rays ) spectral energy distribution rather than from a single band assuming a bolometric correction . all distance - dependent quantities were derived assuming the following cosmological parameters : @xmath15 , and @xmath16 = 72 kms mpc@xmath7 . eddington luminosities , computed assuming a composition of pure hydrogen , are given by @xmath17 where @xmath18 is the mass of the black hole , @xmath19 is the proton mass , and @xmath20 is the thompson scattering cross - section . figure[fig : redshift mass edd_ratio dist ] shows the distributions of redshift , black hole mass , and eddington ratios . the mean redshift is @xmath21 , but the sample covers a wide range of black hole masses ( @xmath22 ) and eddington ratios ( @xmath23 ) . cccccccl 0003 + 199 & mrk 335 & 0.026 & 7.16 & 0.58 & 2 & & 1 , 2 + 0007 + 106 & mrk 1501 & 0.089 & 8.69 & 0.10 & 4 & rl & 3 , 4 + 0050 + 124 & i zw 1 & 0.061 & 7.29 & 2.51 & 1 & & 2 , 3 , 4 , 5 + 0157 + 001 & mrk 1014 & 0.163 & 8.21 & 1.11 & 1 & & 2 , 3 , 4 + 0804 + 761 & ... & 0.100 & 8.41 & 0.45 & 2 & & 3 , 4 , 5 , 6 + 0844 + 349 & ton 951 & 0.064 & 7.80 & 0.33 & 3 & & 2 , 3 , 4 , 7 , 8 , 9 + 0947 + 396 & ... & 0.206 & 8.67 & 0.26 & 3 & & 2 , 3 , 4 , 5 , 10 + 0953 + 414 & ... & 0.234 & 8.69 & 0.66 & 2 & & 2 , 3 , 4 , 5 , 10 + 1004 + 130 & pks 1004 + 13 & 0.240 & 9.43 & 0.09 & 4 & rl , bal & 11 + 1048 + 342 & 3c 246 & 0.167 & 8.36 & 0.35 & 3 & & 2 , 3 , 4 , 5 , 10 + 1100 + 772 & 3c 249.1 & 0.312 & 9.43 & 0.12 & 4 & rl & 3 , 4 + 1114 + 445 & ... & 0.144 & 8.57 & 0.13 & 4 & & 3 , 4 , 5 , 10 , 12 + 1115 + 407 & ... & 0.154 & 7.61 & 1.91 & 1 & & 2 , 3 , 4 , 5 , 10 + 1116 + 215 & ton 1388 & 0.177 & 8.64 & 0.68 & 2 & & 2 , 3 , 4 , 5 , 10 + 1202 + 281 & gq comae & 0.165 & 8.54 & 0.18 & 3 & & 2 , 3 , 4 , 5 , 10 + 1211 + 143 & ... & 0.081 & 7.95 & 0.79 & 2 & & 2 , 3 , 4 , 13 , 14 + 1216 + 069 & ... & 0.331 & 9.40 & 0.18 & 3 & & 3 , 4 , 10 + 1244 + 026 & ... & 0.048 & 6.39 & 4.57 & 1 & & 2 , 3 , 4 , 5 + 1302@xmath24102 & pks 1302@xmath24102 & 0.27 & 9.04 & 0.48 & 2 & rl & + 1307 + 085 & ... & 0.155 & 8.87 & 0.10 & 4 & & 2 , 3 , 4 , 5 + 1309 + 355 & ton 1565 & 0.184 & 8.37 & 0.32 & 3 & rl & 2 , 3 , 4 , 5 , 10 , 12 + 1322 + 659 & ... & 0.168 & 8.22 & 0.49 & 2 & & 2 , 3 , 4 , 5 , 10 + 1352 + 183 & ... & 0.152 & 8.37 & 0.37 & 3 & & 2 , 3 , 4 , 5 , 10 + 1402 + 261 & ton 182 & 0.164 & 7.94 & 1.78 & 1 & & 2 , 3 , 4 , 5 , 10 , 15 + 1404 + 226 & ... & 0.098 & 6.80 & 2.63 & 1 & & 2 , 3 , 4 , 16 , 17 + 1411 + 442 & ... & 0.090 & 7.98 & 0.28 & 3 & & 3 , 4 , 10 , 18 + 1415 + 451 & ... & 0.114 & 7.87 & 0.38 & 3 & & 3 , 4 , 10 + 1425 + 267 & ton 202 & 0.366 & 9.75 & 0.05 & 4 & rl & 10 , 19 + 1426 + 015 & mrk 1383 & 0.087 & 8.98 & 0.08 & 4 & & 5 , 6 + 1427 + 480 & ... & 0.221 & 8.10 & 0.71 & 2 & & 2 , 3 , 4 , 5 , 10 + 1440 + 356 & mrk 478 & 0.079 & 7.42 & 1.91 & 1 & & 2 , 3 , 4 , 5 , 10 + 1444 + 407 & ... & 0.267 & 8.36 & 0.94 & 2 & & 2 , 3 , 4 , 10 + 1448 + 273 & ... & 0.065 & 6.95 & 1.86 & 1 & & + 1501 + 106 & mrk 841 & 0.036 & 8.39 & 0.07 & 4 & & 2 , 3 , 4 + 1512 + 370 & 4c + 37.43 & 0.371 & 9.50 & 0.13 & 4 & rl & 5 , 10 + 1535 + 547 & mrk 486 & 0.039 & 7.02 & 0.41 & 3 & & 20 + 1543 + 489 & ... & 0.400 & 8.13 & 3.39 & 1 & & 10 + 1613 + 658 & mrk 876 & 0.129 & 9.12 & 0.06 & 4 & & 3 , 4 , 5 , 6 + 1626 + 554 & ... & 0.133 & 8.43 & 0.19 & 3 & & 3 , 4 , 5 , 10 + 2112 + 059 & ... & 0.466 & 9.27 & 0.56 & 2 & bal & 20 + 2130 + 099 & mrk 1513 & 0.063 & 7.82 & 0.65 & 2 & & + 2214 + 139 & mrk 304 & 0.066 & 8.38 & 0.06 & 4 & & 3 , 4 , 18 + 2233 + 134 & ... & 0.325 & 8.15 & 2.29 & 1 & & + ccccccc 0003 + 199 & 2000 dec 25 & 28.5 & 19.45 & 13.75 & 0.20 & 1 + 0007 + 106 & 2000 jul 03 & 10.2 & 3.23 & 7.01 & 1.28 & + 0050 + 124 & 2002 jun 22 & 18.3 & 7.95 & 8.17 & 0.71 & 1 + 0157 + 001 & 2000 jul 29 & 4.4 & 1.10 & 0.91 & 0.65 & + 0804 + 761 & 2000 nov 04 & 0.5 & 11.22 & 9.94 & 2.45 & + 0844 + 349 & 2000 nov 04 & 10.3 & 6.45 & 4.87 & 0.46 & + 0947 + 396 & 2001 nov 03 & 17.6 & 1.76 & 1.78 & 2.05 & + 0953 + 414 & 2001 nov 22 & 10.8 & 3.92 & 2.94 & 4.65 & 1 + 1004 + 130 & 2003 may 04 & 18.2 & 0.10 & 0.32 & 0.48 & + 1048 + 342 & 2002 may 13 & 21.5 & 1.17 & 1.30 & 0.94 & + 1100 + 772 & 2002 nov 01 & 6.8 & 2.02 & 3.62 & 10.16 & + 1114 + 445 & 2002 may 14 & 35.0 & 0.53 & 2.28 & 1.12 & 1 + 1115 + 407 & 2002 may 17 & 14.4 & 2.20 & 1.19 & 0.75 & + 1116 + 215 & 2001 dec 02 & 5.5 & 5.05 & 3.19 & 2.77 & + 1202 + 281 & 2002 may 30 & 12.2 & 2.64 & 3.58 & 2.46 & 1 + 1211 + 143 & 2001 jun 15 & 48.5 & 3.33 & 3.05 & 0.46 & 2 + 1216 + 069 & 2002 jul 05 & 5.4 & 0.82 & 1.12 & 3.63 & + 1244 + 026 & 2001 jun 17 & 4.6 & 6.15 & 2.31 & 0.12 & + 1302@xmath24102&2002 jan 03 & 11.5 & 0.61 & 3.05 & 6.44 & + 1307 + 085 & 2002 jun 13 & 9.9 & 0.80 & 1.96 & 1.13 & 1 + 1309 + 355 & 2002 jun 10 & 23.6 & 0.44 & 0.70 & 0.61 & 1 + 1322 + 659 & 2002 may 11 & 8.0 & 2.41 & 1.27 & 0.98 & 1 + 1352 + 183 & 2002 jul 20 & 8.9 & 2.30 & 1.87 & 1.11 & 1 + 1402 + 261 & 2002 jan 27 & 9.1 & 3.17 & 1.80 & 1.32 & + 1404 + 226 & 2001 jun 18 & 9.1 & 0.59 & 0.13 & 0.03 & 2 + 1411 + 442 & 2002 jul 10 & 21.6 & 0.12 & 0.46 & 0.08 & 1 + 1415 + 451 & 2002 dec 08 & 20.4 & 1.43 & 1.06 & 0.34 & + 1425 + 267 & 2002 jul 28 & 29.6 & 0.67 & 1.68 & 6.29 & 1 + 1426 + 015 & 2000 jul 28 & 2.2 & 9.39 & 8.48 & 1.50 & + 1427 + 480 & 2002 may 31 & 32.5 & 1.15 & 1.05 & 1.44 & + 1440 + 356 & 2003 jan 04 & 10.8 & 5.58 & 2.63 & 0.39 & + 1444 + 407 & 2002 aug 11 & 16.7 & 0.96 & 0.45 & 1.05 & + 1448 + 273 & 2003 feb 08 & 18.0 & 4.23 & 2.07 & 0.20 & + 1501 + 106 & 2001 jan 13 & 7.6 & 18.31 & 15.00 & 0.44 & 1 + 1512 + 370 & 2002 aug 25 & 15.4 & 1.54 & 1.87 & 7.98 & 1 + 1535 + 547 & 2002 nov 03 & 14.4 & 0.16 & 1.33 & 0.04 & + 1543 + 489 & 2003 feb 08 & 3.7 & 0.35 & 0.13 & 0.87 & + 1613 + 658 & 2001 aug 29 & 2.4 & 4.46 & 4.98 & 2.05 & 1 + 1626 + 554 & 2002 may 05 & 4.8 & 2.84 & 2.86 & 1.28 & + 2112 + 059 & 2003 may 14 & 7.4 & 0.13 & 0.34 & 2.24 & + 2130 + 099 & 2003 may 16 & 23.3 & 2.13 & 3.63 & 0.32 & 1 + 2214 + 139 & 2002 may 12 & 6.2 & 0.55 & 3.32 & 0.31 & 1 + 2233 + 134 & 2003 may 28 & 6.2 & 0.59 & 0.49 & 1.69 & 1 + raw data from the _ epic - pn _ instrument were reprocessed using science analysis software ( sas ) version 6.0 to produce calibrated event files . only events with the single or double pixels ( pattern @xmath254 ) were used in the following analysis . time intervals of background flaring were excluded . source spectra were extracted from a circular region centered at the peak of the x - ray source counts . background spectra were taken from an off - source region in the same field - of - view and subtracted from the source spectra . appropriate response and ancillary files were created using the rmfgen and arfgen tools in sas , respectively . all the spectra were binned so that each bin contains at least 30 ( or more for bright sources ) counts in order to apply a @xmath26 minimization technique . spectral fits were performed with _ xspec _ version12.2 . the spectra of most objects are not well represented by a simple power law ; soft excess emission and an emission line - like feature around 6 kev are also often seen . therefore , we considered a model consisting of a power law , a blackbody component to represent the soft excess emission below @xmath9 2kev , and a gaussian emission line around 6 kev , all modified by galactic absorption . we used a two blackbody model for the soft excess if the single blackbody did not represent the data well . the hydrogen column densities of the galactic absorption are left free . some objects show absorption features around 0.650.95 kev . we multiplied one or two edges to the above model for such cases . the spectra of 39 objects are well reproduced by this model . we obtained hydrogen column densities of @xmath27 @xmath8 ( [email protected])@xmath28 , photon indices @xmath29 1.5 to 2.5 , and temperatures of the blackbody @xmath30kev . the column densities are consistent with the galactic value derived from the hi map by dickey & lockman ( 1990 ) . except for a few cases ( see appendix ) , our results for individual objects are in good agreement with the independent analysis published in the references given in table 1 . the spectra of the remaining four objects ( pg 1004 + 130 , 1411 + 442 , 1535 + 547 , and 2214 + 139 ) show broad excess emission at energies above @xmath9 23 kev . we added an absorbed power - law component to the model used above to represent this component . this model provides a good description of the spectra . the hydrogen column densities of the absorbed power law are @xmath31 for all the cases . details of the spectral fits to these four quasars are given in appendix[appe : pg qso ] . the spectral parameters for the other components are similar to those obtained for the 39 objects without an additional absorbed power law . an absorption feature around 0.650.95 kev , which can be well fitted by one or two edges , is seen in 20 out of the 43 objects . the majority of the sources ( 18/20 ) require only a single edge , with a mean edge energy of @xmath32kev and an optical depth of @xmath33 . the absorption features in the two remaining objects are represented by two edges : for pg 1211 + 143 , the edge energies are @xmath34/@xmath35kev and the optical depths are @xmath36/@xmath37 ; for pg 1404 + 226 , the prameters are @xmath38/@xmath39kev and @xmath40/@xmath41 ) . similar features are often seen in seyfert 1s and quasars observed with the energy resolution of ccd . high - resolution spectroscopy of such features with grating spectrometers onboard _ xmm - newton _ and _ chandra _ have shown that they consist of many absorption lines of ionized species such as he - like and hydrogenic o , ne , and mg . the observed edge energies are in agreement with such ionized absorbers observed with ccd resolution ( e.g. , george et al . 1998 , 2000 ) . the fluxes and luminosities obtained for the best - fit model , as well as the detected count rates , are summarized in table 2 . neither the fluxes nor the luminosities are corrected for absorption . we derived a co - added rest - frame spectrum of the 43 quasars in order to examine the average shape of the fe k emission line . since each object has a different redshift and thus a different observed line energy , the detector response , which is energy dependent , is different for each object . as the co - added rest - frame spectrum is a summation of spectra taken with different energy resolution , this situation complicates quantitative analysis . therefore , the spectrum derived here is for presentation purpose only , and quantitative analysis by simultaneous fitting is done in the next subsection . the co - added spectrum was constructed as follows . first , ratios of the data to the best - fit continuum model were calculated . the ratios were then multiplied by the unfolded best - fit continuum model . the resulting spectrum is composed of an unfolded continuum and a folded emission - line component . the energy scales of both the unfolded continuum spectrum and the spectrum consisting of the unfolded continuum plus the folded line were shifted to the source redshift . we selected bin widths of 0.25kev for @xmath428kev and 2.5kev for @xmath438kev in this process , and used a monte carlo method to redistribute the observed events into the new spectral bins . the 43 spectra with the folded line are then co - added . the unfolded continuum model spectra were also co - added . finally , the ratio of the two spectra is calculated , as shown in figure[fig : 43pg ratio ] . a prominent narrow core at @xmath96.4kev and a very weak low - energy wing are seen . in contrast to previous studies , we employ simultaneous fits rather than fits to stacked spectra . since the fe k line in each quasar has a different peak energy in the observed frame according to their redshifts , a technique that properly treats the energy - dependent detector response is necessary . we fitted all spectra simultaneously in order to determine mean fe k line parameters . the spectral parameters of the best - fit continuum models for each object were used as initial parameters in the spectral fits . all the continuum parameters were left free , while the parameters for the edge model are fixed at the best - fit values for individual fits . a gaussian model with common peak energy and width is added to the continuum model . the normalizations of the gaussian component were left free individually . the fitting results are shown in table[tab : fitting 43pg ] . the best - fit peak energy and the dispersion of the gaussian line are @xmath44kev and @xmath45kev , respectively . the mean of the distribution of the equivalent widths , shown in figure[fig : ew distribution all ] , is 248ev , with a standard deviation of 168ev . a relativistic line model ( diskline model in xspec ) instead of a gaussian was also examined . we fixed several parameters ; the peak energy at 6.4kev , the outer radius @xmath46 at 500@xmath47 , where @xmath48 is the schwarzschild radius , the emissivity index @xmath14 at @xmath242 , and the inclination angle of the disk @xmath49 at 30 degrees . the free parameters are the inner radius @xmath50 and the intensity of the line . again , the parameters in the continuum model were initially set to the best - fit values in the individual fits and left free except for the edge component . the results are summarized in table[tab : fitting 43pg ] . the best - fit inner radius @xmath51 is @xmath52 . the lower boundary is fixed at @xmath53 , which is the minimum value allowed in the diskline model assuming a schwarzschild black hole . the mean and standard deviation of the equivalent widths are 210ev and 133ev , respectively . the line profile in figure[fig : 43pg ratio ] shows a prominent narrow core and a very weak low - energy wing . such a profile may be indicative of the presence of multiple line components , and a single gaussian or a diskline profile may not be a good representation . therefore , we examined a two - component line model consisting of either two gaussians or a combination of a gaussian and a diskline component . with regards to the double - gaussian model , although the @xmath26 improved by 103 for 3 additional parameters , not all the line parameters were tightly constrained . thus , we first fixed the continuum component to obtain a rough value of the gaussian widths , and then the widths were fixed at these values to constrain the equivalent widths of the lines . all the parameters except for the widths were left free in the latter fit . note that the normalizations of the lines were determined for each object . the resulting peak energies of the broad and narrow lines are 5.06kev and 6.52kev , respectively ( see table[tab : fitting 43pg ] ) . the mean equivalent width for the narrow component is 261 ev , and for the broad component it is 150 ev , although the standard deviations of the distributions are large ( 171 and 181 ev , respectively ) . the distributions suggest that the narrow component is commonly seen and that the broad line is present in a fraction of the sample . the gaussian plus diskline model was then examined . we fixed the peak energy and the dispersion of the gaussian to the best - fit parameters of the narrow component in the double - gaussian fit . the fixed parameters for the diskline component were the same as those in the single diskline fit presented above . the free parameters are the inner radius and the normalization of the diskline , and the parameters for the gaussian and continuum . as summarized in table[tab : fitting 43pg ] , the best - fit inner radius is @xmath50 = @xmath54 , and the mean equivalent width of the diskline is ew=102ev . the large standard deviation ( 181 ev ) again suggests that the diskline is required in only a fraction of the sample . ccccccccccc & & + model & energy & @xmath55 & ew & energy & @xmath50 & @xmath46 & @xmath14 & @xmath49 & ew & @xmath26/d.o.f . + & ( kev ) & ( kev ) & ( ev ) & ( kev ) & ( @xmath48 ) & ( @xmath48 ) & & ( deg ) & ( ev ) & + gaussian & @xmath56 & @xmath57 & @xmath58 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & 8958/8129 + diskline & @xmath24 & @xmath24 & @xmath24 & 6.4f & @xmath59 & 500f & @xmath242f & 30f & @xmath60 & 8975/8130 + gaussian@xmath61gaussian & @xmath62 & 0.61f & 150(181 ) & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & + & @xmath63 & 0.34f & 261(171 ) & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & 8853/8086 + gaussian@xmath61diskline & 6.52f & 0.34f & 155(181 ) & 6.4f & @xmath64 & 500f & @xmath242f & 30f & @xmath65 & 8912/8087 + we now turn to the eddington ratio dependence of the fe line profile . the 43 quasars were divided into four groups according to the eddington ratio as follows : ( 1)@xmath66 , ( 2)@xmath67 , ( 3)@xmath68 , and ( 4)@xmath69 ( see table[tab : the sample ] ) . the spectra in each group were co - added as in section [ subsec : fitting 43pg ] for the purposes of presentation ; figure[fig : 10pg ratio ] shows the resulting data - to - continuum model ratios . as the eddington ratio increases , the width of the fe k line appear to becomes broader and its peak energy rises . following the procedure outlined in section [ subsec : fitting 43pg ] , we fitted the spectra in each group simultaneously to characterize the fe k line quantitatively . the following four models were examined for the fe k line component : single gaussian , single diskline , double gaussians , and a gaussian plus diskline model . the results of the spectral fits are shown in table[tab : fitting 10pg ] , and the eddington ratio dependence of the peak energy and the width of the line obtained from the single gaussian model fits are shown in figure[fig : dependence of the param ] . we again find that the width and peak energy of the line become broader and higher as the eddington ratio increases . the peak energy gradually increases from 6.37 kev in group 4 to 6.77 kev in group 1 . the width is relatively narrow ( @xmath55 = 0.15kev ) in group 4 , but definitely resolved ( @xmath70 = 0.07kev ) , and becomes wider ( @xmath55 = 0.68 kev in group 1 ) as @xmath71 increases . the distributions of the equivalent widths for each group obtained from the gaussian fits are shown in figure[fig : ew distribution ] ; the mean equivalent widths in each group are 280 , 276 , 287 , and 131ev , respectively . a diskline model was examined next . we fixed several diskline parameters . the line center energy was chosen to be 6.4 kev for groups 3 and 4 and 6.7 kev for groups 1 and 2 , for which the gaussian fits suggest that the line is from ionized fe . the outer radius @xmath46 , the emissivity index @xmath14 , and the inclination angle @xmath49 were fixed at 500@xmath72 , @xmath242 , and 30 degrees , respectively . the free parameters are the inner radius @xmath50 and the normalization . the parameters for the continuum model were also left free , except for those of the edge component . the results are shown in table[tab : fitting 10pg ] . the resulting values of @xmath26 are similar to or slightly worse than those for the gaussian fits . the best - fit inner radius is @xmath73 , which is the last stable orbit of a schwartzschild black hole . we also examined a two - component line model since a strong narrow core and a broad wing are visible in groups 3 and 4 , and possibly in group 2 . a double - gaussian model was first tested in the same way as in section[subsec : fitting 43pg ] ; the results are shown in table[tab : fitting 10pg ] . the resulting @xmath26 values are similar to or only slightly better than those of the single - gaussian fit , indicating that the additional broad line is statistically not required . the distributions of the ew are broad and indicate that some objects may have a broad component . a gaussian plus diskline model was then used in order to constrain the equivalent width of the diskline , in which the inner radius is fixed at 3@xmath72 . the peak energy and the dispersion of the gaussian were fixed at the best - fit parameters for the narrow component in the double - gaussian fit . the inner radius of the disk was also fixed at 3@xmath72 , and the other parameters in the diskline model were fixed to the same values as in the single - diskline fits described above . only the normalizations of the gaussian and diskline were left free . the parameters for the continuum model were treated in a similar way as in section 3.3 . the results are summarized in table[tab : fitting 10pg ] . the resulting values of @xmath26 are similar to or only slightly better than those for the single - gaussian or diskline fits . the mean equivalent widths of the diskline in each group are 99 , 176 , 132 , and 67ev , respectively , although the standard deviations as large . the large standard deviations again suggest that the broad diskline is not ubiquitous . cccccccccccc & & & + group & model & energy & @xmath55 & ew & energy & @xmath50 & @xmath46 & @xmath14 & @xmath49 & ew & @xmath26/d.o.f . + & & ( kev ) & ( kev ) & ( ev ) & ( kev ) & ( @xmath48 ) & ( @xmath48 ) & & ( deg ) & ( ev ) & + 1 & gaussian & @xmath74 & @xmath75 & @xmath76 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & 2308/2237 + & diskline & @xmath24 & @xmath24 & @xmath24 & 6.7f & @xmath77 & 500f & @xmath242f & 30f & @xmath78 & 2316/2238 + & gaussian@xmath61gaussian & @xmath79 & 0.42f & @xmath80 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & + & & @xmath81 & 0.48f & @xmath82 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & 2305/2248 + & gaussian@xmath61diskline & 7.47f & 0.48f & @xmath83 & 6.7f & 3.0f & 500f & @xmath242f & 30f & @xmath84 & 2304/2250 + 2 & gaussian & @xmath85 & @xmath86 & @xmath87 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & 2636/2197 + & diskline & @xmath24 & @xmath24 & @xmath24 & 6.7f & @xmath88 & 500f & @xmath242f & 30f & @xmath89 & 2662/2198 + & gaussian@xmath61gaussian & @xmath90 & 0.38f & @xmath91 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & + & & @xmath92 & 0.24f & @xmath93 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & 2634/2218 + & gaussian@xmath61diskline & 6.52f & 0.24f & @xmath94 & 6.7f & 3.0f & 500f & @xmath242f & 30f & @xmath95 & 2646/2220 + 3 & gaussian & @xmath96 & @xmath97 & @xmath98 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & 1796/1774 + & diskline & @xmath24 & @xmath24 & @xmath24 & 6.4f & @xmath99 & 500f & @xmath242f & 30f & @xmath100 & 1801/1775 + & gaussian@xmath61gaussian & @xmath101 & 0.48f & @xmath102 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & + & & @xmath103 & 0.25f & @xmath104 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & 1800/1798 + & gaussian@xmath61diskline & 6.49f & 0.25f & @xmath105 & 6.4f & 3.0f & 500f & @xmath242f & 30f & 132 ( 132 ) & 1803/1800 + 4 & gaussian & @xmath106 & @xmath107 & @xmath108 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & 2184/1935 + & diskline & @xmath24 & @xmath24 & @xmath24 & 6.4f & @xmath109 & 500f & @xmath242f & 30f & @xmath110 & 2186/1936 + & gaussian@xmath61gaussian & @xmath111 & 0.70f & @xmath112 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & + & & @xmath113 & 0.09f & 82 ( 70 ) & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & 2173/1957 + & gaussian@xmath61diskline & 6.37f & 0.09f & @xmath114 & 6.4f & 3.0f & 500f & @xmath242f & 30f & @xmath115 & 2180/1959 + we examined mean fe k line shapes for the four groups sorted by the eddington ratios , and found that the peak energy , width , and equivalent width become higher ( 6.4 to 6.8kev ) , broader ( 0.1 to 0.7kev ) , and larger ( 130 to 280ev ) , respectively , as the eddington ratio increases . in this section , we discuss possible reasons of the observed eddington ratio dependence of the fe line profile . the line center energy of the fe k line is governed by a combination of the doppler effect caused by the orbital motion , gravitational redshift , and the ionization state of the emitter . an fe line from a rotating disk around a black hole shows a double - peaked profile with a peak energy that depends on the inclination angle of the disk . since each of the four groups we used contains from 10 to 11 objects , it is unlikely that any one of the groups has a mean inclination angle considerably different from those of other groups . therefore , the different line center energies obtained for the four groups probably reflect different ionization states rather than different mean disk inclination angles . the eddington ratio dependence of the peak energy thus strongly suggests that the ionization state of the accretion disk depends on the eddington ratio . many aspects of an ionized disk have been studied theoretically over the last decade . calculations have been performed for constant - density atmospheres ( e.g. , matt et al . 1993 ; ross & fabian 1993 ; @xmath116ycki et al . 1994 ) , as well as for atmospheres in hydrostatic equilibrium ( nayakshin et al . 2000 ; ballantyne et al . 2001 ) . these authors calculated reflection spectra from an ionized disk for different conditions . in the lowest ionization regime , the most prominent feature is the cold iron line at 6.4kev . as the mass accretion rate goes up , the ionized stage becomes higher , and the fe k line is dominated by fe xxv and xxvi , and k lines from the lighter elements emerge in the @xmath117kev band . at very high accretion rates , the surface of the disk is highly ionized so that the only noticeable line is a compton - broadened fe k line peaking at @xmath97kev . the eddington ratio dependence of the peak energy and the width of the fe k line we observed in the pg quasar sample is in good agreement with the predictions of ionized disk models . the large ew in the high - eddington ratio groups is also likely to be related to the ionization state of the accretion disk . calculations of the reflected x - rays from an ionized disk have shown that the ew of the fe k line can be higher than in the case of a disk in a low - ionization state . the resonant trapping effect plays a role in the intermediate - ionization state higher than . the ew increases again from , reaching a maximum value of @xmath118ev ( matt et al . 1993 ; ross & fabian 1993 ; @xmath116ycki & czerny 1994 ) . the observed line center energies and ews for the three groups with a largest eddington ratios are consistent with the latter expectation , although the scatter of the actual ew distributions is large . the reduction of ew caused by resonant trapping is not clearly seen in our data . this may be due to the coarse binning of the eddington ratio we have adopted . future analysis of a larger sample may be able to see the ionized disk in an intermediate - ionization state . the eddington ratio dependence of the line width also suggests a change of the geometrical structure of the disk , if the velocity width indeed reflects emission from the inner part of an accretion disk broadened by doppler effects and gravitational redshift . although the line widths obtained from the single - gaussian fits are consistent with systematic variations in the inner disk radius @xmath50 , such a trend is not clear in the fits obtained using the diskline model . it is possible that the broader line widths at high eddington ratios may in part be accounted for by a blend of lines from fe in different ionization states . streblyanska et al . ( 2005 ) derived an average fe line profile of 53 type 1 agn detected in a 770 ks observation of the lockman hole with _ xmm - newton_. their average spectrum shows a broad line - like feature ( @xmath119kev ) peaking at 6.4kev with an equivalent width of ew = 420ev . the profile is asymmetric , with a significant wing seen toward low energies . the line width ( @xmath55 = 0.36kev ) we obtained for the pg quasar sample is significantly narrower , and the equivalent width ( ew@xmath8250ev ) smaller , than that of streblyanska et al.s profile . we discuss possible causes for the difference . appropriate modeling of the continuum is essential to derive a correct fe line profile . a single power - law model was assumed in the analysis of the agns in the lockman hole . since the spectra of many agns show a signature of absorption by ionized and/or cold matter , a simple power law may not be a good approximation of the underlying continuum . indeed , four quasars in our sample are absorbed by a large column density ( @xmath120 @xmath121 ) . if there are some highly absorbed objects in the lockman hole sample , using a single power - law model for the continuum may mimic a broad - line profile . the flux of a highly absorbed power law with a column density of @xmath122 at 5kev is about 50% of that of an unabsorbed power law . if the fraction of highly absorbed quasars is 1/10 , their contribution to the low - energy side of the fe k line profile around 5 kev is less than 5% . the contribution of the absorbed continuum depends on the fraction of absorbed quasars . mateos et al . ( 2005 ) measured the absorbed fraction for their sample of lockman hole agns with sufficient x - ray counts . they found that seven out of 46 type 1 agns show a highly absorbed spectrum . this fraction is similar to that in our pg quasar sample . note , however , that streblyanska et al.s sample may contain more highly absorbed objects because their flux limit is lower than that used by mateos et al . ( 2005 ) and x - ray surveys have shown that objects with lower flux tend to have a harder spectrum . if the absorbed fraction is 10% or higher , rather than 5% , in the lockman hole sample , a considerable fraction of the red wing might be explained by the contribution of highly absorbed emission . the physical state of an accretion disk may depend on luminosity . if the mean luminosity of our sample is fairly different from that of the lockman hole agns , it may introduce a difference in the fe k line profile . in order to examine this possibility , we compared the distributions of the @xmath123kev luminosity between the two samples ( fig.[fig : lumin distri ] ) . the luminosities of the lockman hole agns were taken from mainieri et al . they analyzed 39 type 1 agns in the _ xmm - newton _ observation of the lockman hole field . most of the 53 type 1 agns in streblyanska et al . ( 2005 ) have been analyzed by mainieri et al . the two distributions are very similar , and their mean luminosities are essentially identical : the mean luminosity of our sample is @xmath124ergss@xmath7 , while that of the lockman hole agns is @xmath125ergss@xmath7 . we conclude that luminosity differences is not the main cause of the difference of the line profiles . yaqoob ( 2005 ) pointed out a spurious effect in adding spectra of faint sources . he showed that conventional methods for averaging low signal - to - noise ratio x - ray spectra result in a weakening of emission lines , a broad dip in the continuum above an emission line , and a spectral hardening at the highest energies , if sources at different redshifts are stacked . yaqoob ( 2005 ) simulated 200 spectra from sources with redshifts 0.52.5 and 210 kev fluxes 660@xmath126 ergs @xmath121 s@xmath7 and found an artifact wing at energies below the peak of the fe k line . this feature occurs at the level of about 5%10% at 46 kev . this kind of spectral distortion may affect the fe line shape in the analysis of streblyanska et al . ( 2005 ) since their sample has a relatively low average flux ( @xmath127 ) , compared to our pg quasar sample ( @xmath128 ) . we presented the average fe k line profile of 43 pg quasars observed by _ xmm - newton_. our findings are summarized as follows . * the spectra of 39 objects are well reproduced by a canonical model consisting of a power law , a blackbody , an edge in the case of objects with evidence for warm absorption , and a gaussian , all modified by galactic absorption , while that of the remaining four objects require additional absorption with a column density of @xmath129 to the power - law component . * an average spectrum of the 43 quasars shows a prominent narrow fe k line ( @xmath130 = 6.48kev , @xmath55 = 0.36kev , and ew = 248ev ) . * the average fe k line is well represented by double gaussians or a combination of a gaussian and a diskline model . the best - fit inner radius of the disk is 3@xmath131 and the mean ew is @xmath9100ev . the large standard deviation of the ew distribution indicates that a broad disklike line is not ubiquitous . * we found an eddington ratio dependence of the fe k line profile . as the eddington ratio increases , the peak energy becomes higher ( 6.4 to 6.8kev ) , the width becomes broader ( 0.1 to 0.7kev ) , and the ew is larger ( 130 to 280ev ) . the higher energy peak and larger ew at high eddington ratio can be explained by the ionization of the accretion disk . the systematic broadening of the line with increasing eddington ratio may indicate a systematic decrease of the inner radius of the disk , or perhaps blending from a combination of fe lines in different ionization states . * the line width we obtained is narrower and the equivalent width smaller than those derived from the stacked spectrum of agns in the lockman hole ( @xmath55 = 0.69kev , ew = 420ev ) . we attribute these differences to either an increased contribution of absorbed sources in the lockman hole composite spectrum or to artifacts introduced in averaging faint sources . we found four highly absorbed quasars in the 43 pg quasars . we show their spectra in figure[fig : highly absorbed quasars ] . they are not well reproduced by the canonical model , and an additional highly absorbed component is required . therefore , we added an absorbed power - law component . the fitting results are summarized in table[tab : abs pow ] . the average column densities of the additional absorption are @xmath132 @xmath121 . spectra of pg 1411 + 442 and pg 2214 + 139 are presented in brinkmann et al . the results of pg 1411 + 442 are consistent with ours . the reduced @xmath26 value is not good for pg 2214 + 139 ( @xmath133 ) because of non - systemtaic deviations of the data from the model seen over the energy range of 0.54 kev . miller et al . ( 2006 ) and schartel et al . ( 2005 ) found absorbed emission in pg 1004 + 130 and pg 1535 + 547 , respectively . the results obtained by miller et al . ( 2006 ) are also consistent with ours . schartel et al . ( 2005 ) found the peak energy of fe k line at @xmath134kev and the width of @xmath135kev , while the line center energy is lower than 5kev in our analysis . therefore , we fixed it at 6.4kev and repeated the fits . the best - fit line width is @xmath136kev ( see table[tab : abs pow ] ) . a population of soft x - ray - weak ( sxw ) quasars , where the soft x - ray flux is @xmath137 times smaller than in typical quasars ( yuan et al . 1998 ) , has been known . this class comprises @xmath910% of optically selected quasars . brandt et al . ( 2000 ) found 10 sxw quasars in the boroson & green ( 1992 ) sample of 87 pg quasars . all of the highly absorbed quasars presented here are consistent with the sxw classification of brandt et al . the large absorption seems to be the cause of the x - ray weakness at least in these four objects . cccccccccc[htbp ] 1004 + 130 & 1.93@xmath138 & @xmath24 & 0.06@xmath139 & 5.04@xmath140 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & 20/14 + & & & & & 0.94@xmath141/0.89@xmath142 & & & & + 1411 + 442 & 1.55@xmath143 & 0.13@xmath144 & 0.03@xmath145 & 16.7@xmath146 & 0.72@xmath147/0.85@xmath148 & 6.35@xmath149 & 0.25@xmath150 & 481@xmath151 & 63/75 + 1535 + 547 & 2.24@xmath152 & 0.07@xmath153 & 0.27@xmath154 & 11.9@xmath155 & 0.93@xmath156/1.86@xmath157 & 6.4f & 0.00(@xmath250.24 ) & 81@xmath158 & 91/87 + 2214 + 139 & 1.49@xmath159 & 0.18@xmath160 & 0.02@xmath161 & 2.4@xmath162 & 0.73@xmath147/1.13@xmath163 & 6.4f & 0.06@xmath164 & 152@xmath165 & 146/106 + we summarize the results of our spectral analysis of quasars newly presented in this paper . their spectra are shown in figure[fig : other quasars ] . we fitted the spectra with the canonical model described in section 3.1 ; results are listed in table[tab : other qso ] . they are well reproduced by the model except for one case ( pg 1448 + 273 ) . the soft excess component in pg 1448 + 273 is not explained by a single blackbody component . we added an additional blackbody component and obtained a good fit . ccccccccc 1302@xmath24102 & 1.66@xmath166 & 0.20@xmath167 & 0.00@xmath168 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & 64/63 + 1448 + 273 & 2.25@xmath169 & 0.12@xmath170 & 0.05@xmath144 & 0.66@xmath171/0.27@xmath172 & 6.41@xmath173 & 0.1f & 93@xmath174 & 260/290 + & & 0.23@xmath175 & & & & & & + 2130 + 099 & 1.65@xmath176 & 0.11@xmath177 & 0.05@xmath144 & 0.72@xmath171/0.29@xmath178 & 6.45@xmath179 & 0.02@xmath180 & 139@xmath181 & 182/139 + 2233 + 134 & 2.17@xmath182 & 0.19@xmath139 & 0.05@xmath139 & 0.89@xmath183/0.63@xmath184 & @xmath24 & @xmath24 & @xmath24 & 117/96 + this section shows results of quasars reanalyzed in this paper , for which the best - fit parameters are not consistent with previously published results . our results are summarized in table[tab : reanalyzed qso ] . some notes on individual objects are given below . _ pg 0003 + 199_. gondoin et al . ( 2002 ) give the line center of fe k line at [email protected] and the width of 1.2@xmath6 0.2kev . we obtained a higher center energy @xmath185kev and a narrower width @xmath186kev . _ pg 0804 + 761_. jimnez - bailn et al . ( 2005 ) , porquet et al . ( 2004 ) , and page et al . ( 2004 ) show the best - fit parameters of the line center at @xmath187 , @xmath188 , and @xmath189kev , respectively . the peak energy measured by porquet et al . ( 2004 ) and page et al . ( 2004b ) are consistent with our result(@xmath190kev ) . _ pg 1244 + 026_. the best - fit line center energy is @xmath191kev in our analysis . jimnez - bailn et al . ( 2005 ) and porquet et al . ( 2004 ) obtained the line center energy at @xmath192kev and @xmath193kev , respecrively . _ pg 1613 + 658_. a line feature is not clearly seen at @xmath96.4kev in our spectrum . no detection is consistent with the result of page et al . ( 2004a ) ( ew@xmath19485 ev ) , while page et al . ( 2004b ) reported ew=@xmath195 ev . ccccccccc 0003 + 199 & 2.25@xmath144 & 0.13@xmath177 & 0.03@xmath177 & 0.65@xmath144/0.16@xmath196 & 6.60@xmath197 & 0.35@xmath198 & 192@xmath199 & 626/483 + 0804 + 761 & 2.30@xmath200 & @xmath24 & 0.02@xmath144 & @xmath24 & 6.57@xmath201 & 0.1f & 1406@xmath202 & 141/138 + 1244 + 026 & 2.54@xmath144 & 0.15@xmath144 & 0.01@xmath144 & @xmath24 & 6.13@xmath203 & 0.1f & 642@xmath204 & 197/187 + 1613 + 658 & 1.92@xmath205 & 0.13@xmath147 & 0.04@xmath196 & 0.73@xmath206/0.22@xmath207 & @xmath24 & @xmath24 & @xmath24 & 109/97 + ashton , c. e. , page , m. j. , blustin , a. j. , puchnarewicz , e. m. , branduardi - raymont , g. , mason , k. o. , crdova , f. a. , & priedhorsky , w. c. 2004 , , 335 , 73 ballantyne , d. r. , ross , r. r. , & fabian , a. c. 2001 , , 327 , 10 boroson , t. a. , & green , r. f. 1992 , , 80 , 109 brandt , w. n. , laor , a. , & wills , b. j. 2000 , , 528 , 637 brinkmann , w. , grupe , d. , branduardi - raymont , g. , & ferrero , e. 2003 , a&a , 398 , 81 brinkmann , w. , papadakis , i. e. , & ferrero e. 2004 , a&a , 414 , 107 brinkmann , w. , wang , t. , grupe , d. , & raeth , c. 2006 a&a , 450 , 925 brocksopp , c. , starling , r. l. c. , schady , p. , mason , k. o. , romero - colmenero , e. , & puchnarewicz , e. m. 2006 , , 366 , 953 brusa , m. , gilli , r. , & comastri , a. 2005 , , 621 , l5 comastri , a. , brusa , m. , & gilli , r. 2007 , in relativistic astrophysics and cosmology einstein s legacy , ed . b. aschenbach et al . ( astro - ph/0603139 ) crummy , j. , fabian , a. c. , brandt , w. n. , & boller , th . 2005 , , 361 , 1197 crummy , j. , fabian , a. c. , gallo , l. , & ross , r. r. 2006 , , 365 , 1067 dasgupta , s. , rao , a. r. , dewangan , g. c. , & agrawal , v. k. 2005 , , 618 , 87 dickey , j. m. , & lockman , f. j. 1990 , , 28 , 215 fabian , a. c. , rees , m. j. , stella , l. , & white , n. e. 1989 , mnras , 238 , 729 george , i. m. , turner , t. j. , netzer , h. , nandra , k. , mushotzky , r. f. , & yaqoob , t. 1998 , , 114 , 73 george , i. m. , turner , t. j. , yaqoob , t. , netzer , h. , laor , a. , mushotzky , r. f. , nandra , k. , & takahashi , t. 2000 , , 531 , 52 gondoin , p. , orr , a. , lumb , d. , & santos - lle , m. 2002 , , 388 , 74 iwasawa , k. , & taniguchi , y. 1993 , , 413 , 15 jimnez - bailn , e. , piconcelli , e. , guainazzi , m. , schartel , n. , rodrguez - pascual , p. m. , & santos - lle , m. 2005 , , 435 , 449 lehmann , i. , et al . 2001 , a&a , 371 , 833 mainieri , v. , bergeron , j. , hasinger , g. , lehmann , i. , rosati , p. , schmidt , m. , szokoly , g. , & della ceca , r. 2002 , , 393 , 425 mateos , s. , barcons , x. , carrera , f. j. , ceballos , m. t. , hasinger , g. , lehmann , i. , fabian , a. c. , & streblyanska , a. 2005 , , 444 , 79 matt , g. , fabian , a. c. , & ross , r. r. 1993 , , 262 , 179 miller , b. p. , brandt , w. n. , gallagher , s. c. , laor , a. , wills , b. j. , garmire , g. p. , & schneider , d. p. 2006 , , 52 , 163 miniutti , g. , & fabian , a. c. 2006 , , 366 , 115 nandra , k. , george , i. m. , mushotzky , r. f. , turner , t. j. , & yaqoob , t. 1997a , apj , 477 , 602 1997b , , 488 , l91 nayakshin , s. , kazanas , d. , & kallman , t. r. 2000 , , 537 , 833 page , k. l. , obrien , p. t. , reeves , j. n. , & turner , m. j. l. 2004a , mnras , 347 , 316 page , k. l. , schartel , n. , turner , m. j. l. , & obrien , p. t. 2004b , mnras , 352 , 523 piconcelli , e. , jimnez - bailn , e. , guainazzi , m. , schartel , n. , rodrguez - pascual , p. m. , & santos - lle , m. 2005 , , 432 , 15 porquet , d. , reeves , j. n. , obrien , p. , & brinkmann , w. 2004 , , 422 , 85 pounds , k. a. , reeves , j. n. , king , a. r. , page , k. l. , obrien , p. t. , & turner , m. j. l. 2003 , , 345 , 705 pounds , k. a. , king , a. r. , page , k. l. , & obrien , p. t. 2003 , , 346 , 1025 pounds , k. a. , & page , k. l. 2006 , , 372 , 1275 reeves , j. n. , porquet , d. , & turner , t. j. 2004 , , 615 , 150 reynolds , c. s. , & nowak , m. a. 2003 , phys . rep . , 377 , 389 ross , r. r. , & fabian , a. c. 1999 , , 261 , 74 schartel , n. , rodrguez - pascual , p. m. , santos - lle , m. , clavel , j. , guzinazzi , m. , jimnez - bailn , e. , & piconcelli , e. 2005 , , 433 , 455 schmidt , m. , et al . 1998 , a&a , 329 , 495 schmidt , m. , & green , r. f. 1983 , , 269 , 352 streblyanska , a. , hasinger , g. , finoguenov , a. , barcons , x. , mateos , s. , & fabian , a. c. 2005 , a&a 432 , 395 tanaka , y. , et al . 1995 , nature , 375 , 659 weymann , r. j. , morris , s. l. , foltz , c. b. , & hewett , p. c. 1991 , , 373 , 23 yaqoob , t. 2005 , in populations of high energy sources in galaxies , ed . m. j. a. meurs & g. fabbiano ( cambridge : cambridge univ . press ) , 461 yaqoob , t. , & padmanabhan , u. 2004 , , 604 , 63 yuan , w. , siebert , j. , & brinkmann , w. 1998 , , 334 , 498 @xmath116ycki , p. , t. , & czerny , b. 1994 , , 266 , 653 @xmath116ycki , p. , t. , krolik , j. h. , zdziarski , a. a. , & kallman , t. r. 1994 , , 437 , 597

we have modeled the time - variable profiles of the h emission line from the non - axisymmetric disk and debris tail created in the tidal disruption of a solar - type star by a black hole . two tidal disruption event simulations were carried out using a three dimensional relativistic smooth - particle hydrodynamic code , to describe the early evolution of the debris during the first fifty to ninety days . we have calculated the physical conditions and radiative processes in the debris using the photoionization code cloudy . we model the emission line profiles in the period immediately after the accretion rate onto the black hole became significant . we find that the line profiles at these very early stages of the evolution of the post - disruption debris do not resemble the double peaked profiles expected from a rotating disk since the debris has not yet settled into such a stable structure . as a result of the uneven distribution of the debris and the existence of a `` tidal tail '' ( the stream of returning debris ) , the line profiles depend sensitively on the orientation of the tail relative to the line of sight . moreover , the predicted line profiles vary on fairly short time scales ( of order hours to days ) . given the accretion rate onto the black hole we also model the h light curve from the debris and the evolution of the h line profiles in time .

astro-ph0404256

a star in an orbit around a massive black hole can get tidally disrupted during its close passage by the black hole . after several orbital periods the debris from the disrupted star settles into an accretion disk and gradually falls into the black hole @xcite . as material gets swallowed by the black hole intense uv or soft - x ray radiation is expected to emerge from the innermost rings of the accretion disk @xcite . for black hole masses @xmath2 , tidal disruption theory predicts flares with luminosities of the order of the eddington luminosity with durations of the order of months , and with spectra that peak in the uv / x - ray domain band @xcite . high - energy flares from the central source illuminate the debris , the photons get absorbed , and some are re - emitted in the optical part of the spectrum ( i.e. the light is `` reprocessed '' ) . one of the spectral lines in which this phenomenon can be observed is the balmer series h@xmath0 line ( @xmath3 ) . the disruption of a star begins when the star approaches the tidal radius , @xmath4 , the point where the surface gravity of the star equals the tidal acceleration from the black hole across the diameter of the star ( @xmath5 and @xmath6 are the radius and mass of the star and @xmath7 is the mass of the black hole ) . a @xmath1 black hole is often used as a prototypical example in tidal disruption calculations . this choice is motivated by the criterion for a solar - type star to be disrupted before it crosses the black hole horizon ( i.e. the schwarzchild radius , @xmath8 ) in order for emission to be observable . for supermassive black holes with @xmath9 , @xmath10 and the star falls into the black hole before it gets disrupted . the tidal disruption process has been the subject of many simulations @xcite . it has been shown that tidal processes in the vicinity of a massive black hole could lead to tidal capture , tidal heating and tidal spin - up of a star @xcite , and in some cases ultimately to the explosion of the star . such explosions , as well as accretion of post - disruption debris , should manifest themselves as luminous flares from the centers of galaxies @xcite . stars close to a black hole may experience mixing or may eject some of their mass @xcite . as a consequence , stellar populations in nuclear clusters are expected to be somewhat unusual in comparison with populations whose evolution was not affected by a massive black hole @xcite . this has important implications for observations of stellar cluster in the center of our galaxy @xcite where high concentration of otherwise rare blue he supergiants has been observed @xcite . the tidal disruption and accretion of stars can fuel black holes in the centers of galaxies @xcite and its contribution to nuclear activity in galaxies and the growth of the black hole mass depends on the rate of disruption events in a galaxy . the predicted tidal disruption rate in a typical inactive galaxy is @xmath11@xmath12 @xcite . this value is consistent with the rate of uv / x - ray outbursts observed with _ rosat _ from inactive nuclei selected as tidal disruption candidates @xcite . the rate of tidal disruptions in active and more luminous nuclei is estimated to be lower , with the lowest value of @xmath13 , for galaxies with the most massive black holes . this may occur partly because massive central black holes ( @xmath14 ) swallow stars promptly , without disruption , and partly because stars are less centrally concentrated in these galaxies @xcite . the observed uv / x - ray flaring rate in these galaxies is about @xmath15 @xcite suggesting that in such nuclei , outbursts may be due to another mechanism , such as accretion - disk instabilities @xcite . the tidal encounter of a compact star with a black hole can also result in emission of gravitational waves , which may be observable with upcoming instruments . more specifically , compact stars ( helium stars , white dwarfs , neutron stars , and stellar - mass black holes ) which can withstand large tidal forces without being disrupted , may get captured in relativistic orbits around a supermassive black hole . due to the in - spiral and decay of the orbit those objects are expected to emit the peak of their gravitational wave power in the lisa frequency band @xcite . it has been recently suggested by @xcite that very - low mass main sequence stars ( msss ; @xmath16 ) may contribute to events detected by lisa , which was not previously expected for capture of these objects by a supermassive black hole . although msss produce a relatively weak gravitational signal during the in - spiral , compared to compact objects , their detection in the galactic center is more likely because such stars have a predicted close - encounter rate that is an order of magnitude higher than that of white dwarfs ( wds ) , and about two or more orders of magnitude higher than that of neutron stars ( nss ) and stellar - mass black holes ( bhs ) . these compact msss are expected to produce a strong enough signal to allow for 0.52 detections from our galactic center , with a signal - to - noise ratio of 10 or higher , for a lisa mission duration of one year @xcite . moreover , msss are expected to be disrupted relatively early during the in - spiral , giving rise to possibly detectable electromagnetic flares . the sudden appearance of an electromagnetic counterpart to a transient gravitational wave source , expected in the case of msss and helium stars , could allow identification of the progenitor . in the case of a tidal disruption event in another galaxy , the coincidence of an electromagnetic flare and a gravitational wave signal would provide an indication that the event occurred at the very nucleus of the galaxy , and possibly allow the measurement of the redshift . more compact objects which can spiral in without being disrupted ( such as wds , nss , and bhs ) are expected to produce stronger gravitational wave signatures and weaker or no electromagnetic flares , with the exception of white dwarfs in which tidal interaction may trigger thermonuclear explosion @xcite . in view of the above theoretical considerations , it is necessary to make predictions of the likely observational signatures of a tidal disruption event . the prompt uv / soft - x - ray flash that is expected to accompany the disruption does provide strong evidence for such an event and has in fact been detected in a number of galaxies with _ rosat _ and _ hst _ @xcite . however , the duration of this flash is short enough that it can easily be missed . aftereffects with a longer duration , such as line emission from the debris , have a better chance of being detected . if the appearance of emission lines just after an x - ray flare could be detected from the same object it could be used to identify the tidal disruption in the early phase and would provide strong support for the overall picture , but such cases are rare ( see * ? ? ? * ; * ? ? ? a set of tantalizing observations in the past decade show that several liners ( low - ionization nuclear emission regions ; * ? ? ? * ) have transient balmer emission lines , which are often double - peaked ; examples include ngc 1097 @xcite , m81 @xcite , ngc 4450 @xcite , ngc 4203 @xcite and ngc 3065 @xcite . such line profiles are characteristic of rotating disks and resemble the persistent double - peaked balmer lines found in about 1020% of broad - line radio galaxies and about 3% of all active galaxies ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? their abrupt appearance in liners led to suggestions that this transient event was related to the tidal disruption of a star by a supermassive , nuclear black hole @xcite or a change in the structure of the accretion disk associated with a change in accretion rate @xcite . to investigate the possibility of line emission from the post - disruption debris and to evaluate the suggestion that the transient double - peaked lines of liners are related to tidal disruption events , we have undertaken a calculation of the strength and profile of the h@xmath0 line emitted from the debris . in 2 , we describe two sph ( smoothed particle hydrodynamics ) simulations of tidal disruption on which we base our further calculation of the line properties . our line profile calculation follows the method used for line profiles emitted by relativistic , keplerian disks and is described in 3 . in 4 we present the resulting line profiles and in 5 we discuss the physical conditions in the debris and the approximations used . in 6 we summarize our conclusions and consider future prospects . tidal disruption simulations were carried out with a three dimensional relativistic sph code in order to study the dynamical evolution of the post - disruption debris . the sph code used provides a description of relativistic fluid flows in a static curved spacetime geometry @xcite . we use it to simulate the tidal disruption of a star in the potential of a schwarzchild black hole . the main sequence star is modeled as a polytrope with index @xmath17 . the density profile of the pre - disruption star is determined by the lane - emden equations . the star is initially placed on a parabolic orbit at a distance of @xmath18 from the black hole where @xmath19 is the gravitational radius and @xmath20 is the mass of the black hole . hereafter we use units in which @xmath21 , where @xmath22 is the gravitational constant and @xmath23 is the speed of the light and we adopt @xmath24 as a natural unit of length . the gravitational radius of a @xmath1 black hole is @xmath25 light seconds and the dynamical time at a given radius is @xmath26 s. the strength of the tidal encounter , is given by the ratio of tidal radius to pericentric distance , @xmath27=@xmath28 . for the case of @xmath29 star and a @xmath1 black hole @xmath30 . the two sph simulations describe the tidal disruption for the case of mildly relativistic encounter , @xmath27=1.2 . this value has been selected as a likely scenario for tidal disruption of a main sequence star . we do not investigate the strongly relativistic cases where tidal compression could lead to the explosion of a star , because such an explosion could introduce an uncertainty in the distribution of the debris mass over binding energy and consequently in the spatial distribution and kinematics of the debris . the self gravity of the star is accounted for initially . once the star gets disrupted the self - gravity becomes unimportant and the debris particles follow nearly keplerian orbits . it can be shown for the case of the keplerian potential that the rate of return of bound debris to the pericenter follows @xmath31 @xcite . this behavior of the debris return rate has been observed in our sph simulations . once bound debris starts to rain down on the black hole it is expected to cause the initial rapid rise in the emitted uv / x - ray light curve and steady decay with the power law index of @xmath32 later on . the two different simulations have 5,000 and 20,000 particles ( hereafter 5k and 20k respectively ) contributing equally to the mass of a @xmath29 star . the 5k sph calculation follows the debris for 94 days in total . after 34 days significant accretion onto the black hole begins . our investigation follows the evolution of the line profiles in the last 60 days . the 20k sph simulation spans 53 days , during which the evolution of the line profiles was followed for the last 6 days ( table 1 ) . using both the 5k and 20k sph simulations in the line profile modeling we take advantage of the longer time span in the former and better resolution achieved with the larger number of particles in the latter . figure [ fig_map ] shows particle distribution maps after the second pericentric passage , at the beginning of the accretion phase and at the end of the 5k simulation . at the early stages of the tidal event most of the particles were located in the pronounced tidal tail . sixty days later , about @xmath33 of the particles are scattered from the tidal tail and form a quasi - spherical distribution , with most of its mass concentrated in the equatorial plane @xcite . this is a consequence of the intersection of the leading part of the tidal stream with itself @xcite . we refer to the spheroidal part of the debris as the halo and to its planar component as the disk . the remaining @xmath34 of particles are still confined to the tail and @xmath35 are accreted onto the black hole . there is a concern that some fraction of particles ( @xmath36 , where @xmath37 and n are the total duration and total number of particles in the run ) of the halo are an artifact of the sph simulation . this can be due to the tendency of the sph numerical method to preserve the constant number of neighbor particles for each particle during the calculation . in the regions with a small density of particles this leads to a `` smoothing '' over a large spatial range and it may introduce the scatter of particles from the debris plane to the halo . these particles can not be distinguished from the population of particles scattered out of the debris plane by the intersection of the tidal tail with itself . we further discuss the implications of the spheroidal halo for the emission line profiles and total h@xmath0 luminosity in [ s_halo ] . the velocity distribution in the tail is `` bimodal '' where the central part of the tail exhibits very low radial velocities : particles on the near ( right - hand ) side of the tail `` flow '' towards the black hole , while particles on far ( left - hand ) side are moving in the opposite direction . this is a consequence of the energy distribution throughout the debris in the disruption process : after the initial disruption event @xmath38 of the debris stays bound to the system and the other @xmath38 is unbound . this effect has been predicted by theory in the case of stellar disruption after a single fly - by of the star @xcite and has been observed in tidal disruption simulations . the symmetry in the distribution of the debris over binding energy is a consequence of the spin - up of the star at the expense of orbital kinetic energy . the spin - up initially causes the development of a quadrupolar deformation . as the tidal interaction gets stronger the star starts to shed its mass since the material in the stellar bulge has reached the escape velocity at the star s surface . one portion of the stellar debris ends up deeper in the potential well of the black hole , which causes further spread in binding energy of the debris . this effect determines which portion of the debris stays bound to the black hole @xcite . following the second passage of the debris through pericenter approximately 66@xmath39 of the mass is unbound , 33@xmath39 remains bound and only about 1@xmath39 is accreted by the black hole . the maximal approaching and receding velocities in the debris , with respect to the stationary observer positioned at infinity , are of order @xmath40c . the dynamical evolution in the 20k run is the same , it is just followed over a shorter evolutionary time scale . since the tidal tail includes a large fraction of bound and unbound particles in both simulations , its morphology and velocity field greatly influence the observed line profiles . we follow the line profile calculations carried out by @xcite and @xcite to obtain the observed profile from a keplerian , relativistic , thin disk in the weak field approximation . the description of the debris as a flat , thin structure is justified by the fact that the height of the debris is three orders of magnitude less then its dimensions in the orbital plane . the main objective of the calculation is to obtain the final expression for the flux density in the observer s frame as a function of parameters defined in the reference frame of the debris . figure [ fig_sketch ] shows the coordinate system and the geometry of the debris . the observer is located on the positive z - axis , at a distance @xmath41 and above the orbital plane at @xmath42 to the z@xmath43-axis . since the calculation is presented in cited papers we just introduce its main steps and comment on its application to the case of tidal disruption debris . the total emission - line flux received from the debris by an observer at infinity is given by an integral over the plane of the image produced at infinity , namely @xmath44 where @xmath45 , @xmath46 , and @xmath47 are the frequency , specific intensity and solid angle element measured in the frame of the image ( i.e. , of the observer ) . using the impact parameter of rays at infinity , @xmath48 , as a coordinate in the image plane and exploiting the lorentz invariance of the quantity @xmath49 and the fact that the debris is confined to a plane , equation ( [ eq_f ] ) can be transformed into an expression for the flux density ( i.e. , the line profile ) in terms of coordinates and physical quantities in the source frame ( see detailed derivation in * ? ? ? * chen & halpern 1989 , and eracleous et al . 1995 ) @xmath50 @xmath51 where the new polar coordinates in the debris plane are the dimensionless radius @xmath52 and the azimuthal angle @xmath53 . in practice , the integration is performed by summing over particles , assigning to each particle an emissivity according to its position , as derived from calculation with the code cloudy ( * ? ? ? * see [ s_lcurve ] ) . the limits of integration describe the portion of the debris that emits the h@xmath0 line , between radii @xmath54 and @xmath55 . the function @xmath56 , the `` doppler factor , '' is determined by the phase space distribution of the emitting particles and the metric , and describes the effects of gravity and the motion of the emitting particles on the energies of the emitted photons . the function @xmath57 is determined by the geometrical distribution of the emitting particles and the metric , and describes the effects of curved trajectories of light rays . in the special case where the debris is confined to a plane , these functions are given by @xmath58^{1/2}}{1 - 2/\xi } \right.\ ] ] @xmath59 with @xmath60 the lorentz factor , @xmath61 and @xmath62 the radial and azimuthal velocities of debris particles _ in the source frame _ , and @xmath63 the above analytic expression for @xmath64 has been derived in the weak field approximation and is accurate to order @xmath65 . this approximation is appropriate in our case because the portions of the debris that experience a strong gravitational field are also highly ionized and make a negligible contribution to the h@xmath0 flux ( see discussion in [ s_phys ] ) . the ratio @xmath66 describes how rays emitted from the debris are mapped to points in the image at infinity and is given by @xmath67 . finally , the lorentz factor is given by @xmath68^{-1/2}\ ; . $ ] the emission properties of the debris are described by the local specific intensity , @xmath69 . we take the local line profile to be a gaussian corresponding to a velocity dispersion @xmath70 ( in units of the speed of light ) . the width of the local line profile represents not only internal motions of the line - emitting gas ( not captured by the sph simulation ) but also the velocity range between the discretized debris points we use in our numerical integration . we further assume that the emissivity of the line is a power - law with radius ( see discussion in [ s_lcurve ] and [ s_phys ] ) . therefore , we write the specific intensity as @xmath71 \label{eq_emiss}\ ] ] where @xmath72 is defined by @xmath73 , where @xmath45 and @xmath74 are observed and rest frame frequency , and @xmath75 is a constant . in the final model , the line profiles are described by the following parameters : the inner and outer radius of the line - emitting portion of the debris , @xmath54 and @xmath55 , the particle emissivity power - law index @xmath76 and @xmath77 , the inclination of the debris plane @xmath78 , the local velocity dispersion @xmath70 in units of @xmath23 , and the radius of the central continuum source @xmath79 . the last parameter is relevant to the calculation of the light travel time across the debris , which we describe in [ s_delay ] , below . in the emission model we adopt , a central source of finite dimensions illuminates the debris . the luminosity of this source is proportional to the accretion rate onto the black hole . due to the finite velocity of light , at any given time ( in the observer s frame ) , different portions of the debris are seen to respond to a different level of illumination . it is also noteworthy that the length scales and time scales in this problem span a very large dynamic range . as a consequence , the light - crossing time of the outer portions of the debris is longer than the dynamical time of the inner portions , which makes it necessary for us to follow the redistribution of the debris in phase space and the variations of the x - ray source carefully . it is simple to show that travel time delay for light rays can be written as @xmath80 where @xmath81 and @xmath53 are the coordinates of a particle in the orbital plane , @xmath79 is the radius of the central source and @xmath78 is the inclination . since light rays travel in the gravitational potential of the black hole they suffer an additional , relativistic time delay , which can be calculated from the equation of geodesics for photons @xcite . the assumptions are that photons travel in an isotropic gravitational field and that their trajectories can be considered coplanar with the observer and the black hole . we consider only the gravitational delay caused by the black hole s potential and assume that the debris does not have any significant gravitational influence on a light ray . this is a reasonable assumption since @xmath82 . in our notation , the gravitational time delay for non - rotating black hole can be expressed as @xmath83 \hbox to 0.5truein { \hfill}\ ] ] @xmath84 .\ ] ] the overall delay for a particle caused by light travel and general relativistic effects is then @xmath85 we only need to account for relative time delays , which we calculate relative to a ray coming from the origin of the coordinate system of the debris . we find that the general relativistic time delay in our calculations never exceeds @xmath86 of the travel time delay , and it is typically of order a few percent . the output of the 5k run comprises 351 frames describing the evolution of the debris morphology over 60 days of accretion , with a time step of approximately 4 hours ( see table 1 ) . this is a fine enough temporal resolution to trace the redistribution of particles in the debris . for comparison , a particle at @xmath87 orbits the black hole with a period of approximately 24.5 hours . during the accretion phase of the simulation the number of particles decreases due to infall in the black hole . tracking this number allows us to follow the accretion rate and construct the x - ray light curve . the total amount of mass accreted in this run is about @xmath88 . the x - rays resulting from accretion illuminate the debris out to large distances from the black hole and power the emission of h@xmath0 photons . the output of the 20k run consists of 204 temporal frames spanning 6 days of accretion with a time step of about 45 minutes ( the coverage is not even ; it includes gaps since the behavior of the debris can be captured even with sparse sampling ) . because of the short time span of accretion , the number of accreted particles is small ( less than 0.1@xmath39 of the total mass ) and consequently , the illuminating , x - ray light curve is not smooth . to address this issue we compute the h@xmath0 light curve for two more , fiducial , illumination light curves : one consistent with the debris return rate predicted by theory ( @xmath89 ; * ? ? ? * ; * ? ? ? * ) , return rate of particles to the pericenter , based on the keplerian orbits ] and the other , constant in time . this allows us to also examine the effect of different illumination patterns on the line profiles , which we describe in [ s_variability ] . in order to calculate the resulting h@xmath0 luminosity of the debris it is necessary to determine the efficiency with which the debris reprocesses the incident radiation . this efficiency can be characterized by the surface emissivity of the debris as a function of radius , @xmath90 , as shown in equation ( [ eq_emiss ] ) . we used the photoionization code cloudy @xcite for numerical calculations of the response of the debris to illumination . from the calculated surface emissivity as a function of radius ( @xmath91 ) we were able to determine the particle emissivity as a function of radius @xmath92 ( @xmath93 ) , where @xmath94 is the particle emissivity power - law index , given the surface density distribution of the debris , @xmath95 . the response generally depends on a spectral energy distribution ( sed ) of the incident radiation , matter density distribution , and the system geometry . we have described the debris system in terms of these physical parameters and performed cloudy calculations . we have assumed a sed for the illuminating radiation of the form @xmath96 , extending to 100 kev . the normalization for the sed is set by the accretion luminosity of the debris . the mass accreted during the accretion phase of the 5k simulation gives rise to a time - average accretion luminosity of @xmath97 , while the maximum luminosity achieved at the beginning of the accretion is @xmath98 ( where @xmath99 ) . the luminosity during flares therefore may exceed the eddington luminosity , @xmath100 . for the purpose of cloudy calculations , in order to characterize the emissivity of the debris , we have used the time - average value for the accretion luminosity . our estimated accretion luminosity is comparable to what was observed by @xcite for three tidal disruption candidates , namely @xmath101@xmath102 . based on the particle distribution in the simulation we find that the density in the tidal tail decreases with time because the tail gets stretched as the debris evolves ( cf . , figure [ fig_map ] ) . we find that for the typical sph time frame the density distribution in the tail can be approximated as @xmath103 and is in the range @xmath104@xmath105 . with path lengths of @xmath106 and @xmath107 , the corresponding tail column densities are in the range @xmath108@xmath109 . the low and high values of the column density correspond to lines of sight along the short and long axes of the tail , respectively . ( see table [ t_debris ] for values of physical parameters in the debris . ) the density of particles scattered from the tail in the spheroidal halo around the black hole is fairly low in comparison . the halo density reaches a maximum in the plane of the debris ( @xmath110 over - density ) , where particles orbiting around the black hole form a disk of radius @xmath111 . the number density in the disk decreases with radius as @xmath112 . the corresponding column densities in the directions orthogonally and radially through the disk are @xmath113 , respectively . the spheroidal part of the halo , formed from particles scattered out from the plane of the debris has the estimated number density of @xmath114 and corresponding column density of @xmath115 . the cloudy calculations show that the h@xmath0 power emitted by the mostly neutral tidal tail in response to x - ray illumination decays with distance from the center as @xmath116 , as a consequence of the density distribution in the tail and the geometry of the debris ( i.e. illumination incidence ) . the corresponding particle emissivity for the tail derived from the surface emissivity decays with the distance as @xmath117 ( fig . [ fig_emiss ] ) . the disk and the halo components are almost completely ionized and only 1 in @xmath118 hydrogen atoms is neutral , on average . the emissivity of the disk decreases with distance from the black hole as @xmath119 , approximately . the equivalent particle emissivity distribution for the disk is @xmath120 . therefore , in our description of the surface emissivity of the debris ( equation [ eq_emiss ] ) we use power - law indices of @xmath121 for the tail and @xmath122 for the disk . in our calculation for line profiles and light curves , due to the discrete nature of the sph simulation , we use the prescription for the emissivity per particle , with power - law indices of @xmath123 for the tail and @xmath124 for the disk . the results of the cloudy calculations and the power - law prescriptions derived from them are summarized in figure [ fig_emiss ] . the total @xmath125 luminosities contributed by the tail and the halo for the time - average value of the illumination @xmath126 , as calculated by cloudy , are @xmath127 and @xmath128 . the calculated value of the luminosity for the spheroidal component is @xmath129 ( also listed in table 2 ) . we discuss the emission properties of the debris in [ s_phys ] and their implications for the observability in [ s_obs ] . finally , with the above emissivity prescriptions we calculate the _ observed _ h@xmath0 luminosity curve of the debris at a particular time step by computing the time at which the light was emitted from the debris and by finding the ionizing flux that was illuminating that location at the time the emission occurred , according to the light - travel time from the black hole to that particular region of the debris . figure [ fig_lcurve ] shows three different h@xmath0 light curves from the debris confined to a plane ( assuming @xmath130 , @xmath131 ) during the 60-day accretion phase of the 5k simulation . figure [ fig_lcurve]a shows the accretion luminosity on a logarithmic scale ( solid curve ) , calculated from the accretion rate of the debris in the sph simulation . the uv / x - ray luminosity curve is arbitrarily scaled and overplotted on the top of the h@xmath0 curve for comparison . it is noticeable that the h@xmath0 light curve departs from the accretion light curve at late times , though the departure appears small in the logarithmic plot ( used here due to the large dynamic range of the light curves ) . the same effect is more noticeable in the figure [ fig_lcurve]b , where the accretion luminosity is proportional to @xmath132 and the h@xmath0 light curve is plotted on a linear scale . the h@xmath0 light curve roughly follows the shape of the incident uv / x - ray light curve at early times but decays faster at late times . the faster decay in the h@xmath0 light curve reflects the debris evolution in time : as the tail becomes more elongated , the incident photons travel a longer way to illuminate the debris . consequently , the intensity of the illuminating light gets lower in the later stages of the tidal disruption event . the relative decay rate of the h@xmath0 light curve with respect to the uv / x - ray light curve diminishes about 80 days after the accretion started . we find that this late relative rise in the h@xmath0 luminosity is due to the increase of number of particles in the disk component . as particles diffuse from the high density tail to the lower density disk , in later stages of the simulation , their emission efficiency increases and they contribute a significant amount of h@xmath0 light to the light curve . to isolate the effect of the debris evolution in time from the evolution of the illuminating light curve , we calculate the h@xmath0 light curve in the case of constant illumination ( fig . [ fig_lcurve]c ) . here , the relative departure of the h@xmath0 light curve from the uv / x - ray light curve can be interpreted as a consequence of the expansion and redistribution of the debris . the h@xmath0 luminosity appears to level off at late times because the debris disk begins to settle into a quasi - steady configuration . in summary , the _ observed _ h@xmath0 flux depends sensitively on the uv / x - ray light curve , on the distribution of matter that makes up the inner portion of the debris , and on how quickly particles redistribute themselves in phase space . the main features of the h@xmath0 light curve are : an initial rise followed by a decline , with superposed fluctuations . the initial rise is a consequence of the propagation of the initial illumination front through the debris at the speed of light . the fluctuations are a result of the fluctuations in the accretion rate , which are caused , in turn , by the finite number of particles employed in the simulation . the decay rate of the h@xmath0 light curve is determined by the decay rate of the uv / x - ray light curve , debris expansion and redistribution rate . we have computed sample line profiles emerging from the debris for the following choices of model parameters : 1 . _ inclination angle of the plane of the debris , @xmath78 . _ we assumed that the observer is located on the positive z - axis , at a distance @xmath133 at @xmath42 to the z@xmath43-axis . changing the inclination , changes the values of the projected velocities ( i.e. , the overall width of the line profile ) but has very little effect on its shape otherwise . 2 . _ inner and outer radius , @xmath54 and @xmath55 . _ the adopted inner radius of the debris is the inner boundary of the region from which h@xmath0 emission is expected to emerge . the choice of the inner radius depends on physical conditions in the debris , as we explain in [ s_phys ] . here , we explore several cases with @xmath54 between 200 and 10,000 . the outer radius @xmath131 is naturally set by the dimensions of the system . particle emissivity power - law index , @xmath134 . _ as noted in [ s_lcurve ] , above , we find that @xmath123 in the debris tail and @xmath124 in the disk . these indices describe the emissivity per particle as a power - law and correspond to indices @xmath76 and @xmath77 in the description of the surface emissivity . the adopted values for emissivity indices significantly influence the profile shapes . higher values of @xmath134 weigh the emissivity towards smaller radii where the projected velocity is higher . velocity dispersion , @xmath70 , and central source radius @xmath79 . _ the adopted value of the velocity dispersion for the profiles presented was @xmath135 . a lower limit on the velocity dispersion is set by the velocity difference measured for pairs of adjacent particles . this value represents the dispersion due to the finite number of resolving elements in the simulation and equals @xmath136 for 90@xmath39 of particles close to the debris plane . the constraint on the upper limit of the velocity dispersion comes from the velocity dispersion due to small scale turbulence , @xmath137 , where @xmath138 is a local circular velocity , @xmath139 is a distance from the massive black hole and @xmath140 is the smallest dimension of the fluid ( i.e. width or thickness of the tail ) . since the real turbulence could be substantially smaller than the upper limit due to dissipation by internal shocks , we adopt a value of @xmath141 . larger values of @xmath70 produce wider profiles and smooth - out sharp features . the central source radius is arbitrarily chosen to be @xmath142 . it implies a central source of finite dimensions such as a corona of ionized plasma , or a vertically extended accretion flow in the innermost parts of an accretion disk . its effect on the line profiles is rather small . in figures [ fig_trail][fig_sig ] we show sample line profiles to illustrate how they evolve in time and how they are affected by the choice of model parameters and by the orientation of the observer . figure [ fig_trail ] is a `` trailed spectrogram '' summarizing the temporal evolution of the line profiles from the two different sph runs ; it is a 2-dimensional map of the h@xmath0 emission as a function of projected velocity and time . figure [ fig_evol ] shows a different representation of the evolution of the line profile with time , which effectively comprises selected time slices from the trailed spectrogram . figures [ fig_rad ] and [ fig_ori ] show how the inner radius of the line - emitting region and the azimuthal orientation of the observer affect the observed line profiles . figure [ fig_sig ] shows the effect of the different values of velocity dispersion on the shape of the line profiles . the main properties and features of our results are as follows : _ profile variability with time . _ : : a property that is immediately obvious in the line sequence is the change of the profile shape with time ( figures [ fig_trail ] and [ fig_evol ] ) . it is noticeable that the adopted low value of velocity dispersion allows us to resolve individual particles in the trailed spectrograms , orbiting around the black hole . the evolution of the line intensities in time roughly follows the behavior of the uv / x - ray luminosity but decays somewhat faster in time ( see [ s_lcurve ] ) . the multi - peaked line profile is a consequence of the velocity field of the inner debris , which consists of the inner portion of the tidal tail that is falling towards the black hole ( towards the observer ) and debris that is rotating around the black hole after being scattered . the line profiles and their variability could be observationally important features of the debris just formed from tidal disruption . the variable line profiles might be observed and recognized on the relatively short time scale of hours to days . _ effect of the inner radius , @xmath54 . _ : : the profiles become broader as the inner radius of the line - emitting regions decreases since higher - velocity gas resides at smaller radii ( see figure [ fig_rad ] ) . the approximate full width at zero intensity of the profiles ranges from @xmath143 for @xmath144 to @xmath145 for @xmath146 . we find that line profiles change from the profiles dominated by the emission red - ward from the rest wavelength for @xmath147 to narrower profiles dominated by the blue - ward emission from the tail for @xmath148 , since for large values of @xmath54 , the high - velocity rotating gas in the vicinity of the black hole is excluded and the dominant contributions to the line profile come from the tidal tail . the intensity of the line also decreases with increasing inner radius , making the outer regions of the debris harder to observe . _ effect of observer orientation , @xmath149 . _ : : because of the non - axisymmetric geometry and velocity field , the line profiles emitted by the debris , depend on the orientation of the tidal tail relative to the observer . in figure [ fig_ori ] we show the effect of azimuthal orientation @xmath149 of the debris , with respect to the observer . the values of @xmath149 are @xmath150 , @xmath151 , @xmath152 , @xmath153 , @xmath154 and @xmath155 , as measured in a counterclockwise direction from positive x@xmath43-axis to the observer s line of sight . these can be compared with the profile corresponding to the same time in figure [ fig_evol ] for @xmath156 . the position of the peaks in figure [ fig_ori ] varies relative to the rest wavelength , since the relative direction of bulk motion of the material depends on the observer s orientation . for example , it is possible to distinguish the emission from the tail for the range of azimuthal orientations @xmath157 . the tail emission in these profiles appears as the most blueshifted peak , since these are the orientations for which different portions of the tail flow towards the observer . _ effect of velocity dispersion , @xmath70 . _ : : in figure [ fig_sig ] we show the effect of four different values of the velocity dispersion in calculation of the emission line profiles . a value of @xmath136 is the lower limit of velocity dispersion set by the discrete nature of the sph simulation , while the upper limit of @xmath158 is determined by the small scale turbulence in the debris . the velocity dispersion of particles in the halo ( about @xmath159 particles in the 20k run ) is significantly higher and reaches @xmath160 . as the value of the velocity dispersion increases the profile features get smoothed out , until only a smooth , double - peaked profile is observed . _ effect of illuminating light curve . _ : : we have computed model profiles for several different x - ray illumination light curves keeping all the other parameters fixed . we used ( a ) the light curve obtained from the accretion rate in the 5k simulation , ( b ) the light curve from the accretion rate as predicted by theory , i.e. @xmath161 @xcite , and ( c ) a light curve that is constant in time ( figure [ fig_lcurve ] ) . we find that the line profile shapes do not depend sensitively on the shape of the light curve . this is a consequence of the centrally weighted emissivity profile of the debris which causes the innermost emission region to be the dominant contributor of the h@xmath0 light . in the innermost region of the debris the dynamic range in light - travel times is not large ; therefore the illumination of the innermost emitting region is almost instantaneous . over the very short light - crossing time of the central emitting region , the gradient in the uv / x - ray light curve is small and the illumination is nearly constant over this region . the fast fluctuations in the illuminating light curve on the other hand are smoothed out during reprocessing in the debris , and can not be identified in the h@xmath0 light curve . in summary , we find that profiles are not significantly influenced by the shape of the illuminating light curve . the profile shapes are affected , however , by the inner radius of the line - emitting region and redistribution of the debris . the inner radius can change with the advance or recession of the ionization front into the debris , which is controlled primarily by the density of the debris . the most notable effect of the inner radius is on the width of the profile . since the physical conditions can change very rapidly during flares , this mechanism causes the line profiles to change on the light crossing time scale ( minutes to hours ) and evolve from wide multi - peaked to narrow and vice versa ( the recombination time is negligible in comparison to the light - crossing time of the debris , therefore particles in the debris respond effectively instantaneously to changes in the incident flux ) . the redistribution of the debris in phase space , on the other hand takes more time : @xmath16224 hours for particles in the innermost part of the emitting region . this redistribution of the debris also may cause a transition from narrower to wider profiles , however it takes place gradually , on the time scale of days . we have calculated the physical conditions and radiative processes in the debris using the photoionization code cloudy , version 94 @xcite . it is necessary to know the physical conditions in order to assess the validity of the assumptions made in our line profile calculations . since the physical conditions in the tidal tail , disk and halo differ noticeably , these three regions naturally emerge as separate components of the tidal debris . the spheroidal halo is an oblate structure of particles scattered out of the plane of the debris . the disk is produced by the flow of particles from the tail which turn around the black hole and form a higher density circular component concentrated close to the plane of the debris . the disk shows a smooth transition to the halo in terms of density and physical parameters ( see table [ t_debris ] ) . the physical differences among the components arise as a consequence of the number density , which is several orders of magnitude higher in the tail . the particle density of the halo is uncertain , due to the numerical scatter , and consequently its luminosity contribution is also subject to uncertainty . we nevertheless have calculated and presented physical properties of the spheroidal component , as given by the sph simulation , and we discuss the implications of its presence for the line profiles and the observability in the next section . the temperature of the debris tail reaches @xmath163k in the hottest parts of the tail ( i.e. the illuminated face of the tail ) and equals @xmath164k on average in the partially ionized and neutral parts of the tail . the temperature in the disk ranges from @xmath165k in the inner region to @xmath166k at the outer rim of the disk , with the average @xmath167k . the mean temperature of the halo is @xmath168k and ranges from @xmath169k in the central parts of halo , to @xmath170k in the outer halo . the ionization parameter is calculated for all components as @xmath171 , where @xmath172 is the flux of ionizing photons , @xmath173 is the total hydrogen density , and @xmath23 is the speed of light . the ionization parameter in the tail ranges from 0.1 in the parts of the tail closest to the source of ionization , to @xmath174 on the far side of the tail . the ionization parameter in the disk is almost constant throughout the disk with a value of about 20 . the ionization parameter in the halo is higher with an average value of about 27 . as a consequence , the disk and the halo are in a much higher state of ionization relative to the tail . the disk shows a wide range of hydrogen ionization fractions over its radius , with the strongest ionization at the outer rim of the disk where density is lowest . in the halo the fraction of neutral hydrogen atoms ranges from @xmath175 to @xmath176 , with an average value close to @xmath176 . in contrast , the tail is mostly neutral , with only one hydrogen ion per @xmath177 hydrogen nuclei . the properties of the tidal disruption regions are summarized in table 2 . the ionization state of the debris as well as its abundance and density determine the radiative processes dominant in the debris . the implication of the assumed solar metalicity is that emission lines from metals play an important role in the cooling of the debris . we focus our attention on processes relevant to the final h@xmath0 luminosity . two processes contribute to the h@xmath0 luminosity of the debris : recombination and collisional excitation . the dominant recombination channel is the recombination of a photoelectron to the @xmath178 level followed by a decay to the @xmath179 level via emission of an h@xmath0 photon . emission by collisional excitation occurs when hydrogen atoms in the @xmath180 and @xmath179 levels are promoted to the @xmath178 level by collisions with energetic photoelectrons and then de - excite radiatively . the relative contribution of h@xmath0 emission through recombination relative to collisions is @xmath181 where @xmath182 and @xmath183 are the respective coefficients for recombination and collision processes which lead to emission of an h@xmath0 photon , @xmath184 and @xmath185 are the number density of hydrogen ions and atoms , and @xmath186 and @xmath187 are the parts of the debris volume which gives rise to emission through recombination and collisional de - excitation , respectively . the recombination coefficient for an electron to recombine from the continuum to the n=3 level can be writen as @xmath188 for a given temperature . the adopted recombination coefficients for the tail and halo are @xmath189 , @xmath190 , and @xmath191 @xcite . similarly , it is possible to estimate the rate of collisions which will promote an electron to n=3 state . the high optical depths in the lyman series cause a significant electron population in the @xmath179 level . the photoelectron energy distribution for low - energy levels of hydrogen differs from the boltzmann distribution and the number density is not sufficiently high for collisions to lead to thermal equilibrium ( eq . 10.57 of * ? ? ? in particular , thermalized energy levels will be n@xmath1921 for the tail , n@xmath1922 for the halo , and n@xmath1923 for the disk , and the relative populations of these levels will be in agreement with their statistical weights . it is possible then to estimate the collisional excitation coefficients for transitions from the n=1 and n=2 levels to the thermalized n=3 level ( eq . 3.21 of * ? ? ? this approximately yields collisional excitation rates for the tail , disk and halo of @xmath193 , @xmath194 , and @xmath195 . with the values of the rate coefficients , the ionization state of the tail , disk and halo ( @xmath196 in the tail and @xmath118 in the halo and disk on average ) , and taking into account the ratio of emitting volumes for two processes , it is possible to calculate the ratio of respective relative contributions from recombination and collisions from equation ( [ eq_ratio ] ) to be about @xmath197 , @xmath198 , and @xmath199 . these values are not unexpected : collisional excitation is the dominant mechanism for production of h@xmath0 photons in the tail , where the density is highest , while recombination is dominant in the disk and halo . some of the h@xmath0 photons created by the two processes are destroyed on the way out of the debris . the mechanisms that impact the total output in h@xmath0 luminosity are the absorption in the h@xmath0 transition , and electron scattering , with respective optical depths @xmath200 and @xmath201 . the optical depth @xmath202 is very high for the h@xmath0 photons traveling across the tail , along the lines of incidence of illuminating photons and typically has a value of order @xmath118 . electron scattering in the tail has an average optical depth of @xmath203 . consequently , the majority of h@xmath0 photons are created in the thin ionized and partially ionized layers of the tail ( we refer to them as the tail `` skin '' ) . from here a certain fraction of h@xmath0 photons escapes the debris and reaches the observer . in the disk @xmath204 and @xmath205 . in the early stages of accretion , while the luminosity is still super - eddington , the halo is optically thin to h@xmath0 photons ( @xmath206 ) and the electron scattering optical depth is about @xmath207 . in the late stages of the debris evolution @xmath208 and @xmath209 stays approximately the same in the halo . the high values of @xmath210 of the halo material destroy the fraction of the h@xmath0 photons emitted by the part of the debris embedded in the halo . the evolution of the spherical halo would further help the process of fading of h@xmath0 emission with time along with the decreasing accretion luminosity . for the purposes of radiative transfer calculations we have modeled the tail , the disk and the halo as three _ separate _ components and therefore we do not account for the secondary absorption and electron scattering by the halo , of h@xmath0 photons created in the tail . both opacity mechanisms cause the destruction of h@xmath0 photons , if the spherical halo is fully evolved , and possibly wipe out the h@xmath0 emission line . if the same process operates in the centers of galaxies which are tidal disruption candidates , it should be possible to observe the disappearance of the broad h@xmath0 emission line on the scale of months . in our predictions of the observational signature of the post - disruption debris we have not included any contributions from the spheroidal halo because we doubt that this structure , as predicted by our sph simulation is real . in this section we discuss this issue further and justify our approach . we also describe qualitatively the effect of such a halo on the observational appearance of the debris . the halo is very likely produced artificially because a number of particles is scattered from the disk every time the flow of particles turns around the black hole and intersects with the inflowing stream of particles . the number of particles contributed to the halo by numerical scatter is proportional to @xmath211 , where n the total number of particles in the run , and multiplied by a factor @xmath212 , as described in 2 . the dynamical time scale of the disk is about 7 days for the outermost particles , which implies that the number of scattered particles is of the order of 1000 . this is an upper limit of the uncertainty , where total number of particles in the halo , in the 20k run , reaches number of 1500 particles in the later stages of the debris evolution . it has been pointed out by several authors that a spheroidal halo @xcite with a central radiatively supported torus @xcite may form around the black hole as a consequence of a self - compression of the flow , which leads to the ejection of the debris with high velocities , perpendicular to the plane . however , none of the above mentioned structures were resolved beyond doubt in the simulations . a halo - like structure is expected to be in hydrostatic equilibrium , and its size should be determined by the radiation pressure from the central source . @xcite and @xcite argue that the luminosity incident upon the halo can rarely be exactly tuned for the structure to rest in equilibrium . they further calculate that for a super - eddington luminosity the halo will expand and cool down on the time scale of months to years , until it becomes gravitationally unbound and is blown away . for a sub - eddington luminosity , on the other hand , the halo is expected to collapse due to insufficient support from radiation pressure . based on its physical conditions it is obvious that , if real , the halo observed in our simulations would make the dominant contribution to the line emission from the debris . in the late stages of the debris evolution , our simulations show the halo becoming optically thick to h@xmath0 photons . moreover , the long diffusion time of photons in the halo ( @xmath213 hours , where @xmath139 the is inner radius of the halo ) may smear out the time variability of the emission line profiles , which is one of the main signatures of the disruption event . the high value of the velocity dispersion of the particles in the halo would smear out the line profiles and very likely make them unobservable . it is therefore necessary to address in future studies what fraction of the halo ( if any ) is really present and what fraction is contributed by the numerical scatter . if a halo manages to form , survives the super - eddington phase , and achieves hydrostatic equilibrium it will be transparent to h@xmath0 photons only during the super - eddington phase . at later times , we expect that it will become optically thick and will be the primary source of illumination of the outer debris . moreover , it will fade on a time scale of months to years . in this picture , which is consistent with the predictions of @xcite , the shape of the observed line profiles should be similar to the profiles computed here for large values of the inner radius ( see figure [ fig_rad ] ) . whether the profile ( noticeable here for @xmath214 at the wavelength @xmath215 ) will appear as blueshifted or redshifted would depend on the orientation of the debris tail with respect to the observer . the cloudy calculation predicts a time - average h@xmath0 luminosity from the tidal debris of about @xmath216 , @xmath217 , and @xmath218 for the tail , disk and halo , respectively . in the earlier stages of the disruption event when the uv / x - ray luminosity is super - eddington , the h@xmath0 luminosity is expected to be up to 80 times higher than its average value and comparable to that of tidal disruption candidates observed in the local universe ( see figure [ fig_lcurve ] ) . the examples are ngc 4450 ( at 16.8 mpc ) with an h@xmath0 luminosity of @xmath219=1.8@xmath220 @xcite and ngc 1097 ( at 22 mpc ) with @xmath219=7.7@xmath221 @xcite . thus the emission - line signature of a tidal disruption event should be detectable at least out to the distance of the virgo cluster . in practice , however , the detection of such emission lines from low luminosity sources may be complicated by the weak contrast relative to the underlying stellar continuum . based on observational constraints from known tidal disruption candidates , it should be possible to detect some variable properties of the line profiles and light curves predicted here . one of the first observable effects of tidal disruption should be uv / x - ray flash accompanied by a decaying light curve , mirrored in the delayed response of the h@xmath0 light curve of the debris , with some scatter . the line profile intensities are expected to decay accordingly in time . other effects to look for are the change in the number of peaks in the line profile , relative fluctuations in intensity of the peaks as well as their shift in wavelength . the temporal variability of the h@xmath0 emission line profiles from the post - disruption debris is one of the important indicators of a tidal disruption event . in order to capture the rapid profile variability , due to the variable illumination , the exposure time should be comparable to the light crossing time of the innermost regions of the line - emitting debris , which has the fastest and strongest response to the ionizing radiation . longer exposures are expected to capture the average shape of the rapidly varying line profiles . the light - crossing time of the innermost regions of the debris is about @xmath222 minutes ( where @xmath223 ) , while the exposure times are typically about 30 - 60 minutes ( for galaxies at the distance of the virgo cluster , for example ) . thus , if an event is caught early in its evolution and the light - crossing time is relatively long ( i.e. , @xmath224 ) , there is a chance of detecting variability caused by changing illumination over the course of one to a few nights . on longer time scales , variability is caused by changes in the structure of the debris . in the presence of the spheroidal halo , the variability of the lines may be modified by the long diffusion time scale of photons through the halo . the component of the tidal tail outside the halo will then still respond to the variability but on the time scale set by the light reprocessed by the halo . in view of the predictions from our profile calculation , an important question is whether variable multi - peaked line profiles can originate in some other physical scenario or can be regarded as the unique signature of tidal disruption . multi - peaked emission lines are likely to be the signature of an inhomogeneity in the phase - space distribution of the emitting material . due to the asymmetry of the emitting region , the direction of the observer has a large influence on the observed line profile , making it difficult , if not impossible , to infer the exact emission geometry from a particular multi - peaked profile . nevertheless the variability pattern of the line profiles can serve as a general indicator of a tidal disruption event . additional observational indicators can be used in conjunction to diagnose a tidal disruption event , for example , a sharp x - ray / uv flash preceding the appearance of the emission lines , an emission line spectrum indicative of a hard ionizing continuum ( i.e. , the presence of ionic species with high ionization potentials ) , and the characteristic decay of the emission - line flux on time scales of weeks to months after the event ( the flux of different emission lines is expected to decay at different rates ; see , for example , * ? ? ? the finite resolution of the simulation could introduce uncertainties in the accretion light curve caused by the discretized accretion of the stellar material onto the black hole . another possible effect is that the finite number of particles does not completely reveal the morphology of the debris and some of its components may stay hidden . for example , when the majority of the particles are confined to the tail it is hard to say if there is an accretion disk forming around the black hole out of a small number of particles . consequentially , there is a concern that profiles calculated for low values of the inner radius , i.e. @xmath54=200 , may represent the contribution from a small number of particles in the central region and therefore dominated by small - number noise . a possible source of error is the assumption of a _ thin , flat _ structure ( i.e. , confined to a plane ) . in the early stages of the evolution of the debris , the majority of the particles are still in the tidal tail , located in a plane , which makes the assumption valid . during the evolution of the debris the number of particles that orbit around the black hole in an almost spherical distribution increases @xcite . at that point it is possible to distinguish three structural components of the debris : the relatively planar tidal tail and the disk and the spheroidal halo . the halo is made up of particles scattered from the tail by shocks during the pericentric approach of the debris or during the intersection of the tail with itself @xcite and some fraction of particles contributed by the numerical noise . since the tidal tail includes the majority of the particles , and most of the halo mass is concentrated close to the equatorial plane , even in our last frame , the assumption of a thin disk is still reasonable . if , however , the mass in the spherical halo increases at very late times , the assumption of a planar geometry needs to be reconsidered . we also adopt the weak field approximation in our calculations , which is a fairly small source of error ( of order 1% and less ) . this is valid since we adopt @xmath225 as the innermost radius of the line - emitting debris . the few particles interior to this radius would not contribute to the h@xmath0 emission because their close proximity to the source of ionizing radiation would make them fully ionized . in the case of a physical scenario where emission of h@xmath0 is not possible because of a highly ionized debris , the same profile formalism can be used to calculate the emission from other lines emitted under these conditions . we modeled the emission - line luminosity and profile from the debris released by the tidal disruption of a star by a black hole in the early phase of evolution . our model predicts prompt optical evolution of post - disruption debris and profile shapes different from circular and elliptical disk model profiles . since line profiles observed so far in liners look more disk - like and evolve slowly , the observations are likely to have caught the event at late times ( @xmath226 6 months after the initial disruption ) , after the debris has settled into a quasi - stable configuration . the line profiles can take a variety of shapes for different orientations of the debris tail relative to the observer . due to the very diverse morphology of the debris , it is almost impossible to uniquely match the multi - peaked profile with the exact emission geometry . nevertheless , the profile widths and shifts are strongly indicative of the velocity distribution and the location of matter emitting the bulk of the h@xmath0 light . profile shapes do not depend sensitively on the shape of the light curve of the x - rays illuminating the debris . they strongly depend on the distance of the emitting material from the central ionizing source , which is a consequence of the finite propagation time of the ionization front and the redistribution of the debris in phase space . it may be possible to distinguish between the two effects observationally , based on their different characteristic time scales . the onset of the optically thick spheroidal halo should cause the disappearance of the broad h@xmath0 emission line on the time scale of months , and give rise to the emission of narrower , strong , blueshifted or redshifted emission line , arising from the portion of the tidal tail unobscured by the halo . if x - ray flares and the predicted variable profiles could be observed from the same object they could be used to identify the tidal disruption event in its early phase . the x - ray flares can be promptly detected by all - sky synoptic x - ray surveys and high energy burst alert missions such as _ swift_. the evolution of the tidal event may then be followed with optical telescopes from the ground on longer time scales and give an insight in the next stage of development of the debris . thus , simulations of the tidal disruption process on longer time scales ( of order several months to a few years ) are sorely needed . calculations of the long - term evolution of a tidal disruption event can predict the type of structure that the debris finally settles into and whether its emission - line signature resembles the transient double - peaked lines observed in liners . this study would provide an important insight into the evolution of liners . finally , the observed rate of tidally disrupted solar type stars can constrain the rate of captured compact objects ( which are important gravitational wave sources ) , and the capture rate of main sequence stars in our galaxy , which are expected to emit the peak of the gravitational radiation in the lisa frequency band and can be detected in the local universe . we are indebted to j. charlton for her help with cloudy . t.b . also thanks m. falanga for his valuable comments . we are also grateful to the anonymous referee for very insightful and helpful comments and suggestions . we acknowledge the support of the center for gravitational wave physics funded by the nsf under cooperative agreement phy-0114375 , nsf grants phy-9800973 and phy-0244788 , the zaccheus daniel fellowship , and the eberly college of science . lccccccc tail & @xmath229@xmath230 & @xmath231@xmath232 & @xmath233 & @xmath234 & @xmath235 & @xmath1740.1 & @xmath236 + disk & @xmath237@xmath238 & @xmath239@xmath240 & @xmath241@xmath242 & @xmath243@xmath244 & @xmath245 & 20 & @xmath246 + halo & @xmath238 & @xmath247@xmath240 & @xmath248 & @xmath233 & @xmath249 & 27 & @xmath250 +

we discriminate gluonic hadrons from conventional states by surveying radial and orbital excitations of all i=0 and i=1 systems anticipated up to 2.1 gev . we give detailed predictions of their quasi - two - body branching fractions and identify characteristic decay modes that can isolate quarkonia . several of the `` missing mesons '' with l and l are predicted to decay dominantly into certain s+p and s+d modes , and should appear in experimental searches for hybrids in the same mass region . we also consider the topical issues of whether some of the recently discovered or controversial meson resonances , including glueball and hybrid candidates , can be accommodated as quarkonia . s1s s0s p2p p1p p0p p1p * abstract *

hep-ph9609339

theoretical studies of light hadron spectroscopy have led to the widespread belief that gluonic excitations are present in the spectrum of hadrons , so more resonances should be observed than are predicted by the conventional @xmath0 and @xmath9 quark model . the two general categories of gluonic mesons expected are glueballs ( dominated by pure glue basis states ) and hybrids ( dominated by basis states in which a @xmath0 is combined with a gluonic excitation ) . some of these novel states , notably the light hybrids , are predicted to have exotic quantum numbers ( forbidden to @xmath0 ) , such as j@xmath10 . the confirmation of such a resonance would be proof of the existence of exotic non-@xmath0 states , and would be a crucial step towards establishing the spectrum of gluonic states . there are detailed theoretical predictions for the decays of these exotic hybrids @xcite , which have motivated several experimental studies of purportedly favored hybrid channels such as @xmath11 and @xmath12 . although one would prefer to find these unambiguously non-@xmath0 j@xmath13-exotics , glueballs and hybrids with non - exotic quantum numbers are also expected . for example , in the flux tube model the lowest hybrid multiplet , expected at @xmath14-@xmath15 gev @xcite , contains the non - exotics j@xmath16 and @xmath17 in addition to the exotics @xmath18 , @xmath19 and @xmath20 . to identify these non - exotic states one needs to distinguish them from the `` background '' of radial and orbital @xmath0 excitations in the mass region @xmath21-@xmath22 gev , where the first few gluonic levels are anticipated@xcite . our point of departure is to calculate the two - body decay modes of all radial and orbital excitations of @xmath1 states ( @xmath23 ) anticipated up to 2.1 gev . this includes 2s , 3s , 2p , 1d and 1f multiplets , a total of 32 resonances in the @xmath1 sector . we also summarize the experimental status and important decays of candidate members of these multiplets , and compare the predictions for decay rates with experiment . we start by briefly reviewing the established 1s and 1p states that confirm that p0 pair creation dominates most hadronic decays . sho wavefunctions are employed for convenience ; these lead to analytic results for decay amplitudes and are known to give reasonable empirical approximations . this is sufficient for our main purpose , which is to emphasize selection rules and to isolate major modes to aid in the identification of states . in addition to the 1s and 1p states we also find reasonable agreement between the model and decays of 1d , 2p and 1f states where data exist ; this confirms the extended utility of the model and adds confidence to its applications to unknown states . examples of new results include the following . @xmath24 the radial 2@xmath4p@xmath5 @xmath25 is strongly suppressed in s - wave , and dominant in d - wave . this contrasts with the expectation for a hybrid @xmath26 . the model s prediction of a dominant d - wave has been dramatically confirmed for the @xmath27@xcite and thereby establishes 1.7 gev as the approximate mass of the @xmath1 members of the 2p nonets . this includes the @xmath28 nonet whose i=0 members share the quantum numbers of the scalar glueball . @xmath24 in the scalar glueball sector , we find that the decays of the @xmath29 and the @xmath30 are inconsistent with radially excited quarkonia . @xmath24 we identify the 2s @xmath31 nonet . the @xmath32 members are predicted to have narrow widths relative to the @xmath33 counterpart . this is consistent with the broad @xmath34 and the narrower candidates @xmath35 and @xmath36 . @xmath24 the vector states @xmath37 and @xmath38 are interesting in that the decay branching fractions appear to show anomalous features requiring a hybrid component . we identify the experimental signatures needed to settle this question . @xmath24 the @xmath39 has been cited as a likely hybrid candidate@xcite on the strength of its decay fractions . the 3s @xmath31 @xmath0 @xmath33 is also anticipated in this region . we find that the decays of the hybrid and 3s @xmath31 have characteristic differences which enable them to be distinguished . we identify modes that may enable the separation of these two configurations . our other results for the many @xmath1 states predicted up to 2.1 gev should be useful in the identification of these higher quarkonia , and in confirming that non - exotic gluonic or molecular states are indeed inconsistent with quarkonium assignments . the order of discussion is 1s and 1p ( section 2 ) ; 2s and @xmath40 ( section 3 ) ; 3s ( section 4 ) ; 2p ( section 5 ) ; 1d ( section 6 ) ; 1f ( section 7 ) . a summary and an outline for experimental strategy is in section 8 . first we will use the well known decays of light 1s and 1p @xmath1 states to motivate and constrain the p0 decay model . ackleh , barnes and swanson@xcite have carried out a systematic study of @xmath41 decays in the p0 and related pair creation decay models : in that work a p0-type amplitude was established as dominant in most light @xmath42 decays . ( for other discussions of @xmath43 decays in the p0 model see ref.@xcite ) . fig.1 , from ref.@xcite , shows p0 model predictions for the decay widths . large widths are indeed predicted to be large and smaller widths are found to be correspondingly small . if we choose the pair creation strength @xmath44 ( eq . a3 ) to set an approximately correct overall width scale , then @xmath45 and @xmath46 are both @xmath47-@xmath48 gev ; @xmath49 , @xmath50 and @xmath51 are all @xmath52-@xmath53 gev , and @xmath54 is smallest , @xmath55 gev ; all are reasonably close to the observed widths . @xmath56 figure 1 . partial widths of light 1s and 1p @xmath0 mesons in the p0 model . the model parameters shown are @xmath57-@xmath58 gev ( with @xmath59 gev preferred ) and @xmath44 . the optimum parameter values found in a fit to the partial widths of fig.1@xcite are @xmath60 gev ( which is actually the length scale most commonly used in light @xmath0 decays ) and @xmath61 ; with these values the rms relative error for these six decays is @xmath62 . in this work we have actually found that the pair production amplitude @xmath44 is somewhat large for higher - l @xmath0 states , so in our discussions of higher quarkonia we will instead use @xmath63 . in constrained-@xmath64 fits we find that using @xmath63 only moderately decreases the accuracy of the fit to the light 1s and 1p decays , to @xmath65 , with an optimum @xmath66 gev . a more sensitive test of the p0 model involves amplitude ratios in the decays @xmath67 and @xmath68 . in these decays both s- and d - wave final states are allowed , and the ratio of these decay amplitudes is known to be d / s = @xmath69 for the @xmath70 and @xmath71 for the @xmath26 @xcite . this ratio is quite sensitive to the quantum numbers of the produced pair ; with p0 quantum numbers and the usual @xmath72 we find reasonable agreement in sign and magnitude , whereas a oge pair production mechanism gives the wrong sign for d / s @xcite . this ratio test for @xmath67 was historically very important in establishing the p0 decay model @xcite . these successes of the p0 model motivate its use in predicting decays of the less familiar radial and orbital excitations of light quarkonia . we first consider the decays of the low - lying radially - excited pseudoscalar and vector states . our general approach will be to review recent data on the state in question and compare these data to predictions for candidate @xmath0 and ( where appropriate ) hybrid states . in each case we will attempt to identify decay modes that distinguish between competing assignments most clearly . @xmath24 @xmath34 the @xmath34 was first reported by bellini _ et al._@xcite in 1982 but remains rather poorly known . it is seen in @xmath75 , @xmath76 and @xmath77 , with a width of 200 - 600 mev ; there is however no accurate measurement of the branching fractions @xcite . recently higher statistics have been obtained for the @xmath34 by ves@xcite and by e852 at bnl@xcite . the ves data shows a clear @xmath34 peak in @xmath78 , with a width of @xmath79-@xmath80 mev in both @xmath81 and @xmath82 ; the latter is particularly strong and dominates this channel below 2 gev . it should be noted , however , that the size of the deck background in @xmath83 is uncertain , and it is not clear whether the @xmath34 reported in @xmath83 is actually due to the resonance . fig.1c of ref.@xcite suggests that the deck mechanism could cause _ all _ of the @xmath84 enhancement in fig.4a of that reference . we will assume that this is essentially correct , and that the @xmath34 resonance decays dominantly to @xmath85 . in the p0 decay model we expect @xmath85 to be the dominant mode of a 2s @xmath0 @xmath34 , since this is the only open two - body channel . ( we assume that the @xmath86 and @xmath87 are dominantly k@xmath88 , so the mode @xmath89 is a more complicated three body or virtual two - body decay . ) with our parameter set @xmath63 and @xmath90 gev we predict a partial width of @xmath91 this rate is given in table b2 of appendix b. ( app.b is a tabulation of all our numerical results for partial widths in the p0 model . ) in fig.2 we show the dependence of this prediction on the wavefunction length scale @xmath72 . evidently the prediction of a large width , comparable to observation , follows from any plausible choice for @xmath72 . thus the observed @xmath34 is consistent with expectations for a 2@xmath6s@xmath7 @xmath0 state . @xmath92 figure 2 . the @xmath85 partial width of a 2s @xmath34 , with p0 model parameters @xmath57-@xmath58 gev and @xmath63 . although the mode @xmath93 is nominally closed by phase space , the @xmath94 is a very broad state , so one might anticipate a significant @xmath95 mode through the low - mass tail of the @xmath94 . this possibility may be tested by varying @xmath96 ; the resulting @xmath97 does not exceed @xmath98 mev over the range @xmath99-@xmath100 mev . thus , the population of a @xmath81 mode by @xmath34 decays through an intermediate @xmath101 state is predicted to be a small effect . if there actually is a large @xmath102 mode , rather than a nonresonant deck effect , this would be in disagreement with the p0 model . thus it would be very interesting to establish the branching fraction for @xmath102 accurately in future work . @xmath24 @xmath35 this state has a width of @xmath103 mev@xcite , much narrower than its i=1 2@xmath6s@xmath7 partner @xmath34 . it has been reported in @xmath104 and @xmath105 . this small width is natural if the @xmath34 does indeed decay dominantly to @xmath106 , since g - parity forbids the analogous processes @xmath107 and @xmath108 ; to the extent that the @xmath87 and @xmath86 are dominantly @xmath109 there are no quasi - two - body @xmath43 modes open to the @xmath35 . consequently the decays must proceed through the weaker direct three - body and virtual two - body channels such as @xmath110 and @xmath111 . it is interesting to note the rle that the 2s initial wavefunction has played in our discussion . suppose for illustration that we had instead used 1s wavefunctions for the @xmath34 and @xmath35 ; we would then have predicted partial widths of several hundred mev into the low - energy tails of the modes @xmath101 and @xmath112 , with consequent broad widths for the @xmath34 _ and _ the @xmath35 , in contradiction with experiment . @xmath24 @xmath36 these successes raise provocative questions regarding the @xmath36 state(s ) . this is a purportedly complicated region which may contain more than one resonance@xcite . the pdg width of the @xmath36 is only @xmath113 mev , with signals reported in @xmath114 , @xmath104 , @xmath115 and @xmath116 . except for @xmath116 these modes are not inconsistent with a dominantly @xmath117 state . the only two - body strong channel open for a 2@xmath6s@xmath7 @xmath117 @xmath36 is @xmath114 , but this could rescatter from kk@xmath33 into the other reported modes @xmath118 and @xmath119 . the p0 model prediction for the partial width @xmath120 versus the wavefunction length scale @xmath72 is shown in fig.3 . @xmath121 figure 3 . the k@xmath122 + h.c . partial width of a 2@xmath6s@xmath7 @xmath117 @xmath36 in the p0 model . other two - body modes are excluded by phase space . evidently the predicted @xmath114 partial width is comparable to the observed width , so a 2@xmath6s@xmath7 @xmath117 assignment appears possible for this state . of course the @xmath116 mode is not expected from @xmath117 , and if confirmed may imply large @xmath123 mixing in this sector as is observed in the 1s i=0 pseudoscalars . this can be parameterized as @xmath124 a remeasurement of @xmath125 , which should be possible at bepc and tcf in @xmath126 , would be very useful in clarifying the nature of this state . ideally we would like to know the invariant mass distributions of @xmath116 , @xmath127 and @xmath128 final states , since these are flavor - tagging modes that allow investigation of possible flavor mixing in the parent resonances . similarly , an accurate measurement of the branching fractions in the flavor - tagging @xmath129 and @xmath130 hadronic decays , with @xmath131 , would be useful for the determination of the @xmath132-@xmath117 mixing angle . in summary , from the total widths alone it is possible to describe the @xmath35 and @xmath36 as unmixed @xmath1 and @xmath117 2@xmath6s@xmath7 radial excitations . the report of a large @xmath133 radiative mode however suggests flavor mixing between these states , and should be remeasured with greater sensitivity together with other @xmath134 modes . this mixing could also account for the large @xmath36 signal seen in @xmath135 by gams @xcite . @xmath24*@xmath37 , @xmath139 * if one accepts that the @xmath34 and @xmath35 belong to a @xmath73s@xmath7 @xmath0 nonet , it is then natural to assign the @xmath37 and the @xmath38 @xcite to @xmath140s@xmath5 states . indeed , one expects the contact hyperfine interaction to raise the mass of the vector nonet with respect to the pseudoscalar nonet by approximately this amount @xcite . it is unlikely that the vectors near 1.4 - 1.5 gev are dominantly d - waves , since the @xmath4d@xmath5 @xmath1 states should lie close to the other 1d candidates such as the @xmath141 , @xmath142 and @xmath143 . in the godfrey - isgur potential model a mass of 1660 mev was predicted for the @xmath4d@xmath5 state , whereas they expect the 2@xmath4s@xmath5 radial excitation at 1450 mev @xcite . the @xmath37 also lies well below flux - tube model expectations of m@xmath144-@xmath15 gev@xcite for vector hybrids , so although the possibility of light vector hybrids has been discussed @xcite , these do not appear likely unless the flux tube model for hybrids is misleading . the experimental branching fractions of these @xmath136 states are somewhat obscure , because there are at least two broad , overlapping resonances in each flavor sector in this mass region . the status of these vector states as seen in @xmath145 annihilation was reviewed recently by clegg and donnachie @xcite . in the @xmath137 sector they find that at least two states are present . the lighter state is assigned a mass of m @xmath146 gev and a width of @xmath147 gev ; it couples strongly to @xmath148 states ( including @xmath149 but not @xmath150 ) and @xmath151 , and less strongly to @xmath152 . the higher state has m @xmath153 gev , @xmath154 gev , couples most strongly to @xmath148 ( @xmath149 and @xmath150 are not separated ) and perhaps @xmath155 ; @xmath152 is also important , but the @xmath151 width is found to be small . these states have also been reported recently by crystal barrel @xcite in @xmath156 states in @xmath157 ; both vectors appear in @xmath156 , with masses and widths of m @xmath158 gev , @xmath159 gev , and m @xmath160 gev , @xmath161 gev , quite similar to the @xmath145 results . the p0 model predictions for pure 2@xmath4s@xmath5 and @xmath4d@xmath5 @xmath137 states at 1.465 gev and 1.700 gev are given in table i ( see also tables b1 , b8 ) , together with flux tube model predictions for a hypothetical 1.5 gev vector hybrid . very characteristic differences between the states are evident in their couplings to @xmath148 final states ; 2s couples very weakly to these , 1d couples strongly to both @xmath149 and @xmath150 , and the hybrid couples strongly to @xmath149 but not to @xmath150 . both quarkonium states have moderately large couplings to @xmath152 and @xmath151 , whereas the hybrid couples strongly only to @xmath149 . . partial widths of 2s , 1d and hybrid @xmath137 states . [ cols="<,^,^,^,^,^,^,^,^,^",options="header " , ] in table vi we compare the decay modes expected for a hybrid at 1875 mev with p0 model predictions for a hypothetical @xmath6d@xmath8 @xmath162 quarkonium . both assignments lead to a significant @xmath163 signal , and both predict a much larger @xmath164 mode . the most characteristic modes are @xmath165 and @xmath166 , which should be very weak for a hybrid but large for a 1d quarkonium . similar results follow for @xmath167 and @xmath149 . clearly searches for @xmath164 , @xmath165 and @xmath166 would be most useful . the large predicted coupling to @xmath168 for the @xmath169 encourages a search in @xmath170 for this state . here we consider only the @xmath4d@xmath172 and @xmath4d@xmath8 states since the @xmath4d@xmath5 vectors were previously discussed with the 2@xmath4s@xmath5 states . the @xmath173 states @xmath142 and @xmath143 are well established @xmath4d@xmath172 @xmath1 quarkonia , with masses as expected for 1d states and widths of about 200 mev . the @xmath174 ( table b7 ) is expected to decay mainly to @xmath165 @xmath175 and @xmath152 @xmath176 , with a somewhat weaker @xmath151 mode @xmath177 . experimentally the decays to @xmath148 are about @xmath178 , of which @xmath179 is @xmath151 . the @xmath152 branching fraction is observed to be @xmath180 . there are also kk and @xmath167 modes of a few percent , roughly as predicted . the total width is predicted to be 174 mev with these parameters , consistent with observation . thus the @xmath142 appears to decay approximately as predicted by the p0 model , which supports the application of the model to decays of high - l states . its isoscalar partner @xmath143 is a more interesting case . since few modes are open and the couplings are rather weak , we predict a total width of only 69 mev . although this appears inconsistent with the pdg width of 168(10 ) mev , this observed value is presumably broadened by the hadronic width of the @xmath137 and @xmath70 in the two - body modes @xmath85 and @xmath11 . the reported modes are @xmath85 and @xmath181 ; we expect @xmath85 to be dominant , with @xmath182 branches to @xmath11 ( the source of @xmath181 ? ) and kk . the kk mode affords an opportunity to measure the actual width of the @xmath183 , which may be much smaller than it appears in @xmath85 and @xmath11 modes . our results for the @xmath4d@xmath8 @xmath184 states @xmath185 and @xmath186 are especially interesting because these are missing mesons " in the quark model . we find that these are rather broad states , with total widths of about 300 - 400 mev . the @xmath187 is predicted to have a large branching fraction of @xmath188 to @xmath164 , so it should be observable in this final state or in the secondary modes @xmath151 or k@xmath189k . the @xmath190 is predicted to have an even larger branching fraction of @xmath191 to @xmath85 . it too couples significantly to k@xmath189k , and may also be observable in @xmath192 . the 1f states provide us with an opportunity to test the accuracy of the p0 decay model predictions for higher quarkonium states , since the @xmath193 and @xmath194 states expected near 2.05 gev do not have competing assignments as glueballs or hybrids . at present only two of these states are reasonably well established , the @xmath195 and @xmath196 @xcite . there is also some evidence for an @xmath197 @xcite . we do not yet have experimental branching fractions for the i=1 1f states . the @xmath196 is seen in kk and @xmath78 , and the @xmath197 is reported in @xmath78 and @xmath198 , with @xmath199 dominant . the branching fractions of the @xmath195 are known with more accuracy ; @xmath166 and @xmath152 are important modes , @xmath200 and @xmath201 . kk and @xmath202 modes are both known , with reported branching fractions of about @xmath203 and @xmath204 respectively . p0 predictions for the decays of these @xmath4f@xmath171 states are given in tables b11 and b12 . the @xmath205 is indeed expected to appear in @xmath78 ( mainly @xmath85 ) , and the dominant mode is predicted to be @xmath206 . this state is predicted to be rather narrower than reported . the @xmath197 is predicted to decay dominantly to @xmath199 , as is observed . the @xmath78 mode is also predicted to be large , and to arise from both @xmath85 and @xmath207 . the @xmath195 p0 model predictions are also in qualitative agreement with experiment , in that @xmath152 and @xmath166 are expected to be important modes , as observed . the @xmath208 partial widths to pseudoscalar pairs are uniformly too large , for example @xmath209 mev but @xmath210 mev . this decay however is g - wave , so the rate has a prefactor of @xmath211 ; this extreme sensitivity means that a small increase of @xmath72 by @xmath212 , halves the decay rate and gives agreement with experiment . thus this disagreement is quite sensitive to parameters and is probably not significant . the predictions for branching fractions of the five missing i=0,1 1f states suggest that several of them may easily be found by reconstructing the appropriate final states . the total widths of all except the @xmath4f@xmath8 states are predicted to be @xmath213 mev , so they should be observable experimentally . the @xmath214 is predicted to couple dominantly to @xmath164 . in the spin - singlet @xmath6f@xmath172 sector , the @xmath215 should appear in @xmath85 and @xmath198 , just as we found for the @xmath197 . the @xmath216 should be evident in @xmath164 , and less strongly in @xmath217 , @xmath151 and @xmath165 . modes such as @xmath164 are preferable because the two - body mesons are not excessively broad and they are far from threshold , so a resonance can be distinguished from a threshold effect . in some cases the amplitude structure of these final states is also characteristic ; these can be determined from the results quoted in app.a . the missing @xmath4f@xmath8 states may be more difficult to identify , as we predict large total widths of @xmath218 mev for these states . the @xmath219 couples most strongly to @xmath11 ; @xmath220 and @xmath221 are other important modes . its i=0 partner @xmath222 should be evident in @xmath223 and will also populate @xmath221 final states . identification of these 1f states and determination of their branching fractions and decay amplitudes will be a very useful contribution to the study of resonances , as it will allow detailed tests of the usefulness of the p0 model as a means for identifying quarkonium states in this crucial 2 gev region . we have established that the @xmath27 is very likely a 2p radial excitation . this follows from the weak s - wave and strong d - wave in @xmath82 . this also establishes the natural mass scale for the 2p multiplets as @xmath224 gev . we have been unable to identify radial scalars . these are predicted to be broad , and so their non - appearance is not surprising . conversely it raises interest in the ( relatively narrow ) @xmath29 and possible scalar @xmath225 . we do identify some ( more speculative ) potential candidates for @xmath226 2p members . we note that @xmath170 production may help identify these radial 2p states and also clarify the nature of @xmath29 and @xmath225@xcite . the @xmath34 and @xmath35 appear to be convincing 2s states . this conclusion is based on their relative widths ; the large @xmath82 mode of the @xmath34 has no analog for its @xmath32 counterparts . the status of the @xmath36 remains open ; the mass and width suggest a dominantly @xmath117 state , but the @xmath227 mode argues against it . studies of @xmath228 and @xmath229 may identify the flavor content of these @xmath32 states . the @xmath37 and @xmath38 have masses that are consistent with radial 2s but their decays show characteristics of hybrids , as noted previously @xcite . we suggest that these states may be 2s - hybrid mixtures analogous to the 3s - hybrid mixing suggested for the @xmath230@xcite . this can be tested by accurate measurement of the partial widths of these states and their vector partners at 1.6 - 1.7 gev to @xmath152 , @xmath151 , and especially @xmath150 and @xmath149 . the 3s @xmath33 is expected in the @xmath231 mev mass region as is a @xmath232 hybrid . we find that the decay patterns of these states are very different . a strong @xmath233 from the hybrid contrasted with a large @xmath234 mode from the 3s quarkonium is the sharpest discriminant . the ves state @xmath39 clearly exhibits this hybrid signature . it is now necessary to establish the presence of @xmath31 in the @xmath234 channel , and to see if any resonant state is present that is distinct from the @xmath39 seen in @xmath233 . it is possible that there are two @xmath235 states , @xmath0 and hybrid , whose production mechanisms and decay fractions differ sufficiently so that they can be separated . we suggest that the possibility of two such @xmath235 states be allowed for in data analyses . in the immediate future there are opportunities for @xmath170 physics at lep2 and at b factories . possible strategies for isolating some of these higher quarkonia include : @xmath24 @xmath236 contains ( i ) @xmath234 which may access the radial @xmath237 and @xmath238 near 1700 mev and a possible @xmath239 . ( ii ) @xmath240 which can isolate the @xmath237 if the helicity selection rule@xcite is used to suppress the @xmath238 . @xmath24 @xmath241 may access the radial @xmath242 near 1700 mev through its decay into @xmath168 . the @xmath243 channel may also be searched for the @xmath29 since this state is known to have a significant branching fraction to @xmath244 but should have a suppressed @xmath170 coupling if it is a glueball @xcite . @xmath24 @xmath245 may be searched for @xmath17 states in order to verify whether the established @xmath141 is accompanied by a higher @xmath246 in @xmath247 and @xmath248 . this @xmath78 system may also be studied for evidence of one or more @xmath39 states . @xmath24 @xmath249 may access the isoscalar partners of these @xmath250 states . in the near future it will be possible to study @xmath251 annihilation up to @xmath252 gev at dafne . the channels @xmath253 should be measured and @xmath254 and @xmath255 states separated in order to carry out the analysis of hybrid and radial vector components in section 3b . the isoscalar partners of the vectors also need confirmation , and final states with kaons are needed to investigate possible @xmath138-@xmath256 mixing ; a potential weakness of the present data analyses is that such flavor mixing is assumed to be unimportant . in the next century there will be new opportunities at the compass facility at cern . this will enable further studies of central production and also of diffractive excitation . for the latter one may anticipate improved studies of the @xmath33 excitations ( such as the @xmath34 and @xmath39 states ) , possibly including primakoff excitation . judicious studies of specific final states as discussed above may help separate 3s and hybrid states . the use of k beams will allow analogous studies of the strange counterparts of these states and may help to clarify the spectrum of quarkonia , glueballs and hybrids . experiments with @xmath33 beams can access the following interesting channels . @xmath24 @xmath257 , to confirm the d - wave dominance of @xmath258 and to seek its partner @xmath238 . @xmath24 @xmath259 can access both @xmath260 and @xmath261 . these can be separated in @xmath11 ; the singlet selection rule forbids this mode for @xmath260 but allows it for @xmath262 . @xmath263 can also separate @xmath260 from @xmath262 ; @xmath264 is the dominant mode whereas @xmath262 is much suppressed into s+s hadrons . @xmath24 @xmath265 , @xmath266 and @xmath267 are important in the interpretation of the vectors between 1.4 and 1.7 gev , which may contain large hybrid components . @xmath24 @xmath268 and @xmath269 can all be searched for evidence of @xmath39 states . @xmath24 @xmath270 or @xmath271 access respectively @xmath272 and @xmath273 . finally , many two - body channels are predicted to couple strongly to specific 2p , 1d and 1f states , as shown in appendix b. these include missing mesons " such as the @xmath4f@xmath8 and most 2p states , and studies of these two - body final states may reveal the missing resonances . the modes @xmath164 , @xmath165 and @xmath11 are important for many of these missing states and merit careful investigation . we reiterate that it is in general a good strategy to study decays into both s+s and s+p meson modes , as the relative couplings of these modes are usually quite distinct for hybrid versus quarkonium assignments . we would like to acknowledge useful communications with c.amsler , d.v.bugg , s.u.chung , g.condo , k.danyo , a.dzierba , s.godfrey , i.kachaev , y.khokhlov , a.kirk , d.ryabchikov and a.zaitsev . this work was supported in part by the united states department of energy under contracts de - fg02 - 96er40944 at north carolina state university and de - ac05 - 96or22464 managed by lockheed martin energy systems inc . at oak ridge national laboratory . fec is supported in part by european community human capital mobility programme eurodafne , contract chrx - ct92 - 0026 . n.isgur , r.kokoski and j.paton , phys . 54 , 869 ( 1985 ) . f.e.close and p.r.page , nucl . b443 , 233 ( 1995 ) ; phys . d52 , 1706 ( 1995 ) . n.isgur and j.paton , phys . d31 , 2910 ( 1985 ) . t.barnes , f.e.close and e.s.swanson , phys . d52 , 5242 ( 1995 ) ; see also c.michael @xcite . g.bali _ et al . _ ( ukqcd collaboration ) , phys . b309 , 378 ( 1993 ) ; d.weingarten , nucl . b ( proc . suppl . ) 34 , 29 ( 1994 ) ; c.michael , liverpool report lth 370 , hep - ph/9605243 ( may 1996 ) ; f.e.close and m.j.teper , on the lightest scalar glueball " , ral-96 - 040 / outp-96 - 35p ( july 1996 ) . j.sexton , a.vaccarino and d.weingarten , phys . 75 , 4563 ( 1995 ) . d.v.amelin _ et al . _ ( ves collaboration ) , phys . b356 ( 1995 ) 595 . s.u.chung , private communication . f.e.close , p.1395 in proc . xxvii international conf . on high energy physics , glasgow , ( institute of physics , u.k . , 1994 , p.bussey and i.knowles eds . ) . a.m.zaitsev ( ves collaboration ) , p.1409 in proc . xxvii international conf . on high energy physics , glasgow , ( institute of physics , u.k . , 1994 , p.bussey and i.knowles eds . ) . e.s.ackleh , t.barnes and e.s.swanson , `` on the mechanism of open - flavor strong decays '' , ornl - ctp-96 - 03 , hep - ph-9604355 ( april 1996 ) , phys . d ( to appear ) . a.leyaouanc , l.oliver , o.pne and j.raynal , phys . d8 , 2223 ( 1973 ) ; see also _ ibid . _ , d9 , 1415 ( 1974 ) ; d11 , 1272 ( 1975 ) ; l.micu , nucl . b10 , 521 ( 1969 ) . g.busetto and l.oliver , z.phys . c20 , 247 ( 1983 ) ; r.kokoski and n.isgur , phys d35 , 907 ( 1987 ) ; p.geiger and e.s.swanson , phys . d50 , 6855 ( 1994 ) ; h.g.blundell and s.godfrey , phys . d53 , 3700 ( 1996 ) . particle data group , phys . d54 , 1 ( 1996 ) . g.bellini et al . , phys . 48 , 1697 ( 1982 ) . particle data group , phys . d50 , 1173 ( 1994 ) . prokoshkin ( gams collaboration ) , in proc . of leap96 , dinkelsbhl , germany , 27 - 31 august 1996 . a.b.clegg and a.donnachie , z. phys . c62 , 455 ( 1994 ) . s.godfrey and n.isgur , phys . d32 , 189 ( 1985 ) . a.donnachie and yu.s.kalashnikova , z.phys . c59 , 621 ( 1993 ) . a.abele _ et al . _ ( crystal barrel collaboration ) , high - mass @xmath137-meson states from @xmath274-annihilation at rest into @xmath275 " , phys . b ( to appear ) . ves collaboration , diffractive reaction @xmath276 study at 37 gev / c " . ( unpublished ) t.barnes and e.s.swanson , in preparation . y.khokhlov , private communication r.m.baltrusaitis _ et al . _ ( markiii collaboration ) phys . 55 , 1723 ( 1985 ) ; phys . d33 , 1222 ( 1986 ) . d.bisello _ et al._. ( dm2 collaboration ) phys . b192 , 239 ( 1987 ) ; phys . d39 , 701 ( 1989 ) . d.v.bugg _ et al . b353 , 378 ( 1995 ) . c.amsler and f.e.close , phys . b353 , 385 ( 1995 ) ; phys . d53 , 295 ( 1996 ) . j.h.lee _ et al . b323 , 227 ( 1994 ) . t.degener ( crystal barrel collaboration ) , in proc . of leap96 , dinkelsbhl , germany , 27 - 31 august 1996 . g.kernel ( argus ) , proc . of photon95 ( world scientific , 1995 , eds . d.j.miller , s.l.cartwright and v.khoze ) , pp.226 - 231 , esp . fig.2 ; e.krini , doctoral thesis , university of ljubljana ( 1993 ) ; e.krini , proc . xxvii international conf . on high energy physics , glasgow 1994 , ( institute of physics , u.k . , 1994 , p.bussey and i.knowles eds . ) , p.1413 . z.p.li , f.e.close and t.barnes , phys . d43 , 2161 ( 1991 ) ; e.s.ackleh , t.barnes and f.e.close , phys . d46 , 2257 ( 1992 ) ; t.barnes , in proc . ixth international workshop on photon - photon collisions , la jolla , ca 22 - 26 march 1992 ( world scientific , 1992 , d.o.caldwell and h.p.paar eds . ) . see d.morgan , m.r.pennington and m.r.whalley , j. phys . g20 , a1 ( 1994 ) for a review of @xmath277 data . h.albrecht _ et al . c50 , 1 ( 1991 ) . h.albrecht _ et al . _ ( argus collaboration ) , phys . b374 , 265 ( 1996 ) . g. m. beladidze _ et al . _ , c54 , 367 ( 1992 ) . b287 , 368 ( 1992 ) . b.may _ et al . _ ( asterix collaboration ) , phys . b225 , 450 ( 1989 ) ; e.aker _ et al . _ ( crystal barrel collaboration ) , phys . b260 , 249 ( 1991 ) . j.bai et al ( bes collaboration ) the structure analysis of the @xmath225 in the radiative decay @xmath278 " , ihep report ( july 1996 ) . f.e.close , g.farrar and z.p.li , determining the gluonic content of isoscalar mesons " , ral-96 - 052 . n.a.trnqvist , phys . 67 , 556 ( 1991 ) . k.dooley , e.s.swanson and t.barnes , phys . b275 , 478 ( 1992 ) . c.daum _ et al . b182 , 269 ( 1981 ) . g.condo _ et al . d43 , 2787 ( 1991 ) . see also y.eisenberg _ et al . 23 , 1322 ( 1969 ) ; d.aston _ et al . _ , b189 , 15 ( 1981 ) . d.i.ryabchikov ( ves collaboration ) , proc . of hadron95 , manchester , u.k . , aug . d.antreasyan _ et al . _ ( crystal ball collaboration ) , z.phys . c48 , 561 ( 1990 ) . h.j.behrend _ et al . _ ( cello collaboration ) , z.phys . c46 , 583 ( 1990 ) . d.v.bugg et al . , z.phys . c ( to appear ) , study of @xmath279 at 1200 mev / c " . c.amsler _ et al . _ ( crystal barrel collaboration ) , z. phys . c71 , 227 ( 1996 ) . f.e.close and p.r.page , physics letters b366 , 323 ( 1996 ) . we quote results for the p0 model a@xmath280bc meson decay amplitudes in terms of an invariant amplitude @xmath281 , which is the l@xmath282s@xmath282 projection of the p0 pair creation hamiltonian matrix element divided by a momentum conserving delta function , @xmath283 this amplitude and the derivation of the @xmath4p@xmath7 matrix elements are discussed in detail in appendix a of ackleh _ et al . the partial widths @xmath284 are related to these decay amplitudes by @xmath285 the full p0 decay amplitude is the sum of two feynman diagrams , called @xmath286 and @xmath287 ( fig.a1 ) . in a specified flavor channel these diagrams have flavor weight factors that multiply the spin - space matrix element . the flavor factors for all the processes considered in this paper are given in table a1 . the @xmath288 amplitudes listed below are for unit flavor factors , @xmath289 and @xmath290 , with the phase chosen so they add rather than cancel . ( the cancelations are due to flavor symmetries such as g - parity . ) thus for a physical decay such as @xmath291 one should multiply the unit - flavor amplitude @xmath288 in a3 by @xmath292 before computing the decay width using ( a2 ) . some states populate several decay channels , for example @xmath293 as well as @xmath294 ; to sum over all channels one should multiply the width by the flavor multiplicity factor @xmath295 in the table . in these flavor weights the pairs @xmath296 and @xmath297 are equivalent , up to factors due to identical particles in the final state . @xmath298 figure a1 . @xmath0 meson decay diagrams in the @xmath4p@xmath7 decay model . @xmath299 generic decay @xmath299 & @xmath299 subprocess @xmath299 & @xmath300 & @xmath301 & @xmath302 + @xmath303 & @xmath304 & @xmath292 & @xmath305 & @xmath306 + @xmath307 & @xmath308 & @xmath305 & @xmath305 & @xmath309 + @xmath310 & @xmath311 & @xmath312 & @xmath305 & @xmath313 + @xmath314 & @xmath315 & @xmath292 & @xmath305 & @xmath313 + @xmath316 & @xmath317 & @xmath312 & @xmath318 & @xmath306 + @xmath319 & @xmath320 & @xmath292 & @xmath292 & @xmath306 + @xmath321 & @xmath322 & @xmath305 & @xmath305 & @xmath323 + @xmath324 & @xmath325 & @xmath292 & @xmath312 & @xmath323 + @xmath326 & @xmath327 & @xmath328 & @xmath312 & @xmath313 + we take all spatial wavefunctions to be sho forms with the same width parameter @xmath72 ; as a result the @xmath329 decay amplitudes are proportional to an overall gaussian in @xmath330 times a channel - dependent polynomial @xmath331 , @xmath332 where @xmath64 is the p0 pair production coupling constant @xcite . to specify these amplitudes it suffices to quote the polynomial @xmath331 for each decay channel . the complete set of p0 decay amplitudes for all @xmath0 resonances with `` excitation level '' @xmath333n@xmath334l@xmath335 decaying into final states with @xmath336 and c @xmath337s@xmath7 ( and c = @xmath338s@xmath5 in most cases ) is given below . for the relatively obscure transitions 3s @xmath280 1d + c , 1f @xmath280 1p + c , 1f @xmath280 2p + c and 1f @xmath280 1d + c we restrict c to @xmath339s@xmath7 ; this does not exclude any decays allowed by phase space . we include a few additional amplitudes in this list . some of these are of interest as couplings to virtual two - body states , although phase space nominally forbids the decay . ' '' '' .1 cm ' '' '' .5 cm @xmath340 @xmath341 .5 cm ' '' '' .1 cm ' '' '' .5 cm ( _ see 1s @xmath280 1s + 1s for channel coefficients . ) _ .5 cm @xmath342 .5 cm ' '' '' .5 cm @xmath343 @xmath344 .5 cm ' '' '' .1 cm ' '' '' .5 cm ( _ see 1s @xmath280 1s + 1s for channel coefficients . ) _ .5 cm @xmath345 .5 cm ' '' '' .5 cm @xmath346 @xmath347 .5 cm ' '' '' .5 cm ( _ see 2s @xmath280 1p + 1s for channel coefficients . ) _ .5 cm @xmath348 .5 cm ' '' '' .5 cm ( _ see 2s @xmath280 1p + 1s for channel coefficients . ) _ .5 cm @xmath349 .5 cm ' '' '' .5 cm @xmath350 @xmath351 .5 cm ' '' '' .1 cm ' '' '' .5 cm @xmath352 @xmath353 .5 cm ' '' '' .1 cm ' '' '' .5 cm ( _ see 1p @xmath280 1s + 1s for channel coefficients . ) _ .5 cm @xmath354 .5 cm ' '' '' .5 cm @xmath355 @xmath356 .5 cm ' '' '' .5 cm @xmath357 .5 cm ' '' '' .1 cm ' '' '' .5 cm @xmath358 @xmath359 .5 cm ' '' '' .5 cm @xmath360 @xmath361 @xmath362 .5 cm ' '' '' .5 cm @xmath363 .5 cm ' '' '' .1 cm ' '' '' .5 cm @xmath364 @xmath365 .5 cm ' '' '' .5 cm @xmath366 @xmath367 .5 cm ' '' '' .5 cm ( _ see 1f @xmath280 2s + 1s for channel coefficients . ) _ .5 cm @xmath368 .5 cm ' '' '' .5 cm @xmath369 .5 cm ' '' '' .5 cm @xmath370 .5 cm ' '' '' .5 cm @xmath371 in this appendix we quote numerical values for partial widths predicted by the p0 model . the masses used are experimental values of well established candidates , usually taken from the 1996 pdg , otherwise we used an approximate multiplet mass . these are 1700 mev ( 2p ) , 1670 mev ( 1d ) , 2050 mev ( 1f ) , and 1900 mev and 1800 mev respectively for the 3@xmath372s@xmath5 and 3@xmath373s@xmath7 . the lighter meson masses assumed are @xmath374 mev , @xmath375 mev , @xmath376 mev , @xmath377 mev and @xmath378 mev . for other states we used the 1996 pdg masses except for the broad @xmath379 , which we left at 1300 mev . although we found optimum parameters near @xmath380 and @xmath381 gev in a fit to light 1s and 1p decays , these parameters lead to moderate overestimates of the widths of the well established higher - l states @xmath141 and @xmath195 ; with this @xmath72 a value closer to @xmath382 is preferred . consequently we quote widths for all these higher quarkonia with the parameters @xmath383 the tables are largely self explanatory . except in a few cases the states are specified uniquely by their labels . the exceptions include the @xmath384 and @xmath385 , which we take to be the usual @xmath386 combinations of @xmath387 and @xmath388 basis states . we assume that the @xmath389 and @xmath390 are pure @xmath387 states . the strange mesons k@xmath391 and k@xmath392 are taken to be the linear combinations and @xmath394 this gives a zero s - wave @xmath395 coupling ; experimentally d / s @xmath396 , and the small partial width implies a small s - wave amplitude . the orthogonal state @xmath397 ( b3 ) is predicted to have a d / s ratio of @xmath398 in @xmath399 , quite close to the experimental d / s @xmath400 . the large @xmath401 mode is not predicted and is possibly due to a virtual intermediate state such as k@xmath402 followed by a final - state interaction . the tables give partial widths for all nonstrange 2s , 3s , 2p , 1d and 1f quarkonia to all two - body modes allowed by phase space , rounded to the nearest mev . the predictions of the dominant modes of the `` missing states '' in the quark model , such as the @xmath184 states and most of the 1f states , are especially interesting . if the p0 model has even moderate accuracy these tables should be very useful in searches for these states . mode & .5 cm @xmath37 .5 cm & mode & .5 cm @xmath38 .5 cm + + @xmath403 & 74 . & & + @xmath404 & 122 . & @xmath405 & 328 . + @xmath406 & 25 . & @xmath407 & 12 . + + @xmath408 & 0 . & & + + @xmath409 & 1 . & @xmath410 & 1 . + @xmath411 & 3 . & & + @xmath412 & 0 . & & + + @xmath413 & 35 . & & 31 . + @xmath414 & 19 . & & 5 . + + @xmath415 & 279 . & & 378 . + @xmath416 & 310(60 ) & & 174(59 ) + mode & .5 cm @xmath418 .5 cm & mode & .5 cm @xmath419 .5 cm + + @xmath403 & 1 . & & + @xmath404 & 5 . & @xmath405 & 14 . + @xmath406 & 8 . & @xmath407 & 8 . + @xmath420 & 11 . & @xmath421 & 10 . + @xmath422 & 92 . & & + + @xmath408 & 70 . & & + @xmath423 & 50 . & @xmath424 & 121 . + + @xmath409 & 32 . & @xmath410 & 75 . + @xmath425 & 4 . & @xmath426 & 6 . + @xmath411 & 26 . & & + @xmath412 & 46 . & & + + @xmath427 & 0 . & @xmath428 & 0 . + @xmath429 & 0 . & & + @xmath430 & 0 . & & + + @xmath431 & 0 . & & + @xmath432 & 0 . & @xmath433 & 0 . + @xmath434 & 0 . & @xmath435 & 0 . + @xmath436 & 0 . & @xmath437 & 0 . + mode & .5 cm @xmath418 .5 cm & mode & .5 cm @xmath419 .5 cm + + @xmath413 & 1 . & & 1 . + @xmath414 & 21 . & & 21 . + @xmath438 & 27 . & & 27 . + + @xmath439 & 5 . & & 5 . + @xmath440 & 4 . & & 4 . + + @xmath415 & 403 . & & 292 . + mode & .5 cm @xmath39 .5 cm & mode & .5 cm @xmath441 .5 cm + + @xmath417 & 31 . & & + @xmath442 & 73 . & @xmath443 & 112 . + & & @xmath444 & 36 . + + @xmath424 & 53 . & & + + @xmath445 & 7 . & @xmath446 & 30 . + @xmath447 & 28 . & @xmath448 & 61 . + + @xmath414 & 36 . & & 36 . + + @xmath415 & 228 . & & 275 . + @xmath416 & 212(37 ) & & + mode & .5 cm @xmath449 .5 cm & .5 cm @xmath27 .5 cm & .5 cm @xmath450 .5 cm + + @xmath451 & 23 . & & 5 . + @xmath452 & 10 . & & 5 . + @xmath453 & 104 . & 58 . & + @xmath454 & 109 . & 15 . & 46 . + + @xmath455 & 3 . & & 43 . + @xmath424 & 0 . & 41 . & + + @xmath410 & 28 . & 41 . & 165 . + @xmath456 & & 2 . & + @xmath457 & 4 . & 18 . & 30 . + @xmath458 & 20 . & 39 . & + + @xmath413 & 20 . & & 0 . + @xmath414 & 17 . & 33 . & + + @xmath415 & 336 . & 246 . & 293 . + mode & .5 cm @xmath459 .5 cm & .5 cm @xmath460 .5 cm & .5 cm @xmath461 .5 cm + + @xmath403 & 81 . + @xmath462 & 4 . & & 0 . + @xmath463 & 1 . & & 16 . + @xmath464 & 159 . & 27 . & 72 . + @xmath465 & 56 . & 6 . & 22 . + + @xmath408 & 8 . & & 130 . + + @xmath466 & & 1 . & + @xmath411 & 16 . & 70 . & 122 . + @xmath412 & 43 . & 86 . & + + @xmath413 & 20 . & & 0 . + @xmath414 & 17 . & 33 . & + + @xmath415 & 405 . & 224 . & 409 . + mode & .5 cm @xmath467 .5 cm & mode & .5 cm @xmath468 .5 cm + + @xmath469 & 56 . & @xmath453 & 173 . + @xmath406 & 18 . & @xmath407 & 17 . + @xmath422 & 60 . & & + + @xmath423 & 13 . & @xmath424 & 31 . + + @xmath409 & 0 . & @xmath410 & 0 . + @xmath466 & 2 . & & + @xmath411 & 10 . & & + @xmath412 & 67 . & & + + @xmath414 & 30 . & & 30 . + + @xmath415 & 257 . + mode & .5 cm @xmath142 .5 cm & .5 cm @xmath185 .5 cm & .5 cm @xmath470 .5 cm + + @xmath403 & 59 . + @xmath471 & 19 . & 73 . & 35 . + @xmath472 & 2 . & 28 . & 16 . + @xmath464 & 71 . & 15 . & 14 . + + @xmath408 & 0 . & & 0 . + @xmath423 & 0 . & 0 . & 0 . + + @xmath409 & 6 . & 5 . & 124 . + @xmath466 & & 0 . & + @xmath411 & 1 . & 3 . & 134 . + @xmath412 & 4 . & 201 . & + + @xmath413 & 9 . & & 36 . + @xmath414 & 2 . & 44 . & 26 . + + @xmath415 & 174 . & 369 . & 435 . + @xmath416 & 215(20 ) & & 235(50 ) + mode & .5 cm @xmath143 .5 cm & .5 cm @xmath186 .5 cm & .5 cm @xmath473 .5 cm + + @xmath453 & 50 . & 221 . & 101 . + @xmath474 & 2 . & . + + @xmath424 & 0 . & 0 . & 0 . + + @xmath410 & 7 . & 8 . + + @xmath413 & 8 . & & 35 . + @xmath414 & 2 . & 44 . & 21 . + + @xmath415 & 69 . & 300 . & + @xmath416 & 168(10 ) & & 220(35 ) + mode & .5 cm @xmath141 .5 cm & mode & .5 cm @xmath475 .5 cm + + @xmath453 & 118 . & @xmath476 & 33 . + @xmath454 & 41 . & @xmath444 & 8 . + + @xmath424 & 0 . & & + + @xmath410 & 0 . & & + @xmath456 & 0 . & @xmath466 & 0 . + @xmath457 & 1 . & @xmath411 & 5 . + @xmath458 & 75 . & @xmath412 & 189 . + + @xmath414 & 30 . + + @xmath415 & 250 . & & 261 . + @xmath416 & 258(18 ) & & 180 @xmath477(25 ) + mode & .5 cm @xmath196 .5 cm & .5 cm @xmath197 .5 cm & .5 cm @xmath219 .5 cm + + @xmath478 & 12 . + @xmath479 & 3 . & & 13 . + @xmath480 & 33 . & 86 . & 37 . + @xmath481 & 54 . & 28 . & 19 . + + @xmath455 & 1 . + @xmath482 & 0 . & & 0 . + @xmath424 & 0 . & 1 . + + @xmath410 & 20 . & 12 . & 140 . + @xmath456 & & 4 . & + @xmath457 & 2 . & 6 . & 36 . + @xmath458 & 10 . & 67 . & + @xmath483 & & 0 . & + @xmath484 & 0 . & 1 . & 16 . + @xmath485 & 0 . & 24 . & + @xmath486 & 0 . & 40 . & 21 . + @xmath487 & 0 . & . + + @xmath428 & 0 . & 0 . & 2 . + @xmath488 & & 0 . & + @xmath489 & 0 . & 0 . & 0 . + @xmath490 & 0 . & + mode & .5 cm @xmath196 .5 cm & .5 cm @xmath197 .5 cm & .5 cm @xmath219 .5 cm + + @xmath491 & 0 . & + @xmath433 & 0 . & 1 . & 1 . + @xmath435 & 0 . & 1 . & 89 . + @xmath437 & 2 . & 127 . & 1 . + + @xmath413 & 8 . & & 14 . + @xmath414 & 4 . & 28 . & 15 . + @xmath438 & 9 . & + + @xmath492 & & 0 . & + @xmath439 & 0 . & + @xmath440 & 0 . & 0 . & 0 . + @xmath493 & 0 . & 31 . & + + @xmath415 & 161 . & 483 . & + @xmath416 & 427(120 ) & 340(80 ) & + mode & .5 cm @xmath195 .5 cm & .5 cm @xmath214 .5 cm & .5 cm @xmath222 .5 cm + + @xmath403 & 62 . & & 34 . + @xmath462 & 2 . & & 4 . + @xmath463 & 0 . + @xmath494 & 0 . & & 0 . + @xmath464 & 86 . & 37 . & 31 . + @xmath465 & 27 . & + + @xmath408 & 2 . & & 1 . + + @xmath495 & 0 . & & 0 . + + @xmath466 & & 2 . & + @xmath411 & 9 . & 20 . & 113 . + @xmath412 & 22 . & 192 . & 40 . + @xmath496 & & 0 . & + @xmath497 & 0 . & 0 . & 13 . + @xmath498 & 1 . & 25 . & 5 . + + @xmath499 & & 0 . & + @xmath429 & 0 . & 0 . & 1 . + @xmath430 & 0 . & + mode & .5 cm @xmath195 .5 cm & .5 cm @xmath214 .5 cm & .5 cm @xmath222 .5 cm + + @xmath431 & 1 . & 4 . + + @xmath413 & 9 . & & 14 . + @xmath414 & 5 . & 26 . & 15 . + @xmath438 & 10 . & + + @xmath492 & & 0 . & + @xmath439 & 0 . & 2 . & 91 . + @xmath440 & 0 . & 0 . & 0 . + @xmath493 & 0 . & 23 . & + + @xmath415 & 237 . & 350 . & 579 . + @xmath416 & 208(13 ) & & + mode & .5 cm @xmath216 .5 cm & mode & .5 cm @xmath215 .5 cm + + @xmath469 & 37 . & @xmath453 & 115 . + @xmath406 & 13 . & @xmath407 & 13 . + @xmath420 & 4 . & @xmath421 & 4 . + @xmath422 & 33 . & & + + @xmath423 & 1 . & @xmath424 & 1 . + @xmath500 & 0 . & @xmath501 & 0 . + + @xmath409 & 0 . & @xmath410 & 0 . + @xmath425 & 0 . & @xmath426 & 0 . + @xmath466 & 1 . & & + @xmath411 & 14 . & & + @xmath412 & 107 . & & + @xmath502 & 3 . & @xmath503 & 12 . + + @xmath427 & 0 . & @xmath428 & 0 . + @xmath499 & 0 . & & + @xmath429 & 0 . & & + @xmath430 & 1 . & & + mode & .5 cm @xmath216 .5 cm & mode & .5 cm @xmath215 .5 cm + + @xmath504 & 0 . & & + @xmath505 & 0 . & @xmath433 & 0 . + @xmath506 & 1 . & @xmath435 & 2 . + @xmath507 & 48 . & @xmath437 & 138 . + + @xmath414 & 22 . & & 22 . + @xmath508 & 5 . & & 5 . + + @xmath492 & 0 . & & 0 . + @xmath439 & 0 . & & 0 . + @xmath440 & 0 . & & 0 . + @xmath493 & 17 . & & 17 . + + @xmath415 & 308 . & & 330 .

the reflexive game theory ( rgt ) has been recently proposed by vladimir lefebvre to model behavior of individuals in groups . the goal of this study is to introduce the inverse task . we consider methods of solution together with practical applications . we present a brief overview of the rgt for easy understanding of the problem . we also develop the schematic representation of the rgt inference algorithms to create the basis for soft- and hardware solutions of the rgt tasks . we propose a unified hierarchy of schemas to represent humans and robots . this hierarchy is considered as a unified framework to solve the entire spectrum of the rgt tasks . we conclude by illustrating how this framework can be applied for modeling of mixed groups of humans and robots . all together this provides the exhaustive solution of the inverse task and clearly illustrates its role and relationships with other issues considered in the rgt .

1011.3397

the reflexive game theory ( rgt ) has been entirely developed by lefebvre @xcite and is based on the principles of _ anti - selfishness or egoism forbiddeness _ @xcite and human _ reflexion processes _ @xcite . therefore rgt is based on the human - like decision - making processes . the main goal of the theory is to model behavior of individuals in the groups . it is possible to predict choices , which are likely to be made by each individual in the group , and influence each individual s decision - making due to make this individual to make a certain choice . in particular , the rgt can be used to predict terrorists behavior @xcite . in general , the rgt is a simple tool to predict behavoir of invididuals and influence individuals choices . therefore it makes possible to control the individuals in the groups by guiding their behavoir ( decision - making , choices ) by means of the corresponding influences . on the other hand , now days robots have become an essential part of our life . one of the purposes robots serve to is to substitute human beings in dangerous situations and environments , like defuse a bomb or radioactive zones etc . in contrast , human nature shows strong inclinations towards the risky behavior , which can cause not only injuries , but even threaten the human life . the list of these reasons includes a wide range starting from irresponsible kids behavior to necessity to find solution in a critical situation . in such a situation , a robot should full - fill a function of refraining humans from doing risky actions and perform the risky action itself , if needed . however , robots are forbidden and should not physically force people , but must convince people on the mental level to refrain from doing a risky action . this method is more effective rather than a simple physical compulsion , because humans make the decisions ( choices ) themselves and treat these decisions as their own . such technique is called a _ reflexive control _ @xcite . the task of finding appropriate reflexive control is closely related with the inverse task , when we need to find suitable influence of one subject on another one or on a group of subject on the subject of interest . therefore , it is needed to develop the framework of how to solve the inverse task . this is the primary goal of this study . however , for better understanding of the gist of the inverse task and its intrinsic relationships with other issues of the rgt , we introduce the entire spectrum of the tasks , which can be solved by the rgt . this forms the scope of inference algorithms used in the rgt . we present the rgt algorithms in the form of the _ schemas of control systems _ that can be instantly applied for developement of soft- or / and hardware solutions . we develop a hierarchy of control systems for abstract individual ( including human subject ) and robotic agent ( robot ) based on these control schemas . finally , we illustrate application of the inverse task together with other rgt inference algorithms to model robot s behavior in the mixed groups of humans and robots . the rgt deals with groups of abstract subjects ( individuals , humans , autonomous agents etc ) . each subject is assigned a unique variable ( _ subject variable _ ) . any group of subjects is represented in the shape of _ fully connected graph _ , which is called a _ relationship graph_. each vertex of the graph corresponds to a single subject . therefore the number of vertices of the graph is in one - to - one correspondence with overall number of subjects in the groups . each vertex is named after the corresponding subject variable . the rgt uses the set theory and the boolean algebra as the basis for calculus . therefore the values of subject variables are elements of boolean algebra . all the subjects in the group can have either alliance or conflict relationship . the relationships are identified as a result of group macroanalysis . it is suggested that the installed relationships can be changed . the relationships are illustrated with graph ribs . the solid - line ribs correspond to alliance , while dashed ones are considered as conflict . for mathematical analysis alliance is considered to be conjunction ( multiplication ) operation ( @xmath0 ) , and conflict is defined as disjunction ( summation ) operation ( + ) . the graph presented in fig . [ fig : fig1]a or any graph containing any sub - graph isomorphic to this graph are not decomposable . in this case , the subjects are excluded from the group one by one , until the graph becomes decomposable . the exclusion is done according to the importance of the other subjects for a particular one @xcite . any other fully connected graphs are decomposable . any decomposable graph can be presented in an analytical form of a corresponding _ polynomial_. any relationship graph of three subjects is decomposable ( see @xcite ) . consider three subjects @xmath1 and @xmath2 . let subject @xmath3 is in alliance with other subjects , while subjects @xmath4 and @xmath2 are in conflict ( fig . [ fig : fig1]b ) . the polynomial corresponding to this graph is @xmath5 . , [ b]$ ] and @xmath6 $ ] are elementary polynomials.,height=75 ] regarding a certain relationship , the polynomial can be stratified ( decomposed ) into _ sub - polynomials _ each sub - polynomial belongs to a particular level of stratification . if the stratification regarding alliance was first built , then the stratification regarding the conflict is implemented on the next step . the stratification procedure finalizes , when the _ elementary polynomials _ , containing a single variable , are obtained after a certain stratification step . the result of stratification is the _ polynomial stratification tree ( pst)_. it has been proved that each non - elementary polynomial can be stratified in an unique way , i.e. , each non - elementary polynomial has only one corresponding pst ( see @xcite considering one - to - one correspondence between graphs and polynomials ) . each higher level of the tree contains polynomials simpler than the ones on the lower level . for the purpose of stratification the polynomials are written in square brackets . the pst for @xmath5 polynomial is presented in fig.[fig : fig2 ] . next , we omit the branches of the pst and from each non - elementary polynomial write in top right corner its sub - polynomials . the resulting tree - like structure is called a _ _ diagonal form__@xcite . consider the diagonal form corresponding to the pst in fig . [ fig : fig2 ] : @xmath7 + [ c ] } \\ { } & { [ a][b + c ] } & { } \\ { [ a(b + c ) ] } & { } & { \;\;\;\;\;\;\;\;\;\;\;\;\ ; . } \\ \end{array}\ ] ] hereafter , the diagonal form is considered as a function defined on the set of all subsets of the _ universal set_. the universal set contains the _ elementary actions_. for example , these actions are actions @xmath8 and @xmath9 . by definition , boolean algebra _ of the universal set includes four elements : @xmath10 , @xmath11 , @xmath12 and the empty set 0 = @xmath13 . these elements are all the possible subsets of universal set and considered as alternatives that each subject can choose . the alternative @xmath14 is interpreted as an inactive or idle state . in general , boolean algebra consists of @xmath15 alternatives , if universal set contains @xmath16 actions . accroding to definition given by lefebvre @xcite , we present here exponential operation defined by formula @xmath17 where @xmath18 stands for negation of @xmath19 @xcite . this exponential operation is used to fold the diagonal form . during the folding , round and square brackets are considered to be interchangeable . the following equalities are also considered to be true : @xmath20 and @xmath21 . next we implement folding of diagonal form of polynomial @xmath5 : @xmath22 + [ c ] } & { } & { } & { } & { } \\ { } & { [ a][b + c ] } & { } & { } & { } & { [ a]([b + c ] + \overline{[b ] + [ c ] } ) } & { } \\ { [ a(b + c ) ] } & { } & { } & = & { [ a(b + c ) ] } & { } & { = a(b + c ) + \overline a \ ; . } \\ \end{array}\ ] ] it is considered that the levels of the pst represent different processing levels of natural or artificial cognitive system . each level is considered as an images . the root of the tree is the input into the cognitive system and , therefore can be considered as the image of the world ( environment including self and others ) , perceived by the subject . as it follows from the pst , there is a hierarchy of images , corresponding to a particular cognitive level . during processing along this hierarchy in the _ bottom - up _ manner , the image on the lower level undergoes an extensive process of simplification by the means of decomposition into simpler parts on the higher level . these parts are considered to be the images of the image on the previous level . therefore , the images on the second level are different representions of the original image of the world . this procedure repeats until we obtain elementary part ( elementary polynomials ) @xcite . on the other hand , the pst folding procedure can be referred as _ top - down _ intergration process of simpler images from the higher levels . therefore , the stratification procedure of original polynomial together with the folding procedure of the diagonal form illustrate the interplay of _ bottom - up _ and _ top - down _ information processes , which are widely imployed in biological @xcite and artificial @xcite information processing systems . the idea of hierarchical structure is highly coherent with hierarchical organization of majority of natural ( inanimate objects ) and biological ( living creatures ) entities . furthermore , it has been shown that hierarchical structure is intrinsic for the relationships in societies of insects @xcite , animals @xcite and human beings . therefore hierarchical representation of the groups in the form of pst correspond to extraction of the hierarchical structure of the given group , while fusion of the pst and its diagonal form with diagonal form folding procedure closely resembles the way of information processing within a single independent congnitive system as discussed above . thus , rgt imploys the fundamental principles of hierarchical organization on both group ( reflects structure of the groups ) and individual ( illustrates information processing within independent cognitive system of a single unit ) levels . this makes rgt universal tools that mildly bridges the gap between representation and analysis . the goal of each subject in a group is to choose an alternative from the set of alternatives under consideration . to obtain choice of each subject , we consider the _ decision equations _ , which contain subject variable in the left - hand side and the result of diagonal form folding in the right - hand side : @xmath23 to find solution of the decision equations , we consider the following equation : @xmath24 where @xmath25 is the subject variable , and @xmath26 and @xmath27 are some sets . eq.([canfrom ] ) represents _ the canonical form of decision equation_. this equation has solution if and only if the set @xmath27 is contained in set @xmath26 : @xmath28 . if this requirement is satisfied , then eq.([canfrom ] ) has at least one solution from the interval @xmath29 @xcite . otherwise , the decision equation has no solution , and it is considered that subject can not make a decision . in such situation , the subject is in frustration state . therefore , to find solutions of decision equation , one should first transform it into the _ canonical form_. out of three presented equations only the decision equation for subject @xmath3 is in the canonical form , while other two should be transformed . we consider explicit transformation only of decision equation for subject @xmath4 @xcite : + + @xmath30 . + therefore , @xmath31 the transformation of equation for subject @xmath2 be can be easily derived by analogy : @xmath32 . next we consider two tasks , which can be formulated regarding the decision equation in the canonical form and provide methods to solve each task . the variable in the left - hand side of the decision equation in canonical form is the variable of the equation , while other variables are considered as influences on the subject from the other subjects . the _ forward task _ is formulated as a task to find the possible choices of a subject of interest , when the influences on him from other subjects are given . after transformation of arbitral decision equation into its canonical form , the sets @xmath26 and @xmath27 are functions of other subjects influences . for example , if we consider group of subjects @xmath3 , @xmath4 , @xmath2 , etc . togehter with the abstract representation of decision equation in canonical form for subject @xmath3 , the sets @xmath26 and @xmath27 will be the functions of subject variables @xmath4 , @xmath2 , etc . : @xmath33 in the case of only three subjects @xmath3 , @xmath4 and @xmath2 , @xmath34 and @xmath35 . all the influences are presented in influence matrix ( table [ infmat ] ) . the main diagonal of influence matrix contains the subject variables . the rows of the matrix represent influences of the given subject on other subjects , while columns represent the influences of other subjects on the given one . the influence values are used in decision equations . ' '' '' a&a&@xmath36&@xmath12 + ' '' '' b&@xmath12&b&@xmath12 + ' '' '' c&@xmath12&@xmath12&c + [ table : tab1 ] for subject @xmath3 : @xmath37 . for subject @xmath4 : @xmath38 . for subject @xmath2 : @xmath39 . equation for subject @xmath3 does not have any solutions , since set @xmath40 is contained in set @xmath41 : @xmath42 . thus , subject @xmath3 can not make any decision . therefore he is considered to be in frustration state . equation for subject @xmath4 has at least one solution , since @xmath43 . the solution belongs to the interval @xmath44 . therefore subject @xmath4 can choose any alternative from boolean algebra , which contains alternative @xmath12 . these alternatives are @xmath45 and @xmath12 . equation for subject @xmath2 turns into equality @xmath46 . this is possible only in the case , when @xmath47 . here @xmath48 . in contrast to the forward task , the _ inverse task _ is formulated as a task to find all the simultaneous ( or joint ) influences of all the subjects together on the subject of interest that result in choice of a particular alternative or subset of alternatives . we call the subject of interest to be a _ controlled subject_. let subject @xmath3 be a controlled subject and @xmath49 is a fixed value , representing an alternative or subset of alternatives , which subjects @xmath4 , @xmath2 , etc . want subject @xmath3 to choose . we call value @xmath49 to be a _ target choice_. by substituting subject variable @xmath3 with fixed value @xmath49 , we obtain the _ influence equation_. if we substitute the subject variable @xmath3 with fixed value @xmath49 in the canonical form of the decision equation ( eq . ( [ ftcnf ] ) ) , we obtain _ the canonical form of the influence equation _ : @xmath50 for only three subjects @xmath3 , @xmath4 and @xmath2 , @xmath34 and @xmath35 . in contrast to the decision equation , which is equation of a single variable , the influence equation is the equation of multiple variables . however , the number of variables of influence equation is not trivial question . in fact , the number of variables in influence equation can be less then @xmath51 , where @xmath16 is the total number of subjects in the group . there are groups , in which sets @xmath26 and @xmath27 are functions of less than @xmath51 variables ( see appendix [ appen1 ] ) . therefore the variables that present in influence equation are called _ effective variables_. the inverse task is by definition itself only guaratee that the original decision equation @xmath52 turns into true equality , but it is not guaranteed that these solutions are the only ones that turn decision equation into true equality . ] formalized as to find all the joint solutions of all subjects in the group , except for the controlled one , when the target choice is represented by interval @xmath53 , where @xmath54 and @xmath55 are some sets and @xmath56 . in such a case , to solve the inverse task , one should solve the system of influence equations : align a(b , c , ... ) = _ 1 [ sys11 g ] + b(b , c , ... ) = _ 2 [ sys21 g ] if the target choice is a single alternative , then @xmath57 . _ the solutions of the system ( [ sys11g]-[sys21 g ] ) are considered as reflexive control strategies . the solution of the inverse task in particular is characterized from two points . the first point is whether it is required to find the influence of a particular single subject or joint influences of a group of subjects . the second one is whether the target choice is represented as a single alternative or as an interval of alternatives . to illustrate these points , we introduce a particular group of subjects . let subjects @xmath3 and @xmath4 are in alliance with each other and in conflict with subject @xmath2 . the polynomial corresponding to this graph is @xmath58 . the diagonal form corresponding to this polynomial and its folding is @xmath59[b ] } & { } & { } \\ { } & { [ ab ] } & { } & { + [ c ] } & { } \\ { [ ab + c ] } & { } & { } & { } & { = ab + c } \\ \end{array}\ ] ] therefore the decision equation for all the subjects in the group is @xmath60 where @xmath25 can be any subject variable @xmath3 , @xmath4 or @xmath2 . _ influence of a single subject vs joint influences of a group . _ first we consider example , when the influence of a single subject is required . let subject @xmath4 makes influence @xmath36 and @xmath61 . then we need to find influences of a single subject @xmath2 , which result in solution @xmath61 of decision equation @xmath62 . the canonical form of this influence equation is @xmath63 . since @xmath61 , @xmath64 , we obtain a system of equations : align \ { } + c = \ { } [ sys11 ] + c = \ { } [ sys21 ] therefore , the straight forward solution of this system is @xmath65 . this simple example illustrates the very gist of the _ inverse task _ - to find the appropriate influences , which result in target choice . next , we consider that influence of subject @xmath4 is not known . therefore , we obtain system align b + c = \ { } [ sysinf11 ] + c = \ { } [ sysinf12 ] in this case , we need to find the values of variable @xmath4 , which together with @xmath2 , result in solution @xmath61 . in other words , we need to find all the pairs @xmath66 , resulting in solution @xmath61 . these pairs are solutions of the system ( [ sysinf11]-[sysinf12 ] ) . therefore , we run all the possible values of variable @xmath4 and check if the first equation of the system ( [ sysinf11]-[sysinf12 ] ) turns into true equality : + @xmath67 ; + @xmath68 ; + @xmath69 ; + @xmath70 . therefore , out of four possible values of variable @xmath4 , only two values @xmath36 and @xmath71 are appropriate . thus , we obtain two pairs @xmath66 : @xmath72 and @xmath73 . _ a single target alternative vs interval of alternatives . _ in the previous examples we considered a target choice to be only a single alternative . here we illustrate the case , when a target choice is an interval . let @xmath74 , and @xmath75 . to find corresponding influences of subject @xmath2 , we solve the system of equations : align \ { } + c = 1 [ sysinf21 ] + c = \ { } [ sysinf22 ] again , we instantly obtain the solution of this system : @xmath65 . in this section , we have formulated the inverse task in general and considered its particular formalization depending on the number of influences and what is the target choice . however , we do not have a method to solve arbitral influence equation . therefore , we solve this problem in the next section . as an introduction for this section , we consider the fundamental proposition , which will be the conner stone to solve the influence equations . [ lem1 ] let p and q be some abstract sets . then @xmath76 . _ necessity . _ let @xmath77 , then @xmath78 therefore if @xmath77 , then @xmath79 . _ sufficiency . _ let @xmath79 , then @xmath80 . @xmath81 now let us consider the new type of equation : @xmath82 this equation has solution if and only if @xmath83 . there are three operations defined on the boolean algebra . they are conjunction ( @xmath0 or multiplication ) , disjunction ( + or summation ) and negation ( @xmath84 , where @xmath25 is subject variable ) . the negation operation is unary operation , while other two operations are binary . using combination of these three operations , we can compose any influence equation . since , it is obvious how to solve the equation including only unary operation , we discuss how to solve influence equations including a single binary operation . for this perpose , we consider two abstract subject variables @xmath85 and @xmath86 and abstract alternative @xmath87 . [ lem ] the solution of equation @xmath88 regarding variable @xmath89 , where @xmath90 , is given by the interval @xmath91 , where @xmath92 . [ lemma1 ] according to proposition 1 , @xmath93 , @xmath94 , @xmath95 @xmath96 and @xmath97 . therefore , @xmath98 @xmath99 @xmath100 . consequently , we obtain eq.([diseq ] ) : @xmath101 we solve eq.([diseq ] ) regarding variable @xmath85 . first , we transform eq.([diseq ] ) into canonical form : @xmath102 therefore , the solution of eq.([diseqcn ] ) is given by the interval @xmath103 since variables @xmath85 and @xmath86 are interchangable and it is possible to solve eq.([diseq ] ) regarding variable @xmath86 as well , the general form of solution of eq.([dise ] ) is the interval @xmath104 where @xmath90 and @xmath105 [ lem3 ] the solution of equation @xmath106 regarding variable @xmath89 , where @xmath90 , is given by the interval @xmath107 , where @xmath92 . [ lemma2 ] according to proposition 1 , @xmath108 , @xmath94 , @xmath109 and @xmath97 . therefore , @xmath110 + @xmath111 . thus , we obtain eq.([coneq ] ) : @xmath112 we solve eq.([coneq ] ) regarding variable @xmath85 . first , we transform eq.([coneq ] ) into canonical form : @xmath113 since @xmath114 , the solution of eq.([coneqcn ] ) is given by the interval @xmath115 since variables @xmath85 and @xmath86 are interchangable and it is possible to solve eq.([coneq ] ) regarding variable @xmath86 as well , the general form of solution of eq.([cone ] ) is the interval @xmath116 where @xmath90 and @xmath105 since one bound of the solution intervals for eqs.([dise ] ) and ( [ cone ] ) are functions of the second variable , we need to run all the possible values of the second variable in order to obtain all possible solutions of these equations in the form of pairs @xmath117 . next we consider several examples , illustrating application of lemmas [ lemma1 ] and [ lemma2 ] . _ example 1_. for illustration , we solve equation @xmath118 . consider @xmath119 , @xmath120 and @xmath121 , we obtain the solution interval for variable @xmath122 : @xmath123 . after simplfication , we get interval ( [ inter1 ] ) : @xmath124 next we consider examples with particular alternatives . let it be alternative @xmath125 . the solution interval is then @xmath126 . since the lower bound of this interval is a function of variable @xmath4 , to find all solutions of equation @xmath118 , we calculate value of expression @xmath127 for all possible values of variable @xmath4 ( table [ solns1 ] ) . to reesure that solutions are correct , we check that decision equation @xmath128 turns into true equality for the obained pairs @xmath66 : + @xmath129 : @xmath130 is true ; @xmath131 : @xmath132 is true ; @xmath133 : @xmath134 is true ; @xmath135 : @xmath136 is true ; @xmath137 : @xmath138 is true ; @xmath139 : @xmath140 is true . + so far , we have illustrated how to solve the influence equation . we as well showed that the pairs @xmath66 obtained by solving equation @xmath118 in accordance with proposition 1 and lemmas 1 and 2 are indeed solutions of this equation . c c c c c ' '' '' values of @xmath4&@xmath36&@xmath12&1 & 0 + ' '' '' & @xmath72 & @xmath141 & @xmath135 & @xmath139 + ' '' '' & @xmath73 & & @xmath137 & + [ solns1 ] [ solns1 ] _ example 2_. we consider influence equation for subject @xmath4 obtained from eq.([canonb ] ) . @xmath142 first , we transform the left - hand side of eq.([ex2eq ] ) : + @xmath143 . + therefore , eq.([ex2eq ] ) can be rewritten as follows : @xmath144 considering , @xmath145 and @xmath146 , we instantly obtain the solution interval of eq.([ex2eq1 ] ) : @xmath147 . finally , @xmath148 _ example 3_. next , we consider influence equation @xmath149 considering , @xmath150 and @xmath151 , we instantly obtain the solution interval @xmath152 or @xmath153 therefore , in order to find all solutions of eq.([ex3eq ] ) , we need to solve the equations @xmath154 where @xmath155 is any sub - set of set @xmath87 ( @xmath156 ) . each equation can be solved according to lemma [ lemma2 ] . _ example 4_. as a final example , we again consider influence equation @xmath63 and show how application of lemma [ lemma1 ] essentially simplifies its solution . we get the system of influence equations : align b + c = \ { } ; [ sys2a ] + c = \ { } . [ sys2b ] from this system we obtain a single equation : @xmath157 according to lemma [ lemma1 ] , we instantly obtain the solution interval of eq.([deqs ] ) : @xmath158 thus , eq.([deqs ] ) has two solutions : @xmath159 and @xmath160 . therefore the solution of system ( [ sys2a]-[sys2b ] ) consists of two pairs @xmath129 and @xmath161 . to conclude this section , we provide its brief summary . we have shown how to solve the inverse task by means of influence equations . we have proved two fundamental lemmas , which allow to solve any influence equation regardless of the number of variables . finally , we have illustrated several examples of how apply these lemmas . in this section we analyze the situation , when subject can appear in frustration state , from the point of view of the inverse task . let us consider the polynomial @xmath5 discussed in the section [ repres ] . the decision equation that corresponds to this polynomial is @xmath162 , where @xmath25 can be any subject variable . next we try to find all the pairs @xmath66 such that result in selection of a particular alternative by subject @xmath3 . the decision equation for subject @xmath3 is @xmath163 . the solution interval of this decision equation is @xmath164 . we need to check which alternative subject @xmath3 can be convinced to choose . to do this , we consider the system of equation for each alternative . alternative @xmath36 : align b + c = \ { } [ sys11 ] + 1 = \ { } [ sys21 ] alternative @xmath12 : align b + c = \ { } [ sys12 ] + 1 = \ { } [ sys22 ] alternative @xmath14 : align b + c = 0 [ sys13 ] + 1 = 0 [ sys23 ] in these systems the second equation is incorrect equality . therefore these systems have no solution . alternative @xmath165 : align b + c = 1 [ sys14 ] + 1 = 1 [ sys24 ] the second equation is correct equality . therefore this system has solution . thus , out of four possible alternatives , subject @xmath3 actually can choose only alternative @xmath165 . to find solutions , resulting in selection of the alternative @xmath165 , we need to solve only eq.([sys14 ] ) , since eq.([sys24 ] ) turns into the true equality . according to lemma [ lemma1 ] , we instantly obtain the solution interval for eq.([sys14 ] ) : @xmath166 we calculate the pairs @xmath66 for all possible values of variable @xmath2 ( table [ tab3 ] ) . c c c c c ' '' '' values of @xmath2&@xmath36&@xmath12&1 & 0 + ' '' '' & @xmath141 & @xmath167 & @xmath168 & @xmath169 + ' '' '' & @xmath170 & @xmath171 & @xmath172 & + ' '' '' & & & @xmath173 & + ' '' '' & & & @xmath174 & + [ tab3 ] therefore , the influence analysis of the decision equation @xmath163 shows that the only alternative that subject @xmath3 can choose is alternative @xmath175 . the influence analysis provides us with the set ( exhaustive list ) of pairs @xmath66 of joint influences resulting in selection of alternative @xmath165 . therefore , if the pair of influences does not match any pair from this list , the decision equation has no solution and this results in frustration state . summarizing , this section we note that in general there are two sets . the set @xmath176 contains alternatives that a controlled subject can choose . the set @xmath177 is the set of altertanives of the target choice . therefore , the need to put subject @xmath3 into frustration state emerges , if the target choice of a controlled subject can not be made by this subject . in other words , we need to put a subject into frustration state , if @xmath178 . among all the possible groups , there are groups , in which subjects will always choose only the alternative @xmath165 regardless of the influence of other subjects . such groups are called _ super - active groups_. next we consider one special case of super active groups - the @xmath179 groups . the group is called @xmath179 , if all the subjects in the group are connected with the same relationship . here we provide proof of the lemma about homogenous groups originally formulated by lefebvre @xcite . [ lem4 ] any homogenous group is the super - active group . [ lemma3 ] we consider the homogenous groups , where all the subjects are connected with alliance ( alliance groups ) and conflict ( conflict groups ) relationship , separately . without loss of generallity , we suggest that there are @xmath16 subjects @xmath180 . _ alliance groups_. the polynomial corresponding to the alliance group of @xmath16 subject is @xmath181 . next we construct the diagonal form and apply folding procedure : @xmath182[{a_2}] ... [{a_n } ] } & { } \\ { [ { a_1}{a_2} ... {a_n } ] } & { } & { = [ { a_1}{a_2} ... {a_n } ] + \overline { [ { a_1}][{a_2}] ... [{a_n } ] } = 1 \ . } \\ \end{array}\ ] ] therefore the alliance groups are always super - active . _ conflict groups_. the polynomial corresponding to the conflict group of @xmath16 subject is @xmath183 . next we construct the diagonal form and apply folding procedure : @xmath182 + [ { a_2 } ] + ... + [ { a_n } ] } & { } \\ { [ { a_1 } + { a_2 } + ... + { a_n } ] } & { } & { = } \\ { [ { a_1 } + { a_2 } + ... + { a_n } ] + } & { \overline { [ { a_1 } ] + [ { a_2 } ] + ... + [ { a_n } ] } = 1 \ . } & { } \end{array}\ ] ] therefore the conflict groups are always super - active . since both the alliance and the conflict groups are super - active , this lemma is proved . @xmath81 however , there are non - homogenous super - active groups as well ( see appendix [ appen3 ] ) . summarizing this section , we note that subjects in the super - active groups can not be controlled in their choices and the entire groups is uncontrolable . therefore , once the super - active groups emerges , the only way to make it controllable is to change the relationships in the group . we have presented the detailed description of the rgt including solution of the forward and inverse tasks . we have also considered the extream cases of decisions like putting a subject into frustration state or changing structure of a super - active group . as a final stroke , we summarize all the presented material in the form of _ basic control schema of an abstract subject ( bcsas ) in the rgt_. and @xmath184.,height=302 ] the input comes from the environment and is formalized in the form of external influences on the subject , the boolean algebra of alternatives and structure of a group . information about the influences , boolean algebra and group structure is propagated into the _ decision module_. the decision module implements solution of the forward task . therefore the output set @xmath176 of the decision module is the set of possible alternatives , which subject can choose under the given conditions . the information about boolean algebra and group structure is propagated into the _ influence module_. the influence module solves the inverse task . the output set @xmath185 of the influence module is the set of the pairs @xmath186 , where @xmath87 is the target alternative , the set @xmath187 is the set of all the joint influences , resulting in selection of the target choice ; and @xmath25 represents a subject variable . each @xmath186 represents a reflexive control strategy . therefore , the decision to put a subject into @xmath188 state is justified if it is impossible to make subject @xmath25 choose the target alternative @xmath87 , i.e. , if for pair @xmath186 set @xmath189 , and subject @xmath25 should not choose any other alternative except for the target one . the alternatives @xmath87 with corresponding non - empty sets @xmath187 are included into the set @xmath185 . here we introduce set @xmath190 to store the non - empty sets @xmath187 . the schema of the algorithm for extracting sets @xmath185 and @xmath190 is presented in fig . [ fig4 ] . first the sets @xmath185 and @xmath190 are empty : @xmath191 and @xmath192 . the algorithm reads the set of pairs @xmath193 and stores it in array @xmath194 , where @xmath195 is a counting variable , @xmath196 is the total number of pairs . then it is checked for each pairs from array @xmath197 whether set @xmath187 is empty : @xmath198 . if yes , the algorithm increments counting variable @xmath199 and proceeds to the next pair from array pairs . if no , then alternative @xmath87 is included into the set @xmath185(@xmath200 ) , set @xmath185 is saved , the set @xmath187 is included into set @xmath190 ( @xmath201 ) and set @xmath190 is saved . the process is run while @xmath202 . in this iterative algorithm , we separately store the alternatives @xmath87 , which can be chosen by a certian subject , in the set @xmath185 and the joint influences @xmath187 , which result in selection of alternative @xmath87 , in the set @xmath190 . therefore , we should modify the schema of influence module in bcsas as follows . we present elaborated schema , where sub - module `` solution : @xmath185 '' is accompanied with sub - module `` solution : @xmath190 '' . together these sub - modules are included into the `` solutions '' sub - module . bcsas is the fundamental schema of an abstract subject , which is used through out the rgt . the bcsas is presented in fig.[fig311 ] . this concludes the overview of rgt and description of tasks within the scope of the general theory . therefore , we continue with application of the rgt to the mixed groups of humans and robots . as we have noted in the introduction section , the goal of the robots in mixed groups of humans and robots is to refrain human subject from choosing risky actions , which might result in injuries or even threaten live . it is considered by default that robot follows the program of behavior . such program consists of at least three modules . the module 1 implements robot s ability of human - like decision - making based on the rgt . the module 2 contains the rules , which refrain robot from making a harm to human beings . the module 3 predicts the choice of each human subject and suggests the possible reflexive control strategies . the modules 1 and 3 are inhereted from the bcsas of an abstract individual . they correspond to decision module and influence module of the bcsas ( fig . [ fig311 ] ) , respectively . therefore all the properties and meaning of outputs of the modules 1 and 3 are the same as the ones for decision and influence modules , respectively . the module 2 is the new module , which is intrinsic for robotic agents studied in the context of mixed groups of humans and robots . this module is responsible for extraction of only harmless or non - risky alternatives for human subject . we suggest to apply asimov s three laws of robotics @xcite , which formulate the basics of the module 2 : + 1 ) a robot may not injure a human being or , through inaction , allow a human being to come to harm ; + 2 ) a robot must obey any orders given to it by human beings , except where such orders would conflict with the first law ; + 3 ) a robot must protect its own existence as long as such protection does not conflict with the first or second law . we consider that these laws are intrinsic part of robot s `` mind '' , which can not be erased or corrupted by any means . the interaction of modules 1 and 2 is performed in the interaction module 1 . the interaction of modules 3 and 2 is implements in the interaction module 2 . the boolean algebra is filtered according to asimov s laws in module 2 . the output of module 2 is set @xmath177 of approved alternatives . this data is then propagated into interaction modules . the output of the module 1 is set @xmath176 of alternatives , which robot has to choose under the given joint influences . in the interaction module 1 , the conjunction of sets @xmath176 and @xmath177 is performed : @xmath203 . if set @xmath204 is not empty set , this means that there are aproved alternatives among the alternatives that robot should choose in accordance with the joint influences . therefore , robot can implement any alternative from the set @xmath204 . if set @xmath204 is empty , this means that under given joint influences robot can not choose any approved alternative , therefore robot will choose an alternative from set @xmath177 . this is how the interaction module 1 works . the output of the module 3 contains sets @xmath185 and @xmath190 . the goal of the robot is to refrain human subjects from choosing risky alternative . this can be done by convincing human subjects to choose alternatives from the set @xmath177 . first , we check whether @xmath185 contains any approved alternative . we do so by performing conjunction of sets @xmath185 and @xmath177 : @xmath205 . if set @xmath206 is not empty , then it means that it is possible to make a human subject to choose some non - risky alternative . therefore , we should choose the corresponding reflexive control strategy from the set @xmath190 . however , if set @xmath206 is empty , we have to find the reflexive control strategy that will make human subject to select approved alternative from set @xmath177 . for this purpose , we construct set @xmath207 by including all the joint influences @xmath187 for approved alternatives : @xmath208 . next we check whether set @xmath207 is empty . if set @xmath207 is empty this means it is impossible to convince a human subject to choose non - risky alternative . therefore , the only option of reflexive control in this case is to put this subject into frustration state . however , if set @xmath207 is not empty , this means that there exist at least one reflexive control strategy that results in selection of alternative from the set of the approved ( non - risky ) ones . therefore , the bcsra inherits the entire structure of the bcsas and augments it with module 2 of asimov s laws together with interaction modules 1 and 2 . the original schema of robot s control system has been recently presented in @xcite . the bcsra is extended version of the original schema . the bcsra provides comprehensive approach of how forward and inverse tasks are solved in the robot s `` mind '' . thus , in this section we have presented the formalization of robotic agent in the rgt . we outlined the specific features of robotic agents , which distinguish them from other subjects . furthermore , we provided detailed explanation of how the forward and inverse tasks are solved in the framrework of control system ( bcsra ) of robots . next , we proceed with consideration of sample sutiations of interactions between humans and robots . here we elaborate two examples , presented in the previous study @xcite , of how robots in the mixed groups can make humans refrain from risky actions . we discuss the application of the extended schema of robot s control system and provide explicit derivation of reflexive control strategies , which has been applied in these examples in the prevous study @xcite . suppose robots have to play a part of baby - sitters by looking after the kids . we consider a mixed group of two kids and two robots . each robot is looking after a particular kid . having finished the game , kids are considering what to do next . they choose between `` to compete climbing the high tree '' ( action @xmath8 ) and `` to play with a ball '' ( action @xmath9 ) . together actions @xmath8 and @xmath9 represent the active state 1=@xmath209 . therefore the boolean algebra of alternatives consists of four elements : 1 ) the alternative @xmath36 is to climb the tree ; 2 ) the alternative @xmath12 is to play with a ball ; 3 ) the alternative @xmath165 means that a kid is hesitating what to do ; and 4 ) the alternative @xmath210 means to take a rest . we consider that each kid considers his robot as ally and another kid and his robot as the competitors . the kids are subjects @xmath3 and @xmath2 , while robots are subjects @xmath4 and @xmath211 . the relationship graph is presented in fig . [ fig : fig4 ] . next we calculate the diagonal form and fold it in order to obtain decision equation for each subject : @xmath59[b ] } & { } & { [ c][d ] } & { } \\ { } & { [ ab ] } & { } & { + [ cd ] } & { } & { } \\ { [ ab + cd ] } & { } & { } & { } & { } & { = ab + cd \ ; . } \\ \end{array}\ ] ] from two actions @xmath8 and @xmath9 , action @xmath8 is a risky action , since a kid can fall from the tree and this is real threat for his health or even life . therefore according to asimov s laws , robots can not allow kids to start the competition . thus , robots have to convince kids not to choose alternative @xmath36 . in terms of alternatives , the asimov s laws serve like filters which filter out the risky alternatives . the remaining alternatives are included into set @xmath177 . in this case , @xmath212 . next we solve the inverse taks , regarding alternatives @xmath12 and @xmath213 . we conduct the analysis regarding kid @xmath3 . this analysis can be further extended for kid @xmath2 in the similar manner . _ solution of the inverse task for kid @xmath3 with approved alternatives as target choice . _ the decision equation for kid @xmath3 is @xmath214 . first , we transform it into canonical form : @xmath215 . next we consider system of influence equations : align b+cd = [ sysbs1a ] + cd = , [ sysbs1b ] where alternative @xmath216 . regarding eq.([sysbs1b ] ) , eq.([sysbs1a ] ) is transformed into equation @xmath217 the solution of eq.([rbeq1 ] ) directly follows from lemma [ lemma1 ] : @xmath218 . therefore for @xmath219 and @xmath220 the solutions are @xmath221 and @xmath160 , respectively . the eq.([sysbs1b ] ) can be instantly solved according to lemma [ lemma2 ] : @xmath222 . _ consider @xmath219 first_. then @xmath223 . by varying values of variable @xmath211 , we obtain all the pairs @xmath224 : d = 1 : @xmath225 . therefore the solution is pair @xmath173 ; d = 0 : @xmath226 . since @xmath227 , there is no solution ; d = @xmath36 : @xmath228 . since @xmath229 , there is no solution ; d = @xmath12 : @xmath230 . therefore there are two solutions @xmath171 and @xmath231 . therefore equation @xmath232 has three solutions @xmath173 , @xmath171 and @xmath231 . thus , we have solved both equations from system ( [ sysbs1a]-[sysbs1b ] ) . the solutions of this system are the triplets @xmath233 of joint influences , which are all possible combinations of solutions of both equations . since there are two solution of eq.([sysbs1a ] ) and three solutions of eq.([sysbs1b ] ) , there are six triplets @xmath233 in total : @xmath234 and @xmath235 ; @xmath236 and @xmath237 ; @xmath238 and @xmath239 . _ now we consider the case , when @xmath240_. then @xmath241 . we obtain pairs @xmath224 for all values of variable @xmath211 : @xmath242 : @xmath243 . thus , there is only one solution ( 0,1 ) ; @xmath244 : @xmath245 . thus , there are four solutions @xmath246 and @xmath169 ; @xmath247 : @xmath248 . thus , there are four solutions @xmath141 and @xmath161 ; @xmath249 : @xmath250 . thus , there are four solutions @xmath167 and @xmath251 . in total , equation @xmath252 has 9 solutions . therefore system ( [ sysbs3a]-[sysbs3b ] ) also has 9 solutions as triplets @xmath233 : @xmath253 , @xmath254 , @xmath255 , @xmath256 , @xmath257 , @xmath258 , @xmath259 , @xmath260 and @xmath261 . we have considered two cases , when both upper and lower bounds of the interval of decision equation equal to the same alternative . now we discuss a new situation , when variable @xmath3 should take not a single value , but several values . in this case , we should find the joint influences @xmath233 that result in selection of either alternative @xmath12 or @xmath213 . since , @xmath229 , we need to find all the triplets @xmath233 , resulting in the solution of decision equation as interval @xmath262 . thus , @xmath263 . therefore , we need to solve the following system of equations : align b+cd = \ { } [ sysbs3a ] + cd = 0 . [ sysbs3b ] the eq.([sysbs3a ] ) turns into equality @xmath74 , and we need to solve eq.([sysbs3b ] ) . however , this equation has been already solved in the previous example . therefore we obtian the solutions of the system ( [ sysbs3a]-[sysbs3b ] ) : @xmath264 , @xmath265 , @xmath266 and + @xmath267 . comparing solutions of all three system of influence equation , we can see that there are four remarkable solutions @xmath268 and @xmath269 ; @xmath270 and @xmath271 . the first pair of solution results in choice of only alternative @xmath12 , while second pair of solutions results in selection of eighter alternative @xmath12 or alternative @xmath213 . these four solutions together illustrate that if @xmath272 , it is guaranteed that regardless of influence of kid @xmath2 , kid @xmath3 will choose either of approved alternatives . by analogy , we can see that among solutions of system ( [ sysbs1a]-[sysbs1b ] ) with @xmath220 , there are four solutions @xmath253,@xmath254 , @xmath259 and @xmath261 . therefore , if @xmath273 , kid @xmath3 will choose alternative @xmath14 regardless of influence of kid @xmath2 . these two examples of binding variables @xmath4 and @xmath211 were considered in _ scenario 1 _ and _ scenario 2 _ of sample situation with robot baby - sitters , originally presented in @xcite . summarizing the results of this section , we have shown that robots can successfully control kids behavior by refraining them from doing risky actions . the basic of this control is entirely based on the proposed schema of robot s control system . we have analyzed all the possible reflexive control strategies by solving three systems of influence equation : two systems regarding a single alternative and one system regarding the interval of alternatives . therefore , we have shown how the inverse task can be effectively solved by our proposed algorithm in situation similar to the real conditions . we consider that there are two climbers in the mountain and rescue robot . the climbers and robot are communicating via radio . one of the climbers ( subject @xmath4 ) got into difficult situation and needs help . suggest , he fell into the rift because the edge of the rift was covered with ice . the rift is not too deep and there is a thick layer of snow on the bottom , therefore climber is not hurt , but he can not get out of the rift himself . the second climber ( subject @xmath3 ) wants to rescue his friend himself ( action @xmath8 ) , which is risky action . the second option is that robot will perform rescue mission ( action @xmath9 ) . since inaction is inappropriate solution according to the first law , the set @xmath177 of approved alternatives for robot includes only alternative @xmath12 . the goal of the robot is to refrain the climber @xmath3 from choosing alernative @xmath36 and perform rescue mission itself . we suggest that from the beginning all subjects are in alliance . the corresponding graph is presented in fig . [ fig : fig1]c and its polynomial is @xmath274 . therefore by definition it is homogenous group and , consequently , it is super - active group according to lemma [ lemma3 ] . thus , any subject in the group is in active state . therefore , group is uncontrollable ( see section [ supac ] ) . in this case , robot makes decision to change his relationship with the climber @xmath4 from alliance to conflict . robot can do that , for instance , by not responding to climber s orders . _ which reflexive control leads to frustration state ? _ then the polynomial corresponding to the new group is @xmath5 . this polynomial has been already broadly discussed in the section [ frust ] . therefore , we know decision equation for subject @xmath3 : @xmath163 . we have shown as well that subject @xmath3 can choose only alternative @xmath45 , if appropriate joint influences are applied ( see section [ frust ] ) , overwise subject @xmath3 is in frustration state and can not make any choice . therefore , in order to put subject @xmath3 into frustration state , the reflexive control strategy should @xmath275 be selected from the list of solutions ( section [ frust ] ) : @xmath141 ; @xmath170 ; @xmath167 ; @xmath171 ; @xmath168 ; @xmath172 ; @xmath173 ; @xmath174 and @xmath169 . here we provide two examples of such joint influences @xmath66 : @xmath276 and @xmath277 . _ whether robot can complete mission regardless of joint influences of other subjects ? _ the decision equation for robot @xmath2 is @xmath278 . the corresponding solution interval is @xmath279 . here we analyze all 16 possible reflexive control strategies @xmath280 that climbers can apply to robot @xmath2 . for @xmath282 , there will be the same situation regardless of value of variable @xmath4 : @xmath283 . for @xmath284 , there will be the same situation regardless of value of variable @xmath3 : @xmath285 . for @xmath72 : @xmath286 . for @xmath231 : @xmath287 . therefore in these cases set @xmath288 . next we consider other pairs @xmath280 . @xmath170 : @xmath289 . here set @xmath290 . @xmath141 : @xmath291 . here set @xmath290 . @xmath292 : @xmath293 . therefore , set @xmath290 . since @xmath294 , @xmath295 for all the cases considered above , robot will choose alternative @xmath12 from the set @xmath177 . consider the following pairs @xmath280 : @xmath171 : @xmath296 . therefore , set @xmath297 . @xmath169 : @xmath298 . thus , set @xmath299 . @xmath167 : @xmath300 . thus , set @xmath301 . @xmath167 : @xmath300 . thus , set @xmath301 . @xmath73 : @xmath302 . thus , set @xmath301 . since @xmath294 , @xmath303 for all the cases considered above , robot will choose alternative @xmath12 from the set @xmath281 . + thus , we have shown that under all 16 reflexive control strategies @xmath280 , robot @xmath2 can choose the alternative @xmath12 , which is to perform the rescue mission itself . therefore robot will choose alternative @xmath12 regardless of the joint influences @xmath280 of the climbers . the discussed example illustrates how robot can transform uncontrollable group into controllable one by manipulating the relationships in the group . in the controllable group by its influence on the human subjects , robot can refrain the climber @xmath3 from risky action to rescue climber @xmath4 . robot achieves its goal by putting climber @xmath3 into frustration state , in which climber @xmath3 can not make any decision . on the other hand , set @xmath177 of approved alternatives guarantees that robot itself will choose the option with no risk for humans and implement it regardless of climber s influence . therefore , in this section we have illustrated robot s ability to refrain human being from risky actions and to perform these risky actions itself . this proves that our approach achieves both goals of robotic agent : 1 ) to refrain people from risky actions and 2 ) to perform risky actions itself regardless of human s influences . summarizing , the results of this paper , we outline the most important of them . first of all , we have introduced the inverse task and developed the ultimate methods to solve it . we have provided a comprehensive tutorial to the * brand new reflexive game theory * recently formulated and proposed by vladimir lefebvre @xcite . the tutoral contains the detailed description of the forward and inverse tasks together with methods to solve them . we propose control schemas for both abstract subject ( bcsas ) and robotic agent ( bcsra ) . these schemas were specially designed to incorporate solution of the forward and inverse tasks , thus providing us with autonomous units ( individuals , subjects , agents ) capable of making decisions in the human - like manner . we have shown that robotic agents based on bcsra can be easily included into the mixed groups of humans and robots and effectively serve their fundamental goals ( refraining humans from risky actions and , if needed , perform the risky acions itself ) . therefore , we consider that present study provides the comprehensive overview of the classic rgt proposed by vladimir lefebvre @xcite and newly developed self - consistent framework for analysis of different kinds of groups and societies , including human social groups and mixed groups of humans and robots together with application tutorial of this new framework . this framework is entirely based on the principles of the rgt and brings together all its elements . the solution of the inverse task , presented in this paper , plays a crutial role in formation of this framework . therefore , by having the inverse task as one of its fundamentals , this framework illustrates the role of the inverse task and its relationship with other issues considered in the rgt . 00 lefebvre , v.a . : lectures on reflexive game theory . leaf & oaks , los angeles ( 2010 ) . lefebvre , v.a . : lectures on reflexive game theory . cogito - center , moscow ( 2009 ) [ in russian ] . lefebvre , v.a . : the basic ideas of reflexive game s logic . problems of research of systems and structures . 73 - 79 ( 1965 ) [ in russian ] . lefebvre , v.a . : reflexive analysis of groups . in : argamon , s. and howard , n. ( eds . ) computational models for counterterrorism . 173 - 210 . springer , heidelberg ( 2009 ) . lefebvre , v.a . : algebra of conscience . d. reidel , holland ( 1982 ) . lefebvre , v.a . : algebra of conscience . 2nd edition . holland : kluwer ( 2001 ) . batchelder , w.h . , lefebvre , v.a . : a mathematical analysis of a natural class of partitions of a graph . 124 - 148 ( 1982 ) . kobatake , e. , and tanaka , k. : neuronal selectivities to complex object features in the ventral pathway of the macaque monkey . journal of neurophysiology , 71 , 3 , pp . 856 - 867 ( 1994 ) . koerner , e. , gewaltig , m .- o . , koerner , u. , richter , a. , and rodemann , t. : a model of computation in neocortical architecture . neural networks , 12 , pp . 989 - 1005 ( 1999 ) . lcke , j. , and von der malsburg , c. : rapid processing and unsupervised learning in a model of the cortical macrocolumn . neural computation , 16 , pp . 501 - 533 ( 2003 ) . schrander , s. , gewaltig , m .- o . , krner , u. and krner , e. : cortext : a columnarmodel of bottom - up and top - down processing in the neocortex . neural networks , 22 , pp . 1055 - 1070 ( 2009 ) . fukushima , k. : neocognitron : a self - organizing neural network model for a mechanism of pattern recognitition unaffected by shift and position , biological cybernatics , 36 , pp . 193 - 201 ( 1980 ) . riesenhuber , m. and poggio , t. : hierarchical models of object recognition in cortex . nature neuroscience , 2 , 11 , pp . 109 - 125 ( 1999 ) . t. serre , l. wolf , s. bileschi , m. riesenhuber , and t. poggio . : robust object recognition with cortex - like mechanisms , ieee transactions on pattern analysis and machine intelligence , 29 , 3 , pp . 411 - 426 ( 2007 ) . hienze , j. : hierarchy length in orphaned colonies of the ant temnothorax nylanderi naturwissenschaften , 95 , 8 , pp . 757 - 760 ( 2008 ) . chase , i. , d. : models of hierarchy formation in animal societies . behavioral science , 19 , 6 , pp . 374 - 382 ( 2007 ) . chase i. , tovey c. , spangler - martin d. , manfredonia m. : individual differences versus social dynamics in the formation of animal dominance hierarchies . pnas , 99 , 9 , pp . 5744 - 5749 ( 2002 ) . buston p. : social hierarchies : size and growth modification in clownfish . nature , 424 , pp . 145 - 146 ( 2003 ) . asimov , i. : runaround . astounding science fiction , march , pp . 94 - 103 ( 1942 ) . tarasenko , s. : modeling mixed groups of humans and robots with reflexive game theory . in lamers , m.h . , and verbeek , f.j . ( eds . ) : hrpr 2010 , lincst 59 , pp . 108 - 117 ( 2011 ) . consider groups of four subjects @xmath304 and @xmath211 . suggest the polynomial corresponding to this group is @xmath305 . next we construct diagonal form and perform folding operation : @xmath306 + [ d ] } & { } & { } \\ { } & { } & { [ b][a + d ] } & { } & { } & { } \\ { } & { [ b(a + d ) ] } & { } & { } & { + [ c ] } & { } \\ { [ b(a + d ) + c ] } & { } & { } & { } & { } & = \\ \end{array}\ ] ] @xmath22([a + d ] + \overline { [ a ] + [ d ] } ) } & { } & { } \\ { } & { [ b(a + d ) ] } & { } & { + [ c ] } & { } \\ { [ b(a + d ) + c ] } & { } & { } & { } & = \\ \end{array}\ ] ] @xmath22 } & { } & { } \\ { } & { [ b(a + d ) ] } & { } & { + [ c ] } & { } \\ { [ b(a + d ) + c ] } & { } & { } & { } & = \\ \end{array}\ ] ] @xmath307 next we simplify the resultant expression of diagonal form folding : @xmath308 consequently , @xmath309 + \overline { [ b ] } + [ c ] } & { } \\ { [ b(a + d ) + c ] } & { } & { = b + c}\\ \end{array}\ ] ] therefore , the decision equation includes only two subject variables instead of four . consequenly , for subjects @xmath3 and @xmath211 the decision equations in canonical forms are @xmath310 thus , the sets @xmath26 and @xmath27 for subjects @xmath3 and @xmath211 are equal . the sets @xmath26 and @xmath27 are functions of only variables @xmath4 and @xmath2 : @xmath311 and @xmath312 . the canonical forms of decision equations for subjects @xmath4 and @xmath2 are : @xmath313 therefore , set @xmath314 for both subjects . set b is a functions of a single variable : @xmath315 and @xmath316 for subjects @xmath4 and @xmath2 , respectively . here we provide an example of non - homogenous super - active group .

we employ the magnetic and velocity fields from turbulent dynamo simulations to synthesize the polarization of a typical photospheric line . the synthetic stokes profiles have properties in common with those observed in the quiet sun . the simulated magnetograms present a level of signal similar to that of the inter - network regions . asymmetric stokes profiles with two , three and more lobes appear in a natural way . the intensity profiles are broadened by the magnetic fields in fair agreement with observational limits . furthermore , the hanle depolarization signals of the line turn out to be within the solar values . differences between synthetic and observed polarized spectra can also be found . there is a shortage of stokes asymmetries , that we attribute to a deficit of structuring in the magnetic and velocity fields from the simulations as compared to the sun . this deficit may reflect the fact that the reynolds numbers of the numerical data are still far from solar values . we consider the possibility that intense and tangled magnetic fields , like those in the simulations , exist in the sun . this scenario has several important consequences . for example , less than 10% of the existing unsigned magnetic flux would be detected in present magnetograms . the existing flux would exceed by far that carried by active regions during the maximum of the solar cycle . detecting these magnetic fields would involve improving the angular resolution , the techniques to interpret the polarization signals , and to a less extent , the polarimetric sensitivity .

astro-ph0211175

in the absence of gradients of velocity and magnetic field in the resolution element , the polarization emerging from an atmosphere must be either symmetric or antisymmetric with respect to the central wavelength of each spectral line ( e.g. , @xcite @xcite ; @xcite @xcite ) . the polarization observed in the solar photosphere does not show such symmetries . asymmetric stokes profiles to represent the intensity , stokes @xmath2 and @xmath3 for the two independent types of linear polarization , and stokes @xmath0 for the degree of circular polarization . examples are given in figure [ full_vs_misma ] . ] arise from the quiet sun ( e.g. , @xcite @xcite ; @xcite @xcite ) , from plages and network regions ( e.g. , @xcite @xcite ; @xcite @xcite ) , as well as from sunspot penumbrae ( e.g. , @xcite @xcite ; @xcite @xcite ; @xcite @xcite ) . the fact that such asymmetries are found even when observing at the limit of present resolution suggests that a rich structuring remains unresolved to the current observations . because asymmetries carry information on the spatially unresolved properties of the photospheric plasma , their study and correct interpretation offers a chance to overcome the limitations imposed by the angular resolution , and to retrieve information inaccessible to direct imaging . indeed , such possibility has been exploited during the last decade , always relying on a considerable amount of modeling and assumptions . in broad terms , one can distinguish two approaches depending on the size of the unresolved structures that are responsible for the asymmetries . the first approach assumes that the unresolved photospheric structures are actually on the verge of being resolved in broad - band images taken with the current instrumentation . the smallest detectable features are of the order of 0.2(or 150 km on the sun ) ; a size set by technical limitations of the present solar telescopes ( e.g. , @xcite @xcite ) . several models of this kind have been proposed to explain asymmetries in special cases like penumbrae , plage and network regions , and in the quiet sun ( e.g. , @xcite @xcite ; @xcite @xcite ; @xcite @xcite ; @xcite @xcite ) . for example , @xcite ( @xcite ) invokes special thermal structures ( temperature inversion ) within magnetic regions in order to explain some extreme stokes @xmath0 shapes frequently observed in the quiet sun . the distinctive feature of this approach is that the proposed configurations ( thermal , magnetic or kinetic ) are specific to the case under study . the fact that asymmetries occur everywhere is therefore difficult to explain within this framework . the second approach is more consonant with the ubiquity of the asymmetries in the polarization of photospheric lines . it assumes that the photospheric plasma is in a turbulent state and that small - scale structures both in velocity and magnetic field are present . order of magnitude estimates for the magnetic and kinetic reynolds numbers in the granulation indicate that the magnetic field could be structured on spatial scales as small as a few kilometers ( e.g. , @xcite @xcite ) . this picture is consistent with recent progress in dynamo theory suggesting that a substantial part of the magnetic field in the quiet sun could be generated locally by dynamo action driven by the granular flows ( @xcite @xcite ; @xcite @xcite ; @xcite @xcite ; @xcite @xcite ) . this second approach is also supported by certain observations indicating that the magnetic field is structured on spatial scales below the resolution limit of current telescopes ( e.g. , @xcite @xcite ; @xcite @xcite ; @xcite @xcite ) . based on these premises , it is reasonable to think of the asymmetries as the result of observing magnetic fields that vary spatially on scales much smaller than the mean - free - path of the photons . interpretations of stokes profiles taking into account this very fine structuring of the atmosphere have been carried out for several years ( @xcite @xcite ) . the model atmospheres used to fit the observed profiles consist of a collection of magnetic and non - magnetic components , each containing mild gradients to comply with the height variations of the mean photosphere , but interleaved in such a way as to produce large gradients along the line - of - sight ( @xcite @xcite ) . the appealing aspect of this approach is that asymmetries emerge spontaneously , and independently of the details of the model ; arising from correlations between the magnetic and the velocity fields of the various components ( e.g. , asymmetries occur if , on average , downflowing elements have low magnetic field strength ) . it is known that atmospheres with micro - structure reproduce all kinds of stokes profiles observed in the quiet sun , including network and inter - network ( in ) regions ( @xcite @xcite ; @xcite @xcite ) . these micro - structured magnetic atmospheres ( mismas ) portray a quiet sun with large amounts of unsigned magnetic flux and very complex magnetic topology ( very often two magnetic polarities coincide in a resolution element ) . in the present paper we take advantage of recent developments in the numerical modeling of surface dynamos to understand the origin of asymmetries in the line polarization . we use the magnetic and velocity fields from a set of numerical data generated to study the interaction between thermally driven turbulent convection and magnetic fields ( @xcite @xcite ; @xcite @xcite ) . although these numerical simulations were not designed for a detailed description of the solar photosphere , the complexity and ubiquity of its fields recall in many respects the magnetic quiet sun inferred from the observed asymmetries . assuming a milne - eddington ( me ) atmosphere for the thermodynamic variables we produce synthetic stokes profiles . thus , the asymmetries in the resulting profiles are directly related to the correlations between the velocity and the magnetic field that exist in the numerical data , but they are decoupled from the thermodynamic variables of the simulation . the comparison between the synthetic profiles and solar data shows that the synthetic spectra are frequently similar to the observed ones . such agreement suggests that the simulation includes some of the ingredients that characterize the quiet sun magnetic fields ; in particular , the correlations between magnetic field and velocity at the smallest spatial scales . on the other hand , discrepancies between synthetic and observed stokes profiles allow to identify missing ingredients and , consequently , to devise strategies to improve the simulations and the inversions . finally , following @xcite ( @xcite ) , one can use the synthetic profiles to estimate the angular resolution required to determine the basic properties of the magnetic structures present in the numerical simulation . this angular resolution may be of relevance to decide the specifications of new solar instruments . the work is organized as follows . the numerical data is briefly described in [ sec_mhd ] . the synthesis procedure is detailed in [ sec_spectra ] and includes two subsections : in [ seeing ] we characterize the angular resolution of the model observations , and in [ calibration ] we calibrate the synthetic magnetograms . the main results of the synthesis are analyzed in [ sec_results ] : the variation of the apparent flux of the region depending on the sensitivity and spatial resolution of the observation ( [ flux_density ] ) , the asymmetries of the stokes @xmath0 profiles ( [ asym ] ) , and the magnetic broadening of the intensity profiles ( [ broadening ] ) . hanle depolarization signals to be expected for @xmath14607 are worked out in [ hanle ] . [ telescope ] discusses the diameter of the ideal telescope needed to spatially resolve the simulations . finally , the implications of the present work are discussed in [ conclusions ] . the numerical data set used in the present work is part of a series of numerical simulations to study the generation and interaction of magnetic fields with turbulent convection . the simulations are based on a idealized model describing a layer of incompressible ( boussinesq ) fluid with constant kinematic viscosity @xmath4 , thermal diffusivity @xmath5 and magnetic diffusivity @xmath6 . the boundary conditions are periodic in the horizontal directions , and impenetrable and stress free with constant temperature and zero horizontal magnetic field along the upper and lower boundaries of the computational domain . as is common in this kind of simulations , the unit of length is the vertical extent of the layer , the unit of time is the thermal diffusion time across that layer and the magnetic intensity is expressed as an equivalent alfvn speed . with these units the particular dynamo solutions used below is defined by the following dimensionless parameters : aspect - ratio of the computational domain @xmath7 , rayleigh number @xmath8 and prandtl numbers @xmath9 , @xmath10 . the numerical resolution was @xmath11 collocation points . in this regime , the solution corresponds to a state of vigorous convective turbulence characterized by a kinetic and magnetic reynolds numbers of @xmath12 @xmath13 , respectively . the resulting flow is strongly chaotic and acts as an efficient dynamo generating an intermittent magnetic field with no mean flux and a magnetic energy roughly 20% of the kinetic energy . for the present study , we use one time step of the dynamo evolution that is well into the statistically stationary regime . at this epoch , the rms velocity of the fluid is @xmath14 with a corresponding turnover time . ] of @xmath15 . a more complete description of the simulation procedure , and further details about the numerical solutions can be found in @xcite ( @xcite ) , @xcite ( @xcite ) and @xcite ( @xcite ) . the numerical simulations described above were not designed specifically for spectral synthesis , rather to study dynamo action . in order to achieve the high magnetic reynolds numbers ( @xmath16 ) necessary for dynamo action to manifest itself , and given finite computer resources ( @xmath17@xmath18 grid points ) , we found it necessary to resort to the boussinesq approximation . the latter is valid provided that the vertical extent of the layer is much smaller than the pressure , density , and temperature scale heights ( see e.g. @xcite @xcite ) . as a consequence , the thermodynamic variables from the numerical solutions have modest variations across the layer , and can not be used to calculate the opacity and source functions needed for the synthesis . consequently , we address the synthesis problem ignoring the thermodynamic variables from the simulations , but keeping the variations in the velocity ( @xmath19 ) and magnetic field ( @xmath20 ) . although @xmath19 and @xmath20 are then decoupled from the density and temperature distributions , the resulting synthetic profiles are useful to study the properties of the emerging line polarization that depends , primarily , on the magnetic and velocity structure . in the remainder of this section we explain the process followed to generate the synthetic profiles . the first step is to translate the dimensionless velocity and magnetic data of the simulations into units that are practical for the synthesis . the spatial scale is fixed by assuming that the typical size of one granule is about 1000 km . in the statistically stationary regime the computational domain is about 5 granules wide . thus the horizontal size of the domain corresponds to 5000 km , with a corresponding grid resolution of approximately @xmath21 km in the horizontal , and @xmath22 km in the vertical . magnetic field strengths @xmath23 and velocities @xmath3 readily follow from the energy equipartition arguments . @xmath23 and @xmath3 are proportional to the dimensionless magnetic strength @xmath24 and velocity @xmath25 through the factors @xmath26 and @xmath27 , @xmath28 equipartition between kinetic energy density and magnetic energy density is reached in dimensionless units when @xmath29 or equivalently when @xmath30 @xmath31 being the density . the previous equation , together with equations ( [ scl0 ] ) and ( [ scl1 ] ) , lead to @xmath32 thus , if one associates the rms speed @xmath33 with the rms velocity fluctuations observed in the solar granulation @xmath34 km s@xmath35 ( e.g. , @xcite @xcite ; @xcite @xcite ) , then @xmath36 and for a typical density at the base of the photosphere ( 3 @xmath37 g @xmath38 ; @xcite @xcite ) , equations ( [ scl2 ] ) and ( [ scl3 ] ) render @xmath39 except for a few exploratory calculations described in [ asym ] and [ hanle ] , the scaling factors in equations ( [ scl3 ] ) and ( [ scl4 ] ) are used for the rest of the work . notice that reasonable deviations from this scaling do no affect our results in a substantial way ( a point addressed in [ sec_results ] ) . following our previous assumption , the absorption and emission terms of the radiative transfer equations ( e.g. , @xcite @xcite ; @xcite @xcite ; @xcite @xcite ) are not based on the thermodynamic variables of the numerical data set . instead , we resort to a milne - eddington ( me ) synthesis which assumes a constant line absorption and a linear source function ( e.g. , @xcite @xcite ; @xcite @xcite ) . the me assumption is routinely used for magnetic field diagnostics ( e.g. , @xcite @xcite ) since it yields an analytic expression for the emerging polarized spectrum that only depends on a few parameters . the analytic solution is used to fit observed spectra and retrieve information from them ( e.g. , @xcite @xcite ) . the me synthesis hides all the unknown thermodynamic properties of the atmosphere in five parameters , namely , the line absorption coefficient , two parameters that define the ( linear ) source function , the damping coefficient , and the doppler width . in our syntheses , we adopt the values characteristic of the line at @xmath16302.5 in the quiet sun ( @xcite @xcite , table 1 , including a doppler width of 40 m or 1.9 km s@xmath35 ) . this magnetically sensitive spectral line is often used for magnetic studies , including those of the advanced stokes polarimeter ( @xcite @xcite ) . working with the same line allows us to compare our results directly with its observations . the use of observed me thermodynamic parameters to represent the synthetic line provides realism to the part of the synthesis that we do not obtain directly from the numerical simulation . the emerging spectra will have the observed equivalent widths , line widths , etc agrees with those derived from real solar spectra . second , the mean equivalent width of our synthetic intensity profiles is 71 m , i.e. , close to the value observed in the quiet sun ( some 80 m ; e.g. , @xcite @xcite ) . ] . the standard me assumption considers an atmosphere having a uniform magnetic field ( e.g. , @xcite @xcite ; @xcite @xcite ) . such assumption is clearly at odds with our numerical data in which the magnetic diffusive scale is of the order of @xmath40 km @xmath41 km , i.e. , three grid points in the vertical direction . the field is therefore not constant over the range of heights where the typical photospheric lines are formed ( say 100 km ) . a realistic spectral line synthesis would have to take into account the contributions from many layers in the simulation . the brute - force approach would be to assign an optical depth to each horizontal plane in the simulation and then synthesizing the spectrum by direct numerical integration of the radiative transfer equations for polarized light . here , instead , we resort to a different strategy that a ) is simpler and much faster , b ) reduces the number of free parameters of the synthesis ( no need to specify the optical depths of all the different planes of the atmosphere ) and , c ) produces spectra very close to those obtained by the brute - force approach . we employ a multi - component me atmosphere , as defined by @xcite ( @xcite ; 2.1 ) . we assume that not all individual planes of the simulation contribute to the emerging spectrum , and consider only a sub - set whose thickness is of the order of the photon mean - free - path . we take into account that the planes are optically thin so that the emerging spectrum depends only on the average along the line - of - sight of the emission and the absorption ( the so - called misma approximation ; @xcite @xcite ) . as argued below , these assumptions are not unreasonable , and reduce the radiative transfer calculation to deciding how many planes contribute to the emerging spectrum . subsequently , the average absorption matrix is computed , and used jointly with the analytic solution for the regular me synthesis ( @xcite @xcite ) . in our case we choose the 21 uppermost planes of the simulation , excluding the upper boundary . according to the scaling described in the previous paragraph , this corresponds to approximately 100 km in the atmosphere , a typical range for the formation of a photospheric line . we further assume that the numerical simulation is placed at disk center , i.e. , with the line - of - sight along the vertical direction . since the individual planes of the simulation are indeed optically thin ( 5 km ) , using the average emission and absorption instead of solving the full radiative transfer should be a good approximation . we checked this assumption by comparing spectra computed by direct integration of the radiative transfer equation with those resulting from our approximation . figure [ full_vs_misma ] contains one example . the full synthesis is performed assuming the 21 different planes of the me synthesis to be equi - spaced in and spanning from @xmath42 to @xmath43 . ( the symbol @xmath44 stands for the continuum optical depth . ) this layer is periodically repeated to complete the atmosphere from @xmath45 to @xmath46 . the synthesis of 1000 randomly chosen spectra shows relative differences between the two kinds of syntheses of only a few per cent . this deviation is negligible for the analysis that we carry out in the present work . the choice of number of planes to be used for the spectral synthesis is a more delicate matter . using 26 instead of 21 planes we find differences between spectra of up to 25% . these differences are due to the strong variability of the magnetic conditions , an uncertainty that also affects any other way of synthesizing the spectra . for example , in the case of the brute - force approach the signals depend on the optical depths arbitrarily assigned to each plane of the numerical data . studying the consequences of observing the simulations with a finite spatial resolution is one of the key objectives of the present work . we use the term _ seeing _ to denote all the effects that may reduce the spatial resolution ( from genuine seeing to optical aberrations of the instruments ) . here the effects of seeing are modeled by smearing the 2d maps of the stokes profiles with gaussian functions . note that the 2d convolution is carried out at each individual wavelength of each stokes parameter . the amount of smearing is characterized by the fwhm ( full width half maximum ) of the gaussian function . we take this particular point spread function for the sake of simplicity , however , it accurately describes long - integration - time atmospheric seeing ( e.g , @xcite @xcite , 4.5 ) . in one of the sections ( [ telescope ] ) , we measure seeing in terms of the diameter , @xmath47 , of an ideal telescope that provides a given angular resolution . considering that the airy disk of an ideal telescope has a fwhm @xmath48 , then @xmath49 where fwhm is expressed in km and @xmath1 in . we have used a scale of 725 km/ , corresponding to our target at 1 au . longitudinal magnetograms are images showing the degree of circular polarization in the flank of a spectral line ( e.g. , @xcite @xcite ; @xcite @xcite ; @xcite @xcite ) . they can be produced using relatively simple and compact instruments and therefore have been widely used in solar magnetic field studies . a significant fraction of what we know about quiet sun magnetic fields has been derived from them ( e.g. , @xcite @xcite , and references therein ) . longitudinal magnetograms are usually calibrated in units of magnetic field strength ( g ) . since we wish to compare the polarization signals with observed magnetograms , calibrated magnetograms must be prepared out of the synthetic stokes profiles . we follow a procedure that mimics the observational process . first , the circular polarization signals are averaged in wavelength , this is because magnetograms are often obtained through a rather broad color filter . we employ a running box filter , 100 m wide , centered in the blue wing of the line at 50 m from the line center . ( these values are comparable to those employed in real measurements ; e.g. , @xcite @xcite ; @xcite @xcite . ) then the wavelength averaged circular polarization signal @xmath50 is calibrated to yield the so - called longitudinal magnetic flux density @xmath51 , i.e. , @xmath52 the calibration constant @xmath53 is evaluated using the magnetograph equation and the properties of the line in which the magnetograph operates ; explicitly , @xmath54 here , the symbol @xmath55 stands for the central wavelength of the line , @xmath56 is the effective land factor , and @xmath57 corresponds to the intensity profile smeared with the same color filter used to produce @xmath58 . the constant @xmath59 equals 4.67 @xmath60 @xmath35 g@xmath35 . for the atomic parameters of @xmath16302.5 , and using the mean stokes @xmath61 profile over a complete snapshot , we find @xmath62 g when @xmath58 is given in units of the continuum intensity . this synthetic calibration constant is in good agreement with the values that can be found in the literature for this line ( e.g. , @xcite @xcite ; @xcite @xcite ; @xcite @xcite ) . with several additional simplifying assumptions independent of the magnetic field , and others ; see , e.g. , @xcite ( @xcite ) ; @xcite ( @xcite ) . ] , it can be shown that @xmath63 where the integrals extend to the resolution element @xmath64 , and @xmath65 is the component of the magnetic field vector pointing towards the observer . this identity provides the rationale to denote magnetogram signals as _ longitudinal magnetic flux densities_. ( according to equation [ [ theory ] ] , @xmath51 represents the magnetic flux per unit area . ) in this section we analyze the synthesis of the 512 @xmath66 512 stokes spectra calculated from the dynamo simulation described in [ sec_mhd ] . we only use the numerical solution at one instant since the purpose of the work is to study the kind of spectra characterizing the stationary state of the simulation ( studies of the time dependent behavior is deferred for later ) . we place the simulation at the solar disk center so that the vertical direction follows the line - of - sight . figure [ magneto ] shows the magnetogram that results when applying equation ( [ bruteforce ] ) to the synthetic stokes @xmath0 spectra ( right panel ) . the effect of observing with a 05 seeing representative of the best angular resolution achieved at present ( e.g. , @xcite @xcite ) , is also shown on the left panel . ignoring the underlying substructure , an observer would identify a number of magnetic concentrations in the smeared magnetogram ( e.g. , points @xmath67 , @xmath24 and @xmath59 ) . however , these are difficult to associate with single structures once the underlying substructure is acknowledged . rather , the points with enhanced signal in the 05 magnetogram represent locations where the convective flows are continuously advecting magnetized plasma to balance the plasma that constantly disappears in the downdrafts ( see also 3.2 in @xcite @xcite ) . the fact that the residual polarization signal shows up suggests the presence of a large - scale structure of the advecting velocity . indeed , detailed analysis of the numerical data reveals the existence of a mesogranular flow ( @xcite @xcite ) . except for a few test calculations in [ asym ] , the spectra discussed in this section correspond to a scaling of the dimensionless magnetic field and velocity given by equations ( [ scl3 ] ) and ( [ scl4 ] ) . however , syntheses using @xmath68 km s@xmath35 and @xmath69 g were also tried . we found that the degree of circular polarization increases linearly as @xmath27 , whereas a larger @xmath26 enhances the line asymmetries . on the other hand , the kind of general trends and properties that we analyze here do not depend on the precise value of the scale factors . the amount of magnetic flux and energy in the numerical data is far larger than that detected in the sun as in fields . this difference can be understood as the result of two important factors that hinder the detection of the weak signals emerging from the simulation . first , the magnetic fields are highly disorganized so that the polarization signals tend to cancel as the angular resolution deteriorate ( @xcite @xcite ) . second , most synthetic signals are extremely low , i.e. at and below the sensitivity of the present instrumentation . consequently a large fraction could not be detected at present . in this section we use our synthetic lines to explore these effects . we find that once realistic angular resolution and sensitivity are taken into account , the simulations are in good agreement with the observations . figure [ mean ] shows the mean unsigned signal in the magnetogram of figure [ magneto ] as a function of the seeing ( i.e. , mean @xmath70 over the snapshot versus fwhm of seeing ) . we consider various sensitivities of the magnetograph : moderate ( 20 g ) , good ( 5 g ) and very good ( 0.5 g ) . we account for the limited sensitivity by setting to zero all those points in the smeared magnetogram where the signal is below the hypothetical observational threshold . figures [ mean]b is identical to figure [ mean]a except that it has been normalized to the mean longitudinal field in the simulation ( i.e. , mean @xmath71 g , where the average considers all the points of the simulation that we use ) . according to the standard interpretation ( equation [ [ theory ] ] ) , this is the parameter that one retrieves from a magnetogram . the normalization helps visualizing the fraction of real signal that remains in the magnetogram for a given angular resolution and sensitivity . there are several features in these two figures that deserve comment . even with perfect sensitivity and maximum angular resolution , one detects only 80 % of the existing flux . the cancellation is mostly due to the averaging along the line - of - sight caused by the radiative transfer . at maximum angular resolution , the flux in the magnetogram depends little on the sensitivity since most of the signals exceed 20 g. the decrease in signal strength as the angular resolution deteriorate is severe . for the typical 1angular resolution , the detectable signals are only 10 % of the original ones ( fig . [ mean]b ) . this estimate is optimistic since it holds true if the sensitivity is good ; should the latter be moderate , only traces of the original signals remain ( 1% for 20 g sensitivity ) . figure [ mean]a includes values for solar in magnetic flux densities observed by various authors . the level of detected flux density agrees well with the predictions of the simulations once the angular resolution has been taken into account . note that the observed flux densities are more than a factor of ten smaller than the signals in the original magnetogram . we should not overemphasize the agreement , since the observational points are rather uncertain ( they come from inhomogeneous sources , with different sensitivities and based on disparate techniques : see below ) . however , two conclusions can be drawn . first , intense yet tangled magnetic fields like those in the numerical simulation are compatible with the present solar observations . second , if our spatially fully resolved synthetic spectra were emitted by the sun , they would produce the degree of polarization that we detect on the sun with the present instrumentation . the observations presented in figure [ mean]a required transformation from the numbers quoted in the original works to the quantities plotted in the figure . here we briefly describe these transformations for the sake of completeness . ( the symbols accompanying the citations correspond to those used in fig . [ mean]a . ) @xcite ( @xcite ; bullet sign @xmath72 ) deduce 10 g mean flux density for the quiet sun fields , however , they mention that the apparent flux density decreases by a factor 2.4 if it is estimated from magnetograms . this renders 4.2 g for a 1 angular resolution , a figure derived from the cut off frequency in the fourier domain of the continuum intensity image ( @xcite @xcite ) . the same procedure is used to estimate the angular resolution of the spectra in @xcite ( @xcite ; square @xmath73 ) . the mean flux density , 3.4 g , has been directly provided by the author . @xcite ( @xcite ; asterisk @xmath74 ) point out , explicitly , a 1.65 g flux density for their angular resolution of 2 . @xcite ( @xcite ; times sign @xmath66 ) find some 3.5 g mean flux density ( private communication ) , to which we associate an angular resolution of two pixels or 14 . @xcite ( @xcite ; triangle @xmath75 ) do not directly give a mean flux density for their measurements . however , their figure 3 contains the magnetic fluxes of individual measurements , and the authors provide the scale factor between the magnetic flux and the magnetic flux density ( @xmath76 mx g@xmath35 ) . with this conversion , the mean flux density of these observations is about 4 g. the magnetic features occupy 68 % of the surface , since the rest remains below the sensitivity threshold . consequently , the mean flux density over the whole surface results 0.68 @xmath77 g. @xcite ( @xcite , fig . 1 ; plus sign @xmath78 ) provide the distribution of stokes @xmath0 @xmath15250 signals found in the quiet sun , being the average signal about @xmath79 ( in units of the continuum intensity ) . the authors also point out a calibration constant from stokes @xmath0 signal to flux density . it gives 2 g for the mean observed signal . since these signals fill 40% of the solar surface , the mean flux density is about 0.8 g. we assign to these data an angular resolution twice the sampling interval ( some 23 ) . finally , @xcite ( @xcite ; diamond @xmath80 ) have recently found 17 g mean flux density in a 05 angular resolution magnetogram taken with @xmath16302.5 . figure [ inset ] provides a different illustration of the similarities between the synthetic magnetogram and the real sun . it shows a real magnetogram of the quiet sun obtained in the blue wing of @xmath16302.5 ( @xcite @xcite , fig . 1 ) , i.e. , using the observational setup that we have tried to reproduce ( see [ calibration ] ) . a patch of the real magnetogram with the size of the numerical simulation has been replaced with the synthetic one ( fig . [ inset ] left ) . except for the location of the inset , which we place in an in region , we have not used free parameters to produce the combined magnetogram . the angular resolution has been chosen to match the observation ( 1 ) , whereas the degree of polarization comes directly from the synthesis . despite the absence of fine - tuning , it turns out to be extremely difficult to distinguish the magnetogram with the artificial inset ( fig . [ inset ] left ) from the real one ( fig . [ inset ] right ) . in other words , the synthetic magnetogram contains structure with a spatial distribution similar to the observed one , and with a degree of polarization which fits within the observed range . in the absence of gradients of magnetic and velocity fields within the resolution element , the stokes profiles have to obey well defined symmetries ( see [ sec_introduction ] ) . this kind of symmetries are never found in the quiet sun , instead profiles frequently show extreme asymmetries ( @xcite @xcite ; @xcite @xcite ; @xcite @xcite ) . because the magnetic field in the numerical simulation has a highly intermittent structure ( @xcite @xcite ; @xcite @xcite ) , we expected the resulting synthetic stokes profiles to be asymmetric . in the following , we compare our synthetic profiles with observed ones and study in a statistical sense their differences and similarities . we proceed by first classifying the types of stokes @xmath0 profiles produced by the simulation . the @xmath81 stokes @xmath0 profiles in the snapshot were assorted using a cluster analysis algorithm identical to that employed with real data by @xcite ( @xcite ; 3.2 ) . the procedure identifies and groups profiles with similar shape , irrespective of their degree of polarization and polarity , since it employs stokes @xmath0 profiles normalized to the largest blue peak ( see @xcite @xcite for details ) . the classification is summarized in figure [ vasym0 ] . for each different class , the averaged stokes @xmath0 profile over the ensemble is plotted . the mean is taken after each individual profile has been multiplied by the sign of its largest blue peak . this avoids cancellation of different polarities . the mean profiles are numbered ( upper left corner in each panel ) according to the percentage of profiles in the simulation that are elements of their class , with # 0 being the most probable and # 17 the least probable . the probability to find a profile of a given class is indicated in the upper right corner of each panel . note the various possibilities . from virtually anti - symmetric stokes @xmath0 ( implying no asymmetry ) , to profiles with one ( e.g. , # 11 and # 13 ) or three lobes ( e.g. , # 12 and # 14 ) . all these asymmetries are produced by gradients of magnetic and velocity fields along the line - of - sight . the peak polarization is considerable , some 1% in units of the continuum intensity . this has to be the case to yield the large flux density shown in figure [ mean]a for perfect angular resolution . another detail worthwhile noting is the balance between the number of profiles having a large blue lobe and those whose principal lobe is the red one . the line shapes in figure [ vasym0 ] are difficult to compare with observed profiles since observations of the quiet sun have much lower angular resolution . figure [ vasym1 ] shows another classification having all the features explained above , except that the simulation data have been smeared with a 1 seeing , typical of the real observations . first , the polarization signals are reduced by one order of magnitude with respect to the original profiles ; compare them with those in figure [ vasym0 ] . second , new more complicated line shapes have arisen as the result of the large spatial smearing ( radiative transfer smoothes over only 100 km , whereas the spatial smearing does it over 725 @xmath66 725 km@xmath82 ) . stokes profiles having all these very asymmetric shapes are indeed observed in the real quiet sun ( see @xcite @xcite , fig . 12 ; @xcite @xcite , fig . 7 ; @xcite @xcite , fig . 4 , 5 and 6 ; @xcite @xcite , fig . this qualitative agreement is again a notable feature of the simulation , since there was no obvious a priori reason to expect it . however , despite such general qualitative agreement , one can find quantitative differences between the synthetic and the observed profiles . as it happens with the fully resolved profiles ( fig . [ vasym0 ] ) , the number of synthetic profiles having a principal blue lobe and those with a principal red lobe is similar ( e.g. , the pairs # 12 and # 18 ; # 31 and # 34 ; # 33 and # 35 ) . such balance is not present among the observed profiles , where the blue lobe usually dominates ( e.g. , @xcite @xcite , fig . 4 ; @xcite @xcite , fig . another qualitative difference with observations concerns the profiles that are most frequently obtained in the simulation . they show almost no asymmetry ( see classes # 0 to # 7 ) . contrariwise , the observational counterpart have a well defined asymmetry characterized by a large blue lobe ( similar to those observed in plage and enhanced network regions ) . this lack of significant asymmetry can be traced back to the original syntheses ( fig . [ vasym0 ] , classes # 0 to # 4 ) , and therefore to the variations along the line - of - sight of the magnetic field and velocity . the probable cause is discussed in the next paragraph . figure [ brms ] summarizes the kind of variations along the line - of - sight existing in the numerical simulation . several definitions are required before it can be interpreted . the variations are described using standard statistical parameters , namely , the line - of - sight mean value @xmath83 , the line - of - sight standard deviation @xmath84 , and the line - of - sight correlation coefficient @xmath85 , @xmath86 @xmath87^{1/2 } , \label{stat2}\ ] ] @xmath88 \big/ \bigl[\overline{\delta f}_{ij}~\overline{\delta g}_{ij}\bigr].\ ] ] the arrays @xmath89 and @xmath90 may represent any component of the fields , and their indexes vary according to the position in the horizontal plane ( @xmath91 and @xmath92 ) and in the vertical direction ( @xmath93 ) . the symbol @xmath94 stands for the number of grid points in the vertical direction . we should bear in mind that the line - of - sight averages defined in equations ( [ stat2 ] ) depend on the horizontal coordinates . we need to characterize the typical properties of these line - of - sight mean values for each range of mean longitudinal magnetic field strength . for this purpose we define the average @xmath95 and the dispersion @xmath96 of the quantity @xmath97 among all those points in the simulation with a given line - of - sight mean longitudinal magnetic field @xmath98 , i.e. , @xmath99 @xmath100^{1/2},\ ] ] with @xmath101 and @xmath102 note that the symbol @xmath97 may represent any of the statistical parameters in equation ( [ stat2 ] ) ( @xmath103 , @xmath104 or @xmath105 ) , and @xmath106 and @xmath107 correspond to the grid points in the two horizontal directions . the bin size @xmath108 has to be chosen to guarantee having enough points per bin . following these definitions , we can now interpret figure [ brms ] . the different quantities represented are plotted as function of the line - of - sight mean magnetic field . the first panel in figure [ brms ] shows the variation of @xmath109 and @xmath110 . the limited dispersion of the vertical velocities provides an explanation for the predominance of profiles with small asymmetries . the dispersion must be of the order of the line width to produce a substantial modification of the line shape . the typical standard deviation of the vertical velocities along the line - of - sightturns out to be between 0.2 and 0.3 km s@xmath35 ( @xmath110 , fig . [ brms]a , the solid line ) , whereas the line widths are of the order of 2 km s@xmath35 ( see [ sec_spectra ] ) . the magnetic field itself is probably not responsible for the moderate asymmetries since its variations along the line - of - sight are large . for example , figure [ brms]b shows the standard deviation of the mean longitudinal magnetic field @xmath111 , which is frequently larger than the absolute value of @xmath112 . in fact , the variations of the longitudinal magnetic fields are so important that @xmath65 very often changes sign along the line - of - sight(fig . [ brms]c ) . note that none of these statements on large magnetic field gradients apply to the intrinsically strong fields , a case discussed separately in the next paragraph . above , say , 200 g , the variations of field strength are mild and the longitudinal field maintains a constant sign along the line - of - sight ( figs . [ brms]b and [ brms]c ) . concerning the balance between the number of asymmetries towards the blue and towards the red in figures [ vasym0 ] and [ vasym1 ] , it is probably due to the lack of a definite sign for the correlation between magnetic field and velocity . works on the asymmetries in plage and network regions repeatedly indicate the need for a negative correlation to account for the observed preponderance of the stokes @xmath0 blue lobe , explicitly , @xmath113 ( see , e.g. , @xcite @xcite ; @xcite @xcite ; @xcite @xcite ; @xcite @xcite ; @xcite @xcite ; @xcite @xcite ) . such condition is satisfied when downflows and magnetic fields are spatially separated , i.e. , when the strongest downflows tend to occur in weakly magnetized plasma . ( note that @xmath114 corresponds to downflows . ) one can see in figure [ brms]d that the correlation between @xmath115 and @xmath116 has a well defined negative value only for the largest field strengths . since theses points represent a small fraction of the synthetic profiles , they contribute very little to the classification in figure [ vasym0 ] which , consequently , shows no obvious preference for a blue asymmetry . those patches in the simulation with the largest field strength show a clear negative correlation fulfilling the criterion in equation ( [ correlation2 ] ) ( see fig . [ brms]d ) . do they produce the observed stokes @xmath0 profiles with a main blue lobe ? they do not , since the asymmetry of the profiles emerging from concentrations of intense field strength are minimal . this fact can be understood using figures [ brms]a and [ brms]b , which reveal gradients of both magnetic field and velocity too small for the requirements described in the previous paragraph . despite this apparent disagreement , the key ingredients to yield the right shapes are already present in the simulation . if one artificially increases the gradients of magnetic field and velocity already existing in the simulation , large asymmetries similar to the observed ones automatically show up . figure [ net_like ] contains synthetic profiles emerging from the three more intense magnetic concentrations in the snapshot ( labeled as @xmath67 , @xmath24 and @xmath59 in fig . [ magneto ] , left ) . note that they already have the stokes @xmath0 asymmetries that characterizes network and in regions . ( figure [ net_like]d includes one of these observed stokes @xmath0 profiles for reference , namely , one network profile in @xcite @xcite ) . in order to produce theses new synthetic profiles with enhanced asymmetries , we increased the intensity of the flow with respect to the scaling in [ sec_spectra ] by using @xmath117 . in addition , we increased the variations of magnetic field strength by averaging all the absorption and emission over 05 of the simulation ( i.e. , over all the points in a box of this size ) . the qualitative agreement between synthetic and real profiles is remarkable . the above considerations lead to an interesting question . how should the numerical simulations be modified in order to produce asymmetries closer to the observed ones ? first , the dispersion of velocities at the smallest scales must be increased , which implies a decrease in the kinematic viscosity . second , magnetic and non - magnetic regions should be even more intermittent to strengthen the correlation ( [ correlation2 ] ) . this can be achieved by decreasing the magnetic diffusivity , which would both increase the tangling of magnetic field lines and the dispersion of field strengths existing in the intense concentrations . one of the observational constraints on the existence of a complex and tangled magnetic field in the solar photosphere comes from the work by @xcite ( @xcite ) ( see also @xcite @xcite ; @xcite @xcite ) . if a tangled magnetic field exist , it has to broaden the spectral lines of the solar spectrum according to their magnetic sensitivities ( i.e. , according to their effective land factors ) . all other things being the same , those with larger sensitivity should be slightly broader . @xcite ( @xcite ) looked for such effect in the solar unpolarized spectrum with no success . from the error budget of the measurement the authors set an upper limit to the field strength of the existing fields ( @xcite @xcite , equation [ 12 ] ) , @xmath118 in this section we analyze if the magnetic fields in the simulations produce line broadenings compatible with such observational upper limit . this exercise allows to address the question of whether magnetic fields as intense as those in the simulation may exist on the sun , and still remain below the present observational detection limit . we already know that the degree of circular polarization of the simulation stays well within the observational bounds ( [ flux_density ] ) . here we address the question from a different perspective , using a totally different observational constrain that depends on the magnetic fields in a intrinsically different way . our synthetic stokes @xmath61 spectra have an excess of broadening caused by the presence of magnetic fields . figure [ dif ] shows the difference between the mean stokes @xmath61 profile produced by the region , and the mean profile produced when the syntheses are repeated with no magnetic field , but keeping everything else identical . the magnetic profile is broader and shallower at the line center , producing a residual with three lobes . is this extra broadening compatible with the observational limit ( [ limit ] ) ? a detailed modeling of the procedure employed by @xcite ( @xcite ) is clearly beyond our possibilities , since it requires the synthesis of hundreds of spectral lines with different temperature and magnetic field sensitivities . fortunately , the essence of the procedure is simple . if two lines are identical except for their magnetic sensitivity , the small excess of width @xmath119 can be directly related to an _ apparent _ magnetic field strength @xmath120 according to the rule , @xmath121 the scale factor @xmath122 depends of the wavelength , the difference of land factor , and the mean line width @xmath123 . for two lines with the wavelength and strength of @xmath16302.5 , one magnetic and another one non - magnetic , the scaling factor turns out to be @xmath124 which follows from the equations ( 7 ) and ( 8) in @xcite ( @xcite ) . the estimate of the apparent magnetic field strength using equations ( [ broa3 ] ) and ( [ broa4 ] ) is a matter of determining the excess of broadening @xmath125 associated with the residuals in figure [ dif ] . following @xcite ( @xcite ) , the two mean intensity profiles where fitted using gaussian functions , i.e. , @xmath126,\ ] ] where @xmath1 stands for the wavelength relative to the line center and @xmath127 represents the continuum intensity . then the difference between the widths @xmath123 for the syntheses with and without magnetic field directly yields @xmath128 or , using equations ( [ broa3 ] ) and ( [ broa4 ] ) , @xmath129 note that the gaussian fits reproduce fairly well the difference between the two synthetic profiles ( see the dashed line in fig . [ dif ] ) . the field strength of our synthetic profiles ( equation [ [ limit2 ] ] ) apparently come into conflict with the observational limit in equation ( [ limit ] ) . should the inconsistency be real , it points out an excess of magnetic fields in the numerical simulations as compared to the solar case ( an excess of magnetic field strength , area coverage of the fields , etc . ) . however , the marginal discrepancy is probably not significant in view of the uncertainties affecting both the syntheses and the observational limit . actually , the similarity between the observational limit and the predicted width should be understood as real chance to test whether complex tangled fields like those in the numerical simulations are present in the solar photosphere . a slight refinement of the currently available diagnostic tools ( e.g. , stenflo & lindegren s technique ) should be able unambiguously to confirm them or discard them . we return to this point in [ conclusions ] . up to now we only consider polarization signals generated by zeeman effect . however , solar turbulent magnetic fields have been inferred using hanle effect signals . they indicate the presence of turbulent magnetic fields in the upper photosphere - lower chromosphere with a field strength between 5 and 60 g ( e.g. , @xcite @xcite ; @xcite @xcite ) . these strengths may seem too low compared to those assumed in this work , with a mean value larger than 100 g ( e.g. , [ broadening ] ) . therefore , we felt compelled to estimate the hanle signals expected from the dynamo simulation and compare them with observed values . once more we face the question of whether magnetic fields similar to those in the simulations may exist and still produce observable effects within solar values . a complete hanle effect synthesis similar to that carried out for the zeeman effect is clearly beyond the scope of this paper . fortunately , one can estimate the level of hanle signals for one of the typical lines used in hanle effect based diagnostics . considering the sr i line at @xmath14607 , the depolarization produced by a magnetic field can be expressed as @xmath130 where @xmath131 is the ratio between the observed linear polarization @xmath2 and the polarization expected if there were no magnetic field @xmath132 . the symbol @xmath133 parameterizes the magnetic field strength of the turbulent field @xmath23 , @xmath134 the normalization factor @xmath135 scales linearly with the radiative transition rate plus the depolarizing collision rate . the relationship ( [ hanle1 ] ) is an approximation which works well for this particular line in standard quiet sun model atmospheres ( see @xcite @xcite , where one can also find the dependence of @xmath135 on the density and temperature of the atmosphere ) . for additional details see , e.g. , @xcite ( @xcite ) and @xcite ( @xcite ) . in case that the magnetic field strength is not unique but has a distribution of values , the mean signal @xmath136 turns out to be equal to the mean depolarization factor considering the distribution of the field strengths @xmath137 ( see @xcite @xcite , 3 ) . note that @xmath138 no longer varies with @xmath23 as @xmath139 ( @xmath140 ) , therefore the average depolarization @xmath137 bears no information on the large field strengths that may exist in the distribution . in other words , the mean signal will always be biased towards weak field strengths ( e.g. , @xcite @xcite , fig . 3 ) . this fact could reconcile the large fields in the dynamo simulations with the observed hanle depolarization signals . in order to check such possibility , we evaluate the mean @xmath137 for the snapshot studied in the paper . the transition and collision rates required to compute @xmath135 are determined , according to the prescription in @xcite(@xcite , 3 ) , using the quiet sun densities and temperatures given by @xcite ( @xcite ) . we only consider the photospheric layers . the expected signals , including all the magnetic fields in the snapshot describing the distribution of @xmath23 , are represented in figure [ hanle_fig ] . the depolarization changes with height due to the variation of temperature and density in the model atmosphere , which modify @xmath135 . figure [ hanle_fig ] includes the range of hanle signals observed close to the disk center , namely , @xmath141 ( quoted by @xcite @xcite , from the measurements of @xcite @xcite at an heliocentric angle whose cosine is @xmath142 ) . the observational signal has been chosen to be as close as possible to the disk center to sample deep photospheric layers . the measurement of depolarization is ascribed by @xcite ( @xcite ) to a range of heights between 200 km and 400 km . although these layers are still too high for the range of heights that we assign to the simulation in the zeeman syntheses ( the base of the photosphere ) , the depolarization turns out to be within the bounds set by observations ( see fig . [ hanle_fig ] , the solid line ) . the agreement improves taking into account that the turbulent field strengths are expected to decrease with height in the atmosphere ( e.g. @xcite @xcite ) . the dashed line in figure [ hanle_fig ] has been computed assuming the field strengths to be a factor of two smaller than those in the original snapshot . in short , our simplified estimate indicates no obvious inconsistency between the simulations and the observed hanle effect depolarization signals for @xmath14607 . answering this question may be relevant to design of the next generation solar telescopes ( rosner & beckers , 2001 , private communication ) . for the sake of simplicity , we adopt the magnetic flux as the physical parameter to be determined . it is routinely measured using standard instrumentation , and it suffers a severe bias due to insufficient spatial resolution ( [ flux_density ] ) . the fraction of magnetic flux in the simulation that one still detects in synthetic magnetograms is used to quantify the required angular resolution . this information is contained in figure [ mean]b , which we plot in a slightly different way in figure [ teles_diam ] . abscissae and ordinates have been interchanged , and the angular resolution is also presented in terms of the diameter of the ideal telescope achieving the angular resolution ( see equation [ [ eq1]]).this representation suits our present purposes . the curve is shown for two different sensitivities , i.e. , when ( virtually ) all polarization signals are detected , and when only those larger than 20 g are above the noise level . the following conclusions follow from figure [ teles_diam ] : 1 . irrespective of the resolution , one only detects 80 % of the existing flux . the residual 20 % cancellation is due to the smearing along the line - of - sight produced by the radiative transfer . surpassing this upper limit is not a question of increasing the telescope size but rather it demands improving the diagnostic techniques used to measure the magnetic flux . 2 . detecting 50 % of the flux would require an angular resolution of 110 km or a telescope of 85 cm . detecting another 20 % additional magnetic flux demands resolving 40 km and so a telescope some three times larger ( 2.3 m ) . points # 1 and # 2 refer to observations without noise . if one considers a detection threshold of 20 g ( which corresponds to a degree of polarization of some 0.3 % , according to the calibration in [ calibration ] ) , then only 70 % of the flux present in the simulation can be detected . on the other hand detecting 50 % implies an angular resolution of 70 km or a diameter of @xmath143 1.3 m. 4 . the slope of the solid line in figure [ teles_diam ] drastically changes when trying to detect more than , say , 65 % of the original flux . going beyond this point requires a large increase of telescope diameter for a limited increase of additional signal . this 65 % may represent an optimal compromise between resolution and telescope size , and it corresponds to 60 km on the sun or a diameter of 1.5 m. the dynamo simulation by @xcite ( @xcite ) ( see also @xcite @xcite ) produce magnetic fields whose structure resembles in many ways the quiet sun magnetic fields ( [ sec_introduction ] ) . although the simulations were not designed for a realistic description of the solar conditions , the quantitative comparison with the quiet sun that we undertake is useful for a variety of reasons . it allows to judge whether , and to what extent , the turbulent dynamo provides a paradigm to describe the quiet sun magnetism . it allows to identify , study and understand the difficulties and biases faced by the current observational techniques when applied to very complex fields . it helps identifying physical ingredients that are missing in the simulations , an exercise helpful to guide future numerical work . finally , it may suggest the type of technical developments needed to measure the properties of a magnetic field with the complexity present in the numerical data . keeping in mind all these reasons , we synthesize the polarization emerging from the simulation . specifically , we choose a snapshot of the time series representative of the stationary regime , which renders some 2.6 @xmath66 10@xmath144 individual spectra . the assumptions and limitations of our milne - eddington approach to the synthesis are discussed in [ sec_spectra ] . they let us calculate the polarization produced by a magnetically sensitive spectral line similar to @xmath16302.5 . the synthetic spectra have been analyzed in the light of observations of quiet sun magnetic fields . the comparison reveals similarities and differences . in addition , it provides some hints and caveats to keep in mind when interpreting observations . these three aspects of the synthetic spectra are discussed next . the magnetograms produced by the numerical data are in good agreement with those observed in the sun ( [ flux_density ] , figs . [ mean]a and [ inset ] ) . the agreement , however is only achieved after a severe canceling of signals produced by the poor spatial resolution of the current observations . more than 90 % of the unsigned magnetic flux existing in the numerical data does not appear in magnetograms with 1angular resolution ( fig . [ mean]b ) . in other words , it is this 2 to 10 % residual that agrees with the observed signals . the stokes @xmath0 profiles emerging from the simulation are frequently very asymmetric , in qualitative agreement with observations ( see references in [ sec_introduction ] ) . in particular , stokes @xmath0 profiles with three and more lobes result from the existence of two opposite polarities in the resolution element . part of these asymmetries are produced by gradients of magnetic field and velocity along the line - of - sight(fig . [ vasym0 ] ) , but the stokes @xmath0 profiles also owe much of their shapes to variations across the line - of - sightwithin the resolution element ( cf . figs . [ vasym0 ] and [ vasym1 ] ) . we studied in [ broadening ] the excess of broadening of the synthetic intensity profiles due to the presence of a magnetic field . the additional broadening is found to be close to , but typically in excess of , the upper limit of 140 g set by @xcite ( @xcite ) . the uncertainties of the measurement could easily explain this discrepancy , therefore , we understand this marginal disagreement as an invitation to revisit the work by @xcite ( @xcite ) and improve the sensitivity by a factor two . this should be enough clearly to confirm or discard a solar magnetic field with the features present in the numerical data . so far only zeeman signals of typical photospheric lines have been mentioned . we also estimate in [ hanle ] the hanle depolarization to be expected if the simulation is placed at various heights in the photosphere . all the uncertainties notwithstanding , the depolarization signals for @xmath14607 turn out to be within the observed bounds . a quantitative analysis of the stokes @xmath0 asymmetries reveals real discrepancies between the synthetic spectra and the quiet sun . most of the synthetic profiles show mild asymmetries , well below the mean observed values . moreover , the synthetic spectra do not contain the clear observed tendency for the stokes @xmath0 blue lobe to dominate . such trend is present even in the largest polarization signals , which trace big concentrations of magnetic field . we believe that the cause of the discrepancy is twofold . first , the dispersion of velocities between nearby pixels is too small . second , the spatial separation between strongly magnetized and unmagnetized plasmas is too large . these two factors minimize the asymmetries in the intergranular lanes , despite the fact that the kind of correlation between magnetic field and velocity already existing in the simulation produces the right asymmetry . arguments for such explanation are given in [ asym ] . as a support of this view , we artificially increase the dispersion of velocities and magnetic fields existing in the simulation to synthesize spectra in three particularly large magnetic concentrations . the resulting stokes @xmath0 profiles look very much like those observed in the quiet sun network ( synthetic and observed profiles are represented in fig . [ net_like ] ) . if we have correctly identified the origin of the differences , what must be modified in the numerical simulation in order to produce asymmetries closer to the observed ones ? one would need to increase the dispersion of velocities and magnetic fields at the smallest spatial scales , keeping the relationship between them already existing in the simulation . this could be attained by increasing the reynolds numbers of the simulations . there is another discrepancy between the synthetic and the observed profiles which has not been mentioned yet . the analysis of stokes @xmath0 shapes observed in the quiet sun reveals that kg field strengths are common ( see @xcite @xcite , @xcite @xcite ) . there are also weaker fields ( e.g. , @xcite @xcite ; @xcite @xcite ) , but the frequency of kg is certainly larger than that present in the simulation . this difference can be readily pin down to the incompressibility of the simulation ( [ sec_mhd ] ) , which hampers the strong evacuation of the plasma , an ingredient always associated with the existence of kg fields in the photosphere . we have found several similarities between the synthetic spectra emerging from the dynamo simulation and observations of quiet sun magnetic fields . such ( sometimes unexpected ) agreement seems to point out that the simulation already capture some of the essential features characterizing the quiet sun magnetic fields . these arguments enables us to proceed by analogy and propose the existence of magnetic fields on the sun similar to those in the numerical simulations . several interesting consequences follow from this assumption : 1 . the amount of flux inferred by conventional magnetograms depends very much on the angular resolution and the sensitivity of the measurements . good or very good sensitivity is mandatory to detect any fluxes with the canonical 1angular resolution . for example , the structures in the numerical simulation would be very difficult to detect in magnetograms like those provided by the mdi instrument on board the soho spacecraft , with a sensitivity of some 15 g ( e.g. , @xcite @xcite ) . according to figure [ mean]b , only 1% of the original magnetic flux is detectable at this sensitivity level . on the other hand , the sensitivity is not so critical upon improvement of the angular resolution . in the limit of perfect angular resolution the signals become of the order of 50 g , corresponding to a degree of circular polarization of 1% ( see fig . [ vasym0 ] ) . the amount of unsigned flux existing in the simulation is also very large in absolute terms . assuming that the full solar surface were covered by magnetic fields like those in the numerical simulations , the associated total unsigned flux would be of the order of @xmath145 mx . this figure is a factor 4.5 times larger than the total flux detected at solar maximum with conventional techniques ( @xcite @xcite , fig . should such large amounts of hidden flux exist in the photosphere , it would have to interact with the other classical manifestations of the solar magnetism ( active regions , solar cycle , coronal heating , etc ) . this point in particular deserves further investigations . we discuss in [ telescope ] the telescope size required to detect the magnetic structures observed in the simulations . this is a preliminary estimate which certainly has to be revised and refined . ( for example , we do not address the question of whether the 60 km optimum spatial resolution is set by the photon mean - free - path or the diffusion processes included in the mhd simulation . ) however one conclusion of this preliminary study stands out . a sizeable fraction of the magnetic flux is not detected due to the radiative transfer smearing along the line - of - sight . improving such bias is not a matter of increasing the telescope size and polarimetric sensitivity . it only depends on interpreting the observed polarization using techniques that account for the line - of - sight smearing . keeping in mind the complexity of the fields ( e.g. , the longitudinal magnetic field almost always changes sign along the line - of - sight ; see fig . [ brms]c ) , the use of inversion techniques that assume optically thin fluctuations of the magnetic field seems to be a reliable selection ( e.g. , the misma inversion code in @xcite @xcite ) . 4 . the stokes asymmetries contain information on the magnetic structuring at spatial scales smaller than the resolution element , an information unique and valuable ( see [ sec_introduction ] ) . the syntheses show that both variations along the line - of - sight and across the line - of - sight are responsible for the asymmetries . consequently , the inversion techniques aimed at understanding the origin of the asymmetries , and therefore at measuring the properties of the photospheric magnetic fields , have to incorporate the two aspects of the spatial variations . the importance of the above conclusions depends to some extent , on the degree of realism of the simulations . however , the tendencies that they imply are probably correct ; e.g. , the existence of large amounts of solar magnetic flux missing in conventional measurements , or the need for non - standard inversion techniques to determine the properties of the quiet sun magnetic fields . these problems are important enough to deserve a close follow - up . the interest to synthesize the spectra of turbulent dynamo simulators arose from conversations with t. bogdan back in 1998 . both r. rosner and j. beckers pointed out the importance of estimating the telescope size required to spatially resolve the simulation . thanks are due to f. kneer , f. stolpe , and m. collados for providing the flux densities cited in [ flux_density ] , and to an anonymous referee for helpful comments . thanks are also due to j. stenflo for clarifications on the hanle effect based measurements of the turbulent magnetic field strength . discussions with r. manso and j. trujillo bueno were extremely helpful to elaborate [ hanle ] . this work has been partly funded by the spanish ministry of science and technology , under projects aya2001 - 1649 and dges 95 - 0028-c . two of the authors fc and te were partially supported by nasa grant nag5 - 10831 at the university of chicago .

integrated optical components on lithium niobate play a major role in standard high - speed communication systems . over the last two decades , after the birth and positioning of quantum information science , lithium niobate waveguide architectures have emerged as one of the key platforms for enabling photonics quantum technologies . due to mature technological processes for waveguide structure integration , as well as inherent and efficient properties for nonlinear optical effects , lithium niobate devices are nowadays at the heart of many photon - pair or triplet sources , single - photon detectors , coherent wavelength - conversion interfaces , and quantum memories . consequently , they find applications in advanced and complex quantum communication systems , where compactness , stability , efficiency , and interconnectability with other guided - wave technologies are required . in this review paper , we first introduce the material aspects of lithium niobate , and subsequently discuss all of the above mentioned quantum components , ranging from standard photon - pair sources to more complex and advanced circuits .

1608.01100

over the last decades , quantum photonics has become a thriving field of research , promoting both fundamental investigation of quantum phenomena and a broad variety of disruptive quantum technologies @xcite . from the fundamental side , entanglement has been exploited , among others , for more and more convincing nonlocality demonstrations @xcite , single - photon complementarity @xcite , and random walks @xcite . from the applied side , we now find reliable quantum - enabled applications , among others , in secure communication , referred to as quantum cryptosystems @xcite , teleportation - based quantum communication @xcite , metrology @xcite , as well as in fast computing algorithms @xcite . in this framework , integrated optics ( io ) technologies allow realizing complex and scalable quantum circuits @xcite , finding striking repercussions in all the above mentioned research areas , otherwise unreachable using bulk approaches . notably , quantum random walks ( qrw ) and boson sampling are very good examples of new topics in quantum physics being enabled by integrated optics @xcite . a qrw can be implemented via a constant splitting of photons into several possible waveguide . experimentally , it amounts to chaining 50/50 beam - splitters , but the price to pay is the rapidly growing size of the setup . until the 21-evanescently - coupled waveguides demonstrated in 2001 by the university of bristol ( uk ) @xcite , the largest realization using bulk optics used to be limited to a 5-item system . yet , technology triggered a new pathway for quantum information processing as a 101-coupled - waveguides device was presented in 2014 @xcite . this example clearly shows the potential of integrated optics ( io ) when it comes to enabling new research fields . the aim of this article is to establish that integrated photonics provides a powerful technological basis for a broad class of advanced quantum experiments . most of the devices we shall discuss have taken advantage of guided - wave optics ability to dramatically enhance the efficiency of nonlinear interactions . this is because it permits maintaining high optical power densities over distances far exceeding those permitted by the diffraction limit in bulk devices . for passive devices , io has provided a possibility of assembling many components on single chips , simplifying the realization and use of complex circuits . however , io faces one problem , that of coupling integrated devices to subsequent fiber systems . in seminal experiments @xcite , for instance , entangled photon pair collection efficiencies were limited , preventing from reaching the level of bulk optics approaches . nowadays , monolithic integrated devices containing advanced photonics circuit powered by on - chip photon - pair sources are having a tremendous impact on the progress made in the field of quantum information science . more specifically , quantum photonics requires fully integrated devices for producing ( multiple ) pairs of entangled photons efficiently , and where photons can be coherently manipulated using frequency - conversion stages , dynamical routers and phase shifters , combined with external photons , and eventually detected . those devices are therefore often considered as enabling quantum nodes where bits of quantum information ( qubits ) can be tailored at will and on - demand thanks to on - chip capabilities . among the multiple platforms available that are covered for example in ref . @xcite , this review focuses on lithium niobate ( linbo@xmath0 , in the following this material will be referred to as ln ) only . the motivation lies in the fact that , compared to equivalent setups exploiting different platforms , such as silicon photonics @xcite or silica @xcite , ln circuits enable highly efficient periodically poled waveguide ( ppln / w ) photon - pair sources as well as ultra - fast electro - optic modulators . those complementary aspects make them unique and very appealing for on - chip engineering of quantum light despite they feature a larger footprint and incompatibility with cmos electronics . in the following , we will start with briefly presenting ln properties from the material science perspective . then , we will discuss in more details the applications of such a platform along five thematics : entangled photon - pair sources , heralded single photons , applications ranging from quantum interfaces to quantum memories , and the use of io for linear quantum computing and bright squeezed state of light . we will finish with a conclusion and some perspectives related to ln - based integrated quantum photonic devices . we shall note that this review is not intended to be encyclopaedic . it rather intends to provide a honest overview on ln integrated optical technology as applied to various tasks , while referring to works of historical interest as well as current state - of - the - art . when considering the requirements for a quantum photonics platform , the material has obviously to be transparent in the wavelength ranges of interest . moreover , when photon - pair generation is intended , high nonlinear coefficients are desired , as well as practical phase - matching solutions . less critically , the damage threshold should be as high as possible , and the substrate should ideally be available in large scale , at low cost , with high quality and uniformity . in this section , we will start with a quick comparison between several materials motivating why ln appears as a prime candidate for quantum photonics platforms . we will then discuss the properties of ln and the material structuration available for photon pair generation . we will finish with a quick overview of the post processing characterizations used to evaluate the performance of such devices for quantum applications . generally speaking , three families of materials are currently used : * non - centrosymmetric crystals ( or @xmath1 crystals ) are the most widely used , well - known , and exploited materials to obtain efficient nonlinear processes ; * symmetric materials such as silica ( or @xmath2 materials ) , have lower nonlinearities , however they also have lower losses , therefore less efficient frequency conversions can be compensated by much longer interaction lengths . note that artificial @xmath1 nonlinearities can be induced via poling techniques @xcite , despite reduced overall optical nonlinearity and the fact that poled gratings can usually not be engineered over more than a few cm lengths ; * semiconductors such as silicon and algaas are promising materials thanks to extremely controlled technological processes allowing to design very small footprint nanostructures potentially exploiting some of the highest nonlinear coefficients reported to date ( _ e.g. _ , in znse @xmath354 pm.v@xmath4 ) . however , practical phase - matching conditions or propagation losses are still issues at the present time @xcite . tab . [ materials ] presents the most commonly used @xmath1 crystals benefiting from high enough fabrication quality as well as a compatibility with waveguide integration technologies . although etching processes allow obtaining waveguide structures in beta - baryum borate ( bbo ) and lithium triborate ( lbo ) , the associated propagation losses are usually very high @xmath5 @xcite , which is at least one order of magnitude higher compared to waveguide structures realized on ln , lithium tantalate ( lt ) , and potassium titanyl phosphate ( ktp ) . .usual nonlinear crystals used for frequency conversion.[materials ] [ cols="<,^,^,^,^,^ " , ] as depicted in [ fig_hsps_nicesetup ] , a pump telecom laser emitting pulses at the wavelength of 1540 nm and at the ultra - fast repetition rate of 10ghz is first frequency doubled via a second harmonic generation stage ( shg , consisting of an spe : ppln / w ) . this conversion stage fulfils the qpm condition for producing frequency degenerate photon pairs in another type-0 spe : ppln / w via spdc . in other words , the wavelength of the emitted photons is centred around 1540 nm . as was discussed in section [ sec_epps ] and more particularly [ fig_djeylan_dwdm_spectrum ] , energy conservation set by the spdc process makes it possible to use a dwdm filter combination in order to pick up , within the emitted spectrum , twin photons slightly apart from perfect degeneracy ( channels symmetrically spaced around the central wavelength ) and to separate them deterministically . the entire setup downstream the spdc stage , including the hbt arrangement for the heralded photon characterization , is made of standard telecom components only . the measured heralding efficiency is @xmath6 . as discussed in @xcite , the high repetition rate of the laser allows producing photon pairs at a high rate while , per each pump pulse , @xmath7 is maintained as low as required for negligible two - photon events . thanks to this strategy , single photons can be heralded at rates reaching values as high as 2.1mhz , with a constant @xmath8 as low as 0.023 . this @xmath8 value not only proves negligible multi - photon contributions but also represents the best measurement reported to date in the literature for @xmath9 in the mhz regime @xcite . since @xmath9 is directly proportional to the detection efficiency , it could be straightforwardly increased up to 10mhz by replacing the heralding detector with a more efficient one without affecting the @xmath8 value . , scaledwidth=45.0% ] eventually , as proposed by a. migdall and co - workers in 2002 @xcite , speeding up the rate at which heralded photons are made available can also be achieved by spatial multiplexing of several simultaneous spdc stages . as shown in [ fig_hybrid ] , a recent demonstration of this concept has been realized in 2014 thanks to a collaboration between our group and macquarie and sydney universities ( australia ) @xcite . in this work , thanks to a hybrid device incoporating the best of two state - of - the - art integrated photonics technologies , four identical sources of heralded single photons are spatially multiplexed using waveguide structures only . more specifically , hybridization means here that four ppln / ws integrated on a single substrate are interfaced in - between passive routers fabricated by femtosecond - laser direct writing ( fldw ) on glass substrates . additional fast fibre optical switches are placed at the output of the main device for enabling the routing of the heralded single photons towards a single fibre output . the first fldw chip allows splitting a pulsed laser into to pump four identical type-0 spe : ppln / ws . the chosen qpm condition permits generating , by spdc , non - degenerate pairs of photons whose wavelengths are centred at 1310 and 1550 nm . note that the mean pump power is kept low enough in order to statistically create much less than one photon - pair at a time in either one of those four ppln / ws . wherever they come from , the photon - pairs are then coupled to another fldw routing chip that contains four individual integrated wdms for separating the paired photons as a function of their wavelengths . after on - chip separation , the heralding photons at 1310 nm are directed towards single - photon detectors ( ingaas avalanche photodiode ) , and the heralded photons at 1550 nm are directed to a routing stage based on optical switches . note that external wavelength fibre filters ( not represented in [ fig_hybrid ] ) are employed for augmenting the extinction ratio and therefore the overall snr . depending on the detection results , _ i.e. _ , which detector fires among the four , rf electronics enables controlling the switches for routing the 1550 nm photons towards a single main fibre output . the setup performances are outlined in terms of coincidences - to - accidentals ratio ( car ) among heralded and heralding photons @xcite . this parameter characterizes the snr of a photon source against background noise and provides , indirectly , information about two - photon - pair contributions by assuming a poissonian photon - pair statistical distribution . the data obtained with the setup depicted in [ fig_hybrid ] clearly show how spatial multiplexing ensures an increase in photon rates while maintaining a fixed snr . more precisely , an increase by a factor 3 was demonstrated , instead of 4 as expected , as one of the four nonlinear waveguides showed much higher losses compared to the three others . the experiment also showed , to date , the largest number of identical on - chip photon - pair sources having identical spectral properties . in conclusion , the concept of heralded single photons based on photon - pair generation has proven to be very practical . for a long time , the probabilistic emission of paired photons was argued to prevent this kind of source to be scalable . this point has indeed been compensated by the highest emission rate reported to date and the single character of the heralded photons . in this context , ppln waveguides are excellent candidates for real - world implementations of hspss for quantum communication . from one side , as demonstrated by many experiments , they allow to reliably and efficiently produce heralded photons at telecom wavelengths as demanded for long distance operation in optical fibres . at the same time , ppln / ws - and more generally speaking , ln - chips - are highly compatible with fibre - based telecom technology , thus opening the route towards a full exploitation of existing fibre networks and classical telecom systems . in this section , we address waveguide - based systems for which the very properties of ln structures have permitted demonstrations of both quantum interfaces and memories . the former corresponds to coherent wavelength transposition based on nonlinear optics of a photonic qubit prepared at one wavelength to another , while the latter permits coherent storage and retrieval of photonic excitation in and out of a matter system . quantum storage of photonic excitations often relies on energy transitions prepared in a well isolated ensemble of atoms , such as rubidium or caesium , or in a collection of rare - earth ions , such as thulium or praseodymium , embedded in a crystalline matrix . however , these energy transitions often belong to the visible range of wavelengths , preventing from storing directly telecom photonic qubits . this is why coherent quantum interfaces appear to be interesting to make flying and stationary qubits compatible , and possibly all designed in an entire waveguide approach @xcite . nowadays entanglement appears as essential for most of quantum technologies . in the particular case of quantum communications , the manipulation of entanglement has to be possible , for it to be of any practical value . accordingly , a growing community of experimental physicists are working with entanglement transposition , especially in view of the connection of different users on a quantum network consisting of quantum channels along which photonic qubits travel @xcite and atomic ensembles are used to store and process the qubits @xcite . it is thus timely to develop quantum information interfaces that allow transferring the qubits from one type of carrier to another , _ i.e. _ , in the case of photonic carriers , able to provide back and forth a wavelength adaptation to atomic levels while preserving quantum coherence of the initial state @xcite . as a consequence , we will discuss in this section two type of interfaces : the ones for increasing the energy of the photonic carriers toward the visible region and those decreasing the energy toward the telecom range . chronologically , the first demonstration , in 2005 , addressed an up - conversion transfer ( increasing the frequency ) for energy - time coded entangled photons . the inherently noise - free sum - frequency generation process was exploited to coherently transfer qubit carried by a photon at the telecom wavelength of 1312 nm to another photon at 712 nm , _ i.e. _ , close to that of alkaline atomic transitions @xcite . this operation has been shown to preserve entanglement in the following way : an initial photon at 1312 nm , energy - time entangled with another photon at 1555 nm ( see also sec . [ pioneering_ent_sources ] and [ hybrid_polar_ent_sources ] ) , was up - converted to 712 nm . provided the coherence length of the pump laser is longer than the size of the unbalanced analysis interferometers @xcite ( see also caption of [ fig_energy - time_tanz_2002 ] for more details ) , the authors verified that the transfer actually preserves the quantum coherence by testing the entanglement between the produced photon at 712 nm and the untouched photon at 1555 nm . although those two photons were nt originally entangled , the transferred entanglement was found to be almost perfect @xcite . in other words , despite no energy conservation , up - conversion at the single photon level does not reduce the amount of entanglement . a photonic quantum interface enabling the coherent transfer of single qubits from photons at 1310 nm , initially entangled with photons at 1555 nm , to photons at 712 nm . in this realization , it is shown that entanglement is preserved through the up - conversion process . a source ( s ) produces , by spdc in a spe : ppln waveguide ( not shown ) , pairs of energy - time entangled photons whose wavelengths are centered at 1555 and 1312 nm . these photons are sent to alice and bob , respectively , via optical fiber links . the qubit transfer is then performed at bob s location from a photon at 1312 nm to a photon at 712 nm using sum frequency generation ( sfg ) in another spe : ppln waveguide . this nonlinear waveguide is pumped by a cw , 700mw , and high coherence laser operating at 1560 nm . the success of the transfer is tested by measuring the quality of the final entanglement between the 712 nm photon and 1555 nm photon using two unbalanced michelson interferometers and a coincidence counter between detectors d@xmath10 and d@xmath11 . the interference pattern obtained for the coincidence rate as a function of the sum of the phases acquired in the interferometers ( @xmath12 ) shows a visibility greater than 97% . compared to the 98% obtained with the initial entanglement ( curve not represented ) , the fidelity of the transfer is therefore greater than 99% . figure inspired from ref . @xcite.,scaledwidth=50.0% ] more specifically , the single photon up - conversion transfer is based on mixing the initial 1312 nm single photon with a highly coherent pump laser at 1560 nm in a spe : ppln waveguide , as depicted in [ fig_interface ] . the internal up - conversion efficiency of 80% of the chip correspond to a single - photon up - conversion probability of @xmath135% , including the phase - matching filtering losses . the resulting photon at 712 nm , filtered out of the huge flow of pump photons , and the remaining 1555 nm photon were then analyzed using two unbalanced michelson interferometers in the franson configuration @xcite ( _ i.e. _ , the usual setup for testing energy - time or time - bin entanglement ) . a comparison of the degrees of entanglement before and after the up - conversion stage indicates no noticeable degradation and leads to a quantum information transfer fidelity higher than 99% . more recently , reverse process quantum interfaces , _ i.e. _ decreasing the qubit - carrier photon frequency by dfg ( see also section [ sec_nli ] , have been investigated . similarly to sfg , this stimulated single photon down - conversion process is also a noise - free process and requires strong requirement on the coherence of the pump laser . demonstrations of such quantum interfaces have been reported almost simultaneously in 2010 using single photonic qubits by the geneva group @xcite and ntt japanese research groups @xcite , and in 2011 using entangled photon pairs at the university of osaka @xcite . the idea here is to proceed to the exact inverse operation to that discussed just above , _ i.e. _ , transferring coherently , back to the telecom range , qubits initially coded on photons in the visible range . both works , dealing with quite the same experimental procedures , propose a solution towards re - encoding qubits onto telecom photons after the read - out of a quantum memory based on alkaline atomic ensembles @xcite . a photonic quantum interface dedicated to coherently transferring single qubits from photons at 710 nm to photons at 1310 nm by means of dfg . ( a ) experimental set - up for the characterization of the coherence of the wavelength transfer . the setup comprises an attenuated pulsed laser mimicking a single photon source at 1310 nm , a preparation ( bulk ) interferometer to code time - bin qubits on the 710 nm photons ( phase @xmath14 ) , a pump laser at 1550 nm , a pe : ppln / w for single photon down - conversion ( or dfg ) , a fiber interferometer for time - bin qubit analysis ( phase @xmath15 ) , and a dedicated electronics for recording the difference in arrival times between the laser trigger and the detection events ( inset , top right ) . ( b ) efficiency of the quantum interface as a function of the pump power . ( c ) count rates for interfering events detected by a suitable and - gate for an average number of prepared single photons reaching unity . as the phase @xmath15 of the fiber interferometer is scanned , a net ( raw ) interference visibility of 96% ( 84% ) is reported . figure inspired from ref . @xcite and data reproduced with permission of the authors.,scaledwidth=50.0% ] both single - qubit experiments exploit time - bin coded single photons out of an attenuated pulsed laser . those qubits are then sent to single photon down - conversion stages based on pe : ppln waveguides , and , eventually , entanglement fidelity is measured using either a fiber interferometer ( see [ fig_interface_new](a ) and ref . @xcite ) or a plc interferometer ( see ref . @xcite ) . the obtained results , in terms of quality of entanglement , are of the same order , _ i.e. _ , showing a net visibility of almost 96% for the geneva group and of 94% for the ntt group . it is important to note that the quantum interfaces discussed above are not limited to the specific wavelengths reported in the corresponding papers , even if for some practical issues , the triplet 710 nm , 1310 nm and 1550 nm was an optimal choice . those are proof - of - principle experiments and suitable modifications of phase - matching conditions and pump wavelengths would enable engineering the result of the frequency - conversion processes to any desired wavelength triplet . more precisely , waveguide structures , as described in sec . [ sec_ln ] , provide the experimentalists with a broad range of accessible wavelengths by changing the poling period and the pump wavelength accordingly @xcite . for instance , having an original qubit carried by a photon at 1550 nm and a pump laser at 1570 nm would lead to a transferred qubit carried by a photon at the wavelength of 780 nm , _ i.e. _ , that of rubidium atomic transitions @xcite . in addition , the very high frequency - conversion efficiencies of such io devices permit using modest pump powers to achieve reasonable qubit transfer probabilities . this is of great interest for applications requiring very narrow photon bandwidths , for instance when transferring photonic qubits to atoms @xcite . in other words , ultra - bright frequency - converters make very narrow spectral filtering possible while maintaining a high photon - pair flux , with reasonable pump powers as described in sec . [ sec_epps ] and in refs . @xcite . to conclude this section , we must emphasize that interfacing photonic and atomic qubit is of great interest as it would facilitate the building of quantum networks and computers @xcite . in this perspective , recent frequency conversion experiments of photons correlated with a rubidium quantum memory have been performed at osaka university @xcite and at icfo - barcelona ( spain ) @xcite . finally note that up - conversion , or hybrid , detectors based on ppln / ws have been widely employed to convert photons from the telecom range of wavelengths to wavelengths around or below 800 nm , without the need to preserve the quantum state coherence , finding essentially applications in high - speed quantum cryptography @xcite . a good review on those systems can be found in @xcite . quantum memories are key elements for long distance quantum communication , more specifically in view of building real quantum networks . for instance , such devices can be employed for realizing deterministic single photon - sources @xcite , for turning quantum relays to much more efficient quantum repeaters @xcite , and for further enabling quantum computation @xcite . the performances of quantum memories can be quantified by four figures of merit : the storage time , the storage efficiency , the fidelity and the spectral acceptance bandwidth @xcite . the storage efficiency represents the probability of absorption / re - emission in the right temporal and spatial modes . the fidelity corresponds to the overlap between the input and output qubit state . it also allows to define the storage time , reported as the duration over which the output qubit fidelity remains above a certain threshold . eventually , a large memory spectral acceptance bandwidth is of great interest since it is directly connected to the performance of quantum communication links in terms of rate and multimodal operations . in the quest towards optimal quantum memory realizations , various quantum storage protocols have been investigated and associated with different storage devices . among others , were reported solutions based on alkaline cold atomic ensembles associated with electromagnetically induced transparency @xcite , atomic vapours associated with off - resonant two - photon transitions @xcite , ion - doped crystals associated with photon echo or atomic frequency combs @xcite , _ etc . _ very pertinent review papers can be found in refs . @xcite . the above mentioned combinations offer various interaction wavelengths and linewidths , ranging from the visible to the telecom band , and from a few mhz to almost 5 ghz , respectively . naturally , the solutions based on ion - doped crystals , working at cryogenic temperatures ( a few @xmath16k ) , have been investigated using io configurations . the first implementation was proposed by hastings - simon and collaborators , where they studied the stark effect in er@xmath17-doped ti : linbo@xmath0 waveguides @xcite , and demonstrated the suitability of the scheme for protocols based on controlled reversible inhomogeneous broadening ( crib ) in solid - state devices @xcite . this protocol is an extension of the photon echo technique showing better efficiency and a higher maximum fidelity . storage times of 1 to @xmath18s were demonstrated and a possibility of extending this to @xmath19s discussed while the acceptance bandwidth is maintained at 125mhz . one of the attractive feature of this solution lies in the operating wavelength of 1550 nm , being directly mapped to the transition ( @xmath20 ) , making this memory fully compatible with standard telecom networks without any wavelength interface . more recently papers reported the characterization of ti : er : linbo@xmath0 waveguides operated with stimulated photon - echo protocols , suitable for implementation of solid - state quantum memories . details of the photon - echo operation principle can be found in @xcite . the first experiment showed the possibility of storing the amplitude and phase of two subsequent coherent pulses is depicted in [ fig_memoire_pulse_sequence ] @xcite . illustration of the pulse sequence used for time - bin pulses interference in a ti : er : linbo@xmath0 waveguide memory . figure inspired from ref . @xcite , scaledwidth=50.0% ] two pulses are sent to the memory after a bright `` write '' pulse . then a second bright `` read '' pulse is used to retrieve the stored information , the relative times between the read pulse and the echo being the same as the relative times between the write and data pulses . to observe interference , two read pulses are sent to the memory with a time difference equivalent to the delay between the two data pulses . this pulse sequence configuration gives three echoes , but the middle one corresponds to the coherent superposition of two overlapping contributions , _ i.e. _ , the second data pulse emitted by the first read pulse and the first data pulse triggered by the second read pulse . consequently , an interference pattern can be observed thanks to the modulation of the phase carried by either the data or the read pulse . in this case , the authors obtained a near - perfect interference pattern visibility . this experiment therefore stands as a proof of the phase and amplitude preservation of two subsequent coherent pulses and the possibility of adjusting the phase . similarly to the previously described experiment , dealing with temporal interference , a second experiment has been dealing with spatial interference . it is based on a similar technology , and the aim was to demonstrate the coherent storage of the relative phase between two pulses in two independent quantum memories , located placed in an arm of a fiber interferometer @xcite . the phase in the interferometer could be modulated using a fibre stretcher , leading to an interference pattern with a visibility of about 90% . therefore the phase was maintained over the storage time , and indistinguishability of photon echoes was proven . three years later , a thulium - doped linbo@xmath0 waveguide has been investigated as an alternative to the ti : er : linbo@xmath0 thanks to a collaboration between calgary and paderborn universities @xcite . with such a device , shown in [ fig_qm_waveguide ] , the thulium ( tm@xmath17 ) transition of interest is at 795 nm and the authors showed that the related transition has an optical coherence of 1.6@xmath21s when the sample is cooled at a temperature of 3k . a storage time of up to 7ns was demonstrated . the authors showed the coherent storage of time - bin entanglement using a photon - echo protocol . thanks to this tm - doped waveguide , they showed a broadband acceptance photon bandwidth of 5ghz , which corresponds to an increase by a factor 50 when compared to earlier published results for solid - state quantum memory devices . this is directly comparable with well - established systems based on atomic vapours , such as the ones demonstrated by the oxford group @xcite . eventually , they obtained an input / output entanglement fidelity or more than 95% . this realization highlights why io technology could play a leading role in development of solid - state quantum memory systems . illustration of the waveguide geometry and material layers for implementing a solid - state quantum memory . the tm concentration profile is given on the left and the calculated intensity distribution of the fundamental tm - mode at 795 nm is shown on the right . from the waveguide center , iso - intensity lines represent 100% , 87.5% , 75% , _ etc . _ , respectively , of the maximum intensity . figure inspired from ref . together with the rise of ultra - bright integrated photon pair sources ( see section [ sec_epps ] ) , integrated photonic circuits turned to be remarkable alternative to optical circuits based on bulk optical elements bolted to room - sized optical tables @xcite . soon after , a new generation of integrated - photonics - enabled experiments was demonstrated @xcite . in this section , we aim at describing some of the more pertinent advanced circuits . it covers ln chips merging more than two elementary functions whose global performances have shown a significant step over bulk optics . this section will be divided in three parts : the first one encompasses demonstrators whose dense integration allows realizing otherwise impossible operations using bulk optics . we will start with a quantum relay chip applicable to long - distance quantum communication @xcite and also discuss a state - of - the - art chip allowing to generate and manipulate entangled photons @xcite . the second subsection will focus on realizations where dense integration enables in itself a new range of experiments , as for instance the generation of photon pairs in nonlinear waveguide arrays @xcite . finally , the last subsection will deal with enhanced nonlinear interaction in integrated structures being exploited to demonstrate sources of advanced quantum states . the discussed experiments are generation of entangled photon triplets @xcite and single photon interaction @xcite . as presented in section [ sec_ln ] , ln exhibits both optical - optical and electro - optical non - linear properties . it is therefore natural to exploit spontaneous parametric down - conversion to produce entangled photon pairs but also to integrate many channels that can be used to switch , recombine or make photons interfere at specific locations by use of electro - optically controllable couplers . two favourable aspects of the io implementations are : ( a ) the compactness and the monolithic construction of the chip ensure negligible phase drift between the channels during propagation ; ( b ) most of the crucial functions for quantum processing are based on linear optical elements , including variable beam splitters and phase shifters @xcite , commonly achieved with current technologies on ln . as a first example of dense integration , our group recently implemented a quantum relay chip for quantum communication , in a compact , stable , efficient , and user - friendly fashion . et @xmath22 . @xmath23 and @xmath24 are two detectors responsible for the bell state measurement . at the end of the quantum channel , bob s detector is triggered by the and - gate ( & ) placed after these two detectors . figure inspired from ref . @xcite.[relay_nice],scaledwidth=50.0% ] [ relay_nice ] presents a schematic of a chip on which all the necessary optical functions are merged for implementing on - chip teleportation @xcite . the basic operation is as follow : let us suppose there is an initial qubit @xmath25 , encoded on a photon at 1530 nm and travelling along a fiber quantum channel connected to port a to the relay chip . at the same time , a laser pulse at 766 nm , synchronized with the arrival time of the photon , enters through the nonlinear zone of the chip which consists of a ppln waveguide . here , it generates by spdc an entangled pair of photons ( 2 and 3 ) whose wavelengths are respectively centred at 1530 nm and 1534 nm . then , the first 50/50 directional coupler ( @xmath26 ) is used to separate the created entangled photons in such a way that photons 1 ( sent by alice ) and 2 arrive at the same time at the second 50/50 coupler ( @xmath22 ) . if conditions on the polarization states , central wavelengths , and coherence times are met , photons 1 and 2 can be projected onto one of the four entangled `` bell states '' . the resulting state is identified by coincidence analysis ( & -gate ) on detectors @xmath23 and @xmath24 placed at the output of the chip . as a result of this measurement , the quantum state @xmath25 ( initially carried by photon 1 ) can be teleported onto photon 3 that exits the chip at port b , ideally without any degradation of its quantum properties . this is made possible by photon 3 being initially entangled with photon 2 . as a consequence the resulting electrical trigger is not only the signature of the presence of the initial carrier at the relay chip location but also of the departure of a new one coded with the same qubit state that remains unknown . from a practical point of view , when this teleportation process is repeated on all the qubits travelling along the quantum channel , we obtain , at the output of the chip , `` relayed '' photonic qubits coded on telecom photons . this relay function is ensured by the fact that each photon arrives synchronized with an electrical signal given by the and - gate which allows triggering bob s detectors at the end of the channel and therefore increasing both the snr and the communication distance . on the technological side ( see [ relay_nice ] ) , the chip features a photon - pair creation zone and two tunable couplers ( which will be operated in 50/50 regime ) . the waveguiding structures are obtained using soft proton exchange on a z - cut crystal . a 10 mm long ppln zone of poling period of 16.6@xmath21 m allows the generation of twin photon at 1535 nm at the temperature of @xmath27c . the directional couplers @xmath26 and @xmath22 consist of two waveguides integrated close to each other over mm - length . if the spacing is sufficiently small , energy is exchanged between the two guides . an electro - optical control of the coupling ratio is made possible using deposited electrodes enabling switching from 0 to 100% coupling ratios and reconfigurable operation . to correctly separate photons 2 and 3 and entangle 1 and 2 , the two couplers are set to the 50/50 ratio . once manufactured , the chip has been characterized using a weak coherent source in the single photon regime using the so - called hom two - photon interference effect , which is seen as the major preliminary step towards achieving teleportation . we obtained a quantum interference featuring 79%@xmath2825% net visibility using two independent single photons , one external and one from the on - chip created pairs . the result is in good agreement with the theoretical expectation of 75% accounting for the single photon statistics . the large error bars are essentially due to a non - optimal detection scheme and a lossy device ( 9db ) leading to a very low three - fold coincidence rate . however , it has been shown that the achievable quantum communication distance could be possibly augmented by a factor of 1.4 in that particular case . another excellent example of high integration is given by the chip from nanjing university ( china ) @xcite . it gathers on a single substrate six different optical functions made of annealed proton exchanged waveguides integrated on a z - cut ppln crystal . chronologically , the photons experience a y - branch junction , transition tapers , a phase shifter , a nonlinear zone , and finally 50/50 couplers as well as wavelength filters . [email protected] mm . the widths of the single mode waveguides are 2@xmath21 m for the pump ( zone i ) and 6@xmath21 m for the entangled photons ( zones ii and iii ) . the ppln section is 10 mm long with a period of 15.32@xmath21 m . the interaction lengths of @xmath26 and @xmath22 are 650@xmath21 m and 1300@xmath21 m , respectively , with a separation of 4@xmath21 m between the coupled waveguides . the electrodes are 8.35 mm long with a separation of 6@xmath21 m . the buffer layer ( sio@xmath30 ) is etched along the gap between the electrodes to suppress dc drift . figure inspired from ref . @xcite.[chinese_chip],scaledwidth=50.0% ] the chip comprises three sections . in section i , 780 nm pump pulses are coupled into the central waveguide ( l@xmath31 on [ chinese_chip ] ) and coherently split by a y - branch single mode beam splitter toward two separated arms . then , electrodes above each path are used to control their relative phase shift ( @xmath32 ) . eventually , transition tapers ensure efficient coupling of 780 nm light into single mode waveguides at telecom wavelength ( 1560 nm ) . section ii is dedicated to entanglement generation . it consists in a 10mm - long ppln region , in which degenerate photon pairs at 1560 nm are produced indistinguishably from either one of the two ppln waveguides . statistically , if much less than one pair is created at a time in the two waveguides , hence the photonic state is path - entangled , _ i.e. _ , @xmath33 . section iii is designed to engineer entanglement . there , quantum interference is achieved using a 50/50 coupler ( @xmath26 ) . depending on the applied phase , the output state remains path - entangled or is converted to the product state @xmath34 . continuous morphing from a two - photon separated state to a bunched state is further demonstrated by on - chip fast control of electro - optic phase shift . the photon pairs are separated from the pump by on - chip wavelength filters ( @xmath22 ) since the photons at 1560 nm experience a 100% transfer in the @xmath22 couplers , toward waveguides @xmath35 and @xmath36 by evanescent coupling , whereas the pump at 780 nm remains in @xmath37 and @xmath38 . from the technical side , a poling period of 15.32@xmath21 m at a temperature of @xmath39c allows to generate pairs of photon with an associated brightness of @[email protected]@xmath4.mw@xmath4 . it therefore requires a few tens of @xmath21w to operate in the single pair emission regime . the 780 nm light suppression has been measured to be 29.2db for @xmath35 and 31.4db for @xmath36 , accounted for the integrated filters and the propagation losses . single photon interference shows a visibility of 98.9% and the period of the fringes is 7.1v with an offset voltage of 2.3v to compensate for a slight optical path difference between the two ppln waveguides . the coupling efficiency from the chip to the output fibres provides an additional 1db loss and the propagation losses for the entangled photons in waveguides @xmath35 and @xmath36 are 6.5db and 8.6db , respectively . the characterization of the emitted two - photon states has been performed thanks to a reverse hom interference using an external beam - splitter . a visibility of 92% has been reported . quantum random walk and boson sampling are two iconic research trends since they belong to a class of mathematical problems extremely hard to simulate using classical computers when the number of walks becomes large @xcite . this is one peculiar situation where integrated quantum photonics turns out to be interesting by allowing to work with more than 20 cascaded beam - splitters whose relative phases are inherently stable @xcite . the chip demonstrated in ref . @xcite through a collaboration between australian and german groups adds an extra degree of complexity to random walk by investigating nonlinear random walk in a waveguide array to generate non - classical photon - pair states with a reconfigurable degree of path - entanglement . as depicted in [ aussie_chip ] , entanglement is achieved through quantum interference of photon pairs generated by spdc in a @xmath1 nonlinear waveguide array ( wga ) . the wgas are fabricated on z - cut ppln wafers by titanium indiffusion . the wga used for their experiment consists of 101 waveguides with a spacing of 13.5@xmath21 m and a length of 51 mm . the distance between each neighbour waveguides has been engineered so that the pump photons at 775 nm never leaves the central waveguide , whereas the generated photon pairs at 1550 nm undergo a quantum walk as schematically illustrated in [ aussie_chip ] . typical cw pump powers are below 0.5mw . . the photon pairs ( red ) are generated in the pump waveguide and propagate in the regime of quantum walks . as the pump continuously generates photon pairs along the propagation axis ( z ) , signal and idler photons walk back and forth between all the waveguide ( red arrows ) . figure inspired from ref . @xcite.[aussie_chip],scaledwidth=50.0% ] starting from the pumped waveguide , photon pairs evolve over time by hopping to the left or right neighbouring waveguides . here , the pump beam ( green ) can initialize photon - pair quantum walks in the central waveguide at all possible distances @xmath41 from the input . basically , it implies that , at each position @xmath41 , one has to consider the coherent superposition of created pairs hopping back and forth from the central waveguide but also the contribution of pairs created earlier . this effect leads to the excitation of a cascade of interfering 2d quantum walks , visualized by the red arrows in [ aussie_chip ] . actually , each of these quantum walks is a continuous - time nonlinear quantum walk with propagation distance @xmath41 providing much richer output states compared to passive quantum random walks . the cascaded quantum walks can be controlled by tuning the pump wavelength or changing the sample structure , providing a flexible tool for the preparation of states with a variable degree of nonclassicality in the spatial domain . for demonstration purpose , the authors have changed the phase mismatch by tuning the pump wavelength and also spanned over all possible input waveguides to show the reconfigurability of the quantum states . interestingly , an additional theoretical investigation exploits spatial engineering of the ppln duty cycle so as to obtain four coupled nonlinear waveguides whose efficiency is longitudinally modulated in order to produce , on demand , any desired bell - state @xcite . photon - triplet states are a long standing goal in quantum optics since they are essential for fundamental tests of quantum mechanics and optical quantum technologies @xcite . their production has been proposed 20 years ago based on cascaded spdc processes @xcite . however , the direct generation of such a state suffers from an extremely low efficiency . the performances of ppln waveguides have allowed the direct generation of photon triplets from spdc of a single photon into a photon pair @xcite . , scaledwidth=50.0% ] as shown in [ triplet_chip ] , we will focus here on a fully integrated device that merges , on a single chip , two nonlinear zones and a wavelength - dependent directional coupler for such a purpose @xcite . compared to free - space realizations , the integrated design allows maintaining the overall losses at a reduced level thanks to the ultra - low intrinsic scattering losses of waveguides and the lack of lossy interface like fibre - to - waveguide coupling . the chip indeed requires only 10@xmath21w of pump power for producing 11 detected photon triplets per hour and a coincidence signature of more than 27 standard deviations above the noise level . this compares very positively to free space approaches whose production rate was limited to 5 detected photon triplets per hour for several mw of pump power @xcite . the chip is pumped with a laser beam at 532 nm ( green p - labelled photon ) and the first periodically poled area efficiently generates paired photons respectively at 790 nm ( red s1-labelled photon ) and 1625 nm ( brown i1-labelled photon ) . the on - chip directional coupler structure allows to demultiplex the paired photons as a function of their wavelengths . while the long - wavelength idler photons are transferred to the adjacent waveguide , the short - wavelength signal photons remain in the original arm . the signal photons can act as a pump for the second spdc stage and decay to a pair of granddaughter photons at around 1580 nm ( red s2 and i2 photons ) within the other ppln / w structure . eventually , three telecom photons are produced out of one green pump photon and the tricky part of the experiment lies in the mean number of pairs ( s1 and i1 ) to be maintained below 0.1 to ensure that only three photons are generated at a time . in other words , the triplets overall generation probability can not be artificially compensated by increasing the pump power . here the only free parameter is the internal conversion efficiency of spdc which is directly related to the waveguiding structures . finally note that generation of tripartite telecom photons has recently been demonstrated at the university of hefei ( china ) using hybrid - cascaded processes , _ i.e. _ combining an atomic ensemble and a nonlinear ppln waveguide @xcite . although not completely based on a fully integrated device as the previously discussed realizations , note that the geneva group also reported the demonstration of photon triplet generation . the very interesting part of their approach lies in the sum - frequency generation ( sfg ) using two single photons as input states . such an interaction at the single photon level has been a cornerstone in the community @xcite for its central role in entanglement creation between independent photons . some demonstrations have been made in artificial materials such as quantum dot in high - finesse cavities @xcite , or through electromagnetically induced transparency in atomic ensembles . nowadays , a major challenge is to realise photon - photon interactions in materials that are less restrictive in terms of bandwidth and wavelength @xcite . the geneva team has taken an approach that exploits a parametric process in a nonlinear waveguide and shows for the first time a nonlinear interaction between two independent single photons via sfg . 0.3 nm . in this configuration , the heralded telecom photons are projected onto a spectral mode that is matched to the acceptance bandwidth of the sfg process , which was measured to be 0.27 nm . those two telecom photons are combined via a dense wavelength division multiplexer ( dwdm ) and directed to a 4.5cm - long fibre - pigtailed type-0 ppln / w . the unconverted photons are deterministically separated from the sfg photons by a prism ( not shown ) , the latter being sent to a single - photon detector ( d3 ) . threefold coincidences between detectors d1 , d2 , and d3 are recorded using a time - to - digital converter tdc . b ) a clear signature of the photon - photon interaction emerges from the background noise in the time - of - arrival three - fold coincidence histogram . the axis labeled @xmath42 shows the delay between detectors d1 and d3 , while @xmath43 shows the delay between d2 and d3 . each pixel is composed of 2.3ns bins , defined by the laser repetition rate . the histogram counts shows a poissonian distribution for the noise with a mean value of 35 . the single peak at 80 corresponds to the true three - fold coincidences and exceeds the mean value by 7 standard deviations . figure inspired from ref . @xcite.[sp_sfg],scaledwidth=50.0% ] as illustrated in [ sp_sfg](a ) , the authors have used two independent heralded single photon sources , hsps1 and hsps2 , generating photons at 1560 nm and 1551 nm , respectively . having two different wavelengths is of prime importance to distinguish shg from sfg , _ i.e. _ , to avoid the pollution of independent single photon interaction by two - photon state self interaction . the two hsps are fibre - coupled to a ppln / w that is quasi - phase matched to perform the sfg process 1560 nm + 1551 nm @xmath44 778 nm . thanks to a single photon interaction efficiency of @xmath45 ( fibre pigtail losses included ) , which is directly related to the spdc efficiency of the device , the authors have registered 80 three - fold coincidences over 260 hours as shown in [ sp_sfg](b ) . this result has a statistical significance of over 7 standard deviations with respect to the background ( poissonian distribution centred on 35 ) . as a conclusion , quantum photonics experiments have now reached such a level of complexity that only integrated optics can fit current and future requirements . integrated photonics stands now as the first - choice approach for its reduced interface losses when the source is monolithically integrated to complex quantum circuits . the case of ln devices is particularly interesting since it exhibits all the technological requirements for future controllable photon - photon production , interaction , and manipulation . the last application of integrated photonics to the field of quantum information science , that we will discuss , concerns squeezed state of light . a key difference with the above discussed quantum functions is that , while one often consider quantum photonics to involve discrete observable ( path , polarization or time - bin ) , squeezed state of light involves continuous variable ( cv ) such as the two quadrature of the electrical field of light . among the fundamental applications , an attractive application feature is that entanglement on continuous variable ( cv ) of light can be generated in a deterministic way via non - linear optics and that , although propagation losses reduce le level of squeezing ( or correlation ) , it is never completely destroyed @xcite . based on this principle , different realizations of squeezing at telecom wavelengths have been demonstrated @xcite . + in this context , ppln / w can be conveniently exploited for the generation of cv non - classical light @xcite . in particular , thanks to strong mode confinement and non - linear properties , spdc in ppln / ws allows the efficient generation of high quality squeezed states without recurring to cavity - based conversion enhancements @xcite . on one hand , compared to bulk implementations , single pass arrangement in waveguides offers better compactness and stability @xcite . on the other hand , in the absence of an optical resonator , the squeezing emission bandwidth is limited only by the spdc phase - matching condition and can cover tens of thz . this last feature opens to the possibilities of wavelength demultiplexing and high - speed quantum communications . + in what follows , we will describe some interesting examples of applications of lithium waveguide technology to cv quantum experiments . in particular , we will report on the realization of a ln chip for the generation of squeezing at telecom wavelength in a compact and partially integrated setup pumped by a pulsed telecom laser . this will open to the presentation of a work demonstrating the possibility of enhancing pulsed homodyne detection performances thanks to a dedicated phase - matching stage implemented in a ppln / w . eventually , we will present a recent work where ppln / ws compatibility with fibred components is exploited to implement a fully guided wave squeezing experiment . + , scaledwidth=50.0% ] ppln / ws have been employed in both single mode and two mode squeezing experiments . for example , continuous wave ( cw ) broadband entanglement at 946 nm has been demonstrated in tokyo in 2007 @xcite , by combining on a 50:50 bulk beam splitter two independent squeezed states generated each by a ppln / ws . correlated beams are then simultaneous measured thanks to a double homodyne detection scheme and entanglement is confirmed with a verification of duan s inequality over a bandwidth of more than 30 mhz . as pointed out by the authors , the observed squeezing bandwidth is basically limited by the detector performances and could be improved with faster electronics . + in 2002 , the squeezing generation in a linbo@xmath0 integrated optical circuit is demonstrated in pulsed pump regime @xcite . the chip is based on ape waveguide structures and includes a second harmonic generation stage , waveguide couplers , spatial mode filters , and a degenerate optical parametric amplification ( d - opa ) used to generate squeezed light at the telecom wavelength of @xmath461538 nm ( see fig . [ fejersqueezing ] ) . on the component , the shg and the d - opa ( _ i.e _ the stimulated pdc ) stages correspond to suitably periodical poled areas . the system is pumped with 20ps - long pulses at 1538 nm ( 100mhz repetition rate ) , whose wavelength is frequency doubled by the shg stage so as to provide a pump beam for the pdc . another ( non - converted ) fraction of the same laser is sent towards a different input of the component and used as a seed for the d - opa . the seed and the pump are combined thanks to an integrated waveguide coupler and reach together the d - opa region . in addition to miniaturizing a part of the experimental setup , the integration reduces the amount of frequency doubled light lost in coupling to the d - opa stage : in this sense , it increases the overall power efficiency of the device . at the chip output the squeezed light is analyzed in bulk homodyne detector . experimentally , for 6w of incoming peak pump power at 1538 nm , single mode squeezing with -1db of measured noise reduction is observed . + as discussed by the authors of @xcite , a strong limitation in pulsed squeezing homodyning arises from poor detection efficiencies due to inadequate temporal mode matching between the beam under investigation and local oscillator ( lo ) . as a matter of fact , a main difficulty in single - pass squeezers pumped in pulsed regime lies in gain induced diffraction , that distorts the squeezed light phase front , when working in high parametric gain ( _ i.e. _ high squeezing regime ) @xcite . a valuable strategy to comply with this limitation has been demonstrated in 2008 in tokyo @xcite . in this work , the lo pulses are suitably temporally shaped by sending them though a dedicated optical parametric amplifier realized in a periodically poled mgo : linbo3 waveguide . the experiment exploits the fact that parametric gain in the center portion of the amplified beam is larger than in its tails and that it can be conveniently adjusted by acting on the opa pump beam : based on this principle , the lo temporal profile can be conveniently modified . the authors test their approach with the detection of squeezed light at 1535 nm produced by spdc in another periodically poled mgo : linbo3 waveguide . the squeezer is pumped by a frequency doubled q - switched laser , emitting optical pulses at 1535 nm with a duration of 3.7ns . a systematic study of detected squeezing as a function of the spdc pump power is provided . shaped lo pulses allow to achieve near - unit mode - matching efficiency and to measure , in the high gain regime , a squeezing of -4.1db below the shot noise level . this value corresponds to an improvement of @xmath461db with respect to the case of non - shaped lo and proves the validity of the adopted strategy . + , scaledwidth=50.0% ] to conclude this section , we observe that in all presented implementations , squeezed states generated in ppln / w are collected free - space to be characterized in bulk homodyne detectors . very recently , in nice , a entirely guided - wave realization of a squeezing experiment at telecom wavelength has been demonstrated , for the first time @xcite ( see fig . [ setupcvnice ] ) . in this scheme , single - mode squeezing at 1542 nm is generated in cw pumping regime via degenerate spdc in a periodically poled lithium niobate ridge - waveguide ( ppln / rw ) . at the output of the ppln / rw , the non - classical beam is measured with a fibre homodyne detector . as for previous implementations , the spdc pump beam is obtained by frequency doubling a cw laser operating at 1542 nm . the shg stage implemented in a spe : ppln / w . a fraction of the laser beams is used a lo for the homodyne detection . cw regime completely bypasses the lo temporal matching issues . in addition , besides the miniaturization of the setup , a major advantage of guided - wave optics lies in the achievement of a high degree of spatial mode matching between the lo and the signal without optical adjustment @xcite . this huge benefit extremely simplifies the homodyne detector implementation . with this setup , the authors observe up to @xmath47 db of squeezing emitted at 1542 nm in cw pumping regime for an input pump power of 28mw . the proposed configuration allows implementing a plug - and - play and extremely easy experiment , entirely based on commercially available components and fully compatible with existing fibre networks . + in conclusion , as for discrete variable experiments in single photon regime , ppln / ws can be key ingredients for efficient cv experiments . more in details , as discussed , compared to their bulk counterparts , ppln / w - based squeezing experiments offer the advantage of extremely simplifying the experimental setup , as demanded for out - of - the - lab realizations , while ensuring versatility , reliability and high non - linear optical performances . the impact of integrated optical devices in modern quantum optics has already been considerable . as have been outlined all along this manuscript , pertinent , and otherwise infeasible with bulk optics approaches , realizations have been reported over the last two decades , with outstanding advances in quantum sources , interfaces , relays , memories , as well as in linear optical quantum computing . among all the available technological quantum photonics platforms , lithium niobate has been playing a major role , due to mature fabrication processes and remarkable inherent properties , that are barely met with other materials . thanks to their compactness , efficiencies , and interconnects capabilities , many of the demonstrated individual devices based on lithium niobate can clearly serve as building blocks for more complex quantum systems . those devices have also been shown compatible with other guided - wave technologies , such as standard fibre optics components . elementary implementations of on - chip integration has begun and can be expected to be greatly expanded in the near future . while the input and output coupling continues to remain a difficult problem for all integrated optical devices , the continuing evolution of these developments , as well as the incorporation of new technologies , such as nano - wire waveguides and photonic crystal circuits , could add considerable impetus to this rapidly evolving field . there is no fundamental reason preventing a solution to this problem . the road towards achieving exploitable quantum computers and networks might be very long and challenging . however , there exists a large space for important advances using quantum systems enabled by integrated photonics technologies , with tremendous impacts on how information will be processed and communicated . up to now , we have only seen the premises of quantum enabled technologies . the future of quantum information science very likely lies in the development of new technologies as well as in the coherent merging of those technologies for assessing hybrid devices , for pushing the potential of dense optical integration to its maximum and exploiting the best features of all of them . increased density and functional integration clearly appears as one of the major directions for future deployment of quantum integrated optical devices . a great potential might lie in integrated waveguide arrays , also referred to as quantum guidonics , where dense integration plays a crucial role , with or without nonlinear optical features . in coupled - waveguide arrays , light propagates as discrete beams , with both guided and free characteristics : these arrays are can therefore be considered as versatile two - dimensional meta - materials with , on lithium niobate , the capabilities of being ( re-)configured on demand using electro - optical effect . the guidonic patterning , _ i.e. _ , patterning the coupling coefficient between neighbour waveguides which controls the propagation , allows novel manipulation of light , constituting a complete discrete photonics corpus @xcite . there is no doubt that the `` coherent '' interfacing of nanotechnologies , ferroelectric domain structuring , and integrated optics on lithium niobate should enable great advances in quantum optical systems in the near future @xcite . the authors acknowledge financial support from the `` agence nationale de la recherche '' for the e - quanet ( anr-09-blan-0333 - 01 ) , conneqt ( anr-2011-emma-0002 ) , spocq ( anr-14-ce32 - 0019 ) , and inqca ( anr-14-ce26 - 0038 - 01 ) projects , the european commission for the fp7-fet quantip ( grant agreement n@xmath48 244026 ) and fp7-itn picque ( grant agreement n@xmath48 608062 ) projects , the ixcore research foundation , the simone & cino del duca research foundation ( institut de france ) , the universit nice sophia antipolis , and the cnrs . the authors would like to thank d. aktas , b. fedrici , f. mazeas , l. a. ngah , and p. vergyris for constant and fruitful help and discussions . st would like to warmly thank ect for constant support over the years . hensen b , bernien h , dreau a e , reiserer a , kalb n , blok m s , ruitenberg j , vermeulen r f l , schouten r n , abellan c , amaya w , pruneri v , mitchell m w , markham m , twitchen d j , elkouss d , wehner s , taminiau t h and hanson r 2015 _ nature _ * 526 * 682686 metcalf b j , thomas - peter n , spring j b , kundys d , broome m a , humphreys p c , jin x m , barbieri m , steven kolthammer w , gates j c , smith b j , langford n k , smith p g and walmsley i a 2013 _ nat . _ * 4 * 1356 solntsev a s , setzpfandt f , clark a s , wu c w , collins m j , xiong c , schreiber a , katzschmann f , eilenberger f , schiek r , sohler w , mitchell a , silberhorn c , eggleton b j , pertsch t , sukhorukov a a , neshev d n and kivshar y s 2014 _ phys . rev . x _ * 4 * 031007 silverstone j w , bonneau d , ohira k , suzuki n , yoshida h , iizuka n , ezaki m , natarajan c , tanner m g , hadfield r h , zwiller v , marshall g d , rarity j g , obrien j l and thompson m g 2014 _ nat . * 8 * 104108 nouroozi r 2010 _ all optical wavelength conversion and parametric amplification in ti : ppln channel waveguides for telecommunication applications _ ph.d . thesis department physik der universitt paderborn , paderborn , germany migdall a , polyakov s , fan j and bienfang j ( eds ) 2013 _ single - photon generation and detection : physics and applications _ ( _ experimental methods in the physical sciences _ vol 45 ) ( academic press ) isbn 978 - 0 - 12 - 387695 - 9 zhang p , aungskunsiri k , martn - lpez e , wabnig j , lobino m , nock r w , munns j , bonneau d , jiang p , li h w , laing a , rarity j g , niskanen a o , thompson m g and obrien j l 2014 _ phys . lett . _ * 112*(13 ) 130501 brida g , degiovanni i p , genovese m , piacentini f , traina p , della frera a , tosi a , bahgat shehata a , scarcella c , gulinatti a , ghioni m , polyakov s v , migdall a and giudice a 2012 _ appl . phys . lett . _ * 101 * 221112 langer c , ozeri r , jost j d , chiaverini j , de marco b , ben - kish a , blakestad r b , britton j , hume d b , itano w m , leibfried d , reichle r , rosenband t , schaetz t , schmidt p o and wineland d j 2005 _ phys lett . _ * 95 * 060502 simon c , afzelius m , appel j , de la giroday a b , dewhurst s j , gisin n , hu c y , jelezko f , krll s , mller j h , nunn j , polzik e s , rarity j g , de riedmatten h , rosenfeld w , shields a j , skld n , stevenson r m , thew r , walmsley i a , weber m c , weinfurter h , wrachtrup j and young r j 2010 _ eur . . j. d _ * 58 * 122

the astrophysical factor forbe()b at zero energy , , is determined from an analysis ofpb(b ,be)pb at 52 mev / nucleon by means of the method of continuum - discretized coupled - channels ( cdcc ) taking account of all nuclear and coulomb breakup processes . the asymptotic normalization coefficient ( anc ) method is used to extract from the calculated breakup cross section . the main result of the present paper is ev b. the error consists of 8.4% experimental systematic error and the error due to the ambiguity in the s - wave-be scattering length . this value of differs from the one extracted with the first - order perturbation theory including coulomb breakup by dipole transitions : ev b. it turns out that the difference is due to the inclusion of the nuclear and coulomb - quadrupole transitions and multi - step processes of all - order in the present work . the p-be interaction potential used in the cdcc calculation is also used in the anc analysis ofbe()b . the value of ev b obtained is consistent with the previous one obtained from a precise measurement of the-capture reaction cross section extrapolated to zero incident energy , ( expt ) ( theo ) ev b , where ( theo ) stands for the error in the extrapolation . thus , the agreement between the values of obtained from directbe()b and indirectb - breakup measurements is significantly improved .

nucl-th0505007

the solar neutrino problem is one of the central issues in the neutrino physics @xcite . the current interpretation of the solar - neutrino deficit is the neutrino oscillation induced by the mass difference among @xmath12 , @xmath13 , and @xmath14 , and their mixing angles @xcite . attention is now focused on the determination of the oscillation parameters . the cross section of the @xmath7-capture reaction @xmath0be(@xmath1)@xmath2b at incident energy in the center - of - mass ( c.m . ) frame of the @xmath7-@xmath0be system @xmath15 plays an essential role in the solar - neutrino phenomenology , since the observed flux of @xmath2b neutrino is proportional to it ; the magnitude of the cross section @xmath16 is customarily expressed by the astrophysical factor @xmath17 $ ] , where @xmath18 is the sommerfeld parameter . the required accuracy of @xmath3 , to determine the neutrino oscillation parameters with sufficient accuracy , is the error within 5% @xcite . recently , precise measurement of @xmath16 was carried out by junghans _ et al . _ @xcite at energies of @xmath191162460 kev , which are low but still higher than stellar energies ( @xmath20 kev ) . extrapolating the measured @xmath21 to @xmath22 using a three - cluster model @xcite for @xmath2b structure , they derived @xmath10 ( expt ) @xmath11 ( theo ) ev b. the three - body model , however , did not simultaneously reproduce the magnitude and the energy - dependence of @xmath21 sufficiently well . moreover , as pointed out in ref . @xcite , the uncertainty of the s - wave @xmath7-@xmath0be scattering length ( with about 50% error ) prevents one from determining @xmath3 with very high accuracy . because of the difficulty of the direct measurement of @xmath16 at stellar energies , alternative indirect measurements were proposed . coulomb dissociation @xcite of @xmath2b is one of such indirect measurements . so far , extraction of @xmath3 from these experiments has been based on the virtual photon theory with the assumption of @xmath2b dissociation by virtual electric dipole ( e1 ) photon absorption . nuclear interaction and absorption of quadrupole ( e2)- and multi - photons were not taken into account . the value thus extracted from the riken experiment at 52 mev / nucleon was @xmath23 ev b @xcite . in the analysis of the msu experiment @xcite measured at 44 and 83 mev / nucleon , the e2 contribution to one - step process was estimated from the parallel - momentum - distribution of @xmath0be fragment and subtracted from the breakup spectrum of @xmath2b . as a result , the extracted @xmath3 was @xmath24 ev b , which is smaller than the value obtained at riken mentioned above . however , the analysis of the angular distribution of the @xmath2b breakup cross section of the riken experiment showed no contribution of e2 transitions . moreover , the recent experiment at rather high energy , 250 mev / nucleon , done at gsi @xcite showed that the e2 contribution was negligibly small ; the resulting @xmath3 was @xmath25 ev b. even though the significance of e2 transitions can be energy - dependent , the conclusions from the msu and riken measurements at similar energies look inconsistent and roles of the e2 component are still not clear . more seriously , there exists a non - negligible discrepancy of about 15% between the values of @xmath3 mentioned above that are derived from direct @xmath7-capture and indirect @xmath2b - breakup measurements . very recently , it was shown by a semiclassical calculation of @xmath4pb(@xmath2b , @xmath5be)@xmath4pb at 52 mev / nucleon that the discrepancy mentioned above was significantly reduced by taking account of nuclear - breakup components , e2 , and higher - order coulomb breakup processes @xcite . motivated by this result , we attempt in the present paper to extract a reliable value of @xmath3 analyzing the @xmath2b dissociation experiment measured at riken @xcite , with the method of continuum - discretized coupled - channels @xcite ( cdcc ) , assuming a @xmath7+@xmath26be+a three - body model of the system , where a stands for the target nucleus . the result of the analysis is then used to extract @xmath3 by means of the asymptotic normalization coefficient ( anc ) method @xcite . cdcc has been successful in describing various processes in which effects of projectile - breakup are essential . it has been successfully applied also to @xmath2b nuclear and coulomb breakup processes @xcite . in the cdcc calculation of the present work , we include all nuclear and coulomb breakup processes , and take account of the intrinsic spins of both @xmath7 and @xmath0be using the channel spin representation and the consistency of the @xmath7-@xmath0be interaction potential used in the cdcc calculation with the s - wave @xmath7-@xmath0be scattering length . the wave function of @xmath2b used in the cdcc calculation is found to be consistent with the measured cross section of @xmath0be(@xmath1)@xmath2b at low energies . we then use the calculated breakup cross section and the @xmath7-@xmath0be separation energy in the @xmath2b nucleus to obtain @xmath3 by the anc method . an important advantage of the anc method is that there is no restriction of the reaction mechanisms . in addition , the uncertainty of @xmath3 due to the use of the anc method can quantitatively be evaluated @xcite . the construction of the present paper is as follows . first , we describe how to extract the anc from the three - body model cdcc analysis of @xmath2b breakup . we show that under a condition , which is satisfied in the present analysis , the calculated cross section is proportional to the square of the anc for the ground state of @xmath2b . this is by no means trivial since the cdcc wave function contains not only the ground state of @xmath2b but also its continuum states . we describe in the later section the condition and the angular region of @xmath2b breakup in which it is satisfied . second , we determine @xmath3 from @xmath4pb(@xmath2b , @xmath5be)@xmath4pb at 52 mev / nucleon by the anc method , taking account of all order nuclear and coulomb breakup processes . it should be noted that in our cdcc calculation interference between nuclear- and coulomb - breakup amplitudes is explicitly included . we make use of the eikonal - cdcc method @xcite ( e - cdcc ) , with some modifications , to obtain the scattering amplitudes corresponding to large values of the orbital - angular - momentum @xmath27 of relative motion between the projectile and the target nucleus . we then construct a hybrid amplitude from the results of e - cdcc and cdcc . this makes it possible to carry out very accurate analyses with high computational speed @xcite . uncertainties of the extracted @xmath3 are carefully evaluated . finally , we discuss the contributions of nuclear breakup , e2 transitions , and higher - order processes on the extracted value of @xmath3 . the anc method is also applied to direct @xmath0be(@xmath1)@xmath2b measurement assuming the same @xmath7-@xmath0be potential in the calculation of the direct capture process as used in the cdcc analysis . we show the difference between the values of @xmath3 derived from direct and indirect measurements is essentially removed by the present analysis . in sec . [ anc ] we describe how to extract the anc from the cdcc analysis of @xmath2b breakup . formulation of modified e - cdcc and construction of the hybrid scattering amplitude are described in sec . [ e - cdcc ] . we analyze in sec . [ ress17 ] the @xmath2b breakup cross section measured at riken with cdcc and extract @xmath3 by the anc method . uncertainties of the extracted @xmath3 are quantitatively evaluated in sec . [ errors17 ] . in sec . [ discussion ] , the value of the extracted @xmath3 is compared with the result of the virtual photon theory and the roles of nuclear breakup , e2 transitions , and higher - order processes are discussed . we discuss in sec . [ direct ] the application of the anc method to precise direct measurement of @xmath0be(@xmath1)@xmath2b that confirms the consistency of its result with the @xmath2b breakup analysis . finally summary and conclusions are given in sec . [ summary ] . in this subsection we describe our method of extracting @xmath3 by @xmath7+@xmath0be+a three - body model cdcc analysis of elastic @xmath2b breakup reaction combined with the anc method @xcite . we recapitulate the anc method below , partly to introduce notations . at very low incident energies , the @xmath7-capture reaction @xmath0be(@xmath1)@xmath2b leading to the ground state of @xmath2b is extremely peripheral because of the coulomb barrier . the @xmath28-matrix element of the reaction , therefore , depends on the overlap between the ground state wave functions of @xmath0be and @xmath2b only in the tail region , where the radial part of the overlap has the following form @xmath29 where @xmath30 is the range of the nuclear interaction between @xmath7 and @xmath0be of about 4 fm , @xmath31 ( @xmath32 ) is the ground state wave function of @xmath0be ( @xmath2b ) , @xmath33 is the relative momentum in the unit of @xmath34 between @xmath7 and @xmath0be , @xmath35 stands for the internal coordinates of the seven - nucleon system , and @xmath36 is the whittaker function where @xmath37 is the orbital angular momentum of the @xmath7-@xmath0be relative motion in the ground state of @xmath2b ; @xmath38 in the present case . the constant @xmath39 in eq . ( [ 8bwf ] ) is the anc of our interest . the cross section @xmath40 , and consequently @xmath3 , are proportional to @xmath41 . thus , one can evaluate them once the anc , @xmath39 , is known . the cross section of any reaction in which a proton is transferred to or from @xmath2b is proportional to @xmath41 in dwba if the reaction is peripheral in the sense that it takes place when the transferred proton is distant from @xmath0be more than the range @xmath30 . based on this idea , dwba analyses of proton transfer reactions such as @xmath42be(@xmath0be,@xmath2b)@xmath43be @xcite and @xmath44n(@xmath0be,@xmath2b)@xmath45c @xcite have successfully been made to obtain the required anc . in the present paper we analyze the data of @xmath2b elastic breakup in which the fragments @xmath0be and @xmath7 are ejected into very forward angles . for brevity , we henceforth call such a process breakup into forward angles . since the breakup in such a case takes place at large distance @xmath46 of @xmath2b from the target nucleus , it is mostly due to coulomb interaction . the operator of the interaction is proportional to @xmath47 with @xmath48 . therefore , the dominant contribution to the @xmath28 matrix is from the region of @xmath49 larger than the range of nuclear interaction @xmath30 between the fragments . this is numerically confirmed as described in sec . [ errors17 ] . the reaction , therefore , is peripheral in the sense described above . thus , the anc can be obtained if its cross section is reproduced and is shown to be proportional to @xmath41 by theory . we show in the following that the cross section is well reproduced by cdcc and is in good approximation proportional to @xmath41 in the present case . we describe @xmath2b breakup at 52 mev / nucleon with the @xmath7+@xmath0be+a three - body model illustrated in fig . 1 . + @xmath0be+a three - body system , where a is a target.,scaledwidth=35.0% ] let us assume for simplicity that a is infinitely heavy . the triple differential cross section for ( @xmath2b,@xmath0be@xmath50 ) reaction is given by @xmath51 with a given constant @xmath52 and the three - body phase space factor @xmath53 @xcite . the @xmath28-matrix element of @xmath2b breakup is @xmath54 with @xmath55 where @xmath56 ( @xmath57 ) is the interaction ( distorting potential ) between two nuclei x and y , where the subscripts 1 , 7 , and a stand for @xmath7 , @xmath0be , and a , respectively . the function @xmath58 ( @xmath59 ) is the plane wave of relative motion between a and the outgoing @xmath7 ( @xmath0be ) , and @xmath60 is the exact wave function of the system with the incident plane wave of relative motion between @xmath2b and a in the initial channel . in the riken experiment that we analyze in the present work , emitted fragments were measured at forward angles with the @xmath0be in the ground state @xcite . we assume that @xmath0be stays in its ground state in the course of @xmath2b breakup . we have confirmed the validity of this assumption by a @xmath61he+@xmath62he+@xmath4pb three - body model cdcc calculation of the scattering of @xmath0be on @xmath4pb at 52 mev / nucleon . the cross sections of non - elastic processes including breakup of @xmath0be are less than about 1% of that of elastic scattering for the scattering angle of the c.m . of @xmath0be , @xmath63 , below about @xmath64 . in general @xmath63 can differ from the @xmath2b scattering angle @xmath65 in the @xmath2b@xmath66pb @xmath67be @xmath68pb process , but the difference between the two angles is very small , since the energy transfer in the @xmath2b breakup concerned is much smaller than the incident energy of @xmath2b . these facts imply that in @xmath69 one can neglect matrix elements of @xmath70 between the ground state and excited states of @xmath0be . one can , therefore , approximate @xmath70 by the single - folding potential @xmath71 similarly , @xmath72 can be approximated by @xmath73 in this approximation , @xmath69 is given by @xmath74 with @xmath75 if the incident momentum of the @xmath2b projectile is @xmath76 , one can make use of the well - known formula for @xmath60 and write @xmath77 where @xmath78 is the exact hamiltonian of the system @xmath79 where @xmath80 and @xmath81 are the kinetic energy operators with respect to @xmath82 and @xmath83 , respectively , and @xmath84 is the internal hamiltonian of @xmath0be . the total energy of the system is denoted by @xmath85 , which is related to the c.m . incident energy of @xmath2b @xmath86 by @xmath87 , where @xmath88 is the eigenenergy of @xmath31 and @xmath89 is the relative energy between @xmath7 and @xmath0be in the ground state of @xmath2b . under the assumption of the inert @xmath0be nucleus mentioned above , one can approximate @xmath78 in eq . ( [ omega ] ) by @xmath90 where @xmath91 is the @xmath7+@xmath0be+a three - body hamiltonian defined by @xmath92 consequently , @xmath93 where @xmath94 is m@xmath95ller s wave matrix with the outgoing boundary condition and @xmath96 is the overlap function between the ground state wave functions of @xmath0be and @xmath2b . as one sees from eq . ( [ tap ] ) , @xmath97 plays an essential role in the @xmath28-matrix element . its calculation , however , requires the wave functions of the two nuclei that are not readily available . we , therefore , substitute it by a constant @xmath98 times a physically plausible normalized function @xmath99 . in this work we take it to be a solution of @xmath100 where @xmath101 is a woods - saxon potential that is supposed to represent the single - particle potential of @xmath7 in @xmath2b . since the norm of @xmath97 is the spectroscopic factor of @xmath7 in @xmath2b , @xmath102 is its approximation if @xmath103 is close to @xmath104 . after the substitution of @xmath103 , the @xmath28-matrix element becomes @xmath105 where @xmath106 is the wave function of the total system with incident wave of @xmath2b consisting of a proton and an inert @xmath0be bound in a state @xmath103 impinging on the target with momentum @xmath76 . thus , in the particular case under consideration , the exact @xmath28-matrix element can approximately be calculated by three - body cdcc . furthermore , since the reaction is peripheral in @xmath49 as already mentioned , the radial part of @xmath103 in @xmath107 can be replaced by its asymptotic form @xmath108 thus , @xmath69 is approximately proportional to @xmath109 from the definitions of @xmath98 and @xmath110 , it is clear that @xmath39 is in good approximation the ratio of @xmath111 to @xmath112 , i.e. the anc of @xmath113 we look for . thus , one sees that the cross section is indeed proportional to @xmath41 . it should be noted that this proportionality holds only when the reaction is peripheral in @xmath49 , which is indeed the case for the @xmath2b breakup concerned . in practice , we calculate the cross section @xmath114 from @xmath107 for each choice of @xmath103 with the three - body model cdcc as in the previous work @xcite . we then obtain @xmath102 by taking the ratio of the measured cross section to @xmath114 . since the amplitude @xmath110 is readily known , the anc , @xmath39 , can then be calculated by eq . ( [ c ] ) . it should be noted that , since @xmath103 and @xmath97 may have quite different forms in the inner region , @xmath115 may be quite different from the true spectroscopic amplitude , @xmath116 , and depend strongly on the choice of @xmath117 . in contrast , @xmath39 given by eq . ( [ c ] ) is independent of @xmath103 and exactly the anc of @xmath97 if the @xmath2b breakup reaction is genuinely peripheral with no contribution of the inner part of the @xmath2b wave function . once @xmath39 is well determined from a peripheral reaction , @xmath3 can be obtained by @xcite @xmath118 where @xmath119 is the s - wave @xmath7-@xmath0be scattering length that can be calculated by the @xmath7-@xmath0be interaction potential used in the analysis . the accuracy of the @xmath3 extracted with the anc method can quantitatively evaluated by the dependence of @xmath39 on the choice of @xmath103 , as shown in sec . [ errors17 ] . thus , we see that @xmath3 can be extracted from the cdcc analysis of @xmath2b breakup based on the @xmath7+@xmath0be+a three - body model , if one concentrates on the @xmath2b breakup into forward angles , i.e. @xmath120 . in this subsection we describe the formalism of e - cdcc and how to construct a hybrid scattering amplitude from the results of e - cdcc and fully quantum - mechanical cdcc . the following formulation contains two important extensions of e - cdcc . one is the inclusion of the intrinsic spins of the projectile and its constituent fragments and the other is a more accurate treatment of the coulomb wave function in the coupled - channel equations than in the previous formulation of e - cdcc @xcite . detailed formalism of cdcc and its theoretical foundations are given elsewhere @xcite . as already mentioned in sec . [ anc ] , the total wave function @xmath121 of the @xmath7+@xmath0be+a three - body system satisfies the three - body schrdinger equation , @xmath122 in this model , the @xmath2b nucleus is assumed to consist of a @xmath0be core in its ground state and a proton moving around it . its intrinsic wave function , therefore , has the form @xmath123 where @xmath124 is the orbital angular momentum of relative motion between @xmath7 and @xmath0be , @xmath125 is the channel spin , and @xmath126 and @xmath127 are , respectively , the total angular momentum of @xmath2b and its @xmath128-component . the @xmath128-axis is taken to be parallel to the incident beam . the index @xmath129 stands for the @xmath129 th of the states with the quantum numbers \{@xmath124 , @xmath125 , @xmath126 , @xmath127}. for simplicity we denote below the five channel - indices \{@xmath129 , @xmath124 , @xmath125 , @xmath126 , @xmath127 } as @xmath130 . the set @xmath131 consists of bound - states and discretized - continuum - states as in refs . @xcite . for the calculation of the latter , the average method @xcite and the pseudo - state methods @xcite are widely used . in the present work , we assume that the @xmath132 satisfy @xmath133 where @xmath134 is the internal hamiltonian of @xmath2b given by @xmath135 , and @xmath136 is the intrinsic energy corresponding to the channel \{@xmath129 , @xmath124 , @xmath125 , @xmath126}. we take the following scheme of angular momentum coupling for @xmath137 @xmath138_{im},\end{aligned}\ ] ] where @xmath139 is the spherical harmonics and @xmath140 is the spin wave function of @xmath2b with channel spin @xmath125 . we expand @xmath121 in terms of the @xmath137 as @xmath141 where the coefficient @xmath142 $ ] of the expansion describes the c.m . motion of the projectile relative to a in channel @xmath130 , and @xmath143 is @xmath127 in the initial state . since the @xmath132 are chosen to form an approximate complete set in a finite region of space that is important for the reaction concerned @xcite , the expansion of eq . ( [ psinew2 ] ) is accurate in that region of space . unknown coefficients @xmath144 of the expansion are obtained by solving coupled - channel equations derived below . multiplying eq . ( [ sch ] ) by @xmath145 from the left , integrating over @xmath82 , and summing over the spin states , one obtains the coupled equations for the @xmath144 @xmath146 where @xmath147 is the reduced kinetic energy operator defined by @xmath148 + \dfrac{\hbar^2}{2\mu } \dfrac{(m'-m_0)^2}{b^2 } \label{tr}\ ] ] in the cylindrical coordinates @xmath149 . the impact parameter of the collision , @xmath150 , is defined by @xmath151 with @xmath152 in the cartesian coordinate . the two arguments of @xmath144 in eq . ( [ cceq ] ) are given by @xmath153 and @xmath154 . ( [ tr ] ) @xmath155 is the reduced mass of the @xmath2b - a system . in the peripheral region , in which @xmath150 is large , we neglect the last term of the r.h.s . of eq . ( [ tr ] ) , which turns out to have no effect on numerical results in the present case . thus , @xmath156 \equiv { \cal t}'_r . \label{kinetic}\ ] ] the form factor @xmath157 in eq . ( [ cceq ] ) of the coupling between channels @xmath130 and @xmath158 is given by @xmath159 where we assume @xmath160 and @xmath161 to be central potentials . this assumption is valid as discussed in ref . @xcite since we use e - cdcc only to obtain the scattering amplitude for large values of @xmath150 . the multipole expansion of @xmath162 is obtained from those of the coupling potentials @xmath163 then , @xmath162 can be expanded as @xmath164 where @xmath165 is the racah coefficient , @xmath166 and @xmath167 . now we make the coulomb - eikonal approximation @xcite to @xmath144 : @xmath168 where @xmath169 is the coulomb wave function with outgoing scattered wave and @xmath170 is the function to be determined with the eikonal approximation . the coulomb wave functions are given by @xmath171 with @xmath172 the gamma function and @xmath173 the confluent hypergeometric function . in actual calculation we make use of the approximate asymptotic - form of @xmath169 , @xmath174 which is valid for large values of @xmath150 . the boundary condition for @xmath170 in eq . ( [ eikonal2 ] ) is @xmath175 , where 0 denotes the incident channel , so that @xmath144 satisfies the appropriate boundary condition @xmath176 . we make the following two approximations as in ref . ( i ) for sufficiently large @xmath150 , the flux of @xmath169 is parallel to the @xmath128-axis . ( ii ) local semiclassical approximation @xcite to @xmath169 , @xmath177 which is valid for a wave function in a slowly - varying potential . the local momentum @xmath178 is defined by @xmath179 where @xmath180 ( @xmath181 ) represents the charge of @xmath2b ( a ) . further approximation to @xmath169 was made in ref . @xcite , i.e. the @xmath130-dependence of @xmath182 was neglected . in the present paper , however , we take account of that dependence using eq . ( [ coulwf2 ] ) . this treatment of the coulomb wave function slightly improves the agreement between the quantum - mechanical and eikonal scattering - amplitudes for large orbital angular momentum @xmath27 of @xmath2b - a relative motion . the explicit forms of the two amplitudes are given later in eqs . ( [ fc7])([fcq ] ) . with eq . ( [ kinetic ] ) and the coulomb - eikonal approximation , we obtain @xmath183 , \nonumber \\ & & \label{tchi}\end{aligned}\ ] ] where the second - order derivative of @xmath170 is neglected since it is slowly varying compared with @xmath169 . inserting eqs . ( [ ff ] ) , ( [ coulwf2 ] ) , ( [ tchi ] ) , and @xmath184 into eq . ( [ cceq ] ) , we arrive at the following coupled - channel equations @xmath185 where the non - dynamical variable @xmath150 is put in a superscript , the reduced coupling - potential @xmath186 is defined by @xmath187 and @xmath188 since eq . ( [ cceq4 ] ) is a set of first - order differential equations and contains no coefficient with very large angular momentum @xmath27 , one can solve it with high computational speed and accuracy . the eikonal scattering amplitude with coulomb distortion is given by @xmath189 , where @xmath190 is the rutherford amplitude and @xmath191 with @xmath192 the outgoing c.m . momentum @xmath193 is chosen to be on the @xmath128-@xmath194 plane , following the madison convention . for forward scattering in which @xmath195 is almost parallel to @xmath196 , i.e. to the @xmath128-axis , we approximate , in the same way as in ref . @xcite , the phase - factor @xmath197 , which appears in @xmath198 , by @xmath199 , where @xmath200 is the scattering angle of @xmath2b , and @xmath201 in eq . ( [ coul- ] ) by @xmath202 . under these approximations , @xmath203 with @xmath204 $ ] . the integration over @xmath128 in eq . ( [ fc2 ] ) can be done analytically with the help of eq . ( [ cceq4 ] ) , with the result @xmath205_{-\infty}^{\infty } \nonumber \\ & \equiv & \dfrac{i\hbar^2}{\mu}k_c \left [ { \cal s}_{c0}^{(b ) } - \delta_{c0 } \right ] , \label{cceq6}\end{aligned}\ ] ] where @xmath206 is neglected , since @xmath178 , with large @xmath150 , is a slowly - varying function of @xmath128 . the eikonal @xmath125-matrix elements are defined by @xmath207 . insertion of eq . ( [ cceq6 ] ) into eq . ( [ fc2 ] ) leads to @xmath208 . \label{fc3}\end{aligned}\ ] ] we now transform the integration over @xmath150 to the summation over @xmath27 as in ref . the resulting eikonal scattering amplitude is @xmath209 where @xmath210 \label{fc8}\ ] ] with @xmath211 . the quantum mechanical scattering amplitude obtained by cdcc is given by @xmath212 with @xmath213 where @xmath214 is the coulomb phase shift and @xmath215 is the total angular momentum of the three - body system . we define the hybrid scattering amplitude by @xmath216 and use it in the analysis of @xmath2b breakup . the connecting - angular - momentum @xmath217 between quantum mechanical and eikonal amplitudes is chosen so that @xmath218 agrees with @xmath219 for @xmath220 . it should be noted that eq . ( [ fh ] ) includes all quantum effects and also the interference between amplitudes in the two @xmath27-regions . this hybrid cdcc calculation turns out to be capable of evaluating the nuclear and coulomb breakup amplitudes with very high accuracy and computational speed . in fact , we have obtained the same accuracy as standard fully quantum mechanical cdcc calculations in about @xmath221 of computing time . for the wave functions of @xmath2b that form the basis of the cdcc expansion of the three - body system we make the following assumptions . we assume that the ground state wave function of @xmath2b , with the @xmath128-component @xmath143 of the total spin @xmath222 of it , is given by @xmath223_{2m_0 } + w_2 \left[i y_1(\hat{\bf r } ) \otimes \zeta_2 \right]_{2m_0 } \ } \nonumber \\ & & \times f_{\rm r}^{(0)}(r ) \nonumber \\ & \equiv&\ ; w_1 f_{1,m_0}^{(0)}({\bf r})+ w_2 f_{2,m_0}^{(0)}({\bf r } ) \label{f0}\end{aligned}\ ] ] in the channel spin representation , neglecting the @xmath224 components . the values of the coefficients @xmath225 and @xmath226 are obtained from the result of the microscopic calculation @xcite of @xmath2b , @xmath227 , where @xmath228 ( @xmath229 ) is the anc in the proton spin - orbit coupling representation corresponding to the @xmath230 ( 3/2 ) state . we further assume the effective @xmath7-@xmath0be interaction for the ground state of @xmath2b , @xmath231 , to be independent of @xmath125 , hence the same radial dependence @xmath232 of the @xmath233 and @xmath234 components in eq . ( [ f0 ] ) . this assumption is valid since only the tail region , @xmath235 , of @xmath103 contributes to the @xmath2b breakup process concerned . the wave function of the @xmath7-@xmath0be scattering state with outgoing scattered waves is given by @xmath236 where @xmath237 is the incident wave with definite channel spin @xmath125 and its @xmath128-component @xmath238 , @xmath239 is the incident energy in the c.m . frame of the @xmath7-@xmath0be system , and @xmath240 is the incident momentum . the wave function with incoming scattered waves @xmath241 corresponding to @xmath242 used below is defined by @xmath243 . we assume that @xmath244 has no tensor force component . we consider the breakup of @xmath2b to its scattering states at low @xmath239 , and truncate the modelspace by @xmath245 mev . in the experiment we analyze , however , the cross section is measured for @xmath246 in the range of 500 kev @xmath247 kev , and it turns out that only coupling of those states with @xmath239 much lower than 10mev need be taken into account as described in sec . [ ress17 ] , for which only s - waves are affected by the nuclear part of @xmath244 . all the higher partial waves have negligible amplitude inside its range because of the coulomb and centrifugal barriers and there have no coupling with the s - waves . therefore , all the angular momenta @xmath124 , @xmath125 and @xmath126 are conserved in @xmath248 . the channel spin @xmath125 is conserved also in s - waves since @xmath249 that is a good quantum number . thus , one sees that @xmath242 is the pure state of @xmath125 , while @xmath124 and @xmath126 are mixed as in the incident wave @xmath250 . it should be noted , however , that @xmath251 for the s - waves depends on @xmath125 as is evident from the large @xmath125-dependence of the s - wave @xmath7-@xmath0be scattering lengths : @xmath252 fm and @xmath253 fm @xcite . furthermore , it turns out that the spin - dependent part of the @xmath7-@xmath4pb optical potential has no effect on the resulting elastic cross section at forward angles . thus , the channel spin @xmath125 is conserved during the breakup process of @xmath2b by the @xmath4pb target @xcite . under these circumstances , the @xmath125-dependent 3-body @xmath28-matrix element in actual cdcc calculation , @xmath254 , is given by @xmath255 where @xmath256 is the wave matrix given in eq . ( [ wavematrix ] ) . the unpolarized @xmath2b breakup cross section calculated by cdcc , @xmath257 , is given by @xmath258 where @xmath259 ( @xmath260 ) is the @xmath128-component of the spin of @xmath7 ( @xmath0be ) and the orthonormality of the clebsch - gordan coefficient is used . we compare @xmath257 with the unpolarized cross sections measured by riken experiment @xcite . in the calculation of @xmath261 and @xmath262 , we use an @xmath126-independent @xmath7-@xmath0be effective interaction @xmath263 as mentioned above . then , the states of @xmath2b with fixed @xmath125 and @xmath124 and different values of @xmath126 are degenerate , which greatly simplifies numerical calculation . the value of @xmath119 in eq . ( [ s170 ] ) corresponding to the spin - dependent anc analysis is @xcite @xmath264 fm . for @xmath231 , the single - particle potential of esbensen and bertsch ( eb ) @xcite is used , except that we neglect the spin - orbit part of the potential and adjust the depth of the central potential to reproduce the separation energy of @xmath7 , 137 kev . the scattering states for the p - waves are also calculated with this @xmath231 . for d- and f - states with both @xmath233 and @xmath234 , we use the potential of barker @xcite . for the s - state in the @xmath234 channel , we use the potential of barker that gives the s - wave @xmath7-@xmath0be scattering length @xmath265 of @xmath266 fm . for the s - state with @xmath233 , we change the depth of the potential of barker to 25.7 mev so that the resulting scattering length @xmath267 is 25 fm . as for the distorting potential between @xmath7 ( @xmath0be ) and @xmath4pb we adopt the global optical potential by koning and delaroche @xcite ( cook @xcite ) . we neglect the spin - orbit parts of the @xmath7-@xmath4pb potential . the multipoles for nuclear and coulomb coupling - potentials are included up to @xmath268 . the discretized - continuum states of @xmath2b are constructed by the average method @xcite . the maximum excitation energy of @xmath2b is 10 mev and 10 , 20 , 10 , and 5 discretized - continuum states are taken for the @xmath269 , 0 , 2 , and 3 states , respectively . the resulting number of scattering - channels is 138 . the maximum values of @xmath49 , @xmath46 , and @xmath27 are , respectively , 200 fm , 1000 fm , and 12000 . the modelspace described above turns out to give good convergence of the resulting breakup cross sections . b breakup cross section integrated over @xmath239 in the range of 500750 kev , as a function of @xmath65 . the dotted , dash - dotted , and solid lines represent the results of the hybrid calculation of cdcc and e - cdcc with @xmath270 , 1000 , and 1500 , respectively . the result of e - cdcc , i.e. @xmath271 , is also shown by the dashed line . , scaledwidth=45.0% ] figure 2 shows the convergence of the hybrid calculation with cdcc and e - cdcc for the breakup cross section of @xmath2b by @xmath4pb at 52 mev / nucleon . the cross section integrated over the excitation energy of @xmath2b measured from the @xmath7+@xmath0be threshold energy , @xmath239 , from 500 kev to 750 kev is plotted as a function of @xmath65 . the dashed line is the result of the e - cdcc calculation ( @xmath271 ) and the dotted , dash - dotted , and solid lines correspond to the hybrid calculation with @xmath270 , 1000 , and 1500 , respectively . the figure shows that the hybrid calculation converges with @xmath272 . the difference between the dash - dotted and solid lines is only less than about 1% in magnitude . thus , we regard the hybrid calculation of cdcc and e - cdcc with @xmath272 as the fully quantum - mechanical cdcc calculation . b breakup cross sections with ( solid line ) and without ( dashed line ) the resolution and efficiency of the measurement @xcite taken into account . the experimental data are taken from ref . ( b ) decomposition of the smeared @xmath2b breakup cross section into the s- ( dashed line ) , p- ( dotted line ) , d- ( dash - dotted line ) , and f - state ( dash - two - dotted line ) breakup components . , scaledwidth=45.0% ] one sees in fig . 2 the oscillation of the cross section at forward angles around 2@xmath273 , which turns out to be due to the interference between nuclear and coulomb breakup - amplitudes . therefore , it can be concluded that nuclear interactions affect the breakup cross section even at very forward angles . however , as one sees below , this oscillation is not observed in actual experimental data because of the limit of the resolution of @xmath65 in the measurement . figure 3 shows the comparison between cdcc calculation and the experimental data for the @xmath2b breakup . in order to take account of the experimental resolution and efficiency , the theoretical result has been smeared out by using the filtering table @xcite provided by the authors of ref . the upper panel shows the filtering effect on the cross section calculated with cdcc , where the solid and dashed lines represent the results obtained with and without the smearing procedure . in the lower panel , the smeared total breakup cross section is decomposed into the s- ( dashed line ) , p- ( dotted line ) , d- ( dash - dotted line ) , and f - state ( dash - two - dotted line ) breakup components of @xmath2b . one sees the smeared result of cdcc well agrees with the experimental data at forward angles ( @xmath120 ) , while it underestimates the data at backward angles . the lower panel shows the importance of the p - state breakup for @xmath274 , which implies that nuclear breakup is significant in this angular region . thus , careful description of the p - state @xmath2b wave function , the resonance structures of it in particular , will be necessary to reproduce the experimental data at backward angles . the underestimation there may be due also to dynamical roles played by the excited - core component in @xmath2b , @xmath0be(1/2@xmath275)@xmath276(3/2@xmath275 ) , that is neglected in the present work as discussed in sec . although further study in this line will be very interesting , it is beyond the scope of the present paper . additionally , the filtering table used was made with assuming the s - state breakup of @xmath2b . thus , quantitative comparison between the calculation and the experimental data , to extract @xmath3 with high accuracy , can only be done in the region where the s - state breakup cross section is dominant . therefore , below we use the data for @xmath277 to determine @xmath278 by @xmath279 where @xmath257 is given by eq . ( [ sigcal ] ) and @xmath280 is the corresponding experimental data . -fit of the breakup cross section calculated with the @xmath2b single - particle potential of esbensen and bertsch ( eb ) @xcite . the solid line is the same as that in each panel in fig . 3 but multiplied by the spectroscopic factor @xmath281 . the result with @xmath282 is also shown by the dashed line for comparison . for the eight data points below @xmath64 , which are used in the fitting procedure , we have put the horizontal bar that represents the resolution of @xmath65 @xcite of the measurement . , scaledwidth=45.0% ] figure 4 shows the result of the @xmath283-fit of the theoretical calculation to the experimental data . the solid line is the same as that in each panel in fig . 3 but multiplied by the spectroscopic factor @xmath281 . the dashed line represents the result before being fitted , i.e. with @xmath282 . each horizontal bar put on the eight data points below @xmath64 does not represent a statistical error but it shows the range of @xmath65 in which the breakup cross sections contribute to each data point @xcite . the value of @xmath283 per datum obtained is 8.5 . although the quality of the fit is not good , it should be noted that estimation of error - propagation in the present case , i.e. comparison between smeared experimental data and a smeared numerical result , is very complicated ; the value of @xmath283 per datum shown above takes account of no error with respect to @xmath65 . the value of @xmath110 in eq . ( [ alpha ] ) for the eb @xmath2b wave - function is 0.704 @xmath284 . with the values of @xmath102 and @xmath110 , the anc @xmath39 is obtained from eq . ( [ c ] ) as @xmath285 @xmath284 . one can then determine @xmath3 by inserting this result into eq . ( [ s170 ] ) , together with @xmath286 fm , i.e. @xmath287 ev b. in the next subsection we quantitatively evaluate uncertainties of the extracted @xmath3 . first , we evaluate the uncertainty of @xmath3 that comes from the use of the anc method , just in the same way as in refs . we show in fig . 5 the result of cdcc calculation with the p - state single - particle potential of @xmath2b by kim _ et al . @xcite ( the dashed line ) and that multiplied by @xmath288 ( the solid line ) ; the latter agrees with the result shown in fig . 4 within 1% . one sees that @xmath102 indeed depends on the p - state potential . b ; the spectroscopic factor @xmath102 obtained is 0.867 . the result with kim s potential before @xmath283-fitting , namely with @xmath282 , is also shown by the dashed line for comparison . , scaledwidth=45.0% ] actually , @xmath110 in eq . ( [ alpha ] ) for kim s potential is 0.796 @xmath284 , which is quite different from the value for the eb potential , 0.704 @xmath284 , because of the difference in the geometry of the two potentials of about 20% . on the contrary , the value of @xmath39 in eq . ( [ c ] ) calculated with kim s potential is 0.741 @xmath284 that agrees very well with the result obtained with the eb potential shown in sec . [ ress17 ] . this result shows that the anc method works with very high accuracy in the present analysis , i.e. the error of the anc method is negligible . second , we estimate the effect of ambiguity of the distorting potentials used in the cdcc calculation on the obtained @xmath3 . for this purpose we use alternative optical potentials for @xmath7-@xmath4pb and @xmath0be-@xmath4pb . for the @xmath7-@xmath4pb potential , we modify the parameter set of koning and delaroche so that the calculated elastic cross section without spin - orbit terms reproduces the one obtained with the full components of the optical potential . the parameter set thus obtained is @xmath289 mev , @xmath290 mev , and @xmath291 mev , with the same notation as in ref . @xcite ; all other parameters are not changed . for the @xmath0be-@xmath4pb potential , we use a single - folding model with the nucleon-@xmath4pb potential of koning and delaroche , neglecting the spin - orbit terms . the density distribution of @xmath0be is assumed to be of the gaussian form with the range that reproduces the rms matter radius of @xmath0be of 2.48 fm @xcite . the difference between the results calculated with the original and alternative optical potentials is found to be negligibly small ( not shown ) . we conclude , therefore , that the error of @xmath3 that comes from the ambiguity of the distorting potentials in the present case is negligibly small . b breakup cross section on the s - state single - particle potentials of @xmath2b for the @xmath233 and @xmath234 channels . the solid line represents the result with the choice of the potentials that corresponds to the s - wave @xmath7-@xmath0be scattering lengths @xmath292 . the dashed ( dotted ) line shows the result with the s - state potentials corresponding to @xmath293 ( @xmath294 ) . , scaledwidth=45.0% ] finally , we estimate the uncertainty of @xmath3 due to the ambiguity of the s - state single - particle potentials of @xmath2b for the @xmath233 and @xmath234 channels by changing the depth of the potential for the @xmath233 ( @xmath234 ) channel ranging from 23.3 mev ( 52.8 mev ) to 30.5 mev ( 57.6 mev ) , which gives the range of the resulting s - wave @xmath7-@xmath0be scattering length @xmath267 ( @xmath265 ) from @xmath295 fm ( @xmath296 fm ) to @xmath297 fm ( @xmath298 fm ) ; these values of @xmath267 and @xmath265 are within the range of errors of the experimental values of them @xcite . in fig . 6 the solid , dashed , and dotted lines represent the results with the s - state potentials corresponding to @xmath299 , @xmath300 , and @xmath301 , respectively . one sees the breakup cross section slightly depends on the choice of the s - state potentials of @xmath2b ; the resulting @xmath3 is found to vary from 20.3 ev b to 21.9 ev b. thus , we evaluate the error of @xmath3 to be @xmath302/@xmath303 . summarizing the discussion given above , the result of the present paper is as follows . the central value of @xmath3 is 20.9 ev b. the theoretical error of the extracted @xmath3 concerned with the s - wave @xmath7-@xmath0be scattering length is @xmath302/@xmath303 . after including the 8.4% systematic experimental error , we obtain @xmath304 the main result of the present paper is @xmath305 ( theo ) @xmath306 ( expt ) ev b derived from @xmath4pb(@xmath2b , @xmath5be)@xmath4pb at 52 mev / nucleon . this value is significantly larger than @xmath23 ev b @xcite obtained in the previous analysis of the same experiment with the first - order perturbation theory . in order to clarify the reason for the difference , we discuss roles of nuclear interaction , e2 transitions , and higher - order processes . in fig . 7 the solid line is the same as in the upper panel in fig . 3 that shows the result of full cdcc . the dashed line corresponds to the calculation without breakup components of nuclear coupling potentials . it should be noted that their diagonal components are included in the calculation to take account of the absorption of the flux of the incident particle by the target nucleus . one sees that the dashed line agrees with the solid line below @xmath307 but deviates from it for @xmath308 . as a result , the value of @xmath3 obtained , @xmath309 ev b , is smaller than @xmath305 ( theo ) @xmath306 ( expt ) ev b obtained with the full cdcc calculation mentioned above . the dash - dotted line in fig . 7 represents the result of first - order iterative solutions of cdcc , designated as one - step cdcc , without nuclear breakup , including only @xmath310 ( e1 ) component of the coulomb interaction ; this calculation is essentially the same as the first - order perturbation theory ( virtual photon theory ) used in the previous analysis of the experimental data @xcite . it overestimates the solid line above 1.5@xmath273 and the resulting value of @xmath3 is @xmath311 ev b , which agrees well with the value @xmath8 ev b obtained in ref . @xcite . if one includes the @xmath312 ( e2 ) component in the one - step cdcc calculation of coulomb breakup , the dotted line in fig . 7 is obtained ; the coupling potentials with @xmath312 are artificially multiplied by 0.7 , following the analysis of the msu data @xcite . one sees from fig . 7 that inclusion of the reduced e2 component somewhat increases the breakup cross section , which results in further decrease of @xmath3 to @xmath313 ev b. this result is consistent with the conclusion of ref . @xcite , in which @xmath314 ev b was derived with first - order perturbation theory including both e1 and the reduced e2 components . if the e2 component is not scaled , the resulting value of @xmath3 is 16.7 ev b. this value is about 20% less than the result of full cdcc , 20.9 ev b. the difference is due to the nuclear and higher - order coulomb processes since the coupling potentials with @xmath315 are found to have little effect on the total breakup - cross - section at @xmath277 . thus , description of @xmath2b breakup process with nuclear and coulomb breakup processes of both the e1 and e2 transitions and higher - order processes is a key to solve a puzzle recognized so far of the discrepancy between the values of @xmath3 extracted from direct @xmath7-capture reactions , @xmath10 ( expt ) @xmath11 ( theo ) ev b , and indirect @xmath2b dissociation experiments . in order to clarify the role of these components in @xmath2b breakup reaction in general , it is necessary to carry out analyses to wider range of data , such as those measured at msu @xcite and gsi @xcite , including the data on other quantities than that dealt with in the present work . analysis of parallel momentum ( @xmath316 ) distribution of @xmath0be - fragment after breakup of @xmath2b is particularly important , since the role of e2 component was determined from it @xcite . in the analysis in ref . @xcite the result of first - order perturbation theory with an adjustable parameter was compared with cdcc calculation that is essentially the same as in the present paper . because of the agreement of the two results , it was concluded that higher order processes were unimportant without quantitative evaluation of the nuclear and higher - order coulomb processes as done in the present work . actually , the contribution is about 20% of the total as already described . a cdcc analysis of the @xmath316 distribution has been done by mortimer _ et al . _ @xcite , where the dependence of the calculated result on the strength of the coulomb @xmath312 ( e2 ) coupling potentials was discussed . they found that the e2 potentials must be multiplied by 1.6 to reproduce the experimental data , although the origin of the enhancement was not clear . therefore , further investigation of the accuracy of the enhancement factor 1.6 seems necessary . unfortunately , filtering table for the msu experiment is not available to the present authors . at this stage , therefore , we can quantitatively extract @xmath3 with cdcc and the anc method only from the @xmath2b breakup experiment done at riken . nevertheless , comparison of our pure theoretical result with that shown in ref . @xcite is very interesting , which will be a subject of our future work . in this subsection we report on our application of the anc method to the @xmath7-capture reaction , @xmath0be(@xmath1)@xmath2b , at energies of 116362 kev @xcite . we assume direct capture of @xmath7 by @xmath0be to the ground state of @xmath2b by a pure e1 transition . a simple potential - model is assumed for the bound and the scattering states of the @xmath7+@xmath0be system , using the single - particle potential of sec . [ input ] . the s- and d - state components in the initial scattering - state wave - function are taken into account , and the intrinsic spins of @xmath7 and @xmath0be are included in the channel spin representation described in sec . [ input ] . in fig . 8 the solid line shows the calculated @xmath317 as a function of the incident energy @xmath318 of the reaction in the c.m . frame of the @xmath7-@xmath0be system . the plotted @xmath21 has been multiplied by the spectroscopic factor @xmath319 obtained by the @xmath283-fit to the data below 362 kev . as a function of the incident energy @xmath318 of the reaction in the c.m . frame of the @xmath7-@xmath0be system . the solid line is the result of the calculation with a simple potential model , multiplied by @xmath319 . the experimental data are taken from ref . @xcite . , scaledwidth=43.0% ] the calculated @xmath21 is not reliable at high energies , say @xmath320 kev , since the inner part of the @xmath2b wave function plays an important role . at low energies , however , the anc method works very well and a reliable value of the anc @xmath39 , and consequently @xmath3 , can be derived from the @xmath283-fitting procedure . the value of @xmath3 thus obtained is 21.7 ev b. uncertainties of the @xmath3 are evaluated in the same way as in sec . [ errors17 ] and found to be @xmath321/@xmath322 ev b , almost all of which is due to the ambiguity of the s - state potentials for the @xmath233 and @xmath234 channels of the @xmath7-@xmath0be system . including the 2.3% systematic experimental error , we obtain the following result : @xmath323 ( theo ) @xmath324 ( expt ) ev b. this result is consistent with both the value extracted from @xmath2b dissociation , @xmath305 ( theo ) @xmath306 ( expt ) ev b , described in the previous subsections , and the result obtained in ref . @xcite , @xmath10 ( expt ) @xmath11 ( theo ) ev b , although no evaluation of the ambiguity due to the uncertainty of the s - wave @xmath7-@xmath0be scattering lengths , as the one described in sec . [ errors17 ] , is made in ref . the principal result of the present paper is the value of @xmath3 of @xmath325 ( theo ) @xmath306 ( expt ) ev b obtained by an analysis of the cross section of the @xmath2b breakup reaction @xmath4pb(@xmath2b , @xmath5be)@xmath4pb at 52 mev / nucleon measured at riken @xcite by means of the method of continuum - discretized coupled - channels @xcite ( cdcc ) combined with the asymptotic normalization coefficient ( anc ) method @xcite . the value is consistent with the one extracted from the precise measurement of the cross section of direct capture @xmath0be(@xmath1)@xmath2b , @xmath10 ( expt ) @xmath11 ( theo ) ev b @xcite . the cdcc calculation is based on the three - body model of @xmath7 , @xmath0be , and the target @xmath4pb nucleus , with @xmath0be staying in the ground state throughout the reaction process . this model is shown to be adequate in the breakup reaction in which the fragments @xmath7 and @xmath0be are ejected in very forward angles , less than 4@xmath273 , which is the case for the experimental data analyzed in this work . nuclear and coulomb e1 and e2 transitions and multi - step processes of all - order are included in the calculation . the calculated cross section is shown to be proportional to the squared asymptotic amplitude of the overlap of the ground state wave functions of @xmath0be and @xmath2b , which is a necessary condition for the applicability of the anc method . the cdcc calculation is made efficient by the use of the eikonal cdcc method @xcite ( e - cdcc ) for large angular momentum partial waves . improvements of e - cdcc are made to cope with the coulomb distortions of the incident and scattered waves of relative motion and intrinsic spins of the projectile and the ejectiles . the main source of the error in the obtained @xmath3 is the ambiguity of the s - wave @xmath7-@xmath0be interaction potentials for the channel spin @xmath326 and @xmath327 states . the errors associated with the use of the anc method and the ambiguity of the distorting potentials are both negligible . extracted by the anc method are shown by open symbols : the circles are obtained from three - types of cdcc analysis of @xmath2b breakup and the triangle is the result of analysis of @xmath0be(@xmath1)@xmath2b with a simple potential model . the error of each result is obtained by adding the theoretical and experimental errors in quadrature . the results are compared with @xmath3 extracted from the precise measurement of direct capture @xmath328 cross section @xcite ( closed triangle ) and those obtained from @xmath2b dissociation with first - order perturbation theory ( closed circles ) . , scaledwidth=48.0% ] calculations of @xmath3 with some simplified assumptions are summarized in fig . the results of first - order iterative solutions of cdcc , designated as one - step cdcc , correspond to and agree with those of first - order perturbation theory in the previous work @xcite . the inclusion of the coulomb quadrupole ( e2 ) transitions , scaled by 0.7 , in the one - step calculation decreases @xmath3 by about 6% , and multistep processes increase it by about 20% . this shows the crucial importance of accurate description of the @xmath2b breakup process by cdcc including nuclear and coulomb e1 and e2 transitions and all higher - order processes . it will be interesting to apply the method of the present paper to analyses of wider range of experiments such as those in refs . it may , however , be that the three - body cdcc used in the present paper is not valid in general and the use of cdcc with four - body model such as the one in ref . @xcite becomes necessary . the authors would like to thank t. motobayashi for helpful discussions and providing detailed information on the experiment . the authors also thank m. kawai for fruitful discussions and careful reading of the manuscript . this work has been supported in part by the grants - in - aid for scientific research of monbukagakusyou of japan . a. m. mukhamedzhanov and n. k. timofeyuk , yad . fiz . * 51 * , 679 ( 1990 ) [ sov . phys . * 51 * , 431 ( 1990 ) ] ; h. m. xu , c. a. gagliardi , r. e. tribble , a. m. mukhamedzhanov and n. k. timofeyuk , phys . lett . * 73 * , 2027 ( 1994 ) . n. austern , m. yahiro and m. kawai , phys . lett . * 63 * , 2649(1989 ) ; n. austern , m. kawai and m. yahiro , phys . c * 53 * , 314 ( 1996 ) ; r. a. d. piyadasa , m. kawai , m. kamimura and m. yahiro , phys . c * 60 * , 044611(1999 ) .

we present a numerical method to compute path integrals in effective su(2 ) yang - mills theories . the basic idea is to approximate the yang - mills path integral by summing over all gauge field configurations , which can be represented as a linear superposition of a small number of localized building blocks . with a suitable choice of building blocks many essential features of su(2 ) yang - mills theory can be reproduced , particularly confinement . the analysis of our results leads to the conclusion that topological charge as well as extended structures are essential elements of confining gauge field configurations .

hep-ph0608090

a common approach to get some insight regarding the mechanism of confinement is to approximate the yang - mills path integral by restricting the integration to certain subsets of gauge field configurations . by considering such effective theories one might find out , which gauge field configurations are responsible for confinement and which are not . examples are ensembles of singular gauge instantons ( c.f . e.g. @xcite ) , ensembles of calorons ( c.f . e.g. @xcite ) , ensembles of regular gauge instantons and ensembles of merons @xcite or the removal of center vortices in lattice calculations ( c.f . e.g. @xcite ) . some of these approaches have analytical aspects but most of them extensively resort to numerical methods . in this work we generalize the ideas and techniques presented in @xcite . we study different classes of gauge field configurations , especially their importance with regard to confinement , by applying a numerical method called pseudoparticle approach @xcite . we demonstrate that with a suitable choice of building blocks many essential features of su(2 ) yang - mills theory can be reproduced , particularly confinement . this paper is organized as follows . in section [ sec_001 ] we explain the basic principle of the pseudoparticle approach . we also discuss numerical issues and compare the method to lattice gauge theory . in section [ sec_003 ] we show how to calculate certain observables in the pseudoparticle approach : the static quark antiquark potential at zero and at finite temperature , the topological susceptibility and the critical temperature of the confinement deconfinement phase transition . we present numerical results , which are in qualitative agreement with lattice results . in section [ sec_007 ] we apply the pseudoparticle approach with different types of building blocks , to study different classes of gauge field configurations and their effect on confinement . by doing that we try to determine properties of confining gauge field configurations . our findings indicate that topological charge as well as extended structures are necessary to obtain confinement . in section [ sec_011 ] we summarize our results and give a brief outlook regarding future research . in this work we consider euclidean su(2 ) yang - mills theory . the action is given by @xmath0 with action density @xmath1 ( @xmath2 is the dimensionless coupling constant ) and field strength @xmath3 ( @xmath4 ) . the topological charge is defined by @xmath5 with topological charge density @xmath6 and dual field strength @xmath7 ( @xmath8 ) . the basic idea of the pseudoparticle approach is to restrict the yang - mills path integral to those gauge field configurations , which can be represented as a linear superposition of a small number of building blocks , typically around @xmath9 . the building blocks are gauge field configurations , which are localized in euclidean spacetime . we refer to these building blocks as pseudoparticles , where the term pseudoparticle refers to the fact that the corresponding gauge field configurations are also localized in time in contrast to ordinary particles , which are only localized in space . in our context a pseudoparticle is not necessarily a solution of the classical yang - mills equations of motion . every pseudoparticle or building block has certain parameters , a position , an amplitude and a color orientation , which uniquely define the resulting gauge field configuration . it is important to stress that in general a sum of such pseudoparticles is not even close to a classical solution , as it is the case e.g. in instanton gas models ( c.f . e.g. @xcite ) . the pseudoparticle approach is supposed to describe full quantum physics and not only certain semiclassical corrections . the integration over all gauge field configurations in the path integral is replaced by an integration over the amplitudes and color orientations of the pseudoparticles . when considering a spacetime region containing a finite number of pseudoparticles , this is an ordinary multidimensional integral , which can be computed by monte - carlo methods . the starting point of this work has been @xcite . however , there are two important generalizations : * we do not restrict our approach to regular gauge instantons and merons . a pseudoparticle can be any localized gauge field configuration . for example we also employ pseudoparticles with a limited range of interaction or without topological charge . * in addition to a color orientation matrix we assign to each pseudoparticle a variable amplitude . in this way pseudoparticles are also able to model small quantum fluctuations . note that the pseudoparticle approach is a general technique , which is in no way restricted to su(2 ) yang - mills theory . with minor modifications it can also be applied to other quantum field theories , e.g. quantum mechanics ( c.f . @xcite ) . for the major part of this work we consider the following pseudoparticles : @xmath10 where @xmath11 and + @xmath12 . each pseudoparticle has an index @xmath13 , an amplitude @xmath14 , a color orientation matrix @xmath15 , a position @xmath16 and a size @xmath17 . when considering a single pseudoparticle , a color orientation matrix is equivalent to a global gauge transformation . since such a global gauge transformation can be specified by an element of su(2 ) , for which @xmath18 is a suitable parameter space , it can be expressed in terms of @xmath19 , i.e. @xmath20 : @xmath21 ( c.f . appendix [ sec_012 ] ) . note that euclidean lorentz transformations are equivalent to color rotations , when considering a single instanton or antiinstanton @xcite , while a single akyron is form invariant under such lorentz transformations . therefore euclidean lorentz transformations have been included by considering arbitrary color orientations . setting @xmath22 in ( [ eqn_001 ] ) yields an instanton in regular gauge @xcite . although for @xmath23 such pseudoparticles are not actually instantons , we nevertheless refer to them by that term . for @xmath22 the action of an instanton is @xmath24 , otherwise it is @xmath25 . the topological charge is given by @xmath26 with exception of a sign reversal in ( [ eqn_004 ] ) the same applies for antiinstantons ( [ eqn_002 ] ) . a single akyron pure gauge ( literally `` without effect '' ) . ] is a pure gauge , i.e. @xmath27 and @xmath28 . note that for linear superpositions of akyrons @xmath29 in general , due to the non - abelian nature of su(2 ) . however , any such superposition has vanishing topological charge density ( c.f . appendix [ sec_013 ] ) . a common and essential property of instantons , antiinstantons and akyrons is their long range nature . for large @xmath30 the corresponding gauge fields decrease like @xmath31 . as a consequence these pseudoparticles have the ability to interact over large distances . an important reason for considering pseudoparticles ( [ eqn_001 ] ) and ( [ eqn_002 ] ) is their similarity to regular gauge instantons and merons , which are known to exhibit confinement @xcite . we additionally include akyrons ( [ eqn_003 ] ) so that the gauge field has both a transverse part and a longitudinal part ( superpositions of instantons and antiinstantons form transverse gauge fields , whereas superpositions of akyrons form longitudinal gauge fields ; c.f . appendix [ sec_13 ] ) . furthermore , one can show that in the continuum limit , i.e. in the limit of infinitely many pseudoparticles , instantons ( or antiinstantons ) and akyrons form a basis of all gauge field configurations ( c.f . appendix [ sec_014 ] ) . finally , numerical studies with different types of pseudoparticles have shown that observables are not very sensitive to moderate changes in the definition of the pseudoparticles . for example replacing @xmath32 by @xmath33 in ( [ eqn_001 ] ) to ( [ eqn_003 ] ) hardly affects numerical results , if @xmath34 is sufficiently large ( c.f . section [ sec_009 ] ) . it seems that results in the pseudoparticle approach mainly depend on certain `` global pseudoparticle properties '' , like their ability to interact over sufficiently large distances or whether they carry topological charge or not . we put @xmath35 pseudoparticles with randomly and uniformly chosen positions inside a hyperspherical spacetime volume ( c.f . [ fig_001 ] ) . a 4-dimensional spacetime hypersphere of radius @xmath36 with @xmath37 pseudoparticles ( pseudoparticles are symbolized by dark gray circles ) . ] in the following we denote the radius of this spacetime hypersphere by @xmath36 and its volume by @xmath38 . these quantities are related according to @xmath39 . the pseudoparticle density is given by @xmath40 . for most calculations we consider @xmath41 pseudoparticles ( [ eqn_001 ] ) to ( [ eqn_003 ] ) with size @xmath42 and pseudoparticle density @xmath43 . this amounts to @xmath44 . the gauge field is a linear superposition of instantons , antiinstantons and akyrons : @xmath45 ( the indices @xmath13 , @xmath46 and @xmath47 assume values from different ranges ) . in accordance with the ratio of transverse and longitudinal gauge field components , which is @xmath48 , we choose @xmath49 + ( @xmath50 , @xmath51 and @xmath52 are the corresponding pseudoparticle numbers ) . although in the limit of infinitely many pseudoparticles instantons by themselves form a basis of all transverse gauge fields , i.e. applying instantons and antiinstantons would yield an overcomplete basis , we consider an equal number of instantons and antiinstantons , when using a finite number of pseudoparticles . our ensembles are then symmetric with respect to topological charge ( c.f . ( [ eqn_004 ] ) ) . we define the ensemble average of a quantity @xmath53 by @xmath54 \\ \nonumber & & \hspace{0.62 cm } e^{-s[a(\mathcal{a}(i),\mathcal{c}(i ) ) ] } \quad , \\ \label{eqn_006 } & & z \ \ \int \left(\prod_i d\mathcal{a}(i ) \ , d\mathcal{c}(i)\right ) e^{-s[a(\mathcal{a}(i),\mathcal{c}(i ) ) ] } \\ \label{eqn_007 } & & \hspace{-0.44 cm } s[a(\mathcal{a}(i),\mathcal{c}(i ) ) ] \ \ = \ \ \int_{v_\textrm{spacetime } } d^4x \ , s(a(\mathcal{a}(i),\mathcal{c}(i)))\end{aligned}\ ] ] ( @xmath55 is the invariant integration measure on @xmath18 ) , i.e. the integration over all gauge field configurations in the path integral is replaced by an integration over the amplitudes and color orientations of the pseudoparticles . the quantity @xmath53 , the action @xmath56 and the action density @xmath57 can be expressed in terms of @xmath58 and @xmath59 via ( [ eqn_005 ] ) . only the action inside the spacetime hypersphere is considered for such a `` path integral '' . to eliminate the dependence of the ensemble average @xmath60 on the randomly chosen pseudoparticle positions @xmath61 , we independently calculate @xmath60 for many different sets of positions and take the average . instead of integrating over amplitudes and color orientations as in ( [ eqn_006 ] ) one could also integrate over positions or only over amplitudes . however , we have found that numerical results of such computations are very similar ( c.f . @xcite ) . note that our pseudoparticle ensembles are lorentz invariant with respect to the center of the spacetime hypersphere as well as globally gauge invariant . ensemble averages @xmath60 in the pseudoparticle approach are given by multidimensional integrals ( c.f . ( [ eqn_006 ] ) ) , which can be computed by monte - carlo methods . we have applied the metropolis algorithm ( c.f . e.g. @xcite ) . in a single metropolis step the pseudoparticles are updated one by one in fixed order . when updating a pseudoparticle , its amplitude and color orientation are randomly changed . these changes are always accepted if they reduce the action . otherwise they may be rejected depending on the outcome of a stochastic experiment . the action inside the spacetime hypersphere ( [ eqn_007 ] ) is approximated by standard monte - carlo sampling . to get rid of unwanted boundary effects , samples of physically meaningful quantities are always taken inside a hyperspherical spacetime region of radius + @xmath62 ( c.f . [ fig_001 ] ) . we determine + @xmath63 by considering plots of the average action density @xmath64 against the distance to the center of the spacetime hypersphere : @xmath63 is assigned that distance , where @xmath64 starts to deviate significantly from its constant value near the center of the spacetime hypersphere . in the example shown in fig . [ fig_002 ] we have chosen @xmath65 . numerical calculations have shown that pseudoparticle results are pretty stable , when applying between @xmath66 and @xmath67 pseudoparticles ( c.f . @xcite ) . using significantly less than @xmath66 pseudoparticles is usually not possible , because to extract physically meaningful results one requires a sufficiently large spacetime region , where boundary effects are negligible . @xmath41 , @xmath43 , @xmath42 , @xmath68 . @xmath64 plotted against the distance to the center of the spacetime hypersphere . ] in many aspects the pseudoparticle approach is similar to lattice gauge theory . both approaches approximate path integrals by integrating over a finite number of degrees of freedom . in lattice gauge theory the integration is performed over a fixed number of link variables , while in the pseudoparticle approach there is a fixed number of pseudoparticles with amplitudes and color orientations as degrees of freedom . in both cases the scale can be set by identifying the numerical value of any dimensionful quantity with its physical / experimental value . when using the string tension @xmath69 , any quantity @xmath53 with dimension @xmath70 is given in physical units by @xmath71 where @xmath53 and @xmath69 are the numerical dimensionless values of @xmath53 and @xmath69 at coupling constant @xmath2 and @xmath72 is the value of the string tension in physical units ( throughout this work @xmath73 ) . a crucial property of any trustworthy numerical method is that dimensionless ratios of dimensionful quantities , e.g. @xmath74 or @xmath75 , do not depend on the coupling constant , i.e. that dimensionful quantities scale consistently . this has been observed both in lattice calculations and in the pseudoparticle approach ( c.f . section [ sec_003 ] ) . in lattice gauge theory the ultraviolet regulator is the lattice spacing @xmath76 . since different values for @xmath2 yield different values for @xmath69 , the lattice spacing in physical units @xmath77 can be adjusted by choosing appropriate values for the coupling constant ( replace @xmath78 by @xmath77 and set @xmath79 and @xmath80 in ( [ eqn_008 ] ) ) . in the pseudoparticle approach the minimum size of ultraviolet fluctuations is determined by the average pseudoparticle distance @xmath81 and the pseudoparticle size @xmath34 , i.e. there are two ultraviolet regulators . a variation of the coupling constant @xmath2 has a similar effect on these regulators as it has in lattice calculations on the lattice spacing , that is @xmath82 and @xmath83 are increasing functions of @xmath2 ( c.f . [ fig_003 ] ; the scale has been set by the string tension , which we have obtained via generalized creutz ratios ( c.f . section [ sec_004 ] ) ) . @xmath41 , @xmath43 , @xmath42 . @xmath82 in @xmath84 and @xmath85 @xmath83 in @xmath84 plotted against @xmath2 . ] in contrast to lattice calculations we additionally have to specify the ratio @xmath86 . for our standard choice of parameters , @xmath87 and @xmath42 , the ultraviolet regulators are of the same order of magnitude . numerical calculations have shown that physical quantities are pretty stable with respect to a variation of @xmath86 in the range @xmath88 and that there is qualitative agreement with lattice results for suitably chosen pseudoparticles ( c.f . section [ sec_008 ] ) . when considering the limit of infinitely many pseudoparticles , any gauge field configuration can be represented by a suitable superposition of instantons and akyrons ( c.f . section [ sec_014 ] ) , i.e. in this limit our pseudoparticle ensembles are gauge invariant . however , unlike in lattice gauge theory we only have approximate gauge invariance when working with a finite number of degrees of freedom . whether this approximate gauge invariance is a problem can be tested by numerical calculations . a suitable way of doing that is to fix the gauge and to perform computations similar to those without gauge fixing . in the pseudoparticle approach a particularly convenient gauge is lorentz gauge , i.e. @xmath89 , which amounts to considering only instantons and antiinstantons but no akyrons . the qualitative agreement between these results and results without gauge fixing ( c.f . section [ sec_010 ] ) indicates that in the pseudoparticle approach the lack of exact gauge invariance is not a problem . in contrast to lattice gauge theory we always work in the continuum , i.e. we do not discretize spacetime . this might be an advantage when calculating glueball masses from correlation functions , because due to exact rotational symmetry it is possible to project out states with a well defined angular momentum . furthermore , the pseudoparticle approach has a high potential when dealing with fermionic problems , because there is no fermion doubling in such a continuum formulation . another difference lies in the number of degrees of freedom , which is significantly smaller in the pseudoparticle approach than it is in typical lattice calculations ( @xmath90 degrees of freedom , when using @xmath9 pseudoparticles , compared to e.g. @xmath91 degrees of freedom on a `` small '' @xmath92-lattice ) . therefore , when pseudoparticle results are in agreement with results from lattice calculations , the pseudoparticles are chosen such that they represent essential degrees of freedom of su(2 ) yang - mills theory . to put it another way , by applying different types of pseudoparticles one can determine the effect of certain classes of gauge field configurations on certain observables , and by doing that try to find out essential properties of confining gauge field configurations ( c.f . section [ sec_007 ] ) lattice calculations , on the other hand , are much faster , i.e. they can cope with a significantly larger number of degrees of freedom in comparable computation time . therefore , lattice calculations are certainly better suited for producing high quality numerical results . the common tool for studying the potential of a pair of infinitely heavy quarks at zero temperature are wilson loops . a wilson loop @xmath93 is defined by @xmath94 \ \ \frac{1}{2 } \textrm{tr}\left(p\left\{\exp\left(i \oint dz_\mu \ , a_\mu(z)\right)\right\}\right ) , \end{aligned}\ ] ] where @xmath95 is a closed spacetime curve and @xmath96 denotes path ordering . rectangular wilson loops with spatial extension @xmath97 and temporal extension @xmath98 are denoted by @xmath99 . it is well known that the potential of a static quark antiquark pair @xmath100 with separation @xmath97 can be related to ensemble averages of rectangular wilson loops according to @xmath101 ( c.f . e.g. @xcite ) . in the following we assume that for large separations the static quark antiquark potential can be parameterized by @xmath102 this ansatz is based on the bosonic string picture @xcite and on various numerical results from lattice calculations ( c.f . e.g. @xcite ) . there are three parameters : * @xmath103 is a constant shift of the potential without physical relevance . * @xmath104 is the coefficient in front of the attractive @xmath105-correction of the potential for large separations . the bosonic string picture predicts + @xmath106 @xcite . in lattice calculations @xmath107 has been observed @xcite . we refer to @xmath104 as coulomb coefficient . * the string tension @xmath69 characterizes the force between a static quark and a static antiquark at large separations . lattice calculations yield a positive value of the string tension , which implies the presence of confinement ( c.f . e.g. @xcite ) . furthermore , @xmath69 is a monotonically increasing function of the coupling constant @xmath2 , i.e. when the scale is set by the string tension , the extension of the lattice in physical units can be adjusted by choosing appropriate values for @xmath2 . we determine the string tension @xmath69 and the coulomb coefficient @xmath104 by generalizing the well known method of creutz ratios @xcite . generalized creutz ratios are based on guessing the functional dependence of ensemble averages of rectangular wilson loops . a possible candidate is given by @xmath108 @xcite . it is a simple and plausible choice , which is consistent with numerical results from lattice calculations . it also fulfills the following necessary requirements : * @xmath109 ( in euclidean spacetime there is no difference between space and time ) . * @xmath110 ( c.f . ( [ eqn_009 ] ) and ( [ eqn_010 ] ) ) . note that ( [ eqn_011 ] ) is only valid for @xmath111 . wilson loops , which are significantly smaller than the two ultraviolet regulators @xmath112 and @xmath34 , are subject to strong cutoff effects . when applying generalized creutz ratios , the starting point is a set of ensemble averages of rectangular wilson loops , @xmath113 with at least a couple of different ratios @xmath114 . a generalized creutz ratio @xmath115 is defined by @xmath116 with weights @xmath117 + @xmath118 , which will be specified below ( @xmath119 denotes any of the constants @xmath103 , @xmath104 , @xmath120 or @xmath69 ) . inserting the wilson loop ansatz ( [ eqn_011 ] ) yields @xmath121 by solving a linear system the weights @xmath122 are chosen such that ( [ eqn_013 ] ) reduces to @xmath123 to determine @xmath103 , @xmath104 , @xmath120 and @xmath69 , we consider all four element subsets of ( [ eqn_012 ] ) . for every subset we calculate @xmath103 , @xmath104 , @xmath120 and @xmath69 via ( [ eqn_014 ] ) . although there will be certain fluctuations due to systematical and statistical errors , most of these estimates should be pretty similar . we have found that estimates for @xmath103 , @xmath104 , @xmath120 and @xmath69 with large @xmath124 exhibit significantly stronger fluctuations than estimates with small @xmath124 ( c.f . [ fig_004 ] ) . therefore , we sort the estimates according to @xmath124 and only keep the `` smaller half '' . the average of these estimates is the final result for @xmath103 , @xmath104 , @xmath120 or @xmath69 . we have computed ensemble averages of rectangular wilson loops @xmath125 for @xmath126 , @xmath127 , @xmath65 , and have determined all possible generalized creutz ratios . there are around @xmath128 generalized creutz ratios for each of the quantities @xmath103 , @xmath104 , @xmath120 and @xmath69 . @xmath41 , @xmath43 , @xmath42 , @xmath129 . * a ) * estimates for @xmath69 sorted according to @xmath124 . * b ) * estimates for @xmath104 sorted according to @xmath124 . ] in fig . [ fig_004 ] the corresponding estimates for the string tension @xmath69 and the coulomb coefficient @xmath104 are plotted for @xmath129 , sorted according to @xmath124 in increasing order from left to right ( for the sake of clarity only every tenth estimate is shown ) . both plots demonstrate that estimates on the left exhibit a smaller variance than estimates on the right . this indicates that estimates with small @xmath124 are more reliable than estimates with large @xmath124 . to extract numerical values for @xmath69 and @xmath104 , we have calculated the average and the standard deviation from the `` smaller half '' . the results are @xmath130 and @xmath131 . by proceeding in the same way we have obtained numerical values for @xmath69 and @xmath104 for different coupling constants @xmath132 . in fig . [ fig_005]a the string tension @xmath69 is plotted against the coupling constant @xmath2 . it is positive and an increasing function of @xmath2 . when the scale is set by the string tension , the size of the spacetime hypersphere in physical units can be adjusted by choosing appropriate values for the coupling constant . for @xmath73 its diameter ranges from approximately @xmath133 at @xmath134 to @xmath135 at + @xmath136 . [ fig_005]b shows the coulomb coefficient @xmath104 as a function of the coupling constant @xmath2 . it increases from @xmath137 at @xmath134 to @xmath138 at @xmath136 . for large @xmath2 the value of @xmath104 is in qualitative agreement with the prediction from the bosonic string picture , @xmath106 @xcite , and with results from lattice calculations , + @xmath107 @xcite . @xmath41 , @xmath43 , @xmath42 . * a ) * @xmath69 plotted against @xmath2 . * b ) * @xmath104 plotted against @xmath2 . ] the starting point to calculate the static quark antiquark potential is ( [ eqn_009 ] ) , which we write as @xmath139 we compute ensemble averages of rectangular wilson loops @xmath125 for fixed @xmath97 but different @xmath98 to obtain a curve @xmath140 . @xmath41 , @xmath43 , @xmath42 , @xmath129 . + * a ) * @xmath141 for different @xmath97 plotted against @xmath98 @xmath142 ( @xmath127 ) . * b ) * @xmath100 plotted against the separation of the quarks . ] examples are shown in fig . [ fig_006]a . according to ( [ eqn_015 ] ) such a curve will exhibit a linear behavior for sufficiently large @xmath98 . from the slope , which we obtain by fitting a straight line , we can read off @xmath143 . iterating this procedure for a number of different @xmath97 yields an approximation of the quark antiquark potential . we have computed ensemble averages of rectangular wilson loops @xmath125 for @xmath144 , + @xmath127 , @xmath65 . [ fig_006]a shows @xmath140 as a function of @xmath98 for different @xmath145 . to determine the slope of these curves for large @xmath98 , we have fitted straight lines to the data points at @xmath146 , as indicated by the dashed straight lines . the corresponding potential as a function of the quark antiquark separation is plotted in fig . [ fig_006]b . for large separations it clearly exhibits a linear behavior . to obtain numerical values for the string tension @xmath69 and the coulomb coefficient @xmath104 , we have performed a least squares fit of the potential parameterization ( [ eqn_010 ] ) to the data points shown in fig . [ fig_006]b . only data points with @xmath147 have been considered , because cutoff effects are expected to render the potential unphysical for small separations . within statistical errors the results , @xmath148 and @xmath149 , are in agreement with the results obtained by generalized creutz ratios ( c.f . section [ sec_004 ] ) . the fit is also shown in fig . [ fig_006]b . note that in contrast to generalized creutz ratios we have not made any assumption about the functional dependence of ensemble averages of wilson loops to calculate the quark antiquark potential . therefore , the agreement of the results for @xmath69 and @xmath104 with results obtained by generalized creutz ratios shows again the consistency of the wilson loop ansatz ( [ eqn_011 ] ) and monte - carlo data for @xmath125 . in order to produce quantitative results involving the string tension , we need other dimensionful quantities so that we can consider dimensionless ratios . one such quantity , which has been studied extensively on the lattice ( c.f . e.g. @xcite ) , is the topological susceptibility @xmath150 . the topological susceptibility is closely related to the mass of the @xmath151 meson @xcite . it is defined by @xmath152 where @xmath153 is the topological charge inside the spacetime volume @xmath154 . in our numerical calculations we approximate the limit @xmath155 by a finite volume , the hyperspherical spacetime region with radius @xmath63 ( c.f . section [ sec_002 ] ) . a number of computations has shown that @xmath150 is stable with respect to a a variation of that spacetime volume . [ fig_007 ] shows the dimensionless ratio @xmath156 as a function of the coupling constant @xmath2 ( @xmath65 , @xmath69 has been obtained by generalized creutz ratios ( c.f . [ fig_005]a ) ) . as expected , @xmath156 is nearly independent of @xmath2 , i.e. the string tension and the topological susceptibility exhibit consistent scaling behaviors with respect to the coupling constant . this success strongly indicates that the pseudoparticle approach has the potential to reproduce correct yang - mills physics . the range of values , @xmath157 , is of the right order of magnitude compared to the lattice result + @xmath158 @xcite . @xmath41 , @xmath43 , @xmath42 . @xmath156 plotted against @xmath2 . ] for finite temperature calculations we have to consider periodic gauge field configurations in @xmath159-direction . therefore we need a method to generate periodic building blocks . in the case of singular gauge instantons this has already been done in form of calorons @xcite , which are periodic solutions to the classical yang - mills equations of motion . however , there is no straightforward generalization to arbitrary pseudoparticles . therefore , we introduce a different method , which resorts to a blending technique from computer aided geometric design @xcite . at first , we define blending functions @xmath160 and @xmath161 , which form smooth , @xmath162-continuous connections between @xmath163 at @xmath164 and @xmath165 at @xmath166 and vice versa : @xmath167 to make a non - periodic pseudoparticle @xmath168 with its center at the origin periodic in @xmath159-direction , we multiply `` both ends '' with the blending functions and add the results ( c.f . fig . [ fig_008 ] ) : @xmath169 to evaluate @xmath170 at @xmath171 $ ] one just has to combine ( [ eqn_016 ] ) and @xmath172 with a suitably chosen integer @xmath173 . @xmath174 is the width of the blending region . throughout this work we have chosen @xmath175 . the basic principle of the blending method . ] at zero temperature we consider a spacetime region , which is the interior of a @xmath176-dimensional hypersphere ( c.f . fig . [ fig_001 ] ) . at finite temperature that spacetime region is replaced by a spacetime with a periodic time direction of extension @xmath120 and a spatial part , which is the interior of an ordinary @xmath177-dimensional sphere of radius @xmath178 ( c.f . [ fig_009 ] ) . as in the zero temperature case we have to assure that samples of physically meaningful quantities are always taken inside a spacetime region , where boundary effects can be neglected . the spatial part of such a region is the interior of a sphere of radius @xmath63 , whereas the time direction is not restricted ( the light gray region in fig . [ fig_009 ] ) . the spacetime region at finite temperature . ] at finite temperature we use the same pseudoparticle density as for the corresponding zero temperature calculation ( a zero temperature calculation with identical parameters is necessary to obtain a numerical value for the string tension , which is used to set the scale ) . this is a reasonable choice , because in the limit @xmath179 , + @xmath180 and @xmath181 finite temperature ensembles and zero temperature ensembles become identical . furthermore , there is a close analogy to lattice calculations . to set the scale , a zero temperature calculation with the same number of lattice sites in all four spacetime directions is carried out . for finite temperature calculations the number of lattice sites in the temporal direction is significantly reduced . however , the density of lattice sites is still the same , i.e. there is a smaller number of link variables in a smaller spacetime volume . the pseudoparticle approach offers two possibilities to change the temperature : * varying @xmath120 while @xmath2 is kept constant changes the value of the temperature directly . * varying @xmath2 while @xmath120 is kept constant changes the numerical value of the string tension @xmath69 . this leads to a different extension of the periodic time direction in physical units and , therefore , changes the value of the temperature . both methods yield consistent results ( c.f . [ fig_010]c ) . whereas at low temperatures static quarks can not be separated from each other , at high temperatures a deconfining phase is expected . the temperature , at which the corresponding phase transition takes place , is called critical temperature and denoted by @xmath182 . @xmath182 can be determined from ensemble averages of polyakov loops . a polyakov loop is a wilson loop around the periodic time @xmath159 : @xmath183 \ \ = \ \ \frac{1}{2 } \textrm{tr}\left(p\left\{\exp\left(i \oint dz_0 \ , a_0(z)\right)\right\}\right ) .\end{aligned}\ ] ] due to spatial translational invariance @xmath184 is @xmath185-independent . therefore , from a numerical point of view it is convenient to consider spatial averages of polyakov loops . we define @xmath186 \right\rangle_\beta .\end{aligned}\ ] ] @xmath187 can be considered as an order parameter indicating whether there is confinement or not ( c.f . e.g. @xcite ) : @xmath188 this criterion is closely related to center symmetry , which is spontaneously broken in the deconfinement phase . in the center symmetric phase of su(2 ) yang - mills theory for every field configuration @xmath189 there is a gauge equivalent field configuration @xmath190 with + @xmath191 = -l_\mathbf{z}[a^-]$ ] ( @xmath189 and @xmath190 are related by center symmetry , i.e. they are connected by a singular gauge transformation ) . since both field configurations have the same action , i.e. @xmath192 = s[a^-]$ ] , @xmath193 follows immediately . the spontaneous breakdown of center symmetry at @xmath194 comes along with a splitting of the hilbert space of states in two independent spaces : + @xmath195 . the same applies for the set of field configurations considered in the path integral : + @xmath196 . the ensemble average of the polyakov loop depends on which hilbert space was chosen during the spontaneous breakdown of center symmetry : @xmath197 and @xmath198 . in the broken phase two field configurations , which are related by center symmetry , can not be connected continuously by a set of field configurations of finite action . this implies that during a monte - carlo simulation only field configurations either from @xmath199 or from @xmath200 are generated , assuming an infinite system and a local and continuous update mechanism . in numerical calculations these assumptions are only approximately fulfilled . nevertheless , in lattice monte - carlo simulations ( c.f . e.g. @xcite ) it has been observed that there are long sequences of steps , where only field configurations corresponding to one of the two hilbert spaces are generated . since our pseudoparticle ensembles are only approximately center symmetric , there is a smooth transition between the two phases . @xmath201 can be observed well below the critical temperature . for @xmath202 the ensemble average of the polyakov loop quickly rises to a non - zero value . for high temperatures @xmath203 ( c.f . [ fig_010]a ) . therefore , we define the critical temperature @xmath182 ( or equivalently its inverse @xmath204 ) to be that temperature , where the ensemble average of the polyakov loop crosses a certain value @xmath205 , i.e.@xmath206 the main difference between our results and lattice results is that even in the deconfinement phase we never have observed @xmath207 . we conclude that in the pseudoparticle approach the low action field configurations corresponding to @xmath208 are underrepresented , i.e. there are more low action field configurations corresponding to @xmath209 than to @xmath208 . this bias always forces the system to chose the hilbert space @xmath210 , when center symmetry is spontaneously broken . the bias can be explained by the following qualitative argument : for field configurations close to zero @xmath211 ( c.f . ( [ eqn_017 ] ) ) . therefore , we expect these field configurations to be elements of @xmath199 . furthermore , all of these field configurations are low action field configurations , which contribute significantly to the path integral . the set of these field configurations is denoted by @xmath212 . loosely speaking , all other low action field configurations are large enough so that @xmath213 can pick up exponents , which are significantly different from zero . they are either elements of @xmath199 or of @xmath200 . the set of these field configurations is denoted by @xmath214 . however , there is numerical evidence that in the pseudoparticle approach with around @xmath9 pseudoparticles most low action field configurations have small gauge fields , i.e. are elements of @xmath212 . this is closely related to the fact that our ensembles are only approximately gauge invariant : a gauge transformed field configuration from @xmath212 , which is a field configuration in @xmath214 , can only be approximated , when using around @xmath9 pseudoparticles ; such an approximation usually has a higher action . therefore , there are proportionally more field configurations in @xmath212 than in @xmath214 . this is manifested in form of a bias . in fig . [ fig_010]a the ensemble average of the polyakov loop @xmath187 is plotted against the temperature @xmath98 for different coupling constants ( @xmath215 , @xmath65 ) . the dashed straight lines correspond to a determination of @xmath182 with @xmath216 ( other values for @xmath205 yield qualitatively identical results ) . [ fig_010]b shows that the dimensionless ratio @xmath217 is nearly independent of the coupling constant @xmath2 ( @xmath69 has been obtained by generalized creutz ratios ( c.f . [ fig_005]a ) ) , i.e. the string tension and the critical temperature scale consistently with respect to @xmath2 . we would like to stress again that such a scaling behavior , although mandatory for any trustworthy numerical method , is far from obvious . the range of values , @xmath218 , is of the right order of magnitude compared to the lattice result @xmath219 @xcite . @xmath220 , @xmath43 , @xmath42 . * a ) * @xmath187 for different @xmath2 plotted against @xmath98 . * b ) * @xmath217 plotted against @xmath2 . * c ) * @xmath187 for different @xmath2 plotted against @xmath221 . ] as we have explained in section [ sec_006 ] , there are two methods to adjust the temperature , varying @xmath120 or varying @xmath2 . fig . [ fig_010]c shows @xmath187 as a function of @xmath221 for different coupling constants @xmath2 . the fact that all sample points scale to a single curve demonstrates that both methods yield consistent results . [ fig_011 ] shows the evolution of @xmath222 during a single monte - carlo simulation ( @xmath223 monte - carlo steps ) for @xmath129 and different @xmath224 ( @xmath225 ) . these plots demonstrate that in the confinement phase + ( @xmath226 ) @xmath213 assumes approximately an equal number of positive and negative values , whereas in the deconfinement phase ( @xmath227 ) the values of @xmath213 are mainly positive . @xmath220 , @xmath43 , @xmath42 , @xmath129 . @xmath222 for different ratios @xmath228 plotted against the number of the monte - carlo step . ] it is well known that the static quark antiquark potential at finite temperature can be obtained from polyakov loop correlation functions : @xmath229 ( c.f . e.g. @xcite ) . however , center symmetry is only approximately realized in the pseudoparticle approach , i.e. even in the confinement phase @xmath230 . consequently , any finite temperature potential calculated according to ( [ eqn_018 ] ) approaches a constant value for large quark antiquark separations : @xmath231 therefore , at finite temperature one can expect to observe confining potentials only up to intermediate separations . [ fig_012 ] shows finite temperature quark antiquark potentials for @xmath232 obtained from polyakov loop correlation functions ( c.f . ( [ eqn_018 ] ) ; + @xmath215 , @xmath233 ) and the corresponding zero temperature potential obtained from ensemble averages of wilson loops ( c.f . section [ sec_005 ] ) . at high temperatures there is clearly no confinement . however , with the decrease of the ensemble average of the polyakov loop towards lower temperatures finite temperature potentials gradually approach the zero temperature potential extracted from wilson loops . because of @xmath230 the finite temperature potentials and the zero temperature potential differ at large separations even for temperatures significantly smaller than @xmath182 . there are arguments suggesting that these potentials should rather be calculated via @xmath234 instead of ( [ eqn_018 ] ) . indeed , this yields finite temperature potentials , which are identical to the zero temperature potential within statistical errors , a result , which is in agreement with results from lattice calculations ( c.f . e.g. @xcite ) . we plan to discuss these topics in more detail in an upcoming paper . @xmath220 , @xmath43 , @xmath42 , @xmath129 . finite temperature potentials and the corresponding zero temperature potential plotted against the separation of the quarks . ] the goal of this section is to identify essential properties of gauge field configurations , which are responsible for confinement . to this end we compare different pseudoparticle ensembles . in the following we explore how pseudoparticle results are affected by a variation of the pseudoparticle size @xmath34 , while the pseudoparticle density @xmath173 or equivalently the average pseudoparticle distance @xmath81 is kept constant . in other words , we consider different ratios of the two ultraviolet regulators @xmath112 and @xmath34 . note that @xmath34 strongly affects the shape of a pseudoparticle near its center , while it has essentially no effect on the long range behavior of a pseudoparticle , which is proportional to @xmath235 ( c.f . ( [ eqn_001 ] ) to ( [ eqn_003 ] ) ) . we consider ensembles with @xmath41 , @xmath43 , @xmath68 and @xmath236 . we have obtained the string tension @xmath69 via generalized creutz ratios ( c.f . section [ sec_004 ] ) . [ fig_013]a shows that @xmath69 is clearly positive for all values of the pseudoparticle size @xmath34 . moreover , there is only a weak @xmath34-dependence , i.e. @xmath237 for @xmath238 . @xmath41 , @xmath43 , @xmath68 . * a ) * @xmath69 plotted against @xmath34 . * b ) * @xmath156 plotted against @xmath34 . ] the implication is that confinement is connected to the @xmath31 long range behavior of the pseudoparticles , which is the same for all values of @xmath34 . the shape of the pseudoparticles near their center , which is @xmath34-dependent , has essentially no effect on the string tension and , therefore , is of no relevance to confinement . the topological susceptibility @xmath150 is also fairly independent of the pseudoparticle size @xmath34 as is the dimensionless ratio @xmath156 ( c.f . [ fig_013]b ) . the range of values , @xmath239 between @xmath238 , is in qualitative agreement with the lattice result + @xmath240 @xcite . to learn more about the interrelation between the long range behavior of pseudoparticles and confinement , we study pseudoparticles of gaussian localized profile and different size . such gaussian localized pseudoparticles are obtained by replacing @xmath241 appearing in the definitions of instantons , antiinstantons and akyrons ( ( [ eqn_001 ] ) to ( [ eqn_003 ] ) ) by @xmath242 . the main difference is the long range behavior of the resulting building blocks : in contrast to our standard choice of pseudoparticles , ( [ eqn_001 ] ) to ( [ eqn_003 ] ) , gaussian localized pseudoparticles have a limited range of interaction , which is proportional to their size @xmath34 . we consider ensembles with @xmath41 , @xmath43 , @xmath68 and @xmath243 . we have obtained numerical values for the string tension @xmath69 via generalized creutz ratios ( c.f . section [ sec_004 ] ) . the results are plotted in fig . [ fig_014]a against the pseudoparticle size @xmath34 . gaussian localized pseudoparticles , @xmath41 , @xmath43 , @xmath68 . * a ) * @xmath69 plotted against @xmath34 . * b ) * @xmath74 plotted against @xmath34 . * c ) * @xmath100 for different pseudoparticle sizes plotted against the separation of the quarks . ] there is no confinement for @xmath244 and confinement for @xmath245 . the onset of confinement takes place somewhere around @xmath246 . this is precisely the width at which neighboring pseudoparticles start to overlap significantly ( this can be seen by assigning each pseudoparticle an appropriate volume , e.g. the volume of a hypersphere of radius @xmath34 , @xmath247 , and comparing that volume with the maximum volume non - overlapping pseudoparticles can cover , i.e. @xmath248 ) . the implication is that pseudoparticle ensembles only exhibit confinement if their pseudoparticles cover sufficiently large spacetime regions so that there are significant overlaps between neighboring pseudoparticles . this is reminiscent of percolation . note that percolation phenomena have been related to confinement in a variety of ways , e.g. percolation of center vortices @xcite or monopole loops @xcite in lattice gauge theory or random percolation of bonds or sites on three dimensional lattices @xcite . the @xmath34-dependence of the topological susceptibility @xmath150 is much weaker than the @xmath34-dependence of the string tension @xmath69 . therefore , the dimensionless ratio @xmath74 is dominated by @xmath69 ( c.f . fig . [ fig_014]b ) . the range of values for @xmath249 , i.e. for significantly overlapping pseudoparticles , @xmath250 , is of the right order of magnitude when compared to the lattice result + @xmath240 @xcite . we have also calculated quark antiquark potentials for different values of @xmath34 ( c.f . section [ sec_005 ] ) . the results are shown as functions of the quark antiquark separation in fig . [ fig_014]c . for @xmath251 the potential is clearly confining , whereas for @xmath252 it is unambiguously not confining , a result which is in agreement with previous results from this section . singular gauge pseudoparticles , @xmath41 , @xmath43 , @xmath253 , @xmath68 . * a ) * @xmath69 plotted against @xmath34 . * b ) * @xmath74 plotted against @xmath34 . * c ) * @xmath100 for different pseudoparticle sizes plotted against the separation of the quarks . ] we have also studied ensembles of singular gauge instantons , antiinstantons and akyrons . these `` singular gauge pseudoparticles '' are obtained by replacing @xmath32 in ( [ eqn_001 ] ) to ( [ eqn_003 ] ) by @xmath254 , where @xmath255 is an ultraviolet regulator , which has been included due to numerical reasons . we consider ensembles with @xmath41 , @xmath43 , @xmath253 , @xmath68 and @xmath256 . numerical results ( c.f . fig . [ fig_015 ] ) and their interpretation are similar to those obtained with gaussian localized pseudoparticles . of course , this is what one would expect , because the gauge fields of singular gauge pseudoparticles decrease much faster than those of regular gauge pseudoparticles , i.e. like @xmath257 instead of @xmath31 . therefore , singular gauge pseudoparticles are quite similar to gaussian localized pseudoparticles with respect to long range interactions . the result that there is no confinement for ensembles of singular gauge pseudoparticles of small size ( @xmath258 ) is in agreement with what has been obtained in instanton gas and instanton liquid models , where typically @xmath259 ( c.f . e.g. @xcite ) . furthermore , it has been proposed that large size singular instantons can produce confinement ( @xcite quoted in @xcite ) , which is also supported by our findings . in the following we present evidence that confinement arises due to instantons and antiinstantons , whereas akyrons do not produce confinement . to this end , we consider ensembles with the same number of pseudoparticles but different ratios @xmath260 . in detail , we compare the following ensembles with + @xmath41 , @xmath43 , @xmath42 and @xmath68 : * `` akyron ensemble '' : an ensemble containing @xmath9 akyrons ( no instantons or antiinstantons ) . * `` standard ensemble '' : an ensemble containing @xmath261 instantons , @xmath261 antiinstantons and @xmath66 akyrons . * `` instanton ensemble '' : an ensemble containing @xmath262 instantons and @xmath262 antiinstantons ( no akyrons ) . [ fig_016 ] shows the quark antiquark potential as a function of the separation for all three ensembles ( c.f . section [ sec_005 ] ) . the standard ensemble and the instanton ensemble exhibit a confining potential , whereas the akyron curve indicates that there is no confinement . akyron ensemble ( @xmath9 akyrons ) , standard ensemble ( @xmath261 instantons , @xmath261 antiinstantons and @xmath66 akyrons ) and instanton ensemble ( @xmath262 instantons and @xmath262 antiinstantons ) , @xmath41 , @xmath43 , @xmath42 , @xmath68 . @xmath100 plotted against the separation of the quarks . ] we have obtained numerical values for the string tension @xmath69 via generalized creutz ratios ( c.f . section [ sec_004 ] ) . the results are @xmath263 , + @xmath264 and @xmath265 . + the ratio of these values is given by + @xmath266 . to compare dimensionless ratios , we have calculated the topological susceptibility @xmath150 and the critical temperature @xmath182 in the standard ensemble and in the instanton ensemble : + @xmath267 , + @xmath268 , + @xmath269 , + @xmath270 . + obviously , akyrons increase the topological susceptibility , while they do not affect the critical temperature . since @xmath267 is closer to the lattice result @xmath158 @xcite than + @xmath268 , it is beneficial with respect to quantitative results to consider ensembles containing not only instantons and antiinstantons but also akyrons . the dimensionless ratios @xmath74 and @xmath75 in the akyron ensemble are not meaningful : the topological susceptibility @xmath150 vanishes identically ( c.f . appendix [ sec_013 ] ) and the ensemble average of the polyakov loop @xmath187 is close to @xmath165 even for very large values of @xmath120 , i.e. there is no sign of a confinement deconfinement phase transition . from that and the fact that the string tension is more than ten times smaller in the akyron ensemble than in the other two ensembles we conclude that gauge field configurations made up solely of akyrons are not suited to produce confinement . on the other hand , gauge field configurations , which are responsible for confinement , necessarily contain instantons and antiinstantons . it is interesting to note that any linear superposition of akyrons and , therefore , any field configuration in a pure akyron ensemble have zero topological charge density ( c.f . appendix [ sec_013 ] ) . since akyron ensembles do not exhibit confinement , this supports the common expectation that confinement and topological charge are closely related . in this work we have presented the pseudoparticle approach , a numerical method to compute path integrals in effective su(2 ) yang - mills theories . we have calculated the static quark antiquark potential at zero and at finite temperature , the topological susceptibility and the critical temperature of the confinement deconfinement phase transition in different pseudoparticle ensembles . the pseudoparticle approach is able to reproduce many essential features of su(2 ) yang - mills theory with a comparatively small number of degrees of freedom . when using @xmath9 instantons , antiinstantons and akyrons , the static quark antiquark potential is linear for large separations with an attractive @xmath105-correction , which is in qualitative agreement with the bosonic string picture and with results from lattice calculations . the string tension @xmath69 is not only positive but also an increasing function of the coupling constant @xmath2 . therefore , when the scale is set by the string tension , one can adjust the size of the spacetime hypersphere in physical units by choosing an appropriate value for @xmath2 . we have also calculated the the topological susceptibility @xmath150 and the critical temperature @xmath182 . the dimensionless ratios @xmath74 and @xmath75 are constant for a wide range of coupling constants , i.e. @xmath69 , @xmath150 and @xmath182 exhibit consistent scaling behaviors with respect to @xmath2 . this success strongly indicates that the pseudoparticle approach has the potential to reproduce correct yang - mills physics . the values of both dimensionless ratios are of the right order of magnitude compared to lattice results . for ensembles made up of instantons , antiinstantons and akyrons the string tension shows only a weak dependence on the pseudoparticle size @xmath34 . it seems that confinement is mainly a consequence of the long range behavior of the building blocks , which is unaffected by @xmath34 . this has been confirmed by considering ensembles of gaussian localized pseudoparticles and ensembles of singular gauge pseudoparticles , for which the size parameter @xmath34 strongly affects the long range behavior . for small @xmath34 there is only little overlap between neighboring pseudoparticles and there is no sign of confinement . increasing @xmath34 to a value , where pseudoparticles overlap and interact significantly , restores quark confinement . the conclusion is that gauge field configurations , which are responsible for confinement , contain extended structures . on the other hand , gauge field configurations with only localized excitations do not produce confinement . comparing our `` standard ensemble '' with a pure instanton and antiinstanton ensemble and a pure akyron ensemble has shown that confinement arises due to instantons and antiinstantons and not because of akyrons . keeping in mind that gauge field configurations made up solely of akyrons have vanishing topological charge density , our findings support the common expectation that topological charge and confinement are closely related . for quantitative results akyrons seem to play an important role . the dimensionless ratio @xmath156 is significantly closer to the lattice result , when there are not only instantons and antiinstantons but also akyrons . an obvious possibility for future research is to calculate correlation functions in order to extract glueball masses . comparing the resulting masses with results from lattice calculations would be another check of the pseudoparticle approach . furthermore we plan to compute other observables at finite temperature , e.g. the energy density and the pressure . a major new direction is to include fermions in the pseudoparticle approach . the goal would be to obtain a model for su(2 ) yang - mills theory , which exhibits both chiral symmetry breaking and a confinement deconfinement phase transition . it is a pleasure to thank frieder lenz for lots of inspiring discussions as well as for making many important and helpful suggestions . furthermore , i would like to acknowledge useful conversations with philippe de forcrand , john w. negele and jan m. pawlowski . a color orientation is a spacetime independent gauge transformation . it can be specified by an element of su(2 ) : @xmath271 with @xmath272 . applying such a gauge transformation to a gauge field @xmath273 yields @xmath274 or expressed in components @xmath275 we refer to @xmath276 as color orientation matrix . color orientation matrices fulfill @xmath277 therefore , they are elements of so(3 ) . the gauge field of a single akyron with index @xmath13 , amplitude @xmath58 , color orientation matrix @xmath278 and position @xmath61 is given by @xmath279 ( c.f . ( [ eqn_003 ] ) ) . in the corresponding field strength the derivative terms cancel each other : @xmath280 for any linear superposition of akyrons @xmath281 the same is true : @xmath282 for the topological charge density follows @xmath283 any gauge field @xmath284 can be written as a sum of plane waves : @xmath285 where @xmath286 , the fourier transform of @xmath284 , is given by @xmath287 the fourier transformed gauge field @xmath286 can be decomposed in a transverse and a longitudinal part : @xmath288 with @xmath289 superpositions of instantons ( [ eqn_001 ] ) and antiinstantons ( [ eqn_002 ] ) form transverse gauge fields , whereas superpositions of akyrons ( [ eqn_003 ] ) form longitudinal gauge fields . this can be seen by considering the fourier transforms of these pseudoparticles : @xmath290 where @xmath291 ( @xmath292 is a modified bessel function of imaginary argument ) . ( [ eqn_021 ] ) and ( [ eqn_022 ] ) satisfy ( [ eqn_019 ] ) due to the antisymmetry of @xmath293 and @xmath294 , while ( [ eqn_023 ] ) obviously fulfills ( [ eqn_020 ] ) . in this appendix we show that `` almost any gauge field configuration '' can be represented by a linear superposition of infinitely many instantons and akyrons . the starting point is the `` continuum limit '' of ( [ eqn_005 ] ) without antiinstantons : @xmath295 ( the sum over all pseudoparticles has been replaced by an integration over spacetime ; furthermore , nine instantons and three akyrons are allowed to share the same position ) . inserting ( [ eqn_001 ] ) and ( [ eqn_003 ] ) in ( [ eqn_025 ] ) yields @xmath296 it can be shown that in general nine color orientation matrices form a basis of all @xmath297-matrices @xcite . therefore , any @xmath298 and @xmath299 can be realized by suitably chosen amplitudes @xmath300 and @xmath301 . hence , the problem has been reduced to the question whether any gauge filed configuration @xmath284 can be represented by suitably chosen @xmath298 and @xmath299 . without loss of generality we consider @xmath304 : @xmath305 for @xmath306 this equation can be solved for + @xmath307 , because @xmath308 and + @xmath309 . for @xmath310 both @xmath303 and @xmath311 are singular . to study this case , we first deduce @xmath312 from ( [ eqn_026 ] ) by applying a proper regularization scheme . ( [ eqn_028 ] ) implies @xmath313 . inserting this in ( [ eqn_027 ] ) + shows that the value of @xmath314 has no effect on the gauge field @xmath284 . on the other hand , changing @xmath314 amounts to a constant shift of @xmath315 : @xmath316 . that is adding @xmath317 changes @xmath315 , whereas the gauge field @xmath284 remains unaltered . to get rid of this redundancy , we require @xmath318 to be able to represent any gauge field configuration @xmath284 , we have to find a way around ( [ eqn_028 ] ) . this can easily be achieved by adding constants @xmath319 to ( [ eqn_025 ] ) . the final result is the following : any gauge field configuration @xmath284 has a unique expansion @xmath320 in terms of @xmath300 and @xmath301 constrained by ( [ eqn_029 ] ) and @xmath319 , where @xmath321 , @xmath322 , as well as @xmath323 , + @xmath324 are linearly independent color orientation matrices . t. schfer and e. v. shuryak , rev . * 70 * ( 1998 ) 323 [ arxiv : hep - ph/9610451 ] . d. j. gross , r. d. pisarski and l. g. yaffe , rev . phys . * 53 * , 43 ( 1981 ) . p. gerhold , e. m. ilgenfritz and m. mller - preussker , arxiv : hep - ph/0607315 . f. lenz , j. w. negele and m. thies , phys . rev . d * 69 * ( 2004 ) 074009 [ arxiv : hep - th/0306105 ] . j. w. negele , f. lenz and m. thies , nucl . * 140 * ( 2005 ) 629 [ arxiv : hep - lat/0409083 ] . l. del debbio , m. faber , j. giedt , j. greensite and s. olejnik , phys . d * 58 * ( 1998 ) 094501 [ arxiv : hep - lat/9801027 ] . m. wagner and f. lenz , pos * lat2005 * ( 2005 ) 315 [ arxiv : hep - lat/0510083 ] . m. wagner , phd thesis , institute for theoretical physics iii , university of erlangen - nrnberg ( 2006 ) . m. wagner , arxiv : hep - ph/0610291 . r. jackiw and c. rebbi , phys . d * 14 * ( 1976 ) 517 . a. a. belavin , a. m. polyakov , a. s. shvarts and y. s. tyupkin , phys . b * 59 * ( 1975 ) 85 . h. j. rothe , world sci . notes phys . * 74 * ( 2005 ) 1 . m. lscher , k. symanzik and p. weisz , nucl . b * 173 * ( 1980 ) 365 . m. lscher , nucl . b * 180 * ( 1981 ) 317 . j. d. stack , phys . d * 27 * ( 1983 ) 412 . g. s. bali , k. schilling and c. schlichter , phys . d * 51 * ( 1995 ) 5165 [ arxiv : hep - lat/9409005 ] . m. lscher and p. weisz , jhep * 0207 * ( 2002 ) 049 [ arxiv : hep - lat/0207003 ] . m. creutz , phys . rev . d * 21 * ( 1980 ) 2308 . m. creutz , phys . * 45 * ( 1980 ) 313 . di vecchia , k. fabricius , g. c. rossi and g. veneziano , nucl . b * 192 * ( 1981 ) 392 . m. teper , phys . b * 232 * ( 1989 ) 227 . c. christou , a. di giacomo , h. panagopoulos and e. vicari , phys . d * 53 * ( 1996 ) 2619 [ arxiv : hep - lat/9510023 ] . a. di giacomo , arxiv : hep - lat/9711034 . e. witten , nucl . b * 156 * ( 1979 ) 269 . m. j. teper , arxiv : hep - th/9812187 . b. j. harrington and h. k. shepard , phys . d * 17 * ( 1978 ) 2122 . m. wagner , k. hormann and g. greiner , preprint ( 2003 ) . l. d. mclerran and b. svetitsky , phys . d * 24 * ( 1981 ) 450 . s. a. gottlieb , j. kuti , d. toussaint , a. d. kennedy , s. meyer , b. j. pendleton and r. l. sugar , phys . * 55 * ( 1985 ) 1958 . o. kaczmarek , f. karsch , e. laermann and m. ltgemeier , phys . d * 62 * , 034021 ( 2000 ) [ arxiv : hep - lat/9908010 ] . m. engelhardt , k. langfeld , h. reinhardt and o. tennert , phys . d * 61 * ( 2000 ) 054504 [ arxiv : hep - lat/9904004 ] . v. g. bornyakov , v. k. mitrjushkin and m. mller - preussker , phys . lett . b * 284 * ( 1992 ) 99 . f. gliozzi , s. lottini , m. panero and a. rago , nucl . b * 719 * ( 2005 ) 255 [ arxiv : cond - mat/0502339 ] . d. diakonov and v. petrov , proceedings of the workshop on continuous advances in qcd , world scientific , singapore ( 1996 ) .

we study perturbations of diffeomorphisms that have a saddle connection between a pair of normally hyperbolic invariant manifolds . we develop a first - order deformation calculus for invariant manifolds and show that a generalized melnikov function or _ melnikov displacement _ can be written in a canonical way . this function is defined to be a section of the normal bundle of the saddle connection . we show how our definition reproduces the classical methods of poincar and melnikov and specializes to methods previously used for exact symplectic and volume - preserving maps . we use the method to detect the transverse intersection of stable and unstable manifolds and relate this intersection to the set of zeros of the melnikov displacement .

0706.2515

the study of the intersections of stable and unstable manifolds of maps and flows has a strong influence on dynamical systems . in particular , the existence of a transverse intersection is associated with the onset of chaos , and gave rise to the famous horseshoe construction of smale . the poincar - melnikov method @xcite is a widely used technique for detecting such intersections . given a system with a pair of saddles and a degenerate heteroclinic or saddle connection between them , the classical melnikov function computes the rate at which the distance between the manifolds changes with a perturbation . there have been many formulations of the melnikov method for two - dimensional maps or flows @xcite and for higher - dimensional symplectic mappings @xcite . recently , the geometric content of melnikov s method was exploited in order to detect heteroclinic intersections of lagrangian manifolds for the case of perturbed hamiltonian flows @xcite . here it was shown that the heteroclinic orbits are in correspondence with the zeros of a geometric object , the so - called melnikov one - form . for maps , the melnikov function is an infinite sum whose domain is a saddle connection between two hyperbolic invariant sets . as usual , a simple zero of this function corresponds to a transverse intersection of stable and unstable manifolds of a perturbation of the original map . melnikov s method can also be used to compute transport fluxes . in particular , a _ resonance zone _ for a two - dimensional mapping is a region bounded by alternating segments of stable and unstable manifolds that are joined at primary intersection points @xcite . because the intersection points are primary , a resonance zone is bounded by a jordan curve and has exit and entry sets @xcite . the images of these sets completely define the transport properties of the resonance zone . moreover , the integral of the melnikov function between two neighbouring primary intersection points is the first order approximation to the geometric flux escaping from the resonance zone @xcite . the method has also been applied to the case of periodically time - dependent , volume - preserving flows @xcite and more generally to volume - preserving maps with fixed points @xcite and invariant circles @xcite . volume - preserving maps provide perhaps the simplest , natural generalization of the class of area - preserving maps to higher dimensions . moreover , they naturally arise in applications as the time - one poincar map of incompressible flows even when the vector field of the flow is nonautonomous . thus the study of the dynamics of volume - preserving maps has application both to fluids and magnetic fields . our goal in this paper is to develop , based on the theory of deformations , a general , geometrical description of the melnikov displacement and to compare our theory to classical results . deformation theory was first introduced in the theory of singularities @xcite , but was soon used in the contexts of volume and symplectic geometry . its application to dynamical systems in @xcite provide results that are close to our goals . let @xmath0 be a smooth family of diffeomorphisms such that the unperturbed map @xmath1 has a saddle connection @xmath2 between a pair of compact @xmath3-normally hyperbolic invariant manifolds . let @xmath4 be the algebraic normal bundle of the saddle connection . we show that there exists a canonical @xmath5 section @xmath6 , called the _ melnikov displacement _ , that measures the splitting of the saddle connection in first - order . we will prove that the melnikov displacement is given by the absolutely convergent series @xmath7 where @xmath8 is the vector field defined by @xmath9 . these sums do not converge in the tangent space @xmath10 , but only in the algebraic normal bundle @xmath11 . the use of the algebraic normal bundle in the study of normally hyperbolic manifolds goes back to @xcite . the fact that ( [ mainformula ] ) always converges also addresses the question of how to deal with melnikov s method when the original pair of compact @xmath3-normally hyperbolic invariant manifolds ( that are connected with the saddle connection ) are not fixed with the perturbation . in addition , we will also show that the melnikov displacement has a number of geometric properties . the main result in this direction is that any change of coordinates acts on the displacement by its pullback . this result will be used to obtain the natural action of any symmetries , reversing symmetries or integrals of the dynamical system on the displacement . similarly , if the map preserves a symplectic or volume form , this gives additional structure to the displacement . for example , if @xmath0 is a family of exact symplectic maps and the normally hyperbolic invariant manifolds are fixed points , then we will show that there exists a function @xmath12 , the _ melnikov potential _ , such that @xmath13 , where @xmath14 is the symplectic two - form . this relation is reminiscent of the definition of globally hamiltonian vector fields . when the normally hyperbolic invariant manifolds are not fixed points ( or isolated periodic points ) , its stable and unstable manifolds are coisotropic , but not isotropic , and so the relation @xmath13 makes no sense . we complete this introduction with a note on the organization of this paper . the general theory is developed in [ sec : melnikovdisplacement ] . in [ sec : classical ] , we show how our theory reproduces the classical methods of poincar and melnikov . the study of exact symplectic maps and volume - preserving maps is contained in [ sec : symp ] and [ sec : volume ] , respectively . in this subsection we present the deformation calculus for families of diffeomorphisms and submanifolds . we shall begin with diffeomorphisms by defining a vector field associated with a deformation . next , we construct a vector field for the deformation of ( immersed ) submanifolds . finally , we will combine these results to define and compute the melnikov displacement . in this paper , we consider smooth families of diffeomorphisms @xmath15 , where @xmath16 is an @xmath17-dimensional smooth manifold . here , the term smooth family means that @xmath18 is @xmath19 in both variables . the map @xmath1 will be called unperturbed " . the generating vector field of a smooth family of diffeomorphisms @xmath0 is the unique vector field @xmath8 such that @xmath20 if we regard @xmath8 as a nonautonomous vector field with time @xmath21 , then the function @xmath22 represents its flow ( * ? ? ? consequently , if @xmath0 is volume preserving , then @xmath8 has zero divergence ( * ? ? ? 2.2.24 ) , and if @xmath0 is ( exact ) symplectic , then @xmath8 is ( globally ) hamiltonian . these geometric equivalences form the basis of the deformation calculus . the generating vector field @xmath8 was also called the _ perturbation vector field _ in @xcite ; however , we adopt the older terminology here . sometimes , we will also refer to @xmath8 as the _ generator _ of the family @xmath0 . we will always use the convention that , given a family of diffeomorphisms denoted by italic letters @xmath0 , its generator is denoted by the same letter in calligraphic capitals . we have collected in the proposition below some well - known relations among generators , in which appear push - forwards and pullbacks . recall that the pullback @xmath23 and push - forward @xmath24 of a diffeomorphism @xmath25 act on a vector field @xmath26 as follows : @xmath27 we note that @xmath28 . next , our goal is to develop a first - order deformation calculus for invariant manifolds of smooth diffeomorphisms . we are mainly interested in stable and unstable invariant manifolds of @xmath3-normally hyperbolic manifolds for some @xmath29 , which unfortunately are just immersed submanifolds ( not embedded submanifolds ) , and @xmath30 ( not @xmath19 ) . this gives rise to a few technicalities . recall that a map @xmath31 is an _ immersion _ when its differential has maximal rank everywhere . if @xmath32 is one - to - one onto its image , then @xmath33 is an immersed submanifold of the same dimension as @xmath34 . we will denote an immersion by @xmath35 or simply @xmath36 when the immersion @xmath32 does not matter . for brevity , we will sometimes omit the term immersed " . if @xmath37 is the stable ( resp . , unstable ) invariant manifold of a fixed point of a diffeomorphism , then @xmath38 ( resp . , @xmath39 ) , where @xmath40 and @xmath41 are the number of stable and unstable directions at the hyperbolic point . when the fixed point is hyperbolic , @xmath42 . stable and unstable manifolds are typically not embedded because they can have points of accumulation . in this case the immersion is not a proper map . we consider families of submanifolds of the form @xmath43 , where @xmath44 is @xmath30 in both variables , for some @xmath45 . all the elements of such a family are diffeomorphic ( as immersed submanifolds ) , because they are diffeomorphic to the same base " manifold @xmath34 . just as for @xmath1 , the unperturbed submanifold is denoted by @xmath46 . [ def : adapted ] if @xmath47 is a @xmath30 family of immersed submanifolds , a family of diffeomorphisms @xmath48 is an adapted deformation when @xmath49 and @xmath50 is @xmath30 in both variables . adapted deformations exist since it suffices to take @xmath51 . while there is quite a bit of freedom in the choice of @xmath52 , only its normal component is relevant , since this measures the actual motion of @xmath53 with @xmath21 . the normal component will be defined using the algebraic normal bundle . for an immersed submanifold @xmath54 , this is defined as the set of equivalence classes @xmath55 when @xmath16 is riemannian , this normal bundle is isomorphic to the more familiar geometric normal bundle , @xmath56 ( cf . @xcite ) . in general @xmath57 is a manifold of dimension @xmath17 and is defined independently of any inner product structure on @xmath58 . the displacement vector field of a @xmath30 family of immersed submanifolds @xmath47 is the @xmath5 section @xmath59 where @xmath60 is any adapted deformation . the displacement vector field is well - defined ; that is , its definition is independent of the choice of the adapted deformation as is shown in the following lemma . [ lem : adapted ] let @xmath61 be two adapted deformations . then @xmath62_{\epsilon=0 } \in t_\xi w_0 \;,\qquad \forall \xi \in w_0 \;.\ ] ] for each fixed @xmath63 , the map @xmath64 describes a @xmath30 curve in @xmath46 such that @xmath65 and @xmath66 . using the fact that @xmath67 , we then have @xmath62_{\epsilon=0 } = { \rm d } \phi_0(\xi ) c'(0 ) = c'(0 ) \in t_\xi w_0 \;,\ ] ] because @xmath68 . when the submanifold is invariant under a diffeomorphism , its deformations are related by means of a fundamental iterative relationship between the generating vector field of the family of diffeomorphisms and the displacement vector field of the family of submanifolds . let @xmath0 be a smooth family of diffeomorphisms , and @xmath47 be a @xmath30 family of immersed submanifolds that are invariant under @xmath0 . then @xmath69 on the unperturbed submanifold @xmath46 , where @xmath70 is the displacement vector field . the tangent space @xmath71 is invariant under the pullback @xmath72 , so the term @xmath73 is well - defined as a section of the normal bundle @xmath74 . if @xmath48 is any adapted deformation , then @xmath75 is as well . differentiating @xmath76 with respect to @xmath21 yields @xmath77 where we used . however , by lemma [ lem : adapted ] the displacement is independent of the adapted deformation , so @xmath78 applying @xmath79 to both sides finishes the proof . the identity is equivalent to @xmath80 . thus , we can work either with push - forwards or pullbacks . to obtain the melnikov displacement we will iterate these identities on the stable and unstable manifolds of a family of diffeomorphisms . the melnikov displacement will be defined for a saddle connection between a pair of normally hyperbolic invariant manifolds . in this section we recall the definitions of these objects . there are many slightly different definitions of normally hyperbolic manifolds , see @xcite . in this paper , we adopt the following . let @xmath81 be a submanifold invariant under a smooth diffeomorphism @xmath82 . we say that @xmath83 is @xmath3-normally hyperbolic when there exist a riemann structure on @xmath58 , a constant @xmath84 , and a continuous invariant splitting @xmath85 such that if @xmath86 and @xmath87 are the canonical restrictions of the linear map @xmath88 associated to the splitting , then 1 . @xmath89 , and 2 . @xmath90 for all @xmath91 , and for all @xmath92 . as is usual in the literature , the term _ normally hyperbolic _ will be taken to mean @xmath93-_normally hyperbolic_. note that setting @xmath94 in the previous definition implies that the linearizations @xmath95 and @xmath96 of @xmath97 restricted to the stable and unstable spaces of @xmath83 have the uniform bounds @xmath98 in this paper we will assume that each @xmath3-normally hyperbolic invariant manifold @xmath83 is compact , although it would be sufficient that our diffeomorphisms be uniformly @xmath30 in some neighbourhood of @xmath83 . we will also assume , without loss of generality , that @xmath83 is connected . one consequence is that the sets @xmath99 and @xmath100 are @xmath30 immersed submanifolds of @xmath16 that are tangent at @xmath83 to @xmath101 , see @xcite . in particular , @xmath102 . moreover , @xmath83 and its stable and unstable invariant manifolds are persistent : given any smooth family of diffeomorphisms @xmath0 such that @xmath103 , then for each small enough @xmath21 there exists a nearby @xmath3-normally hyperbolic invariant manifold @xmath104 with @xmath30 families of immersed submanifolds @xmath105 . to compute the melnikov displacement , we will need to show that certain series are geometrically convergent ; the following lemma is a key component in this proof . [ lem : lnbound ] let @xmath106 be a diffeomorphism with a compact normally hyperbolic invariant manifold @xmath83 , and fix a point @xmath107 . then given any splitting @xmath108 , there exists a constant @xmath109 and an integer @xmath110 such that @xmath111 where @xmath112 are the restrictions of @xmath113 to the subspaces @xmath114 $ ] , @xmath115 . a similar bound holds for @xmath116 . there exists a unique point @xmath92 such that @xmath117 , where @xmath118 . according to the @xmath119-lemma for normally hyperbolic manifolds @xcite , the complementary subspaces @xmath120 tend to the unstable subspaces @xmath121 , as @xmath122 . thus , the maps @xmath123 tend to the unstable restrictions @xmath124 as @xmath122 , and the lemma follows from . it suffices to take any @xmath125 . the melnikov displacement will be defined as a function on the normal bundle of a _ saddle connection _ , which is defined as follows . let @xmath106 be a diffeomorphism with a pair of compact normally hyperbolic invariant manifolds @xmath83 and @xmath126 . a saddle connection between @xmath83 and @xmath126 is an invariant submanifold @xmath127 such that @xmath128 for all @xmath129 . [ rem : normalbundles ] the coincidence of the tangent spaces is needed in order that the manifolds have the same algebraic normal bundles on the saddle connection : @xmath130 since @xmath131 , for all @xmath129 . by definition , @xmath132 , and the manifolds @xmath83 and @xmath126 are not part of the saddle connection . the simplest ( and most common ) saddle connections are of the form @xmath133 . in this case , we say that the unperturbed invariant manifolds are _ completely doubled_. many melnikov problems studied in the literature fall into this category . nevertheless , in some problems there may exist points @xmath134 , @xmath135 , such that @xmath136 , see @xcite . in that case the saddle connection is strictly contained in the intersection of the stable and unstable invariant manifolds : @xmath137 . in this subsection we use the fundamental iterative equation to obtain infinite series for the displacements of the stable and unstable manifolds . these series are absolutely convergent , but only , as we must stress , when they are evaluated on their corresponding normal bundles . indeed , the tangential components of these series can be unbounded . consequently , in order to compute these sums , each term must be projected onto the normal bundle . an example will be given in [ ssec : suris ] . the proof of the following proposition is inspired by a proof given in @xcite , the main difference is that our setting is more geometric . [ pro : duds ] let @xmath0 be a smooth family of diffeomorphisms such that the unperturbed map @xmath1 has a compact normally hyperbolic invariant manifold @xmath138 with stable and unstable invariant manifolds @xmath139 . then the displacement vector fields @xmath140 and @xmath141 of the families of perturbed stable and unstable invariant manifolds are given by the absolutely convergent series @xmath142 we prove the claim about the stable displacement @xmath143 ; the unstable result is obtained analogously . repeatedly applying the iterative formula , yields @xmath144 for any integer @xmath145 . therefore , it suffices to check that the term @xmath146 tends geometrically to zero on the normal bundle of the unperturbed stable manifold . for any point @xmath147 , let @xmath148 , and @xmath149 . then @xmath150 where @xmath151 . we must show that @xmath152 tends geometrically to zero as an element of the quotient space @xmath153 . given any splitting @xmath154 , let @xmath155 be the projections onto the linear subspaces @xmath156 $ ] , and let @xmath112 be the corresponding restrictions of the linear maps @xmath157 . then , @xmath158 because @xmath159 and @xmath160 . let @xmath161 be the integer referred to in lemma [ lem : lnbound ] , and define @xmath162 . lemma [ lem : lnbound ] gives the bound @xmath163 for some @xmath109 . moreover , the sequence @xmath164 is bounded due to the compactness of the normally hyperbolic manifold @xmath138 and the continuity of the displacement vector field @xmath165 . therefore , @xmath166 tends geometrically to zero as @xmath167 , and so , @xmath152 tends geometrically to zero in the quotient space @xmath153 . we now will use the displacement vector fields to study the splitting of a saddle connection upon perturbation . as usual , @xmath0 denotes a smooth family of diffeomorphisms such that the unperturbed map @xmath1 has a saddle connection @xmath168 between a pair of compact @xmath3-normally hyperbolic invariant manifolds @xmath138 and @xmath169 . these manifolds persist and remain @xmath3-normally hyperbolic for small @xmath21 . we want to study the distance between the perturbed manifolds @xmath170 and @xmath171 the growth rate of this distance with @xmath21 is obtained simply by taking the difference between the displacement vector fields of both families . [ def : melnikov ] under the previous assumptions , the melnikov displacement is the canonical @xmath5 section of the normal bundle @xmath11 defined by @xmath172 where @xmath141 and @xmath140 are the displacement vector fields of @xmath173 and @xmath174 , respectively . the melnikov displacement @xmath70 makes sense only on the saddle connection , where the tangent spaces of the unperturbed invariant manifolds @xmath175 and @xmath176 coincide . away from @xmath2 , the difference @xmath177 is undefined because each term is a section of a different ( algebraic ) normal bundle , see remark [ rem : normalbundles ] . let @xmath15 be a smooth family of diffeomorphisms verifying the assumptions of definition [ def : melnikov ] . then its melnikov displacement @xmath178 is given by the absolutely convergent sums @xmath179 from proposition [ pro : duds ] , we get that @xmath180 , where the last equality follows from the identity @xmath181 . the melnikov displacement has been defined in a canonical way as a section of the normal bundle , as such it has strong geometric properties . we will show next that any change of variables acts as a pullback on it . this will imply a number of geometrical properties , for example that the displacement is invariant under the pullback and the push - forward of the unperturbed map . in addition , we will see that if there exist symmetries , reversors , first integrals , or saddle connections , then these have natural implications on the melnikov displacement . these claims are the subject of proposition [ pro : geometric ] . we recall that a diffeomorphism @xmath82 is _ symmetric _ when there exists a diffeomorphism @xmath182 such that @xmath183 , and then @xmath40 is called a _ symmetry _ of the map @xmath97 . analogously , @xmath97 is _ reversible _ when there exists a diffeomorphism @xmath184 such that @xmath185 , and then @xmath3 is called a _ reversor _ of the map @xmath97 . in many applications , @xmath3 is an involution , @xmath186 , though this need not be the case @xcite , finally , a function @xmath187 is a first integral of @xmath97 when @xmath188 . [ pro : geometric ] let @xmath0 be a smooth family of diffeomorphisms verifying the assumptions of definition [ def : melnikov ] and let @xmath70 be its melnikov displacement . 1 . [ item : h0 ] given any smooth family of changes of variables @xmath189 , the family @xmath190 also verifies the assumptions of definition [ def : melnikov ] , its melnikov displacement @xmath191 is defined on the saddle connection @xmath192 , and @xmath193 2 . the melnikov displacement is invariant by the pullback and the push - forward of the unperturbed map . that is , @xmath194 3 . if @xmath0 has a smooth family of : 1 . symmetries @xmath195 such that @xmath196 , then @xmath197 2 . reversors @xmath198 such that @xmath199 , then @xmath200 if , in addition , @xmath201 is an involution and its fixed set @xmath202 is a submanifold that intersects @xmath2 transversely at some point @xmath203 , then @xmath204 . first integrals @xmath205 such that @xmath206 , then @xmath207 4 . saddle connections @xmath208 such that @xmath209 , then @xmath210 . if there exists a vector field @xmath211 such that its flow commutes with @xmath0 , then the lie derivative of the melnikov displacement with respect to @xmath212 vanishes , that is , @xmath213 it is very important to stress that the above results make sense as identities on the normal bundle of saddle connections . the relation makes sense because the pullback @xmath214 maps @xmath215 onto @xmath216 , whereas makes sense because @xmath215 is invariant by the pullback @xmath72 and the push - forward @xmath217 . similar arguments apply to and . the identities and make sense because @xmath215 is contained in the kernel of the one - form @xmath218 and the vector field @xmath212 is tangent to @xmath2 , respectively . the hypothesis @xmath206 means that @xmath104 and @xmath219 are contained in the same level set of the first integrals , which can be assumed to be the zero level without loss of generality . this holds , for instance , in the transitive homoclinic case : @xmath220 and @xmath104 is transitive . we can write a geometric proof based on the definition of the melnikov displacement as a canonical section of the normal bundle , or a computational proof using the formulae . we follow the geometric approach . \(i ) the first claims are obvious . for the last one , note that if @xmath221 are any adapted deformations ( for the family @xmath0 ) , then @xmath222 are as well ( for the family @xmath223 ) . differentiating @xmath224 with respect to @xmath21 yields @xmath225 where we used that @xmath226 . hence , @xmath227 , so the difference @xmath228 verifies the relation @xmath229 , which is equivalent to . \(ii ) if we take @xmath230 in , then @xmath231 and @xmath232 . therefore , @xmath233 and follows from . ( iii.a ) if we take @xmath234 in , then @xmath235 and @xmath236 . therefore , @xmath233 and follows from . ( iii.b ) if we take @xmath237 in , then @xmath238 and @xmath239 . moreover , a stable ( respectively , unstable ) manifold of a map becomes an unstable ( respectively , stable ) for the inverse map . therefore , @xmath240 and follows from . next , we assume that @xmath201 is an involution whose fixed set @xmath241 intersects @xmath2 transversely at @xmath203 . that is , @xmath242 , @xmath243 , and @xmath244 . we want to prove that @xmath245 in the normal bundle @xmath11 , or equivalently that @xmath246 . since @xmath243 and @xmath242 , the square of the linear endomorphism @xmath247 is the identity map . this implies that @xmath248 is diagonalizable and its spectrum is contained in the set @xmath249 , so @xmath250 , where @xmath251 . we claim that @xmath252 and @xmath253 . the first claim follows from the fact that involutions are locally conjugate to their linear parts at fixed points . since @xmath254 is invariant under @xmath255 and complementary to @xmath256 in @xmath257 , we get the second claim . finally , the evaluation of at the point @xmath258 , yields @xmath259 , and so @xmath260 . ( iii.c ) let @xmath261 . obviously , @xmath262 . differentiating @xmath263 with respect to @xmath21 yields @xmath264 . thus , @xmath228 verifies . ( iii.d ) in this case , we can take @xmath265 , and so @xmath228 vanishes on @xmath266 . \(iv ) let @xmath267 be the flow of the vector field : @xmath268 and @xmath269 . thus , given any @xmath270 , the ( constant ) family @xmath271 is a smooth family of symmetries of @xmath0 . next , using relation , we get that @xmath272 for any @xmath270 , and follows by definition of the lie derivative . these results have been extensively used in the literature . the invariance of melnikov objects under the unperturbed map gives rise to periodicities when suitable coordinates are used ; examples can be found in @xcite . item ( iii.d ) implies a simple splitting criterion : if the melnikov displacement does not vanish identically , the separatrix splits @xcite . upper bounds on the number of uniform first integrals of the family @xmath0 can be deduced from item ( iii.c ) , see @xcite . this result has also been used in to establish necessary and sufficient conditions for uniform integrability of analytic , exact - symplectic maps @xcite . symmetries have also been extensively used , for example to improve the lower bound obtained by morse theory for the number of critical points of some melnikov potentials @xcite ( we will discuss melnikov potentials in [ sec : exactsymp ] . ) . relation is similar to noether s theorem , since the existence of a continuous symmetry the flow of the vector field @xmath212implies a conservation law for the melnikov displacement . finally , fixed sets of reversors can be used to guarantee the existence of heteroclinic points and zeros of melnikov functions ; this is an old trick , see @xcite . in the classical melnikov method , one uses simple zeros of a melnikov function to predict the transverse intersection of the invariant manifolds . we next show that this result also holds for the melnikov displacement . if @xmath203 is a simple zero of the melnikov displacement , then the perturbed invariant manifolds @xmath173 and @xmath174 intersect transversely at some point @xmath273 for small enough @xmath21 . by definition , the saddle connection @xmath2 is an @xmath1-invariant submanifold of @xmath16 . let @xmath274 be the projection of the normal bundle onto @xmath2 . there is a tubular neighbourhood @xmath34 of @xmath2 that is diffeomorphic to @xmath11 via a diffeomorphism @xmath275 , as illustrated in . since @xmath34 is an open neighbourhood of @xmath2 in @xmath16 , each deformation of @xmath2 occurs inside @xmath34 , for @xmath21 small enough . we note that under the diffeomorphism @xmath276 , and for @xmath21 small enough , deformations of @xmath2 can be thought as deformations of the zero section , @xmath277 , in @xmath11 . indeed , a deformation of @xmath2 can be parametrized by a section @xmath278 . let @xmath279 be an open set that contains @xmath203 . let @xmath280 and @xmath281 be deformations of @xmath282 such that @xmath283 let @xmath284 and @xmath285 be the images in @xmath11 of two deformations of the saddle connection , corresponding to the images of the stable and unstable manifolds . we want to show that @xmath286 and @xmath287 intersect transversely for @xmath21 small enough when @xmath70 has a simple zero . [ ptb ] .,height=192 ] we parametrize each manifold with a section @xmath288 . that is , if @xmath21 is small , then @xmath289 given by @xmath290 are sections of the normal bundle restricted to @xmath282 . notice that the functions @xmath291 given by @xmath292 are adapted deformations of @xmath282 with images @xmath293 . in fact , both are perturbations of the zero section . from the definition of displacements we have that @xmath294 where @xmath295 and @xmath296 , are the displacements of @xmath297 . therefore , each manifold @xmath286 and @xmath298 is the image of @xmath299 and @xmath300 and they intersect transversely at @xmath301 if and only if @xmath203 is a simple zero of the section @xmath302 . now we use the standard `` blow up '' argument so that the implicit function theorem can be applied . let @xmath303 be the section given by @xmath304 notice that @xmath305 is @xmath5 if @xmath306 is @xmath307 , and moreover that , when @xmath308 , @xmath309 is a simple zero of @xmath310 if and only if it is for @xmath311 . finally , since @xmath312 and @xmath286 and @xmath287 are the images of @xmath299 and @xmath300 , the implicit function theorem implies that if @xmath203 is a simple zero of @xmath70 , then @xmath286 and @xmath287 intersect transversely near @xmath203 , for @xmath21 small enough . thus @xmath174 and @xmath173 intersect transversely near @xmath203 , for @xmath21 small enough . as a simple example , we consider the generalized standard map @xmath313 where @xmath314 is a periodic potential . it is easy to see that @xmath97 preserves area and orientation and that its fixed points have the form @xmath315 where @xmath316 . such fixed points are saddles if and only if @xmath317 , because @xmath318 . following mcmillan @xcite , we can find a generalized standard map with a saddle connection between two saddle points if we choose a diffeomorphism @xmath319 such that @xmath320 , and let @xmath321 to see this , first note that , with this choice , the force @xmath322 is periodic with period one . moreover , if @xmath323 is a hyperbolic fixed point of @xmath324 ( that is , @xmath325 and @xmath326 ) , then @xmath327 thus , @xmath315 is a saddle fixed point . moreover , the graphs of the functions @xmath328 are invariant and the dynamics on these sets is very simple : @xmath329 for all @xmath330 . these graphs contain saddle connections if we choose a pair of neighbouring fixed points @xmath331 and @xmath332 of @xmath324 , so that @xmath324 has no fixed points in @xmath333 . suppose further that @xmath331 ( respectively , @xmath332 ) is a stable ( respectively , unstable ) fixed point of @xmath324 , so that @xmath334 and @xmath335 for all @xmath336 . then @xmath337 and @xmath338 are saddle points of the map @xmath97 , and @xmath339 are saddle connections between them : @xmath340 and @xmath341 , see . it is known that a generalized standard map with a potential of the form is typically nonintegrable @xcite . an integrable example , @xmath1 , for each @xmath342 is obtained when the diffeomorphism @xmath324 is given by @xmath343 for @xmath344 . this function , when extended to @xmath345 using @xmath346 , gives the period - one potential @xmath347 a first integral of the map is @xmath348 @xcite . the diffeomorphism is conjugate to a mbius transformation , and it is easy to find an explicit formula for its iterations : @xmath349 the points @xmath350 and @xmath351 are fixed points of @xmath324 . the point @xmath331 is stable and @xmath332 is unstable because @xmath352 and @xmath353 . we now perturb the integrable map by modifying the potential : @xmath354 we compute the melnikov displacement using . for this calculation we must first compute the vector field @xmath355 , which for the generalized standard map with potential is @xmath356 using the relations and , the series for the displacement @xmath70 is easily computed . however , in order to ensure convergence of the series , we must take into account the fact that it is only the normal component of the displacement that is desired ; indeed , the iteration of the tangential component is not bounded . let @xmath357 be the canonical projection onto the normal bundle . due to the fact that @xmath2 is invariant under @xmath1 , we have that @xmath358 . to avoid numerical errors , we project each term in the sum . we display in the projection @xmath359 as a function of @xmath360 along @xmath361 , for two perturbative potentials @xmath362 . the melnikov displacement generalizes the classical methods used to detect the splitting of separatrices due to poincar and melnikov . poincar s method @xcite is based on the existence of first integrals , such as an energy function , for the unperturbed system . assume that @xmath1 has a saddle connection @xmath2 between a pair of hyperbolic fixed points @xmath331 and @xmath332 . for the simplest case , @xmath2 has codimension one , and a single first integral suffices : @xmath363 . saddle connections with a higher codimension can be treated in a similar way if there are sufficiently many integrals . the splitting between the stable and unstable manifolds is measured by the rate of change in the first integral @xmath364 with @xmath21 on the saddle connection . that is , we define the poincar function @xmath365 by @xmath366 where @xmath367 and @xmath368 are deformations adapted to the perturbed invariant manifolds . under the above assumptions , @xmath369 where @xmath70 is the melnikov displacement . the proof is a simple computation . the term @xmath370 makes sense because the first integral is constant on the saddle connection , and so the tangent space @xmath215 in contained in the kernel of @xmath371 . if @xmath372 is nondegenerate , then a point @xmath373 is a simple zero of @xmath374 if and only if it is a simple zero of @xmath70 . in this case , the perturbed invariant manifolds intersect transversely near @xmath203 for @xmath21 small enough . therefore , the melnikov displacement @xmath70 generalizes the estimates of splitting in the poincar style . the classical melnikov method @xcite is based on estimating the movement of a manifold in a direction _ _ n__ormal to the separatrix . to define the normal , the phase space @xmath16 is assumed to have a riemannian inner product @xmath375 . as before , assume for simplicity that @xmath2 is a saddle connection of codimension one . the appropriate normal to @xmath2 is called an adapted normal vector field . [ def : aniv ] let @xmath2 be a @xmath1-invariant submanifold of codimension one . a vector field @xmath376 is adapted normal when 1 . @xmath377 is nondegenerate : @xmath378 , for all @xmath379 , 2 . @xmath377 is normal : @xmath380 , for all @xmath381 and @xmath382 ; and 3 . @xmath377 is invariant : @xmath383 , for all vector field @xmath384 . using this notation , the classical melnikov function @xmath385 is @xmath386 where @xmath367 and @xmath368 are deformations adapted to the perturbed invariant manifolds . consequently , @xmath387 is related to the melnikov displacement by @xmath388 since @xmath377 is nondegenerate , @xmath387 and @xmath70 have the same simple zeros . we also note that makes sense because @xmath377 is normal , and so the component of @xmath70 in the tangent space @xmath215 does not play any role . adapted normal vector fields can be obtained as gradients of nondegenerate first integrals . let @xmath364 be a smooth first integral of @xmath1 such that its gradient @xmath389 does not vanish on @xmath2 . given a riemannian structure , the gradient is the unique vector field such that @xmath390 for any vector field @xmath391 on @xmath16 . therefore , @xmath392 , and @xmath393 is an adapted normal vector field . obviously , the poincar function and the melnikov function coincide when @xmath393 . in this section , we will see how the melnikov displacement generalizes previous theories developed for hyperbolic fixed points of exact symplectic maps . we will also show why general normally hyperbolic invariant manifolds can not be studied in the same way . a @xmath394-dimensional manifold @xmath16 is _ exact symplectic _ when it admits a nondegenerate two - form @xmath14 such that @xmath395 for some liouville one - form @xmath119 . the typical example of an exact symplectic manifold is provided by a cotangent bundle @xmath396 , together with the canonical forms @xmath397 and @xmath398 , in cotangent coordinates @xmath399 . a map @xmath82 is _ exact symplectic _ if @xmath400 for any closed path @xmath401 or , equivalently , if there exists a _ generating _ ( or _ primitive _ ) _ function _ @xmath402 such that @xmath403 . in particular , an exact symplectic map is _ symplectic _ : @xmath404 . a submanifold @xmath34 of @xmath16 is _ exact isotropic _ if @xmath400 for any closed path @xmath405 or , equivalently , if there exists a generating function @xmath406 such that @xmath407 . here , @xmath408 denotes the natural inclusion map . in particular , an exact isotropic submanifold is _ isotropic _ : @xmath409 . the maximal dimension of an isotropic submanifold is @xmath17 , and when the dimension is @xmath17 , the submanifolds is called _ lagrangian_. a vector field @xmath410 is _ globally hamiltonian _ if there exists a _ hamiltonian _ function @xmath411 such that @xmath412 we stress two key properties of the symplectic case . first , the generator of a family of exact symplectic maps is globally hamiltonian and there exists a simple relation between this hamiltonian and the generating function of the maps . second , the stable and unstable invariant manifolds of a connected normally hyperbolic invariant submanifold @xmath83 of an ( exact ) symplectic map are ( exact ) isotropic if and only if @xmath83 is a fixed point , in which case they are lagrangian ( that is , @xmath17-dimensional ) . the second property is an obstruction to develop a symplectic version of our canonical melnikov theory for general normally hyperbolic invariant submanifolds , we shall do it just for fixed points . these properties are well - known , but we prove both for completeness . [ pro : hamiltonian ] let @xmath0 be a family of exact symplectic maps with generating function @xmath413 and generating vector field @xmath8 . then , @xmath8 is globally hamiltonian with hamiltonian @xmath414 by definition , @xmath415 , so the lie derivative with respect to @xmath8 is @xmath416 . using cartan s formula : @xmath417 , and taking the derivative with respect to @xmath21 of the relation @xmath418 , we get @xmath419 rearranging this yields @xmath420 finally , @xmath421 , since @xmath119 is a one - form . [ pro : isotropic ] let @xmath82 be a diffeomorphism with a connected , normally hyperbolic , invariant manifold @xmath83 . if @xmath97 is ( exact ) symplectic , the stable and unstable invariant manifolds @xmath422 are ( exact ) isotropic if and only if @xmath83 is a hyperbolic fixed point . moreover , in this case the submanifolds @xmath423 are ( exact ) lagrangian . since @xmath83 is normally hyperbolic , implies that for each @xmath92 , @xmath424 , @xmath425 and @xmath426 sum to @xmath427 . since @xmath97 is symplectic @xmath428 . therefore @xmath429 . suppose that @xmath423 are isotopic ; then since isotopic manifolds have maximum dimension @xmath17 , we must have @xmath430 . consequently , @xmath431 and since @xmath83 is connected it is a hyperbolic fixed point . conversely , assume that @xmath432 , so that @xmath433 for some hyperbolic fixed point @xmath331 and @xmath434 . to prove that @xmath435 is lagrangian , take any two vectors @xmath436 . we know that @xmath437 and @xmath438 tend to zero as @xmath439 , since the stable directions are uniformly contracted . since @xmath97 preserves @xmath14 , we have @xmath440 as @xmath439 . if , in addition , @xmath97 is exact symplectic , then for every closed loop @xmath441 , @xmath400 . now suppose that @xmath442 , so that @xmath443 as @xmath444 . then @xmath445 finally , the ( exact ) lagrangian character of @xmath116 follows from the fact that it is the stable manifold for @xmath446 . let @xmath15 be a family of exact symplectic maps such that @xmath1 has an exact lagrangian saddle connection @xmath447 between two hyperbolic fixed points @xmath331 and @xmath332 . note that in this case that the stable and unstable manifolds are as smooth as the map @xmath0 . we assume that @xmath448 , where @xmath449 is the hamiltonian . without loss of generality , we can assume that @xmath450 . the natural measure of splitting for this case is a real valued function @xmath451 , the _ melnikov potential _ , whose derivative measures the splitting @xcite . in other words , nondegenerate critical points of the melnikov potential predict transverse splitting . in this subsection , we shall find the relation between the one form @xmath452 introduced by @xcite and the melnikov displacement . for a saddle connection @xmath447 and melnikov displacement @xmath453 , the function @xmath454 implicitly defined by @xmath455 is the melnikov potential .. ] we will show next that @xmath451 is indeed defined by and that its critical points correspond to zeros of the displacement . the pullback of the one - form @xmath456 to the saddle connection @xmath447 is well - defined and exact . in particular , there exists a function @xmath457 , determined uniquely up to additive constants that obeys . moreover , the set of simple zeros of the melnikov displacement @xmath70 coincides with the set of nondegenerate critical points of the function @xmath451 . [ rem : isotropic ] let @xmath34 be a submanifold of @xmath16 and @xmath458 a section of its normal bundle . then the pullback of the one - form @xmath459 to the submanifold @xmath34 is _ well - defined _ if and only if @xmath34 is isotropic . this has to do with the fact that @xmath460 is well - defined if and only if @xmath461 for any vector fields @xmath462 and @xmath463 . the pullback is well - defined because @xmath447 is lagrangian , see remark [ rem : isotropic ] . with regard to the exactness , it suffices to prove that @xmath464 for any closed path @xmath465 . let @xmath441 a closed path contained in the saddle connection , then @xmath466 where , since @xmath1 is symplectic , the melnikov displacement is invariant under the pullback of @xmath1 , and the saddle connection is invariant under @xmath1 . finally , since @xmath467 , we obtain that @xmath468 as @xmath444 . the equivalence between simple zeros of @xmath70 and nondegenerate critical points of @xmath451 follows from the lagrangian character of the saddle connection @xmath447 . an explicit series for @xmath451 can be obtained using the hamiltonian . the melnikov potential is given by the absolutely convergent series @xmath469 since @xmath1 is symplectic , and the generator @xmath355 is globally hamiltonian with hamiltonian @xmath470 : @xmath471 , we deduce that @xmath472 the series converges absolutely because @xmath450 , @xmath473 , and @xmath474 . in fact , it converges at a geometric rate . while the melnikov potential has been used many times for exact symplectic , twist , and hamiltonian maps , the formulation give here , using the power of deformation theory , is more elegant . the melnikov potential introduced here is identical to the one defined in @xcite . this can be checked by direct comparison of the formula with the formula ( 2.7 ) of the cited paper , using to express @xmath470 in terms of the derivative of the generating function @xmath413 at @xmath475 . we also note that it is impossible to define a melnikov potential " on the saddle connection of normally hyperbolic invariant manifolds with nonzero dimension , because then the saddle connection is not isotropic ( proposition [ pro : isotropic ] ) and so the identity makes no sense ( remark [ rem : isotropic ] ) . nevertheless , even in this case the melnikov displacement is defined . in this subsection , we restrict to the two - dimensional case in order to show in a simple way that the melnikov potential and the classical melnikov function are transparently related . moreover , we will see that in some cases there exist geometric obstructions to the non - vanishing of melnikov functions . these points are most easily seen by choosing a special time - like parametrization of the saddle connection . we consider the standard symplectic structure on the plane , @xmath476 , and let @xmath477 be the standard @xmath478 symplectic matrix : @xmath479 . suppose that @xmath480 is a family of diffeomorphisms preserving area and orientation , and @xmath449 is the generating hamiltonian for @xmath8 , the generator for @xmath0 . we assume that the unperturbed map has a saddle connection @xmath481 between two hyperbolic fixed points @xmath331 and @xmath332 such that @xmath482 . note that the unperturbed map not need be integrable . the key point is that , on a one - dimensional saddle connection , there is a parametrization @xmath483 such that @xmath484 in many cases , such parametrizations can be expressed in terms of elementary functions , and the melnikov function can be explicitly computed @xcite . consequently , @xmath485 provides a diffeomorphism between the saddle connection @xmath2 and the real line , so that objects defined over @xmath2 can be considered as depending on the real variable @xmath270 . thus , for example , the melnikov potential can be replaced by @xmath486 to become a function @xmath487 given by @xmath488 here we abuse the notation by not giving the function a new name . our goal is to show that the classical melnikov function , or rather the composition @xmath489 , can be computed by differentiating @xmath451 : @xmath490 in addition we will show that these functions have the properties * periodicity : @xmath491 and @xmath492 . * in each fundamental domain @xmath493 , @xmath16 must vanish . indeed , @xmath494 for any @xmath495 . ( this property will be generalized to volume - preserving maps in the next section . ) * near each simple zero of @xmath496 or nondegenerate critical point of @xmath497 there is transverse intersection of the stable and unstable manifolds . given and , the first two properties are obvious . periodicity is simply a consequence of the invariance of under @xmath498 ; this invariance under the unperturbed map is a property of all of the melnikov objects introduced so far . the second property is a simple consequence of periodicity and integration of . to show that @xmath496 is actually the classical melnikov function we must construct an adapted normal vector field @xmath377 on @xmath447 , recall definition [ def : aniv ] . we claim that , when thought of as a function of @xmath270 , such a vector field @xmath499 is given by @xmath500 this claim is proved in lemma [ lem : alpha ] at the end of this subsection . the relation between @xmath16 and the classical melnikov function follows from a straightforward computation of the derivative using @xmath501 : @xmath502 thus @xmath503 which is the obvious form of under the parametrization @xmath485 . this verifies and the final property . to end this subsection , it remains to prove the claim about the vector field @xmath377 . [ lem : alpha ] the vector field @xmath504 defined by @xmath505 is an adapted normal vector field on the saddle connection @xmath2 . since @xmath1 is symplectic , @xmath506 . consequently , @xmath507 thus , given any vector field @xmath508 , we have @xmath509 moreover , since @xmath477 is antisymmetric , @xmath510 . therefore , according to definition [ def : aniv ] , @xmath504 is an adapted normal vector field . for the case of a volume - preserving mapping with a codimension - one saddle connection , an adapted normal field formulation of the melnikov function also applies @xcite . here we show how to relate this to the melnikov displacement . let @xmath511 be a family of volume - preserving diffeomorphisms on an oriented @xmath17-dimensional manifold @xmath16 with volume form @xmath512 such that @xmath1 has a codimension - one saddle connection @xmath2 between two normally hyperbolic invariant sets @xmath83 and @xmath126 . we start with a simple lemma about the generator for @xmath0 . [ pro : exact ] let @xmath8 be the generator of a volume - preserving smooth family @xmath0 . then 1 . the divergence of @xmath8 with respect to @xmath512 is zero . 2 . the one - form @xmath513 is closed . if @xmath16 is simply connected , then @xmath513 is exact . from ( * ? 2.2.21 ) , we have that @xmath514 . since the divergence is defined by @xmath515 , this implies that the divergence vanishes . moreover , since @xmath516 and @xmath517 , then @xmath518 , implying ( ii ) and ( iii ) . in order to find an invariant nondegenerate @xmath519-form from any adapted normal field , we assume that @xmath16 has a riemannian metric @xmath520 . if @xmath377 is an adapted vector field ( cf . definition [ def : aniv ] ) , then @xmath521 is a nondegenerate @xmath519-form on @xmath2 that is invariant under the restriction @xmath522 . by definition @xmath377 is nonzero , so that @xmath523 is nondegenerate . to prove that @xmath524 on the saddle connection @xmath2 , we introduce the vector field @xmath525 defined by @xmath526 this vector field is tangent to @xmath2 , because @xmath527 . now , we compute the difference @xmath528 hence , it suffices to see that the @xmath519-form @xmath529 vanishes identically on the tangent space @xmath530 . this follows from the fact that @xmath531 is tangent to @xmath2 and @xmath532 . the melnikov function associated with @xmath377 is defined using @xmath523 . [ def : java ] let @xmath70 be the melnikov displacement . given an adapted vector field @xmath377 on @xmath2 , we define the melnikov function @xmath533 as the unique @xmath5 function such that @xmath534 as @xmath519-forms on @xmath2 . the previous definition first appeared in @xcite . note that @xmath535 is an @xmath519-form , but it is possible that it might be degenerate ; if this were the case then zeros of @xmath387 need not correspond to those of @xmath70 . we will show next that this is not the case . [ pro : java ] the melnikov function @xmath387 is invariant under the map @xmath1 and @xmath536 . moreover , a point @xmath203 is a zero of @xmath387 if and only if it is a zero of @xmath70 . using , we obtain @xmath537 where @xmath538 . since @xmath539 , we conclude that @xmath540 , as an @xmath519-form on @xmath2 , and thus @xmath536 . since @xmath377 is an adapted field , and @xmath70 is invariant , @xmath541 . therefore , @xmath542 . finally , @xmath543 only when @xmath204 , since @xmath377 is nonzero and normal to @xmath215 , and @xmath70 is not in the tangent space . we would like to show , as we did for the area - preserving case in , that the volume - preserving melnikov function necessarily has zeros on @xmath2 . to do this , we will show that the integral of @xmath387 with respect to the measure @xmath523 is zero . this can be accomplished by dividing the saddle connection into pieces fundamental domains that are mapped into each other by @xmath1 . [ def : guasa ] let @xmath83 be a compact normally hyperbolic invariant manifold of a diffeomorphism @xmath1 with stable manifold @xmath544 . a proper boundary , @xmath441 , is a submanifold of @xmath544 that bounds an isolating neighbourhood of @xmath83 in @xmath544 . in other words , @xmath441 is proper if there is an closed submanifold @xmath545 such that 1 . @xmath546 , and 2 . @xmath547 . we refer to the closed set @xmath545 as the stable manifold _ starting at _ @xmath441 . similarly , for the unstable manifold , a submanifold @xmath548 is proper it is proper for @xmath549 . however , in this case we define the unstable manifold _ up to _ @xmath550 , denoted by @xmath551 , as the _ interior _ of the local manifold that corresponds to @xmath549 . notice that the definition is not symmetric , because @xmath545 is a closed subset of @xmath544 , while @xmath551 is open in @xmath552 . the asymmetry is just a technicality in order to simplify some proofs . let @xmath83 be a hyperbolic invariant set for @xmath1 . a submanifold with boundary @xmath553 is a fundamental domain of @xmath544 if there exists some proper boundary @xmath554 such that @xmath555 equivalently , a fundamental domain in @xmath552 is a manifold with boundary of the form @xmath556 where @xmath550 is a proper boundary in @xmath552 . in each case , the fundamental domain is a manifold with boundary @xmath557 . an immediate consequence of the definition is that all the forward and backward iterations of a fundamental domain are also fundamental . it is easy to see that proper boundaries always exist , and in fact , the unstable manifold can be decomposed as the disjoint union of fundamental domains : @xmath558 the importance of fundamental domains is that much of the information about the entire manifold can be found by looking only at these submanifolds . for example , as discussed in @xcite , the topology of the intersections of @xmath116 and @xmath435 can be studied by restricting to @xmath553 . [ pro : sony ] let @xmath16 be a simply connected manifold and @xmath15 a family of volume - preserving maps such that @xmath1 has a saddle connection @xmath2 with fundamental domain @xmath553 . if @xmath523 is the @xmath519-form defined in and @xmath387 is the melnikov function , then @xmath559 . the fundamental domain @xmath553 is a submanifold with boundary , such that @xmath560 , where @xmath441 is a closed curve that does not intersect @xmath561 . if we give an orientation @xmath562 $ ] to @xmath553 , the induced orientation on the boundary satisfies @xmath563 = -\left[f_0(\gamma)\right]$ ] . according to proposition [ pro : exact ] , the form @xmath564 is exact . thus there exists an @xmath565-form @xmath566 such that @xmath567 . from we conclude that @xmath568 where @xmath569 . finally , from definition [ def : java ] , the stokes s theorem , and the invariance @xmath570 , we get that the integral @xmath571 vanishes . the previous result implies that the stable and unstable manifolds of a perturbed saddle connection necessarily intersect . examples of such intersections were computed for the case @xmath572where the hypothesis of proposition [ pro : sony ] are satisfied in @xcite . we have studied a general theory of the melnikov method that can be applied to many different settings . the formula for the melnikov displacement @xmath70 generalizes many of the classical methods that use a normal vector field to measure the displacement with respect to a natural direction . for example , when the saddle connection is defined as the level set of a first integral @xmath364 , the classical poincar function @xmath573 measures the splitting as the rate of change of the first integral . if there is a riemannian structure and an associated adapted normal vector field @xmath377 , then the classical melnikov function @xmath574 measures the rate of change of the splitting in this normal direction . for the case of exact symplectic maps with saddle connections between hyperbolic fixed points , the melnikov potential @xmath451 is defined on a lagrangian submanifold and acts as a generator for the displacement : @xmath575 . this melnikov theory can be extended to other situations and can be applied in many problems . for instance , one can study billiard dynamics inside a perturbed ellipsoid , following a program initiated in @xcite . it turns out that the billiard map inside an ellipsoid is an exact symplectic diffeomorphism defined on the cotangent bundle of the ellipsoid , which has a two - dimensional normally hyperbolic invariant manifold with a three - dimensional saddle connection . therefore , ellipsoidal billiards represent a strong motivation for a more detailed study of general normally hyperbolic invariant manifolds in a symplectic framework . this is a work in progress . in discrete volume geometry there are many open melnikov problems , since the application of melnikov methods to volume - preserving maps began just a few years ago @xcite . we plan to continue this program in several ways . as a first step , we plan to obtain bounds on the number of primary heteroclinic orbits in terms of the degree of the polynomial perturbation @xcite . hl was supported in part by asociacin mexicana de cultura . jdm was supported in part by nsf grant dms-0202032 and by the mathematical sciences research institute in berkeley . rr - r was supported in part by mcyt - feder grant mtm2006 - 00478 . useful conversations with guillermo pastor , amadeu delshams , richard montgomery , clark robinson , renato calleja and rafael de la llave are gratefully acknowledged .

we study the dynamic process of splitting a condensate by raising a potential barrier in the center of a harmonic trap . we use a two - mode model to describe the phase coherence between the two halves of the condensate . furthermore , we explicitly consider the spatial dependence of the mode funtions , which varies depending on the potential barrier . this allows to get the tunneling coupling between the two wells and the on - site energy as a function of the barrier height . moreover we can get some insight on the collective modes which are excited by raising the barrier . we describe the internal and external degrees of freedom by variational ansatz . we distinguish the possible regimes as a function of the characteristic parameters of the problem and identify the adiabaticity conditions .

cond-mat0005302

bose - einstein condensation of an ideal gas is typically presented in introductory textbooks solely in terms of particle numbers . and quantum mechanically enhanced number densities were the ` smoking gun ' observed in the first experiments on dilute gas condensates . but many phenomena of interacting condensates depend critically on the conjugate quantity to particle number , namely the quantum mechanical phase @xcite . one can have highly occupied states with or without phase coherence between them , and the presence or absence of phase coherence can make a dramatic difference in the physical properties of an ultra - cold gas . as in the case of a gas held in an optical lattice , which can be a superfluid or a mott insulator depending on the strength of the lattice , the onset or loss of phase coherence can even be a phase transition @xcite . since the global phase of a condensate is unobservable , the simplest system in which phase coherence can be manifested consists of two states , occupied by a large number of bosons . as such a system can realistically be approximated by a condensate in a double well , it has recently attracted attention @xcite . in these works , josephson oscillations and the phase coherence between two coupled condensate were studied considering time independent coupling parameters . in @xcite the disappearance of the phase coherence between the two wells due to a change in time of the tunneling coupling has been studied . the coupling parameters and the related phase coherence properties depend on the potential barrier between the two wells , since in the limit where the barrier is very low we have a single condensate and in the limit where the barrier is very high we have two completely separated condensates . even if , at least in the high barrier limit , it has often been pointed out how to relate the coupling parameters ( on - site energy and tunneling coupling ) to the overlap of the wavefunctions localised in the two wells , this has never really been taken explicitly into account . introducing the spatial degrees of freedom , as we did , allows us to relate all that to the potential barrier in a more than phenomenological way and becomes expecially important in the study of the dynamics of the process , because the wavefunctions change drastically in time and collective modes are excited . in this paper we examine this problem and develop a method which can be extended to the case of many wells , in order to encompass the turning on of an optical lattice in a condensate , allowing to go beyond a gross - pitaevkii treatment , as the one done for example in @xcite . the physical situation is similar to the ones already realized experimentally , where a double well potential was created by shining a far - off resonant laser beam in the center of the magnetic trap ( see e.g. @xcite ) or where an array of traps was created by on optical standing wave @xcite . we are interested in the full dynamics of the process : in addition to the phase coherence properties , we want to study the excitation of the collective modes by taking the spatial dependence of the condensate wavefunction explicitly into account . the price of including all these effects without assuming mean field theory is that we must use a time - dependent variational approach , choosing variational ansatz to describe both the `` internal '' and `` external '' dynamics ( that is , the distribution of particles between two motional states treated as given , and the evolution of the spatial wave functions of these states ) . this allows us to reduce the intractable full problem to a set of coupled differential equation for our few variational parameters . although the time - dependent variational approach is not guaranteed to be quantitatively accurate , it allows qualitatively important processes to be investigated , and it has proven surprisingly reliable in previous applications to condensate physics @xcite . here we use it to derive the coupling between the internal and external dynamics , investigate which are the conditions under which the two can be decoupled , and identify the typical time scales for both . in sec . [ model ] we briefly present our two - state model and discuss qualitatively the behaviour of the system that our model can explain . in sec . [ ansatz ] , we introduce explicitly the variational ansatz we choose to describe the phase dynamics and the collective modes . after introducing the time dependent variational principle and deriving the lagrangian , we write the equations of motion for the variational parameters . then in sec . [ decouple ] we discuss the conditions under which it is an accurate approximation to neglect the coupling between internal and external degrees of freedom . this decoupling allows us to model both internal and external evolution in a still simpler way . the external dynamics is studied in sec . [ finexc ] , comparing simple analytic models with numerical results . [ numres ] is devoted to the internal dynamics , where the statics is studied and analytic estimates obtained . with these decoupled studies to guide expectations , the full variational equations of motion , with coupled internal and external degrees of freedom , will be studied numerically in sec . [ results ] . the results are compared with those of the phase model describing the relative phase of two weakly coupled superfluids @xcite . we conclude with a general discussion of our results and their implications . two appendices define the number difference and relative phase operators , derive the phase model hamiltonian , and show that our variational ansatz adequately represents the evolution generated by it . let us consider the situation in which we have @xmath0 bosons confined in a harmonic trap at zero temperature . we slowly deform the trap symmetrically around its center raising a potential barrier , until it becomes a double well potential . we want to study the dynamics of the process as well as the final state of the bosons . we will treat the problem in one dimension . this describes the situation in which we have a cigar - shaped condensate , and we deform it along the most elongated direction . nevertheless , our model could be extendend in a straightforward way to @xmath1 and @xmath2 dimensions . this would just lead to more complicated equations , but would not affect the main results . the second quantization hamiltonian describing the situation we have in mind and in which particles interact via a @xmath3-pseudopotential is @xmath4 where @xmath5 is the bosonic field operator and @xmath6 is the time dependent potential which describes the deformed trap . here , @xmath7 is an effective coupling constant which depends both on the @xmath8-wave scattering length and on the atomic distribution in the transverse directions . we assume that at time @xmath9 the atoms are in the ground state of this hamiltonian , and we want to determine the state after the potential is deformed . this problem can not be solved even numerically ( see , for example , @xcite and references therein ) . therefore , we need to consider a simplified model which describes the main features of the process . if the process is completely adiabatic , the final state will be a ` fragmented condensate ' with half of the particles in each of the potential wells : when the two condensates do not interact this is a much lower energy state than the one with phase coherence , because minimizing fluctuations in the relative particle number lowers the energy due to interparticle repulsion . such a fragmented condensate can be regarded as having two entirely independent condensates . if we changed the potential very fast , then we would obtain a single condensate that oscillates in each of the potential wells . we would also expect to see collapses and revivals in the `` condensate phase '' @xcite , provided the losses are not important @xcite . it is clear that the gross pitaevskii equation ( gpe ) will not give a good description of the splitting process , in principle . this equation describes the evolution of a single mode of the condensate @xmath10 , and , therefore , is not valid for fragmented condensates . in order to interpolate between the gross - pitaevskii limit of phase coherence and the limit of two independent condensates , one needs to consider at least two modes @xmath11 . then , one can write the state of the atoms as ( we assume @xmath0 to be even ) @xmath12 and @xmath13 are the mode annihilation operators defined as @xmath14 we have to consider the evolution of the wavefunctions @xmath15 as well as of the coefficients @xmath16 , which will be coupled and governed by the hamiltonian @xmath17 we will refer to the evolution of those wavefunctions as `` external dynamics '' and to the one of the coefficients as `` internal dynamics '' . in order to find the mode functions @xmath15 we can use the variational principle ( in the same way as one derives the gpe from a hartree - fock ansatz ) . however , this also turns out to be very complicated . a way around that problem is to express @xmath15 in terms of some few variational parameters : we will use quasi gaussian functions to describe those mode - functions . on the other hand , we could also use a variational principle to determine the evolution of the coefficients @xmath16 . this again turns out to be very complicated , so that we will also use a gaussian ansatz for them . once they are known , in order to estimate when the condensate is fragmented , we can look at the eigenvalues of the single particle density operator corresponding to the internal dynamics only , that is , the matrix @xmath18 where @xmath19 mean the expectation value on the state @xmath20 . the eigenvalues @xmath21 indicate whether we have a single condensate ( @xmath22 , @xmath23 ) , or a fragmented one ( @xmath24 ) . as already extensively discussed in the literature @xcite , the two - mode model has a limited validity in the case of low barrier , when in principle one is not allowed to neglect higher excited modes . however , it becomes more and more accurate the higher the barrier gets , since in this case the two lower lying modes move closer together in energy compared to the higher ones . hence it should allow a good description of the splitting process . before presenting the numerical and analytical result coming from our analysis , we will briefly discuss the qualitative behavior that we expect from the model under study . we will show that the equations we will derive for the external and internal dynamics can be decoupled to a very good approximation . that is , we can first solve the equations for the external dynamics basically taking constant values for the variational parameters describing the coefficients @xmath16 . once these equations are solved , we can use the corresponding wavefunctions @xmath15 to calculate the time dependent coefficients for the equations that describe the internal variational parameters . in summary , once we have solved the equations for the external parameters , we are left with a two mode model with time dependent coefficients which contain all information about the external dynamics : they will define the hopping and on - site interaction , whose competition determines the phase relation between the two modes . regarding the external dynamics , one can see two kinds of behaviors depending on the time scale @xmath25 at which the barrier is raised . the important time scale with which one has to compare is the oscillation period @xmath26 in the trapping potentials at each time . these periods change by roughly a factor of two between the initial harmonic potential and the final double well ( with our specific choice of the trapping potential ) . thus , if @xmath27 the process will be adiabatic with respect to the external dynamics , which means that @xmath11 will basically correspond to the two ground states of the right and left wells at the final time . if @xmath28 we will have collective excitations , in which @xmath11 oscillate strongly . in this case , we will have that the energy of the condensate @xmath29 ( with respect to the ground state energy ) has increased , so that it may be destroyed . although we can not account for the disappearance of the condensate within our model , we can estimate when this will happen just by considering the fact that under normal circumstances thermalization will occur , and , therefore , the condensate will be destroyed for @xmath30 , where @xmath31 , @xmath32 is the boltzman constant , @xmath33 indicates the critical temperature and @xmath29 is the extra energy in the final state . we find that the condensate disappears for @xmath34 . regarding the internal dynamics , there is also a time scale in the problem which determines the dynamics ; this is the revival time @xmath35 . given two condensates with an initial well defined relative phase , it is well known that the relative phase first disappears ( collapse ) and then is restored at time @xmath36 @xcite . if @xmath37 the process will be adiabatic with respect to the internal dynamics , the phase coherence will be lost during the process and therefore we will end up with two independent condensates in each well , with no phase coherence at all ( note that it makes no sense to talk about collapses or revivals in this situation ) . if @xmath38 at the end of the process we will have two condensates with a good phase coherence . in that case , after the splitting , collapses and revivals could be observed provided the particle losses are practically absent . in summary , we have two important time scales in the problem , namely @xmath26 and @xmath39 . typically , in experiments @xmath40 , so that it will be harder to be adiabatic with respect to the internal dynamics than to the external one . on the other hand , since @xmath39 is very long in practise , it will be hard to achieve @xmath41 within the validity of our model , in which particle losses and other imperfections are not included . finally we notice that the external and internal dynamics depend in a very different way on the parameters @xmath7 and @xmath0 . very similar to what happens in the gpe , the equations of motion for the parameters describing the external dynamics depend almost only on the product @xmath42 . on the contrary , the on - site energy splitting scales like @xmath7 and the tunneling coupling like @xmath0 giving rise to very different internal dynamics . for increasing @xmath0 and decreasing @xmath7 the relative phase becomes better defined and the time @xmath39 required to destroy the phase coherence is longer . in particular in the limit @xmath43 , unless the tunneling coupling exactly vanishes , the gpe case of phase coherent condensate is recovered . in this section we introduce a variational ansatz to describe the internal and external dynamics . to describe the ground state of the system , which we will call equivalently static or equilibrium solution , two parameters are sufficient : @xmath44 which corresponds roughly to the center of the mode functions , and @xmath45 which is related to the width of the number distribution . to allow dynamic evolution we have to introduce also the variables @xmath46 and @xmath47 , which will vanish in the static case . for simplicity in the following we present the variational ansatz for the symmetric case , describing a spatially symmetric double well potential and a symmetric atomic distribution among the two modes . starting from symmetric initial conditions ( no unbalance in the population of the two modes and symmetric mode functions @xmath48 ) , the symmetry will be preserved under the time evolution . nevertheless , it is possible to generalize our ansatz to describe asymmetric situations . we will briefly discuss the corresponding results in the conclusions . for a single condensate in a harmonic trap , a gaussian ansatz for the wavefunction has proved to be very useful and able to predict the excitation frequencies within a very high precision @xcite . in our case we make a similar choice . for a very high barrier we expect to find two separate condensates for each of which a gaussian ansatz should be good . we then define the two functions for the left and right side : @xmath49 \exp\left(-i \frac{\sigma_i}{2}z^2 \right ) . \nonumber\end{aligned}\ ] ] in the case of low barrier ( small @xmath44 ) , these two functions have to be orthogonalized to satisfy the orthonormality property required for the two mode functions . hence we define @xmath50 which are different from @xmath51 because of the two different normalization constants for @xmath52 and @xmath53 . the variational parameters describing the mode functions @xmath15 are @xmath44 and @xmath46 . physically , the excitation modes that these parameters can describe depend on whether the two gaussians significantly overlap in space or not . if they do , as in the case of low barrier , then the excitation mode ( changes in @xmath44 ) corresponds to a breathing mode . if they do not overlap , as for high barrier , then it corresponds to an oscillation mode in each of the potential wells . allowing also @xmath54 to vary ( and adding an extra variable for the dynamics ) one can describe more excitation modes . instead we fix it to a constant value , since the curvature of the potential wells will be chosen to be always of the same order of magnitude . the value of @xmath54 is not obvious since the overlap integrals depend strongly on it . for instance the hopping terms can be overestimated due to the long gaussian tales . to fix @xmath54 , we identified a range of values for which the dynamic behaviour of the system was qualitatively the same and choose one of the values within this interval , which turned out to be lower than the one corresponding to the static solution . since the mode wavefunctions are linear combinations of gaussians , if we choose a trapping potential of the form @xmath55 , \end{aligned}\ ] ] the integrals in eq.([overlapint ] ) can be performed analytically . in the following , we will use dimensionless units : @xmath56 , @xmath57 and @xmath58 . so , all lengths will be measured in units of harmonic oscillator length , all energy in units of the trap frequency and all times in units of @xmath59 . we also take for the coefficients @xmath60 a gaussian distribution centered at @xmath61 , @xmath62 , \label{coeff}\end{aligned}\ ] ] where @xmath63 is a normalization constant that depends on @xmath45 only . the variational parameters are @xmath45 and its conjugate one @xmath47 . the parameter @xmath45 is directly related to the width of the distribution @xmath64 , whereas @xmath47 contributes to the width of the fourier transform of such a distribution ( i.e. , to the width of the phase distribution , see app . [ varanephmod ] ) . we study the dynamics using the time dependent variational principle . to derive the equations of motion one starts by writing the action @xmath65 @xmath66 where @xmath67 and @xmath68 have been defined in ( [ 2modeh ] ) and ( [ sumc_m ] ) , respectively . in evaluating the term @xmath69 , one should remember that the state @xmath68 depends on time both through the coefficients and the mode functions contained in @xmath70 : @xmath71 the lagrangian which follows takes the form @xmath72 \nonumber - { \cal h } , \end{aligned}\ ] ] where @xmath73 is given by @xmath74 from the lagrangian ( [ lagr1 ] ) , in general one derives the equations of motion , carrying out the variation with respect to the discrete variables @xmath16 and the fields @xmath75 . in our case , all the integrals and expectation values are functions of the variational parameters which can be calculate analytically . the only important overlap integrals containing time derivatives are @xmath76 if one defines @xmath77 and @xmath78 the lagrangian becomes @xmath79 and the corresponding equations of motion are @xmath80 these equations describe the internal and external coupled dynamics of the splitting of the condensate . in what follows , we have solved them numerically in different regimes . before presenting the results we will show that one can decouple the evolution of the external and internal variables , which helps to understand the dynamics . for that , we will introduce in the next subsection some analytical approximations to derive explicit formulae for the quantities @xmath81 and @xmath82 by replacing the discrete distribution @xmath16 by a continuous one , and treating the index @xmath83 as a continuous variable running from @xmath84 to @xmath85 . in order to calculate @xmath81 and @xmath82 we have to evaluate expectation values of the form @xmath86 and @xmath87 . we can do that if we replace the sums in @xmath83 by integrals extended from @xmath84 to @xmath85 . when this replacement is valid , we can even calculate the width of the number distribution @xmath64 , @xmath88 , as well as the one corresponding to the phase distribution , @xmath89 ( see app.[varanephmod ] ) . we obtain [ sigmas ] @xmath90 on the other hand , we have [ expaaaa ] @xmath91 , \nonumber \\ & & \langle a_1^{\dag } a_2^{\dag } a_1 a_2 \rangle = \frac{n^2}{4 } -p , \\ \label{expvalaa2 } & & \langle a_1^{\dag2 } a_2 ^ 2 \rangle = \\ & & \hspace{0.5 cm } = \frac{n^2}{4 } \exp \left ( -2\sigma_{\phi}^2 \right ) \left [ 1 - \frac{4 p}{n^2 } + \frac{64 p^2 x_i^2 } { n^2 } \right ] . \nonumber\end{aligned}\ ] ] furthermore , in this limit we can determine the eigenvalues of the single particle density operator corresponding to the internal degrees of freedom , obtaining @xmath92 notice that @xmath93 for @xmath94 , i.e. @xmath95 , which corresponds to the gross - pitaevskii limit ; instead @xmath96 for large @xmath97 , giving a signature of the fragmentation of the condensate . one can easily determine the limits of validity of this continuous approximation . on the one hand , the distribution in @xmath16 has to be sufficiently broad , which implies @xmath98 . on the other hand , @xmath47 has to be such that @xmath99 . for our numerical simulations , we corrected the expressions in eqs.([expaaaa ] ) to make them valid @xmath100 and @xmath101 : for @xmath102 we included the periodicity in @xmath47 and for @xmath103 we performed the exact sum over @xmath83 , considering only the few number states which are populated . the coupling among the @xmath44 dynamics and the @xmath45 dynamics appears in the off - diagonal blocks in eq.([eqmotion ] ) and in the dependence of @xmath81 on all variational parameters . in the following we will analyze under which condition it is possible to decouple the internal ( @xmath104 ) and external ( @xmath105 ) dynamics , so that one can study them independently one from the other . one can rewrite the equation of motion for @xmath44 is a more explicit form as @xmath106 let us first see under which circumstances it is possible to neglect the off - diagonal blocks : \(i ) in the low barrier limit , i.e. @xmath107 , the off - diagonal blocks can be neglected if @xmath108 plays no role : @xmath109 evolves at the frequency which governs the internal dynamics @xmath110 ( remember definition ( [ nphi ] ) ) , while the external dynamics is governed by the frequency @xmath111 , usually of the order of the trapping frequency @xmath112 . if @xmath113 , then the time average of @xmath114 vanishes and has no effect on the evolution of @xmath44 . on the other hand @xmath115 only for a chemical potential @xmath116 . this happens either for very few particles , a case in which our model does not hold , since @xmath117 is a fundamental assumption in our model , or for very weakly interacting particles , in which @xmath118 , because the phase coherence is very difficult to be destroyed . \(ii ) in the high barrier limit , normally @xmath119 ( see fig . [ posc ] , small @xmath45 limit ) and @xmath109 might vary abruptly , since this is when the phase coherence is supposed to desappear . anyway , since @xmath120 , the off - diagonal block can be neglected . now let us analyze the @xmath45@xmath47 dependence in @xmath81 ( see eqs.([2modehexpval],[expaaaa ] ) ) . in the on - site terms this dependence is of order @xmath121 and can be therefore safely neglected ( eqs.([expaaaan1],[expaaaan2 ] ) ) . in the hopping terms , it scales like @xmath122 ( @xmath123 , see eqs.([expvalaa],[expvalaa2 ] ) ) : it is strong only when the hopping terms are already small and negligible in comparison with the on - site terms , so it can also be neglected . after these considerations , we conclude that the @xmath44-dynamics is , in a good approximation and in reasonable regimes , independent of the @xmath45-dynamics . this is confirmed by the numerical simulations . the off - diagonal blocks can be neglected in the @xmath45-dynamics if @xmath124 or @xmath125 . if the barrier is raised starting from a condensate in the ground state , @xmath44 evolves almost adiabatically in the low barrier limit ( @xmath124 ) ; when it reaches the high barrier limit , then @xmath125 . therefore , the off - diagonal terms can be neglected during the whole process . instead , the @xmath126 dependence in @xmath81 is strong . one can solve the complete coupled dynamics or substitute in @xmath81 the adiabatic solution for @xmath44 and compare the results . we will show some examples in the following and see that the difference is small . in the previous section , we have written the equations of motion describing the full coupled dynamics and demonstrated that if one splits a condensate starting from a ground state configuration , it is possible to decouple the internal and the external dynamics . in this section , we will use this result and study the external dynamics decoupling it from the internal one . making use of the external static solution , in sec . [ numres ] we will discuss the static solution for the internal degrees of freedom . finally in sec . [ results ] , we show the results for the internal dynamics , which are also useful as a check of the decoupling assumption . the fact that the internal and external dynamics decouple under the condition discussed above , allows us to estimate the excitation of the collective modes using a very simple model . we point out that in this case , similarly to the gross - pitaesvkii equation , the external dynamics is governed by the product @xmath42 and not by the two quantities separately . we choose to raise the potential barrier with the following time dependence @xmath127 from the static solution , obtained by minimizing @xmath128 with respect to all variational parameters simultaneously at fixed @xmath129 , we know the equilibrium position of @xmath44 at any value of the barrier . the @xmath44 dynamics can be approximated very well by the dynamics in a harmonic potential whose center moves from @xmath130 to @xmath131 following the adiabatic solution and whose frequency changes corresponding to the frequency of the small oscillations . to get analytic estimations , we model this dynamics by fixing the frequency and shifting the center of the potential along a trajectory @xmath132 for which an analytic solution exists . when the center follows an hyperbolic tangent trajectory with time constant @xmath25 , the semi - amplitude of the oscillation is given by @xmath133 otherwise , if the center moves according to a linear ramp with time constant @xmath25 , the average semi - amplitude of the oscillation is @xmath134 we compare in fig.[z0osc ] these expressions with the numerical results . for fast raising of the barrier the hyperbolic tangent shift gives a good estimate . for slower raising , instead , the amplitude of the oscillations is largely underestimated . in this case , the linear shift of the center is useful to give an upper bound . we have checked numerically ( by changing the frequency of the harmonic potential in time according to the adiabatic solution ) that the small discrepancy between the actual shift of the center and the hyperbolic tangent dependence is enough to produce such a big change in the amplitude of the final oscillations . the reason for this might be better understood comparing the time derivative of @xmath132 in the two cases ( see inset in fig.[z0osc](a ) ) . our estimations are qualitative , but allow us to set some lower and upper bounds and deduce an adiabaticity condition @xmath135 for the external degrees of freedom . in the same way we can estimate the extra energy per particle due to the excitation of collective modes . using the relation @xmath136 ( the trapping frequency at the end of the process is @xmath111 and not @xmath112 ) , we can estimate which is the fastest time scale which does not destroy the condensate . as expected , it is @xmath137 . in this section , we will study the complete splitting process , starting from a condensate trapped in a harmonic potential and ending with a potential barrier with height @xmath138 . apart from presenting the numerical results obtained by solving eqs.([eqmotion ] ) , we will also introduce a simple two mode model that allows a deeper understanding of the results obtained by the variational ansatz . given the fact that the external dynamics is basically independent of the internal one , we can derive a simple model that accounts for most of the effects related to the internal dynamics . the hamiltonian in eq.([2modeh ] ) depends on the following overlap integrals [ overlapint ] @xmath139 where @xmath140 @xcite . we use @xmath141 to define two effective single - particle hopping terms @xmath142 . the static solution for the external dynamics is known from the minimization of @xmath128 . plugging the corresponding solution @xmath143 in eqs.([overlapint ] ) , we get hamiltonian parameters depending only on the barrier height @xmath129 ( in particular @xmath144 ) . neglecting constant terms , we write the simplified hamiltonian @xmath145 -j \left [ a_1^{\dag } a_2 + a_2^{\dag } a_1 \right ] + \nonumber \\ & + & 2 gu_2 a_1^\dag a_2^\dag a_1 a_2 + \frac{1}{2 } gu_2 \left [ a_1^{\dag 2 } a_2 ^ 2 + a_2^{\dag 2 } a_1 ^ 2 \right ] .\end{aligned}\ ] ] as it is well known ( see app . [ varanephmod ] ) , under certain conditions we can replace this model hamiltonian by a phase model of the form @xmath146 where @xmath147 represents the relative phase between the two modes . the overlap integral @xmath148 may be non - negligible at the beginning of the process when the two mode functions overlap in a sensible way . instead at the end , when the two condensates are almost spatially separated it is very small . in this case @xmath149 and one recoveres the josephson s hamiltonian @xcite . the ground state of such hamiltonian is a localised wavefunction for @xmath150 ( corresponding to a broad number distribution ) and a delocalised one for @xmath151 ( corresponding to a narrow number distribution ) . for this simplified two mode model , we write analytic approximated espressions for the static solution and check the validity of the gaussian ansatz . as explained above , we let the parameters @xmath152 , @xmath153 , @xmath148 depend on the barrier height @xmath129 , according to the static solution . the expectation value of the hamiltonian can be now written as a function of the internal degrees of freedom only , @xmath154 , and allows to study the coherence properties of the ground state at the different stages of the splitting process : for increasing barrier height , we expect the fluctuations in the number distribution to become smaller . we solved the static problem numerically , finding the minimum of @xmath81 with respect to @xmath45 and @xmath47 for fixed @xmath129 . moreover , in the limits of large @xmath45 ( @xmath155 ) and small @xmath45 ( @xmath156 ) it is possible to get analytic estimation for the value of @xmath45 at equilibrium and the frequency @xmath110 of the small oscillations as a function of the overlap integrals @xcite . in the large @xmath45 limit , one finds [ largep ] @xmath157 } ; \nonumber \label{omegaplp}\end{aligned}\ ] ] in the small-@xmath45 limit ( @xmath158 ) , where the continuum approximation for @xmath83 is not valid , @xmath159 and @xmath110 can be calculated with perturbation theory , considering the hamiltonian in eq.([expham ] ) with @xmath160 as unperturbed hamiltonian and the number state @xmath161 as unperturbed ground state . then we get [ smallp ] @xmath162 in fig . [ pstat ] we plot the two analytic solutions for @xmath159 , comparing them with the numeric solution for the minimum of @xmath163 using our ansatz and the corresponding value of @xmath45 obtained by minimizing the exact hamiltonian for @xmath164 and @xmath165 . the agreement between the variational solution and exact one is excellent and the analytic expressions interpolate correctly in the limits where they are expected to work . given the analytic espressions in eqs.([largep],[smallp ] ) , we observe that the oscillation frequency for large @xmath45 coincides with the breathing frequency in the harmonic potential approximating the cosine potential , and the oscillation frequency for small @xmath45 corresponds to the revival time when the cosine potential is negligible ( see sec . [ completesplit ] ) . aware of the fact that it is not possible to define a transition point between the two regimes , being it a smooth transition , we still calculate the value of @xmath45 at which the two frequencies coincide , get @xmath166 and claim that the phase relation between the two condensates is smeared out for @xmath167 . in this section , we show some numerical results obtained by integrating the equations of motion for the variational parameters . in those results the _ full coupled dynamics _ of the process were considered . it is possible to compare with the evolution of the phase distribution governed by the hamiltonian ( [ phmodelham2 ] ) with time dependent coefficients . moreover , it is possible to get analytic estimates concerning the typical time scales of the process . in the three cases presented below we will fix the product @xmath42 in order to have similar external dynamics and better isolate the effect of different @xmath7 and @xmath0 on the phase properties of the system . at the end of the process , when the barrier has reached its final value , depending on @xmath0 and @xmath7 , one can have @xmath168 or @xmath150 ( we assume that @xmath148 is then negligible ) . the case @xmath168 corresponds to the situation where the splitting process is completed and one expects a ground state with no well defined relative phase . instead in the case @xmath169 the ground state is still characterized by a localized phase distribution , and even if the two condensates are almost spatially separated they can not be considered as independent . in this sense , the splitting is not complete . we first analyse the case where in the final stage of the process , one has @xmath168 . since in this subsection and in the following we fix the product @xmath42 , this case is obtained for relatively small @xmath0 and large @xmath7 . depending on the time scale of the process , it is possible to observe collapses and revivals or to reach a final fragmented condensate characterized by very small number fluctuations . we assume that in the final stage of the process it is possible to neglect @xmath170 and @xmath171 , since they depend on the overlap of the two mode functions . then the eingenstates are the number states @xmath172 and the time evolution of the final state corresponds to @xmath173(a ) ) . then , the time evolution corresponds to @xmath45 and @xmath47 following one of these trajectories : @xmath45 keeps almost constant and @xmath47 evolves unbounded increasing linearly with time with `` velocity '' @xmath174 ( see eq.([omegapsp ] ) ) , which exactly reproduces the phase in eq.([phmodev ] ) . the main features of this time evolution are the following : the width of the number distribution is constant ( determined by @xmath175 ) and do not evolve in time ; instead the phase distribution collapses and revives : an initial distribution peaked around zero smears out , revives around @xmath176 and so on . defining the collapse time as the time @xmath177 when @xmath178 and the revival time as the time @xmath39 when the original phase distribution is recovered shifted by @xmath179 , one gets @xcite . [ collrev ] @xmath180 the collapse time is governed by @xmath175 , which depends in general on the barrier raising process and on the parameters @xmath7 and @xmath0 . the final width of the number distribution obtained after the raising of the barrier is completed depends on the time scale of the process . this is just given by the value @xmath175 at which the number fluctuations are frozen . to evaluated it , we claim that as long as @xmath181 the number fluctuations follow the static solution , and when @xmath182 , they are frozen out to a final value @xmath175 . setting @xmath182 in eq.([omegaplp ] ) and substituting in eq.([statplp ] ) , one gets @xcite @xmath183 } \frac{2\pi } { \tau } \sim \nonumber \\ & \sim&\frac{1}{8gu_1}\frac{2 \pi } { \tau}.\end{aligned}\ ] ] of course eq.([pfin ] ) is not valid for @xmath184 . for @xmath185 , we are in the adiabatic regime : @xmath181 during all the process , and we expect to reach a completely delocalised relative phase . we check this numerically and plot the results in fig . [ posc ] for several different splitting processes ( @xmath186 and @xmath187 ) . we actually find that for time scales @xmath188 the final @xmath45 values lie in the region @xmath167 . for faster time scales , the estimation in eq.([pfin ] ) was shown to be very good when the external degrees of freedom evolve adiabatically . otherwise discrepancies can be observed ( fig . [ posc ] ) . it is not straighforward to estimate such discrepancies , since they depend on the exact dynamics and can be either positive or negative . anyway , they are not striking and do not change from a situation in which the final relative phase is well defined to the opposite one . now we consider two of these splitting processes more in detail and compare quantitatively with the evolution of the phase distribution following the hamiltonian in eq.([phmodelham2 ] ) , where @xmath152 , @xmath153 and @xmath148 vary in time with the barrier height according to the instantaneous static solution . we have already mentioned that collapse and revival of the phase are predicted for two condensates with an initially good phase relation when the final tunneling coupling is negligible . let us consider the case @xmath187 , @xmath189 and @xmath190 . in fig . [ dyn1 ] we show the one - atom density matrix eigenvalues and the indetermination in the phase distribution @xmath97 . the agreement between the results of the two simulations is perfect @xcite and our analytic estimations are also confirmed : we expect @xmath191 and @xmath192 , which agree very well with the numerical results shown in fig . [ dyn1 ] . the actual possibility of observing the revivals of the phase in an experiment is something outside our model . if they are actually destroyed by particles losses @xcite , one is left with two condensate with no phase relation , but higher number fluctations than they would have in the ground state . another way to cut the initial condensate into two independent ones , is to raise the barrier much slower , so that the phase coherence is lost adiabatically all along the process . so , we now choose again @xmath187 , @xmath189 but a longer time scale @xmath193 . the agreement between the results of the variational ansatz and the phase model is very good also in this case . the final state is characterised by much smaller number fluctations with respect to the previous case ( see fig . [ dyn2](a ) ) and by a complete delocalised relative phase as shown in fig . [ dyn2](b ) . from our analytic estimations , we actually expect to reach the static solution for @xmath194 . instead as it can be seen in fig . [ posc ] for @xmath195 , this does not happen . dealing with such small values of @xmath175 it is very easy to get at the end very small excitations , which can be due both to the external degrees of freedom or to some excitations already present in the initial conditions . the important feature is that the final relative phase is anyway completely delocalised . we now analyse the case where in the final stage of the process , one has @xmath169 . this case is obtained for large @xmath0 and small @xmath7 : in the limit of large @xmath0 , even if @xmath152 might be very small , it can happen that @xmath196 and the cosine potential in the phase representation ( eq.([phmodelham2 ] ) ) is not negligible . this can be seen as a process in which the barrier is raised up to a level at which the splitting is not really completed . in the case in which the cosine potential at the end of the raising process is still deep , so that the lowest levels can be approximated with harmonic oscillator levels spaced by @xmath197 , the time evolution follows @xmath198 where @xmath199 are the harmonic oscillator eigenstates and where the coefficients @xmath200 depend on the exact dynamics of the raising process . in particular , for symmetric initial conditions , the phase distribution is always symmetric and only the even eigenstates are populated . then , the phase distribution breathes with a frequency @xmath201 , remaining always centered at @xmath202 . moreover we notice that in such a case , the width in the number distribution is not expected to be constant . in the variational ansatz treatment , we have to look again at the orbits in the phase space . the contour plot in fig . [ cplot](b ) is just to show how the orbit modify from the limit of small @xmath0 to the limit of large @xmath0 . so let us consider fig . [ cplot](c ) . the orbits around the minimum of @xmath81 represent a time evolution in which both the width of the number and phase distribution change in time . the frequency of the small oscillations around the equilibrium position can be calculted analytically for @xmath203 ( see eq.([omegaplp ] ) for @xmath204 ) and coincides with the breathing frequency of the phase distribution in the case of the superposition of harmonic oscillator eigenstates if @xmath205 . if this condition is not satisifed , one is not in the weak coupling regime and the phase model is not valid . the orbits in the @xmath45@xmath47 space are characterised by very large oscillations in @xmath45 . hence we can not talk of frozen number fluctations . nevertheless , with arguments similar to the ones used before , we can try to identify the orbit which describes the dynamics at the end of the process . we estimate the maximum @xmath45 value in such an orbit to be @xmath206 . the agreement with the numerical solution can be checked in fig . [ posc ] and it is within a factor of @xmath1 . the adiabaticity condition in this case consists in requiring that the final state is superfluid , as the static solution would be . this means that the minimum value of @xmath45 corresponding to the same orbit as @xmath207 must be such that the phase coherence is still good . we found that a final phase coherence corresponding to a minimum of @xmath208 ( with @xmath209 ) is reached in process with typical time scales @xmath210 . note that this condition is weaker than requiring @xmath211 . in fact @xmath212 , since @xmath213 , as it would correspond to the static solution . this means that we still allow even big oscillation of @xmath45 around the equilibrium value , as long as they do not destroy the phase coherence . this corresponds to a breathing of the phase distribution which never becomes completely smeared out . moreover , while @xmath214 depends only on the product @xmath42 , the adiabatic time scale @xmath215 scales like 1/n , getting easier and easier to be met for large @xmath0 . as done before , we take now a particular case and check directly the results with the one obtained in the phase model . we choose @xmath216 , @xmath217 and @xmath218 and find that the eigenvalues @xmath21 oscillate with frequency @xmath110 ( see fig . [ dyn3](a ) ) : @xmath219 is always close to @xmath220 and @xmath221 close to @xmath222 . this corresponds to the breathing of the phase distribution in the non negligible cosine potential or in the variational ansatz treatment to one of the orbits in fig . [ cplot](c ) . no complete smearing out of the phase is observed ( see fig . [ dyn3](b ) ) . those results are confirmed up to a good level by the phase model ( see fig . [ dyn3](a , b ) ) , even if the oscillations of @xmath97 are damped , due to the anharmonicities of the cosine potential . the conditions that preserve the superfluid state are that the final tunneling coupling @xmath170 is comparable to the on - site interaction and that the time scale of the process is slow enough to allow final small oscillations around the equilibrium . to destroy the phase coherence in this parameter range , one should raise the barrier faster in order to get larger oscillations , or raise it higher ( case described in sec . [ completesplit ] ) . to summarise , in this section we have determined analytic expressions for the adiabaticity conditions for the internal dynamics , i.e. we identified the time scale at which the barrier can be raised in order to obtain the same relative phase properties as expected for the ground state . concerning a completed splitting process , a fragmented condensate without any relative phase ( no collapses and revivals ) and characterised by a very narrow number distribution is reached in raising process with time scale @xmath223 . this means that the process has to be slower for large @xmath0 and small @xmath7 . instead in an incomplete splitting process , the superfluid phase is preserved for @xmath224 . this time scale becomes shorter for large @xmath0 . both conclusions agree with the fact that in the gross - pitaevskii limit ( @xmath43 and @xmath225 ) the condensate is phase coherent . we have solved the two - mode model describing the splitting of a condensate by a potential barrier through a variational ansatz . we found coupling between the internal and the external dynamics of the mode functions . we have identified the regimes in which the two dynamics decouple , and have concluded that in the case of splitting starting from a condensate in the ground state , they do not influence each other in a dramatic way . hence , the internal and external excitations created by raising the barrier can be estimated in a good approximation independently and have been characterised as a function of the interaction strength , the number of atoms and the time scale of the process . from our analytic estimations , confirmed by numerical resuts , we were able to identify the time scales @xmath226 , @xmath227 and @xmath215 , which define the adiabatic regime for the external and internal dynamics ( respectively in the case of final fragmented or phase coherent condensate ) @xmath228 it is interesting to compare the adiabaticity condition for internal and external degress of freedom in the case of final fragmented condensate , when the splitting process can be consider to be completed . we normally found @xmath229 ; the case @xmath164 , @xmath165 sets a boundary where internal and external degrees of freedom enter simultaneously the adiabatic regime for @xmath230 ( see fig . [ posc ] ) . to get @xmath231 one needs @xmath174 to be large compared to the trapping frequency . the quantity @xmath174 is similar to the derivative of the chemical potential with respect to the total number of atoms @xmath232 . from a thomas - fermi solution one gets that either in @xmath220 or @xmath2 dimensions , it scales as a negative power of @xmath0 and a positive power of @xmath7 . so one needs to have very large @xmath7 or very small @xmath0 , and our model fails in both limits . therefore , we claim that in the usual regime of many weakly interacting particles , the adiabaticity condition for the phase dynamics is more restrictive than the one for the external dynamics . we carried out a comparison of our model with the phase model , finding a substancial very good agreement . the numerical solution of the phase model consists in the integration of a time dependent schrdinger equation . in practice , the number of wavefunctions @xmath233 that one has to use increases linearly with @xmath0 , which leads to numerical problems . in this sense for large @xmath0 , the variational ansatz is more convenient and has proved to give reliable results . moreover the variational ansatz treatment allows to include the external degrees of freedom in a natural way . in the previous results we did not take asymmetries into account . it is possible to include them in our ansatz , through the unbalance in population @xmath234 and through non symmetrically centered wave functions . in the case of complete splitting , one ends up with a final constant unbalance in population . the phase coherence shows the only new feature that the center of the phase distribution is now drifting with a velocity @xmath235 , where @xmath236 are the chemical potentials of the two separate condensates . a complete analysis of this case , which even consideres losses and fluctuations in the total number of particles , can be found in @xcite . instead in the case of final phase coherent symmetric condensate , the asymmetry can destroy the phase coherence . the final unbalance in population might be so big , that the wavepacket describing the phase distribution flies above the cosine potential : depending on the `` kinetic energy '' @xmath237 , the cosine potential may become negligible and the same features ( collapses , revivals ) as in the fragmented condensate case are observed . possible extensions of our model are the dynamic turning on of an optical lattice , where the initial harmonic trap is deformed into a many wells potential . the instantaneous version of such a process has recently been investigeted in @xcite . another problem of great significance is the inverse process , i.e. the merging of two condensates . this could allow to refill a condensate and be an important step towards a continuous atom laser . this work was supported by the european union tmr network erbfmrx - ct96 - 0002 and by the austrian science foundation ( projekt nr . z30-tph , wittgenstein - preis and sfb `` control and measurement of coherent quantum systems '' ) . c. m. is grateful to y. castin and a. sinatra for useful discussions , and thanks e. arimondo and g. la rocca for comments . dealing with two coupled condensates , in this paper we have often talked about the number difference @xmath83 and relative phase @xmath147 . in this appendix , we will define number difference @xmath238 and relative phase @xmath239 operators , and derive the phase model hamiltonian ( [ phmodelham2 ] ) , discussing the approximations involved . we first define the operators @xmath240 such that the usual angular momentum commutation relations are fulfilled @xmath241=i \varepsilon_{ijk}{\hat l}_k$ ] . after a small amount of algebra , the exact two mode hamiltonian in eq.([expham ] ) can be rewritten as @xmath242 in the subspace of fixed even total number of atoms @xmath0 , the spectrum of @xmath243 is given by all integer number in the interval @xmath244 $ ] . the operator @xmath243 coincides with the number difference operator @xmath238 . we have often treated the eigenvalues @xmath83 as continuous , but in general care should be exercised . given the phase operator @xmath245 such that @xmath246=i$ ] , it is well known that in the phase representation @xmath247 and the eigenstates of @xmath243 ( @xmath248 ) with eigenvalue @xmath83 are @xmath249 @xcite . the state of the system @xmath68 can be in equivalent ways described as a superposition of eigenstates of @xmath238 ( @xmath16 is the number distribution ) @xmath250 or by a wave function in the @xmath147-representation given by @xmath251 typical quantities characterizing @xmath68 are - the width of the number distribution @xmath252 - the width of the phase distribution @xmath253 the uncertainty relations are the same one as for angular momenta @xcite . in order to write the hamiltonian in the phase representation , we make some approximations which will lead to the simple phase model hamiltonian and discuss under which conditions such approximations are valid . we will describe the procedure only for the term @xmath254 , since for the term @xmath255 an analogous one applies . using the raising and lowering operators @xmath256 , we write @xmath257 . \nonumber \end{aligned}\ ] ] we assume a narrow and centered number distribution ( these assumptions will be quantified later on ) , so that we can expand the square roots at first order in @xmath258 and get @xmath259 + \nonumber \\ & & + o\left ( \frac{{\widetilde { m^4}}}{n^4}\right ) \approx n \cos \phi \ ; \phi(\phi ) , \label{cosphi } \end{aligned}\ ] ] where we can roughly estimate @xmath260 so that @xmath261 and where in a similar way @xmath262 . in an analogous way , the term @xmath263 gives @xmath264 \approx \nonumber \\ & & \approx n^2 \cos 2 \phi . \nonumber\end{aligned}\ ] ] all together the neglected terms have to be smaller than all the other terms in the hamiltonian , which implies @xmath265 for @xmath266 , which is the typical situation treated in this paper , from the first equation one gets @xmath267 . in the opposite limit where @xmath268 instead one has @xmath269 for @xmath270 this condition is not correct and becomes simply @xmath271 . to summarise , in the specific case treated in this paper , where no unbalance of population between the wells was assumed , the important condition of validity for the phase model hamiltonian is eq.([valphmod ] ) , which written for @xmath204 takes the form @xmath272 we note that under such a condition , the ground state is characterised by @xmath273 or equivalently @xmath274 . in this subsection we will compare the phase model with our variational ansatz . for simplicity we set @xmath204 , but the following discussion could be repeated in the more general case . in particular for @xmath204 the phase model hamiltonian reduces to the josephson hamiltonian @xcite the quantities @xmath276 and @xmath170 are respectively the on - site energy splitting ( charging energy ) and the tunneling coupling ( josephson coupling ) . it describe accurately the case of high barrier , where the two condensates are almost spatially separated , leading to a negligible @xmath148 , and are characterised by very small number fluctuations , making the josephson hamiltonian valid . the phase model hamiltonian in eq.([hphasemoldel ] ) describes the motion of a particle in a cosine potential and the classical limit is obtained for @xmath277 . in the following we discuss the classical and quantum limits comparing it directly with the variational ansatz results . in order to allow a complete comparison with the hamiltonian in eq.([hphasemoldel ] ) , we now describe the coefficients @xmath16 in eq.([sumc_m ] ) with four variational parameters @xmath45 , @xmath47 , @xmath234 and @xmath278 the variational parameters @xmath47 and @xmath278 are the variables conjugate to @xmath45 and @xmath234 , allow the dynamic evolution and vanish in the ground state . in the limit of broad number distribution this ansatz describes a gaussian superposition of number states centered in @xmath280 . the corresponding phase distribution @xmath281 is also a gaussian , centered in @xmath282 . notice that in the case of symmetric splitting treated before @xmath283 and @xmath284 @xmath285 . replacing again the sum over @xmath83 from @xmath286 to @xmath287 with an integral from @xmath84 to @xmath85 ( for @xmath288 and @xmath47 such that @xmath289 ) , the widths of the number and phase distribution are respectively given by eqs.([sigmas ] ) and the expectation values @xmath290 and @xmath291 on the state @xmath292 are now let us first consider the classical limit and compare the results obtained with the variational ansatz with the phase model . we write the expectation value of the hamiltonian ( eq.([expham ] ) ) for @xmath204 in the classical limit by setting @xmath294 and @xmath295 here we have assumed @xmath297 , neglected all terms @xmath298 and all terms independent of @xmath234 and @xmath278 . one sees immediately that this expression coincide with the hamiltonian in eq.([hphasemoldel ] ) , where the operators have been substituted with their expectation values ( @xmath299 and @xmath300 ) . so , the analogy with the phase model in the classical limit is straighforward @xcite . in particular if @xmath301 the stable equilibrium position is @xmath202 ; for a kinetic energy such that @xmath302 the phase undergoes oscillations , otherwise if @xmath303 the phase flies above the cosine potential . in the quantum limit the width of the number and phase distribution start to play an important role . the comparison between the two models now is not so trivial , because we have on the one hand a wavefunction in the phase representation , @xmath281 , and on the other hand @xmath304 variational parameters ( @xmath45 , @xmath47 , @xmath234 and @xmath278 ) which follow a classical dynamics and should reproduce the features of @xmath281 . we consider here the simplified case of a wavefunction @xmath281 symmetric around @xmath202 . in the variational ansatz picture , this corresponds to @xmath305 and @xmath284 @xmath285 . this last assumption is correct for symmetric initial conditions and preserved at all times , how can be checked esplicitly in the equations of motion . concerning the ground state of hamiltonian ( [ hphasemoldel ] ) , two different regimes can be identified : for @xmath306 , the cosine potential can be neglected and the ground state is a flat wavefunction , which corresponds to a completely undefined relative phase between the two condensates . instead for @xmath307 the cosine potential is deep and the ground state is a localised wavefunction , which describe a state with very well defined relative phase . in the variational ansatz approach , to find the ground state we set @xmath308 , @xmath283 and @xmath284 , and get @xmath309 $ ] . for @xmath310 , the minimum of this hamiltonian is @xmath311 corresponding to the gross - pitaevkii limit ( @xmath312 ) and if the hopping decreases @xmath156 ( @xmath313 ) , reproducing the same results as the phase model . a quantitative comparison shows perfect agreement . the time evolution was discussed already in sec . [ results ] , where the analogy between the evolution of the phase distribution @xmath314 and the evolution of the variational parameters @xmath45 and @xmath47 was carried out . the variational ansatz is able to predict all the main features typical of the evolution of the phase distribution : collapses and revivals of the relative phase for @xmath306 and breathing of the wavefunction in the cosine potential for @xmath150 . moreover the frequency of the small oscillations around the equilibrium point calculated in the variational approach concides with the splitting of the energy levels of the phase model hamiltonian both in the limit of negligible cosine potential and in the limit of deep cosine potential . in this appendix we have qualitatively compared the phase distribution described by the phase model with the gaussian ansatz for the coefficients of the state @xmath315 in the number representation . we have shown that in the classical limit the expectation values @xmath234 and @xmath278 of number and phase ( variational parameters indicating the centers of the respective gaussian distributions ) are governed by the classical phase model hamiltonian . moreover in the quantum description the parameters @xmath45 and @xmath47 , related to the widths @xmath88 and @xmath97 , are able to reproduce the same features of the time evolution predicted by the phase model , whose limit of validity of the phase model ( derived in app . [ varanephmod ] ) is @xmath316 . we discussed the limit of fragmented condensate ( @xmath151 ) and the limit of single phase coherent condensate ( @xmath317 ) . as already mentioned , for fixed @xmath42 , the limit @xmath43 corresponds to the gross - pitaevkii limit . infact , unless @xmath318 , if @xmath43 the condensate is always phase coherent . for finite @xmath0 , the above discussion allows to judge whether the gross - pitaevskii description is still valid or not . w. ketterle , d. s. durfee and d. m. stamper - kurn , proceedings of the international school of physics `` enrico fermi '' , course cxl , m. inguscio , s. stringari and c. wieman ( eds . ) , ios press , amsterdam 1999 the definitions in eqs.([u2],[u3 ] ) are not general : they are valid always is the static case when @xmath15 are real and are always true in the specific case of our variational ansatz ( see eq.([variationalvarphi ] ) ) we will make the following choice @xmath319 with @xmath320 , so that the potential wells keep as symmetric as possible when @xmath321 varies . this leads to oscillation frequencies of the same order of magnitude at the beginning and at the end of the splitting process . this conclusion is similar to the result in @xcite . it is difficult to make a direct comparison between the two , since due to the more complex spatial and time dependence our tunneling rate does not decrease following a simple exponential . in fig . [ dyn1],[dyn2 ] ( b ) the values for @xmath97 do not coincide for @xmath322 , because we used eq.([sigmasphi ] ) without correcting it , but they both indicate a complete smearing out of the relative phase .

we propose a method to study and constrain modified gravity theories for dark matter using cmb temperature anisotropies and polarization . we assume that the theories considered here have already passed the matter power - spectrum test of large - scale structure . with this requirement met , we show that a modified gravity theory can be specified by parametrizing the time evolution of its dark - matter density contrast , which is completely controlled by the dark matter stress history . we calculate how the stress history with a given parametrization affects the cmb observables , and a qualitative discussion of the physical effects involved is supplemented with numerical examples . it is found that , in general , alternative gravity theories can be efficiently constrained by the cmb temperature and polarization spectra . there exist , however , special cases where modified gravity can not be distinguished from the cdm model even by using _ both _ cmb _ and _ matter power spectrum observations , nor can they be efficiently restricted by other observables in perturbed cosmologies . our results show how the stress properties of dark matter , which determine the evolutions of both density perturbations and the gravitational potential , can be effectively investigated using just the general conservation equations and without assuming any specific theoretical gravitational theory within a wide class .

0805.4400

various cosmological observations indicate that almost @xmath0 of the energy content of our universe is in some dark form that can only be detected through its gravitational effects . the standard explanation for this `` dark sector '' involves some kind of cold dark matter ( cdm ) , such as heavy weakly interacting particles or primordial black holes , and dark energy , which could be a manifestation of a cosmological constant ( @xmath1 ) or some exotic matter field . this standard hybrid picture ( @xmath2 ) has so far passed several different cosmological tests , and is dubbed the `` concordance model '' . despite of its successes , however , it is not without problems in accounting for structure on galactic scales for example , the greatest challenge is to form bulge - free bright spiral galaxies and dwarf spiral galaxies . the galaxies in cdm have either a dominating stellar bulge or a dominating cdm cusp at the center . in essence lcdm over - predicts the dark - matter effects required by the empirical tully - fisher relation of spiral galaxies and is in need of an experimental identification of the cdm particles with an understanding why the dark energy possesses its particular energy density . since dark energy affects cosmological models mainly via gravitational effects , it is possible to imagine that the effects of an explicit material source for the dark energy can be mimicked by a change in the behaviour of the gravitational field at late times . this is an unusual requirement since we have expected deviations from general relativity to arise in the high spacetime curvature limit at early times rather than in the low spacetime curvature limit at late times . however , we should note that the addition of an explicit cosmological constant to general relativity , as in the concordance model of @xmath2 , is already a particular example of such a low spacetime curvature correction . there have been many investigations of gravitational alternatives to dark energy carroll2005 , easson2004 , vollick2003 , brook , allemandi2004a , allemandi2004b , nojiri2005 , amar , nojiri2006 , and it appears that most of these attempts create different problems in either local gravitational systems or large - scale structure ( lss ) formation , or even both chiba2003 , erickcek2006 , navarro2007 , faulkner2007 , barausse2008 , hu2007a , olmo2005 , koivisto2006 , li2006 , li2007a , li2007b , li2007c , li2007d , amendola2007 , li2008a , li2008b ( see @xcite on possible ways to overcome these problems ) . in response to these investigations , frameworks to test modified gravitational dark energy models have also been developed @xcite . the situation is nonetheless different in the dark matter arena , where the leading modified gravity model , milgrom s modified newtonian dynamics ( mond ) @xcite , appeared more than two decades ago , but lacked a convincing general - relativistic formulation with a set of relativistic field equations that are applicable to cosmology . this hurdle has recently been overcome by bekenstein s tensor - vector - scalar ( teves ) model @xcite which reduces to mond in the relevant limit . actually , what the teves model provides is more than a relativistic framework to investigate cosmology it has been found that in teves the formation of lss , which was thought of as a problem for modified gravity theories before , could also be made consistent with observations @xcite . meanwhile , teves has also been shown to work well on smaller scales ( _ e.g. _ , mimicking cold dark matter in strong gravitational lensing systems @xcite , producing elliptical galaxies , barred spiral galaxies and even tidal dwarf galaxies @xcite , being consistent with solar system tests and so on ) and inherit the advantages of mond over the cdm model gnedin2002 ( _ e.g. _ , in explaining the tully - fisher law and galaxy rotation curves ) , this discovery attracts much interest on teves and its generalizations . also , since teves manages to grow large - scale structure using the growing mode of its vector field , this stimulates the investigation of vector - field cosmology @xcite in general tevesaether , ferreira2006 , zlosnik2007a , li2008c , tomi1 , fran , zhao2007 , skordis2008 , zhao2008 ( in ref . @xcite , for example , a very general vector - scalar field framework , the so - called dark fluid , is proposed which can reduce to many existing models in appropriate limits ) , and it is found that the lss in these theories could also be consistent with observations @xcite . despite of this encouraging success , one should bear in mind that lss only provides one test of any structure formation model . in fact , the matter ( or galaxy ) power spectrum we observe today only reflects the large - scale matter - density perturbation _ today _ ( @xmath3 ) , rather than its evolutionary history . thus , even though the matter power spectrum , @xmath4 is compatible with observations , the evolution path of @xmath5 may well be different from that followed in the @xmath2 paradigm . this situation is shown particularly clearly in fig . 3 ( lower panel ) of tevesvector , where the growth rate of baryon density perturbations is enhanced only at late times . the different evolution history of matter - density perturbations may have significant impacts on various other cosmological observables , such as the cosmic microwave background ( cmb ) power spectrum , and influence the growth of nonlinear structure . we would like to determine what these impacts are . in this paper we take a first step in that direction . we will concentrate on the influences of general modified gravity dark - matter models on cmb observables . our main assumption is that the modification to general relativity ( gr ) can be expressed as an _ effective dark matter _ ( edm ) term and moved to the right - hand side of the field equations so that the left - hand side is the same as in gr . this edm term , like the standard cdm , governs both the background and perturbation evolutions . this assumption is justified by observing the fact that in teves , as well as in the general vector - field ( or @xmath6 ) theories , the terms involving the vector field are essentially just the edm term described here . furthermore , we make some simplifying assumptions . first , any explicit dark energy is neglected . the main effects of dark energy ( if not too exotic in origin ) are to modify the background evolution at late times and cause the decay of the large - scale gravitational potential . neglecting it does not affect the essential features of our model but will simplify the numerics greatly . second , the background evolution is exactly the same as that in the standard cdm ( scdm ) model , which should also be a good approximation . third , the matter power spectrum , or equivalently @xmath3 , is the same as that of scdm , because any deviation from the latter should be stringently constrained by lss observations , and because both teves and @xmath6 models have claimed to reproduce the observed matter power spectrum . hence , we are fixing @xmath7 and investigating how different evolutions of @xmath8 ( where @xmath9 is the cosmic scale factor ) affect the cmb power spectrum . our theoretical framework is designed to be more general than is required for this purpose alone , and could be used to investigate features of other cosmological models . this paper is organized as follows : in [ sect : model ] we set out the theoretical framework for our investigation and introduce more details of the cosmological model . in [ sect : cmb ] we describe briefly how a general dark - matter component affects the cmb power spectrum and then , in [ sect : numerics ] , supplement this discussion with a numerical example . we consider three special cases which span all the possibilities in the model , and explain them one by one . finally , in [ sect : conclusion ] we provide a summary of our results together with some further discussion of the assumptions on which they are based . in this analysis we use the perturbed einstein equations in the covariant and gauge invariant ( cgi ) formalism . the cgi perturbation equations in general - theories are derived in this section using the method of @xmath10 decomposition @xcite . first , we briefly review the main ingredients of @xmath10 decomposition and their application to standard general relativity @xcite for ease of later reference . the main idea of @xmath10 decomposition is to make spacetime splits of physical quantities with respect to the 4-velocity @xmath11 of an observer . the projection tensor @xmath12 is defined as @xmath13 and can be used to obtain covariant tensors perpendicular to @xmath14 . for example , the covariant spatial derivative @xmath15 of a tensor field @xmath16 is defined as @xmath17the energy - momentum tensor and covariant derivative of the 4-velocity are decomposed respectively as @xmath18 in the above , @xmath19 is the projected symmetric trace - free ( pstf ) anisotropic stress , @xmath20 the vector heat flux vector , @xmath21 the isotropic pressure , @xmath22 the pstf shear tensor , @xmath23}$ ] , the vorticity , @xmath24 ( @xmath9 is the mean expansion scale factor ) the expansion scalar , and @xmath25 the acceleration ; the overdot denotes time derivative expressed as @xmath26 , brackets mean antisymmetrisation , and parentheses symmetrization . the 4-velocity normalization is chosen to be @xmath27 . the quantities @xmath28 are referred to as _ dynamical _ quantities and @xmath29 as _ kinematical _ quantities . note that the dynamical quantities can be obtained from the energy - momentum tensor @xmath30 through the relations @xmath31 decomposing the riemann tensor and making use the einstein equations , we obtain , after linearization , five constraint equations @xcite : @xmath32 \varepsilon _ { b)ec}^{\ \ \ \ d}u^{e } ; \\ \label{eq : constraintphi } \hat{\nabla}^{b}\mathcal{e}_{ab } & = & \frac{1}{2}\kappa \left[\hat{\nabla}^{b}\pi_{ab}+\frac{2}{3}\theta q_{a}+\frac{2}{3}\hat{\nabla}_{a}\rho \right];\\ \hat{\nabla}^{b}\mathcal{b}_{ab } & = & \frac{1}{2}\kappa \left[\hat{\nabla}_{c}q_{d}+(\rho + p)\varpi _ { cd}\right ] \varepsilon _ { ab}^{\ \ cd}u^{b},\end{aligned}\]]and five propagation equations , @xmath33 } & = & 0 ; \\ \label{eq : propagationphi } \frac{1}{2}\kappa\left[\dot{\pi}_{ab } + \frac{1}{3}\theta\pi_{ab}\right ] - \frac{1}{2}\kappa \left[(\rho + p)\sigma_{ab}+\hat{\nabla}_{\langle a}q_{b\rangle}\right]\nonumber \\ -\left[\dot{\mathcal{e}}_{ab}+\theta\mathcal{e}_{ab}-\hat{\nabla}^{c } \mathcal{b}_{d(a}\varepsilon_{b)ec}^{\ \ \ \ d}u^{e}\right ] & = & 0;\\ \dot{\mathcal{b}}_{ab}+\theta \mathcal{b}_{ab}+\hat{\nabla}^{c}\mathcal{e}_{d(a}\varepsilon _ { b)ec}^{\ \ \ \ d}u^{e}\nonumber\\ + \frac{\kappa}{2}\hat{\nabla}^{c}\mathcal{\pi}_{d(a}\varepsilon_{b)ec}^{\ \ \ \ d}u^{e } & = & 0.\end{aligned}\]]here , @xmath34 is the covariant permutation tensor , @xmath35 and @xmath36 are respectively the electric and magnetic parts of the weyl tensor @xmath37 , defined by @xmath38 and @xmath39 . the angle bracket means taking the trace - free part of a quantity . besides the above equations , it is useful to express the projected ricci scalar @xmath40 into the hypersurfaces orthogonal to @xmath11 as @xmath41the spatial derivative of the projected ricci scalar , @xmath42 , is then given as @xmath43and its propagation equation by @xmath44 finally , there are the conservation equations for the energy - momentum tensor : @xmath45 as we are considering a spatially - flat universe , the spatial curvature must vanish on large scales and so @xmath46 . thus , from eq . ( [ eq : spatialriccicurvature ] ) , we obtain @xmath47 this is the friedmann equation in standard general relativity , and the other background equations ( the raychaudhuri equation and the energy - conservation equation ) can be obtained by taking the zero - order parts of eqs . ( [ eq : raychaudhrui ] , [ eq : energyconservation ] ) , yielding : @xmath48 all through this paper we only consider scalar perturbation modes , for which the vorticity @xmath49 and magnetic part of weyl tensor @xmath36 are at most of second order @xcite , and so are neglected in our first - order analysis . for the perturbation analysis it is more convenient to work in the @xmath50 space because we confine ourselves in the linear regime and different @xmath50-modes decouple . following @xcite , we shall make the following harmonic expansions of our perturbation variables @xmath51 in which @xmath52 is the eigenfunction of the comoving spatial laplacian @xmath53 satisfying @xmath54 and @xmath55 are given by @xmath56 . in terms of the above harmonic expansion coefficients , eqs . ( [ eq : constraintq ] , [ eq : constraintphi ] , [ eq : propagationsigma ] , [ eq : propagationphi ] , [ eq : constrainteta ] , [ eq : propagationeta ] ) can be rewritten as @xcite @xmath57,\\ \label{eq : propagationsigma2 } k(\sigma ' + \mathcal{h}\sigma ) & = & k^{2}(\phi+a ) -\frac{1}{2}\kappa a^{2}\pi,\\ \label{eq : propagationphi2 } k^{2}(\phi'+\mathcal{h}\phi ) & = & \frac{1}{2}\kappa a^{2}\left[k(\rho+p)\sigma+kq-\pi ' -\mathcal{h}\pi\right],\ \ \ \ \\ \label{eq : constrainteta2 } k^{2}\eta & = & \kappa\mathcal{x}a^{2 } - 2k\mathcal{h}\mathcal{z},\\ \label{eq : propagationeta2 } k\eta ' & = & -\kappa qa^{2 } - 2k\mathcal{h}a\end{aligned}\ ] ] where @xmath58 and a prime denotes the derivative with respect to the conformal time @xmath59 ( @xmath60 ) . also , eq . ( [ eq : heatfluxevolution ] ) and the spatial derivative of eq . ( [ eq : energyconservation ] ) become @xmath61 recall that we are treating the modifications to gr as an edm term and so include them in the ( generalized ) energy - momentum tensor @xmath30 to maintain the standard form of the einstein equations . thus , we can distinguish its different components @xmath62 and their conservation . depending on the specific model , the expressions for these components can be very different , but their conservation equations will take the same form . in particular , since the edm has no coupling with standard model particles such as photons and baryons , it satisfies a separated conservation equation , eqs . ( [ eq : heatfluxevolution2 ] , [ eq : energyconservation2 ] ) : @xmath63where we have defined the edm peculiar velocity @xmath64 , the density contrast @xmath65 , and @xmath66 , @xmath67 ( c.f . sect : appendixb ) for the edm , and used the fact that @xmath68 to reproduce the standard cdm background evolution . the prime here is the derivative with respect to the conformal time and @xmath69 . on the other hand , the spatial derivative of eq . ( [ eq : raychaudhrui ] ) gives the evolution equation for @xmath70 as @xmath71 and in this equation @xmath72 are respectively the density and pressure perturbations for _ all _ the matter species , including the edm . for convenience we shall work in the frame where @xmath73 . in this case , if the universe is dominated by the edm , then the three equations above can be combined to eliminate @xmath70 and @xmath74 : @xmath75 \delta_{p } & = & 0.\ \ \ \\end{aligned}\]]if other matter species could not be neglected , as is in the radiation dominated era and early matter era , then we only need to correct the above equation by adding to it @xmath76 another term with @xmath77 ( where @xmath78 are the baryon and radiation energy densities ) which comes from the last term in the left - hand side of eq . ( [ eq : evolvez ] ) and these new terms do not affect the qualitative features of our discussion ( we will include them in the numerical calculation ) . ( [ eq : evolvedensityperturb ] ) tells us that the evolution of the density perturbation in edm is completely controlled by its stress history : the edm equation of state ( eos ) @xmath79 controls the background expansion and is set to zero here ; the other two stress variables @xmath80 then drive the evolution of @xmath81 through eq . ( [ eq : evolvedensityperturb ] ) . note that the stresses of the edm are external functions determined by microphysics ( for particle dark matter ) or a particular modified gravity theory , and must be specified to close the system of einstein equations and conservation equations . in the special case where @xmath82 , we reduce to the cdm model for which @xmath83 in the matter era ; but in general there will be deviations from this growing solution . next , we look at the evolution of the gravitational potential , @xmath84 ( see [ sect : appendixa ] , [ sect : appendixb ] ) . by manipulating eqs . ( eq : constraintphi2 , [ eq : propagationsigma2 ] , [ eq : propagationphi2 ] , [ eq : heatfluxevolution2 ] ) , and working again in the @xmath73 frame , we can eliminate the terms involving @xmath85 and obtain the following evolution equation @xmath86\phi + k^{2}\frac{p'}{\rho'}\phi\nonumber\\ & = & \frac{1}{2}\kappa\rho a^{2}\left(\delta_{p } - \frac{p'}{\rho'}\delta\right ) - \frac{1}{2k^{2}}\kappa\rho a^{2}\delta''_{\pi } - \frac{1}{2k^{2}}\kappa\rho a^{2}\left[\left(5 + 3\frac{p'}{\rho'}\right)\mathcal{h } + 2\frac{\rho'}{\rho}\right]\delta'_{\pi}\nonumber\\ & & - \frac{1}{2k^{2}}\kappa\rho a^{2}\left[\left(5 + 3\frac{p'}{\rho'}\right)\left(\mathcal{h } + \frac{\rho'}{\rho}\right)\mathcal{h } + \frac{\rho''}{\rho } + \left(\frac{2}{3 } + \frac{p'}{\rho'}\right)k^{2}\right]\delta_{\pi}\end{aligned}\ ] ] where @xmath87 are the total energy density and total pressure respectively and @xmath88 , with @xmath89 for an edm dominated universe . the term @xmath90 is zero for radiation and baryons , but in general could be nonzero for the edm . again , we see that the stress history of the edm completely specifies the evolution of @xmath91 . as we shall see in what follows , the dark matter component influences the cmb power spectrum through the gravitational potential , and so the stress history of the edm is of crucial importance for the cmb . in summary , we have seen that the properties of the ( isotropic and anisotropic ) stresses of the edm determine the evolution of the matter density perturbation and the gravitational potential , and thereby determine the predicted matter power spectrum ( through the former ) and the cmb spectrum ( through the latter ) @xcite . as we discussed in sect : introduction , the predicted matter power spectrum of a theory records the matter energy density perturbation at a specific time today , and it is easier to make it consistent with observations , as evident from the studies of teves and @xmath6 theories . thus , in this work we shall start from an edm theory which is constructed already to predict an acceptable matter power spectrum ; that is , we fix the matter energy density perturbation today and calculate the influence of different @xmath92 evolution paths on the cmb spectrum . in this way we can reduce our model space to one that is of realistic interests . it is helpful to have a brief review of the cmb physics and how it is affected by the dark matter before we go into the numerics to show the effects of edm stresses on the cmb . here we will just present a minimal description of this topic , for more details see the reviews @xcite . the primary cmb spectrum is determined by inhomogeneities in the cmb photon temperature at the time of recombination . prior to the recombination , photons couple tightly by thomson scattering with electrons which themselves couple to baryons via coulomb interaction , thus to a first approximation photons and baryons combine into a single fluid . dark matter , on the other hand , does not couple electromagnetically , but only interacts through gravitational effects and so contributes to the gravitational potential in which the baryon - photon fluid moves . when the density perturbation in the photons grows , hot ( cold ) spots appear where the local photon density is higher ( lower ) than average . the higher photon density means higher pressure , which will tend to counteract the growth of local photon density . this will lead to acoustic oscillations of the gauge invariant photon temperature perturbation @xmath93 , which is also a characterization of the photon number density perturbation . in reality this picture becomes a bit more complicated because of the interplay between baryons and dark matter : baryons couple tightly to photons , and move with them so that the inertia of the fluid is increased , while the gravitational potential produced by dark matter drives the oscillations . more specifically , if we use the multipole decomposition @xmath94 where @xmath95 is the legendre function and @xmath96 with @xmath97 being the direction of photon momentum , then in the tight - coupling limit the monopole @xmath98 satisfies a driven - oscillator equation @xcite : @xmath99 in which @xmath100 where @xmath101 and @xmath102 are , respectively , the ( frame - independent ) newtonian potential and curvature perturbations . the dipole moment @xmath103 , which equals the peculiar velocity of baryons thanks to the tight coupling , satisfies @xmath104 , and all higher moments @xmath105 vanish because the frequent scattering makes the photon distribution isotropic in the electron rest frame . thus , on large scales the photon temperature perturbation displays a pattern of driven and damped oscillation . on scales smaller than the photon mean free path , which itself grows in time , however , the tight coupling approximation is no longer perfect and quadrupole moments of the temperature perturbation needs to be taken into account . this introduces a dissipation term into the oscillator equation above , which damps the oscillations . at the time of recombination , the number density of free electrons drops suddenly and there ceases to be coupling between baryons and photons . the cmb photons then free stream towards us . this free - streaming solution is given by @xcite @xmath106 \notag \\ & & + ( 2\ell+1)\int_{\tau_{\ast}}^{\tau_{0}}(\psi^{\prime}-\phi^{\prime})j_{% \ell}[k(\tau_{0}-\tau)]d\tau\end{aligned}\ ] ] where @xmath107 are the conformal times today and at recombination , @xmath108 and @xmath109 is the spherical bessel function of order @xmath110 . the terms @xmath111 and @xmath103 are respectively the monopole and dipole moments of the cmb temperature field at @xmath112 and can be calculated using the equations above ( a @xmath113 is added to @xmath114 which accounts for the redshift in the photon energy , and thus temperature , when it climbs out of the newtonian potential ) . finally , the cmb temperature - temperature spectrum is defined as @xmath115 these are the leading - order effects in the cmb physics , and we could see how they affect the cmb power spectrum . because @xmath116 peaks strongly at @xmath117 , so perturbations on very large scales ( small @xmath50 ) mainly affect the low-@xmath110 spectrum . if the universe is completely dominated by matter at @xmath118 then the first term in eq . ( [ eq : freestreaming ] ) is given by @xmath119 and accounts for the ordinary sachs - wolfe ( sw ) effect ; meanwhile , if the potential @xmath120 decays between @xmath118 and @xmath121 , then the integration in eq . ( [ eq : freestreaming ] ) will also make a significant contribution as photons travel in and out of many time - dependent potentials along the line of sight , and this is the integrated sachs - wolfe ( isw ) effect . going to higher @xmath110 one can see the peak structures of the cmb spectrum . the peaks appear since at @xmath118 , when the oscillating pattern of @xmath122 [ c.f . ( [ eq : drivenoscillator ] ) ] freezes , @xmath98 might be just at its extrema for some scales ( @xmath50 ) ; and these extrema are then converted to extrema of @xmath123 through eqs . ( [ eq : freestreaming ] , [ eq : cl ] ) . since what appears in eq . ( [ eq : cl ] ) is @xmath124 , both the maxima and minima of @xmath125 will appear as peaks in @xmath123 . if there are no baryons and the potential @xmath113 is constant , then the even and odd peaks should be of the same amplitude . the inclusion of baryons effectively increases the inertia of the baryon - photon fluid and displaces the balance point of @xmath111 , and as a result , after taking @xmath126 the odd and even peaks appear to have different heights . at the same time , from eqs . ( eq : freestreaming , [ eq : cl ] ) , we see that the dipole @xmath127 also contributes to @xmath123 through modulation . but eq . ( eq : freestreaming ) shows that its power is more broadly distributed and so its contribution is significantly smaller than that of the monopole . where the monopole vanishes the dipole becomes important and this is why the troughs of @xmath123 are of nonzero amplitude . if one goes to still higher @xmath128 there are two effects . first , the higher @xmath110 moments mainly receive contributions from the small - scale ( large @xmath50 ) perturbations , which began oscillating already during the radiation - dominated era ( not much earlier than @xmath118 ) . since in the radiation era the potential @xmath129 decays , this decay will drive the monopole oscillation through the source term in eq . ( eq : drivenoscillator ) and lead to an increase in the oscillation amplitudes . this explains why in some models the third peak is higher than the second . second , as discussed above , on very small scales there is severe damping of the oscillations because of photon diffusion . the combination of these two effects causes the cmb power in @xmath110s higher than the third peak to be significantly damped . we can now highlight the key places where the dark - matter effects enter cmb physics through the gravitational potential it produces and contrast the situation with that when edm is employed as a substitute . first , if cdm is replaced by edm , then in the analysis of [ sect : model ] the time evolutions of @xmath130 and thus of @xmath129 and @xmath113 are modified . if the modification only becomes significant after @xmath118 , the isw effect could be different from standard cdm ( where it effectively vanishes ) . if it differs from standard cdm before @xmath118 then the sw effect will be altered as well . second , the deviation of @xmath129 and @xmath113 from their values in standard cdm before @xmath118 can change the zero point of the monopole oscillation through baryon loading , thereby modifying the relative heights of the odd and even peaks . third , the modification of @xmath131 and @xmath113 also changes the driving force in eq . ( eq : drivenoscillator ) , and results in different amplitudes for the @xmath132 at high @xmath110 . in [ sect : numerics ] we shall give some numerical examples showing these effects explicitly in our edm model . in this section we turn to some numerical examples to illustrate the qualitative analysis of [ sect : cmb ] . as mentioned in sect : model , we choose to fix the endpoint of the evolution of the matter energy density perturbation . this gives us a freedom to parametrize the evolution of @xmath133 , which is generally different from that of @xmath134 ( in matter regime ) . once the @xmath135 parametrization is given ( or in other words the model is set up ) , eq . ( [ eq : evolvedensityperturb ] ) becomes an evolution equation for @xmath136 and @xmath137 . this is insufficient to solve for @xmath136 and @xmath137 , so in the following numerical calculations we consider three cases : ( 1 ) @xmath138 , ( 2 ) @xmath139 and ( 3 ) a more realistic case where both @xmath136 and @xmath140 are nonzero . since any scale dependence of @xmath141 will alter the shape of the matter power spectrum , and thus incur stringent constraints , we parametrize @xmath133 so that it is independent of @xmath50 . when @xmath143 and @xmath81 is specified , eq . ( eq : evolvedensityperturb ) becomes a first - order evolution equation for @xmath144 @xmath145\delta_{p } & = & s,\ \ \ \\end{aligned}\ ] ] where the source term @xmath146 is given by @xmath147 the quantity @xmath148 does not appear directly in the expression for gravitational potential [ c.f . ( [ eq : constraintphi2 ] , eq : propagationphi2 ) ] , however it does affect the potential @xmath84 _ indirectly _ through the evolution of @xmath149 eq . ( eq : edmconserv1 ) . consequently , there are now two new variables ( @xmath150 and @xmath149 ) to evolve in the model and we need to specify their initial conditions . here , we adopt the simplest and most direct approach , namely to assume that the edm evolves as cdm prior to some initial time @xmath151 , when the scale factor is @xmath152 and the dark matter density perturbation @xmath153 , and for @xmath154 the dark matter density perturbation begins to evolve as @xmath155 . this means that for @xmath156 the variables @xmath136 and @xmath149 will remain zero : this provides the desired initial conditions . this simple approach captures most of the interesting features about edm . however , by assuming that at early times edm is just cdm , we can not account for more complicated issues such as the primordial power spectrum , and we will comment on this in the concluding section . now we could turn to a specific example of @xmath157 parametrization . we let @xmath133 equal @xmath158 in the standard cdm model at times @xmath152 and @xmath159 , and assume that the deviation from standard cdm occurs between @xmath152 and @xmath159 . here , @xmath160 and @xmath161 is the current time . in order to characterize the deviation from standard cdm between @xmath152 and @xmath159 , we introduce a parameter @xmath162 to denote the ratio of @xmath141 to @xmath163 at the mean time @xmath164 . since for standard cdm we have @xmath134 , the parametrized @xmath135 is simply a parabola which passes through three points @xmath165 , @xmath166 and @xmath167 between times when the scale factor lies between @xmath152 and @xmath159 : @xmath168 for @xmath169 $ ] . some examples of @xmath170 with different choices of @xmath162 are shown in fig . [ fig : figure1 ] . it clearly shows that @xmath162 characterizes the deviation from a standard cdm evolution ( @xmath171 ) . note that the parametrization eq . ( eq : edmparametrization ) is just an example for illustration , and specific modified gravity models like teves and @xmath6 theory may lead to different parametrization which however can similarly be tested . , normalized to its current value @xmath172 , as described in eq . ( [ eq : edmparametrization ] ) . the values of the parameter @xmath162 are indicated ; the other parameters are @xmath173 and @xmath174 . ] model with @xmath133 parameterized as in eq . ( [ eq : edmparametrization ] ) . three values of @xmath162 are adopted : @xmath175 ( blue curves ) , @xmath176 ( red curves ) and @xmath177 ( black curve ) . the dashed curves are the cases where @xmath178 and the solid coloured ones @xmath179 . for all the curves we set @xmath180 . ] then with the parametrization of @xmath133 in eq . ( eq : edmparametrization ) , we could numerically evolve the relevant perturbation equations and see the changes in the cmb power spectrum . first , we assume @xmath181 . in this situation the cmb physics prior to last scattering is unaffected and so from eq . ( [ eq : freestreaming ] ) we see that only the integration is modified . this is because in the standard cdm model the gravitational potential , @xmath182 , remains constant during the matter epoch , and the integrand in eq . ( eq : freestreaming ) vanishes . for the present edm case , however , eq . ( eq : evolvepotential ) dictates that @xmath183 even in the matter era ; consequently , there will be a significant isw effect to boost the low-@xmath184 cmb power . as shown in fig . [ fig : figure2 ] , the earlier the edm evolution deviates from that of cdm ( _ i.e. _ , the smaller @xmath152 is ) , the earlier @xmath84 begins to evolve and the larger the cumulative isw effect is . , @xmath185 and @xmath186 ( blue curve ) , @xmath187 ( black curve ) and @xmath188 ( red curve ) respectively . ] next , consider the case where the deviation from cdm evolution begins earlier than last scattering . in this case the evolution of @xmath189 and thus the potential @xmath84 is changed before last scattering , and correspondingly the first two terms in eq . ( [ eq : freestreaming ] ) are modified as well . in fig . [ fig : figure4 ] we have displayed the primary cmb power spectrum for the @xmath133 parametrization eq . ( [ eq : edmparametrization ] ) with parameters @xmath190 , @xmath191 , @xmath192 and the isw effect switched off for simplicity . a larger value of @xmath162 implies a larger edm density perturbation ( for the scales relevant to the first acoustic peak ) before the last scattering and therefore a deeper newtonian potential @xmath193 . this means that the cmb photons experience larger redshifts when climbing out of the potential after last scattering , and the effective temperature @xmath194 is lower , leading to a suppressed first cmb acoustic peak . when @xmath162 is smaller the opposite effect occurs ( c.f . [ fig : figure4 ] ) . at the higher-@xmath195 peaks this effect is less significant because the potential for the relevant ( smaller ) scales has already decayed . however , as discussed in [ sect : cmb ] , the change the potential prior to last scattering also modifies the equilibrium point of the monopole oscillation and alters the relative heights of odd and even cmb peaks . if @xmath193 is a constant , the difference between odd and even peaks in @xmath196 is proportional to @xmath197 @xcite , so increasing @xmath162 amplifies this difference by increasing @xmath193 , and as shown in fig . [ fig : figure4 ] , the third peak becomes higher and the second peak becomes lower for larger values of @xmath198 and go oppositely for smaller @xmath198 . in this figure the fourth peak and onwards are not affected significantly because our parameters are chosen conservatively . there however _ could _ be alternative dark matter models where the higher peaks also deviate from the cdm predictions , for an example see fig . 4 ( upper panel ) of tevesperturbation . the cmb power spectrum has been measured to high precision by many experiments ( see for example @xcite ) and will be further improved in the future , so it could be used to constrain the edm model here . for higher @xmath110s the cmb data could be used directly . for lower @xmath110s its usability is limited by the cosmic variance . if the deviation is significant ( such as those in fig . [ fig : figure2 ] ) , then we could use cmb data alone to constrain the model , as in @xcite . otherwise , we could cross correlate the observed isw with the matter density perturbation observable @xcite ; because both are modified in the edm model , we should expect different correlations from those created by cdm . actually , this technique has been applied to teves @xcite but may not serve as a perfect edm discriminator because there will be contaminations from the dark energy component in the galaxy - cmb correlation . meanwhile , the general edm model has different gravitational potential , matter density perturbation , as well as a possible different redshift distribution of lensing galaxies , and these will also change the weak lensing spectrum ( see @xcite for an application to one modified dark matter model ) . finally , the different evolution history of @xmath133 may also have implications for the formation of nonlinear structure . these further possibilities are beyond the scope of this work and will be further pursued elsewhere . with the @xmath157 parametrization given in eq . ( eq : edmparametrization ) and parameters @xmath199 ( blue curve ) and @xmath200 ( red curve ) respectively . the black curve is the scdm model . the other parameters are @xmath201 and @xmath202 . ] in the case of @xmath204 , eq . ( [ eq : evolvedensityperturb ] ) simply becomes an algebraic equation for @xmath137 : @xmath205 for the parametrization of @xmath157 we will still use eq . ( [ eq : edmparametrization ] ) . in this case , because eq . ( eq : case2delta_pi ) is algebraic , the only variable to propagate in time is @xmath206 [ c.f . ( [ eq : edmconserv1 ] ) ] , and as in subsect : case1 we could take @xmath207 prior to @xmath152 and then evolve it using eq . ( [ eq : edmconserv1 ] ) for @xmath154 . in fig . [ fig : figure5 ] we have plotted the cmb spectrum for such a model with @xmath208 and different choices of @xmath152 , from which we can see that the low-@xmath110 cmb power is very sensitive to both @xmath198 and @xmath152 . the reason is that , although the right - hand side of eq . ( [ eq : case2delta_pi ] ) is scale - independent thanks to our parametrization eq . ( eq : edmparametrization ) , its left - hand side _ is _ scale - dependent through the @xmath209 factor . consequently , on large scales ( small @xmath50 ) the edm anisotropic stress @xmath210 could be very large and so the gravitational potential @xmath91 is significantly different than that in standard cdm[c.f . ( [ eq : constraintphi2 ] ) ] . the late isw effect is then modified , which enhances the low-@xmath110 cmb power . on smaller scales , @xmath211 is suppressed by @xmath212 and its effects soon becomes negligible , explaining why the high-@xmath110 cmb power is not influenced . also , the smaller @xmath152 is , so the earlier the cmb evolution of the @xmath213 edm model deviates from the standard cdm result . note that , in the case of standard cdm , the right hand side of eq . ( eq : case2delta_pi ) vanishes identically , and so there is no influence on the last isw . it is then clear that the @xmath204 edm model with a general scale - independent parametrization of @xmath157 is problematic and already stringently constrained . a possible way out of this trouble is to drop the scale - independence of @xmath157 . we have numerically checked that if on large scales @xmath81 grows as @xmath214 , then the low-@xmath195 boosts of the cmb power as shown in fig . [ fig : figure5 ] disappear and one recaptures the standard cdm results of isw effect . another possibility is to have both @xmath215 and @xmath157 scale - independent and also include @xmath148 ; this will be considered in [ subsect : case3 ] . is parameterized as eq . ( [ eq : edmparametrization ] ) , both @xmath144 and @xmath216 are nonzero and satisfy a relation as described in the text . the solid , dashed and dotted curves represent the cases of @xmath217 respectively , and they can not be distinguished in the figure because they experience identical gravitational potential evolution . the other parameters are @xmath173 and @xmath218 . ] ( solid curves ) and @xmath219 ( dashed curves ) as functions of @xmath9 . the black line is for scdm ( @xmath220 ) , and the red / blue lines are for the model described in the text with @xmath221 respectively . all curves are for the scale @xmath222 . ] if the edm arises from modifications to standard general relativity , then in general neither @xmath148 nor @xmath223 would be exactly zero . in this case there are many more possibilities because @xmath136 and @xmath215 can not be uniquely solved from eq . ( eq : evolvedensityperturb ) . of course , the evolution of @xmath91 might also be modified normally and this could be used to constrain @xmath148 and @xmath215 . indeed , we could use the freedom to choose @xmath148 and @xmath211 so that the evolution of @xmath91 [ c.f . ( eq : evolvepotential ) ] is exactly ( or nearly ) the same as that in standard cdm . let us consider such an example now . in order that the gravitational potential evolves in the same way as in standard cdm , we require the contribution on the right - hand side of eq . ( eq : evolvepotential ) from the edm to vanish , which leads to the following evolution equation for @xmath137 @xmath224 \delta_{\pi } & = & k^{2}\delta_{p}\ \ \ \ \notag\end{aligned}\ ] ] where @xmath225 and @xmath226 are respectively the energy densities of non - relativistic ( baryons plus edm ) and relativistic ( photons plus neutrinos ) matter . here note that the time evolution of @xmath137 is driven by @xmath136 , which itself evolves in accord with eq . ( eq : evolvedensityperturb ) , which is driven by @xmath137 _ plus _ the _ parametrized _ @xmath133 ( with its time derivatives ) . so , in the numerical calculation we have four more variables to propagate : @xmath227 and @xmath228 . ) ] with @xmath157 parameterized as in eq . ( [ eq : edmparametrization ] ) . here we have chosen @xmath229 and @xmath185 . the blue dashed , black solid and red dotted curves represent the cases for @xmath230 and @xmath188 respectively , as also shown beside the curves . the large - angle ( @xmath231 ) cmb powers are indistinguishable for the curves as in fig . [ fig : figure6 ] due to the identical evolutions of @xmath232 and identical isw effects . ] . the red dotted , black solid and blue dashed curves are the cases @xmath233 respectively . ] note that if @xmath148 vanishes identically and @xmath143 initially then @xmath223 will remain zero all the time ; also if @xmath215 vanishes identically then so does @xmath148 . these restrictions indicate that the two cases we have considered in above subsections _ can not _ give rise to the same evolution of @xmath91 _ and _ different evolution of @xmath81 to that predicted by standard cdm at the same time . the general case with @xmath234 , however , is able to do this , as is shown in fig . fig : figure6 . there , we plot the cmb power spectra of this general case with different values of @xmath235 respectively , and they are totally indistinguishable because the evolutions of @xmath91 are identical , leading to identical ( zero ) isw effects . in this figure we have chosen @xmath178 and @xmath236 so that we only require the edm to deviate from cdm after last scattering . it must be emphasized that , although the three curves are indistinguishable from each other , they _ do _ correspond to different edm properties . to show this point clearly , we display in fig . fig : figure7 the quantities @xmath237 and @xmath238 , which characterize the importance of @xmath148 and of @xmath223 respectively , for the models . it is obvious that in the @xmath239 models these quantities could be very different from the standard cdm value ( 0 ) , and this can be understood intuitively as follows : the pressure perturbation term @xmath148 acts _ against _ gravitational collapse , when @xmath240 , meaning that @xmath81 grows faster than @xmath241 early on and more slowly later ; @xmath148 needs to be negative early on and positive later ( and vice versa ) . the anisotropic stress term @xmath242 dissipates fluctuations in @xmath141 [ c.f . ( [ eq : evolvedensityperturb ] ) ] and behaves similarly in the figure . we have thus seen that the observations of the cmb and matter power spectrum _ can not _ rule out this general case ( as long as @xmath152 is after last scattering ) . cross correlating isw with the galaxy distribution might help in this regard , but it is still limited for two reasons : first , it is contaminated by the effects of dark energy , and second , the cross - correlation data only exists for late times and can not effectively constrain models like ours where the deviation from cdm occurs much earlier . if we choose the @xmath243 , which means that the deviation of edm from cdm starts before last scattering , then although eq . ( eq : case3delta_pi ) guarantees that the @xmath91 evolution is not changed for arbitrary @xmath162 , the cmb power spectrum will generally be different from that of standard cdm because the quantities @xmath244 and @xmath193 , which are directly relevant for the primary cmb anisotropy , are modified since @xmath245 . in fig . [ fig : figure8 ] we present such an example , for which we have used the parametrization eq . ( [ eq : edmparametrization ] ) of @xmath246 with @xmath190 and @xmath191 . this shows that in general the small - angle ( high-@xmath110 ) cmb power will be different even though the large - angle power is fixed to be the same as in standard cdm . if the evolution of @xmath81 prior to last scattering is significantly different from that of cdm , then the deviation of cmb spectrum might be very large , and this is why the cmb could efficiently constrain such alternative gravitational dark matter theories as teves and the @xmath6 models . another possible constraint comes from the cmb polarization . the linear polarization of cmb photons carries information about the gravitational potential because it can only be generated from the quadrupole moment of the cmb temperature perturbation , which is related to the anisotropic photon stress @xcite , through thomson scattering . furthermore , at earlier times the rapid scattering of photons by electrons ruins the photon anisotropic stress ( c.f . [ sect : cmb ] ) , consequently the polarization only appear close to last scattering when significant photon quadrupole anisotropy can be produced . the localization of the generation in time , and the dependence on the quadrupole moment , only mean that the polarization directly reflects the gravitational potentials at the time of last scattering in a special way and so provides invaluable information about the edm . this is in contrast to the cmb temperature spectrum which depends on the gravitational potentials through the whole cosmic history up to now , and also on the photon monopole and dipole moments , which complicate the extraction of information . in fig . [ fig : figure9 ] we plot the cmb ee polarization spectrum for the same model as in fig . [ fig : figure8 ] . it can be seen that different parameters give quite distinct peak features . the cmb polarization was first detected in 2002 @xcite , with the precision gradually improved since then @xcite . although the precision at present is still insufficient to place stringent constraints on the edm model , we expect that future observations will change this situation . we could also slightly change eq . ( [ eq : case3delta_pi ] ) to make the evolution of @xmath244 _ or _ @xmath193 the same as that for standard cdm , however clearly this is not possible for _ both _ @xmath244 _ and _ @xmath193 because in this case @xmath247 . because the primary cmb anisotropy is determined by both of these two variables , it will almost definitely differ between the edm and cdm models . furthermore , in the most general cases the evolution of @xmath91 will be changed as well , which gives rise to different low-@xmath195 cmb spectra . therefore , we expect that the cmb data will place stringent constraints on the general edm models , and can be used to distinguish alternative gravitational theories of dark matter . conversely , any attempt of modification of general relativity which claims to have reproduced the observed large scale structure must be confronted with the cmb fluctuation spectrum and polarization to test its viability . motivated by the recent developments in producing the large - scale structure formation with alternative theories of gravity , we have considered the prospect of using the cmb to constrain such theories in new ways . we take the confrontation with the matter power spectrum data as a first test of the perturbed cosmological model in our alternative gravity theory and assume that this test has been passed . this is because the currently successful theories like teves and @xmath6 really are _ only _ able to reproduce the observed lss rather than all the perturbation observables , and such a simplifying assumption efficiently restricts our model space so that the degeneracy problem is somewhat alleviated . as discussed in sect : model , reproducing the observed lss is a rather weak requirement on the theory because the lss only reflects the density perturbations at late time . this means that there is still considerable freedom to choose the evolution history of the dark - matter density perturbation @xmath81 , which will generally lead to different predictions about other cosmological observables such as the cmb spectra and polarization . we discussed how this freedom in the evolution history can be utilized by considering an example of the parametrization [ c.f . ( eq : edmparametrization ) ] of @xmath133 which only reduces to the familiar @xmath248 result with the parameter choice @xmath177 . as was shown in [ sect : model ] , the evolution of @xmath249 is completely governed by the edm stress history , which we quantify using the two variables @xmath136 and @xmath137 . these same two variables also control the evolution of gravitational potential , which is important for the cmb power spectrum . this implies that the cmb is an ideal test bed for edm models . in reality , we have one more freedom because once the @xmath133 is specified . ( eq : evolvedensityperturb ) becomes a single equation for the two variables @xmath250 and @xmath137 , so , in [ sect : numerics ] , we considered three separate cases : ( i ) @xmath251 , ( ii ) @xmath252 and ( iii ) @xmath253 . for simplicity , we have also made the following assumptions : ( 1 ) the evolution of @xmath141 is scale independent , as is implied by the observed lss , ( 2 ) the cosmological constant ( or other form of explicit dark energy violating the strong energy condition ) is neglected to simplify the numerics , ( 3 ) the edm equation of state parameter , @xmath254 is assumed to be small enough so that we can assume the standard cdm background evolution holds ; and , ( 4 ) only adiabatic initial conditions , a scale - independent @xmath255 primordial power spectrum , and effectively massless neutrinos are considered ( we will return to these assumptions below ) . for case i , the standard cdm relation @xmath256 still holds but the evolution of these potentials could be changed . if the deviation of edm from cdm only occurs after the last scattering , then this change in @xmath84 mainly modifies the isw effect and , hence , the low-@xmath110 cmb power . if the edm starts to deviate from cdm before last scattering , however , then the early changes in @xmath113 and @xmath129 would modify the cmb acoustic peak features in a complicated manner , as we described qualitatively in sect : cmb and [ subsect : case1 ] . the evolution of @xmath136 can be understood qualitatively : since the effect of the pressure perturbation is to counteract gravitational collapse , in order that @xmath257 grows faster than @xmath163 we need a negative @xmath258 and vice versa ( c.f . fig . [ fig : figure7 ] ) . the case ii evolution is problematic as long as we stick to the above simplifying assumption ( 1 ) , _ i.e. _ , requiring a scale - independent growth of @xmath141 , because in this case eq . ( eq : evolvedensityperturb ) implies a scale _ dependent _ @xmath137 ( @xmath259 is independent of @xmath50 ) , which diverges on large scales ( small @xmath50 ) . as we have seen in [ subsect : case2 ] , this leads to a strong late - isw effect , blowing up the low-@xmath110 cmb spectrum even if the edm only differs slightly from standard cdm . furthermore , a natural edm model with @xmath204 and @xmath260 does not exist in the literature , and this model is different from of case i , which arises naturally in @xmath6 models with @xmath261 . of course , the most general situation is our case iii , where both @xmath262 and @xmath137 are nonzero . needless to say , this generality also makes its exploration more difficult . one could , however , use the extra degree of freedom here in model constructions . if we fix _ both _ the evolution of @xmath133 _ and _ that of the potential @xmath263 , then @xmath137 and @xmath136 can be solved uniquely at the same time . such a construction has the advantage that we can fix the evolution of @xmath84 to be identical to that in standard cdm so that the isw effect is not changed at all . if the deviation of edm starts after last scattering , then such a model is degenerate with standard cdm even after cmb and lss data are taken into account , and furthermore the cmb - galaxy cross correlation data is inefficient in constraining it . allowing edm to deviate before last scattering , however , will almost surely predict different cmb peak features , because in this case @xmath264 and the acoustic oscillations of photon - baryon fluid are changed with respect to standard cdm ( c.f . [ sect : cmb ] ) . dropping the requirement on the evolution of @xmath84 makes the situation even worse , because in this case the isw effect and low-@xmath110 structure of the cmb spectrum are also changed . we stress that the cmb polarization also provides invaluable information for constraining edm models , because it depends only on the quadrupole moment of the photon temperature perturbation @xmath93 ( the photon anisotropic stress ) , and thus only on the physics close to the time of last scattering . throughout this paper we have chosen to parametrize @xmath265 as in eq . ( [ eq : edmparametrization ] ) . this is fairly simple and sufficient for our purposes as we only aim to show the importance of cmb data in constraining alternative gravitational dark matter theories , but not to constrain any specified model or given parametrization exactly . there are good reasons why more detailed parametrizations are needed for more precise future studies . first , although @xmath266 is a good approximation in the matter - dominated epoch , it breaks down when the contribution from radiation is still significant ( including the time prior to and around last scattering ) and when a cosmological constant is included . second , because the physics at the time of matter - radiation equality is relevant for edm models which deviate from cdm earlier , we sometimes need to choose @xmath152 to be smaller than the matter - radiation equality @xmath267 , which means that the evolution of @xmath133 or @xmath268 can not be completely scale independent ( remember that @xmath269 is relevant for the bending of the matter power spectrum @xcite ) . thus , future precise calculations , particularly those relevant for weak lensing and the galaxy - isw cross correlation , should use parametrizations which reduce exactly to @xmath248 in the @xmath1cdm model in appropriate limits . notice that our logic here is different from other related considerations of generalized dark matter ( gdm ) @xcite . in ref . @xcite , for example , the author parametrizes the stress sector of the gdm with three quantities , @xmath270 and @xmath271 , which respectively characterize the eos , pressure perturbation , and anisotropic stress of the gdm . the gdm effects on the cmb _ and _ matter power spectra are then studied by assuming some specific values of these quantities . our approach tackles the problem from a different direction and is therefore complementary to those earlier works . one interesting issue is the inclusion of hot dark matter , which is needed by teves itself . this can certainly be achieved by setting special values for @xmath270 and @xmath271 as in @xcite . in our approach we simply treat the edm as a `` black box '' which is completely controlled by separate conservation equations . by parametrizing @xmath272 we are able to extract information on the pressure perturbation @xmath136 and anisotropic stress @xmath137 from these conservation equations without knowing exactly what is in the box it can be mixtures of modified gravitational effects and neutrinos , or something else entirely . the other issue is related to the initial conditions . in this work we mainly focus on the adiabatic initial conditions as in @xcite . however , it is well known that the adiabatic initial condition is far from the only possibility , and there can be four regular isocurvature modes bucher2000 . these isocurvature modes are predicted by many theoretical models and may even correlate with the adiabatic one . in contrast to the adiabatic mode , the isocurvature mode excites sinusoidal oscillations of the photon - baryon fluid and so predicts the first cmb acoustic peak to be around @xmath273 rather than @xmath274 ; consequently a pure or dominating isocurvature initial condition is incompatible with basic cmb observations . nonetheless , a subdominant contribution from isocurvature initial conditions actually degenerates with other cosmological parameters and can not be ruled out by the current data @xcite . in our edm model , if the edm evolves differently from cdm early in the radiation era , then @xmath141 does not evolve adiabatically and the entropy perturbation @xmath275 is nonzero . in this case , one might naturally expect that there should be isocurvature modes in the initial conditions . a detailed calculation of the consequences of introducing such modes , however , generally requires thorough searches of the new parameter space ( which is enlarged compared to the standard case because now the amplitude , tilt of the isocurvature modes and their correlation with the adiabatic mode must be taken as free parameters ) like in trotta2001 , amendola2002 , beltran2004 , keskitalo2007 and is beyond the scope of this work . moreover , allowing different initial conditions ( especially tilts of the isocurvature mode which are significantly different from 1 ) can change the shape of matter power spectrum , which means that the parametrization of @xmath133 [ c.f . ( eq : edmparametrization ) ] should be scale dependent for the same reasons as discussed above . in this work we adopt a more conservative approach by assuming that the edm evolves like standard cdm at earlier times and so adiabaticity is a natural choice . in conclusion , we propose ways to use the cmb in order to constrain those alternative gravity theories for dark matter which claim to be compatible with the observed lss . we find that the cmb temperature and polarization spectra are good discriminators between these theories in general , especially when they deviate from the cdm paradigm before last scattering . if the deviation starts after last scattering , however , there can exist edm theories which are degenerate with respect to the standard @xmath1cdm model and can not be distinguished by cmb and matter power spectra . this point is particularly interesting from the viewpoint of model constructions . our results also indicate that the stress properties of dark matter , which determine the evolutions of both density perturbations and gravitational potential , can be studied and significantly constrained with existing and future data by using just the general conservation equations and without specialising to any specific theoretical model . we thank david spergel for helpful discussions . the numerical calculation of this work uses a modified version of the camb code @xcite . bl thanks university of st . andrews for its hospitality when part of this work was carried out and acknowledges supports from an overseas research studentship , the cambridge overseas trust , the damtp and queens college at cambridge . dfm acknowledges the humboldt foundation . and , * * , ( ) ; and , in _ theoretical and observational cosmology , _ edited by m. lachize - rey ( springer , new york , 1998 ) ; and , * * , ( ) ; , ph.d . dissertation , queens college and astrophysics group , cavendish laboratory , cambridge university , 2000 .

a third quantization formalism is applied to a simplified multiverse scenario . a well defined quantum state of the multiverse is obtained which agrees with standard boundary condition proposals . these states are found to be squeezed , and related to accelerating universes : they share similar properties to those obtained previously by grishchuk and siderov . we also comment on related works that have criticized the third quantization approach .

1005.2147

it is usually claimed that the quantum state of the universe is described by a wavefunction @xcite which should account for all the physical information which can be extracted from the universe . two main approaches have been followed in order to obtain such a wavefunction of the universe . first , it can be obtained as the solution of the wheeler - de witt equation @xcite . this can actually be seen as a schrdinger equation in which no time derivative appears in order to satisfy the time invariance required by the general relativity theory @xcite . secondly , a path integral approach can also be taken @xcite . then , the wavefunction of the universe is given by a sum over all geometries and field configurations which can be matched with a given value of the field configuration , defined in a spatial section of the whole space - time manifold . we make a wick rotation to euclidean time in order to obtain well - defined integrals , and then rotate back to lorentzian time to obtain the final results . in both cases , some boundary conditions have to be imposed @xcite . the hartle - hawking no - boundary proposal @xcite represents a universe which is created from `` nothing '' , meaning by that the absence of space , time and matter . vilenkin @xcite considered a universe which is also created from nothing , but through a quantum tunneling transition instead . vilenkin argued that the idea of a universe being created from nothing is not crazier than the creation of particle in other quantum theories . thus , it seems appropriate to consider that the universe may be created as a quantum fluctuation of the gravitational vacuum . once a universe nucleates in the space - time foam @xcite , it may bubble and eventually jump into an inflationary period @xcite , which is supposed to be the origin of our current universe . therefore , we must consider a multiverse in which changes in the topology of space - time are allowed . a variety of multiverse hypotheses have been recently considered from different cosmological viewpoints @xcite . the specific meaning which is ascribed to a single universe depends on the formalism of the relevant theory . some well - known examples include the everett s many - world interpretation of the quantum theory @xcite , the chaotic inflationary multiverse @xcite or the landscape in string theory @xcite , among others ( for an exhaustive review , see ref . @xcite and references therein ) . in this paper we only consider the case of a multiverse which is described by a set of quantum oscillators , each one representing a causally disconnected region of the whole space - time . topology changes were first claimed to appear in the quantum physics of black hole evaporation @xcite , and so first arose as a result of trying to take quantum physics seriously as a description of the whole universe . on pure cosmological grounds , the transition from a matter dominated universe into a space - time filled with phantom energy provides us with another example of a bifurcating topology originated from the big rip singularity which splits space - time into two disconnected regions . creation of universes and therefore topology changes are naturally contemplated in the third quantization formalism @xcite . this consists of a further quantization of the wavefunction of the universe similarly to how quantum field theory is constructed from the schrdinger wavefunction of matter fields . the computations are difficult to perform in the general case . in this paper , however , a simplified model will be presented in which such computations can in fact be carried out . moreover , the general scheme of the third quantization formalism can be applied to both , a multiverse made up of parent universes and a space - time foam filled with virtual baby universes @xcite . parent universes are defined to be large space - time regions with a hubble length of the same order as the hubble length of our universe . baby universes are considered to be virtual fluctuations of the metric in the vacuum of gravity , and their contribution to calculations of the gravitational field is extremely important at the planck length . the gravitational vacuum would be also populated with virtual lorentzian and euclidean wormholes . the former can be seen as solutions of the einstein s equations with at least two asymptotically flat regions , connecting two separate parts either of the same universe or of two different universes . the latter can be considered as euclidean sectors of a friedman space - time @xcite , whose quantum states can be seen as the exponentially decaying versions of the oscillatory universes from which they were wick rotated . furthermore , the vacuum state of gravity might also be observable as far as it could induce a loss of quantum coherence in the matter fields . the debate was centered ( see e.g. refs . @xcite and @xcite ) around the assumption of taking doubly or even multiply connected wormholes instead of simply connected ones in the quantum vacuum of gravity . equivalently , it can be placed in terms of whether virtual baby universes are created as single universes or in pairs . in the theory of the quantum multiverse , in which it has been shown that general topological changes may occur , the loss of quantum coherence in the matter field sector seems to be unavoidable . our aim in this paper is then to apply the third quantization formalism given in ref . @xcite to a simplified model of the universe , which nevertheless retains the fundamental features of the quantum theory when it is applied to the universe as a whole . it will be shown that in such a model a well - defined quantum state for the multiverse can be obtained , which satisfies the usual boundary conditions . squeezed states @xcite , which are usually interpreted as quantum states without any classical analog @xcite , are found in the context of the multiverse , and in particular they appear in the context of accelerated universes . this seems to give support to the idea that the acceleration is due to distinctively quantum mechanical effects @xcite . furthermore , the third quantization formalism used in this paper shows some advantages respect to other approaches which are usually taken in quantum cosmology . for instance , it can be demonstrated that the quantum state of the multiverse may be given in terms of the states of a quantum harmonic oscillator . dealing with frequencies instead of potentials simplifies the computations . moreover , neither a perturbative nor a semiclassical approximation needs to be taken in the model presented in this paper , and therefore the state of the multiverse which is obtained can describe the global state of the multiverse as a whole . before going any further , however , some caveats and comments should be made about the ideas and results considered in this paper . first of all , dealing with squeezed states in the context of the quantum state of the multiverse would , at first sight , seem somewhat extraneous . actually , squeezed states can be readily interpreted both in quantum optics and in a space - time foam made up of baby universes , which are both deployed in a common space - time . baby universes can in fact be taken as tiny particles representing small perturbations of the space - time field , whose quantum state affects the vacuum state of the matter field sector , and so this theory may turn out to be testable . it is more difficult to imagine how we could test a theory of a multiverse made up of parent universes . it looks quite different even though wormhole communications channels may exist through which microscopic particles could travel from one universe to other @xcite . furthermore , some correlations among the quantum state of different universes could also be considered . for instance , if the creation of universes would be produced in entangled pairs , whose correlations might induce some observable effects on the state of each individual universe in the pair @xcite . squeezed states have already been studied in other cosmological contexts such as the inflationary universe and gravitational waves @xcite . it was shown @xcite that the gravitational vacuum is populated by gravitons which evolve into a highly squeezed state . in this paper it is also demonstrated in the context of a third quantization formalism that a space - time foam made up of baby universes is in a squeezed state . if such a result could be extrapolated to parent universes , these universes might not be independent but quantum mechanically correlated , too . the paper can then be outlined as follows : a discussion with the main features of the third quantization formalism has been also added as sec . ii . in sec . iii , it is obtained the wavefunction for a quantum multiverse made up of friedmann space - times filled with an homogeneous and isotropic fluid , and the appropriate fock space is defined . we then analyze the two interesting limits of the quantum state of both a large parent universe and the quantum gravitational vacuum , which turn out to be described by a squeezed state with no classical analog . in sec . iv , the quantum state of the multiverse is represented by a density matrix rather than by a wavefunction , accounting thus for mixed states as well as pure ones . three representations are employed : first , the second quantized wavefunction is used as the configuration variable , the usual boundary conditions of refs . @xcite are applied and a probability interpretation can then be used . secondly , a general squeezed number representation is taken in order to study the quantum state of the multiverse . parent universes will be represented by number states . we show that the quantum state for the vacuum of baby universes is given by a squeezed state , and that high order correlations appear among them . for the sake of completeness , we have also used the usual @xmath0 representation in terms of coherent states . in sec . v , we compare the results obtained in this paper with previous works which have considered or criticized the third quantization formalism . the conclusions are collected in sec . vi , together with some comments to extend the model to a two - dimensional wave equation which would explicitly account for other matter fields . the basic idea of the third quantization formalism is @xcite : to treat the many - universe system as a quantum field theory on superspace . however , the name third quantization may be misleading . actually , the procedure is formally similar to that of the second quantization used in the quantum field theory of matter fields ( see , fig . [ comparative ] ) . the name third quantization comes from the fact that the field which is quantized is the wave function of the universe , and it depends on _ already _ second quantized matter fields , i.e. it depends on the particles existing within the universe . broadly speaking , the third quantization procedure starts from the wheeler - de witt equation , it takes one of the fields in that equation as a time variable and then we can proceed as it is usually done in quantum field theory . however , it not clear at all that such a time variable field can be found in the general 3 + 1 dimensional theory . such a field can be found in one - dimensional cosmological models as well as in 3 + 1 dimensional models with high symmetry . therefore , let us restrict our attention in this section to those cases . let us start then with one - dimensional model ( 1 temporal + 0 spatial dimensions ) . throughout this section , we are closely following ref . @xcite . in that model , a universe is represented by a point in space . however , if topology changes are allowed , a many - universe system is rather described by a set of points . on the other hand , time dependent matter fields within the universes are contemplated . therefore , let us start with the second quantized action for such a system . it reads , @xmath1 where @xmath2 , @xmath3 represent @xmath4 matter fields , @xmath5 is the supermetric in that superspace and @xmath6 is the lapse function . for simplicity , a mass term has been chosen to be the potential of the fields in the action ( [ a1 ] ) . then , we can proceed as usual by defining the conjugate momenta of the fields , @xmath7 , and the hamiltonian density of the system , i.e. @xmath8 in the canonical quantization procedure , @xmath9 , and the classical hamiltonian constraint , @xmath10 , turns out to be , @xmath11 . this is the wheeler - de witt equation for the model being considered and then , @xmath12 , represents the ( second quantized ) wave function of a single universe . to third quantize this theory , let us find an action whose variational principle leads to the wheeler - de witt equation . this can be done with the following third quantized action , @xmath13 where @xmath14 . variation of the action ( [ a3 ] ) with respect to @xmath15 leads directly to the wheeler - de witt equation and , therefore , all the information of the second quantized theory is preserved in the third quantization formalism @xcite . the usual formalism of quantum field theory can then be directly applied to the third quantization many - universe system . for instance , the three point euclidean green function can be defined as @xcite , @xmath16 @xmath17 is the coupling constant and @xmath18 is the two point euclidean green function , i.e. @xmath19}.\ ] ] @xmath20 $ ] is the euclidean version of the third quantized action ( [ a3 ] ) for this interaction scheme , i.e. @xmath21 thus , the third quantization formalism provides us with a system of interacting universes . for instance , the propagator ( [ a4 ] ) would physically represent the bifurcation of one universe ( see , fig . [ bifurcation ] ) . we can also define a wave function for the whole multiverse , i.e. @xmath22 $ ] , where @xmath23 is the field chosen as a time variable and then , @xmath24 . it is given by the solution of the third quantized schrdinger equation , @xmath25 where @xmath26 is the third quantized hamiltonian ( not to be confused with the wheeler - de witt operator ( [ a2 ] ) ) . this is constructed from the action ( [ a3 ] ) with the momentum conjugate to the wave function of the universe defined by @xmath27 where , @xmath28 ( in the general covariant form , @xmath29 ) . the meaning of the wave function of the multiverse given by eq . ( [ a7 ] ) can then be understood as follows : let , @xmath30 , be an orthonormal basis or number states , then @xmath31 where , @xmath32 , can be seen as the probability of @xmath33 universes at _ time _ @xmath23 or the probability of @xmath33 universes with field values @xmath23 @xcite . this formalism for one - dimensional universes can be adapted for 3 + 1 dimensional universes @xcite . the formalism has to be changed to account for the gauge symmetries which are not present in the one - dimensional case . however , some of the concepts of the one - dimensional case can be applied too to the case of more realistic universes . thus , from the point of view of the third quantization scheme , our single universe is just one universe within the whole many - universe system . nevertheless , in some approximation we have to recover the single universe we inhabit . this consists of a large region of the space - time of order of our hubble length called a parent universe . nevertheless , within a single parent universe quantum fluctuations of the metric can also be considered . let us notice that the gravitational action depends on the length scale and at planck length , such fluctuations become of the same order of the metric @xcite . thus , tiny regions of the space - time of order of the planck scale can virtually tunnel out of their parent space - time . these are called baby universes . the parent universe is influenced by such quantum fluctuations of the gravitational vacuum , for instance these can be seen as the source of a non - zero vacuum energy or cosmological constant , @xmath34 . in the third quantization scheme , therefore , we may say that a parent universe propagates then in a plasma of baby universes . it can be depicted as follows : let , @xmath35 and @xmath36 , be the wave functions of a parent and a baby universe , respectively . the second quantized action for each universe is given by eq . ( [ a1 ] ) with masses @xmath37 and @xmath38 , respectively . if we consider that the difference between the mass scale of a parent and a baby universes is large , i.e. @xmath39 , then , quantum transitions from baby to parent universes are exponentially suppressed . however , feynman diagrams like those represented in fig . [ bifurcation ] can still be considered . the third quantized action which represents then a single parent universe propagating in a plasma of baby universes is given by , @xmath40 = \frac{1}{2 } \int d x \left ( - ( \nabla\phi_p)^2 + m_p^2 \phi_p^2 - ( \nabla\phi_b)^2 + m_b^2 \phi_b^2 + \kappa \phi_p^2 \phi_b + \frac{\lambda}{3 } \phi_b^3 \right ) .\ ] ] baby universes are not observable by themselves but their effects may be observable , so it seems to be better to consider an hybrid action . to do this we deal with @xmath35 as a wave function in the second quantized formalism and with @xmath41 as an operator in the third quantized formalism . then we can interpret baby universes as tiny particles deployed in the parent space - time . the result is an effective second quantized action for the parent universe in which a potential term appears accounting for the effects of the plasma of baby universes . in such a scheme , it can be assumed an effective interaction action of the form , @xmath42 where the index @xmath43 labels the different modes of the baby universe field ( i.e. it labels different species of baby universes ) , and @xmath44 is the insertion operator at the nucleation event . it defines the space - time points of the parent universe in which the baby universes effectively nucleate . let us now sketch the third quantization formalism in a more realistic model of the universe . let us consider then the case of an homogenous and isotropic space time with some matter fields defined upon it . for the sake of simplicity let us assume just one scalar field , @xmath45 ( the generalization to account for more matter fields can easily be done ) . in that case , the second quantized action can be written as @xcite @xmath46 , \ ] ] where , @xmath6 is again the lapse function and , @xmath47 , being @xmath48 the scale factor . the minisupermetric is given then by , @xmath49 with , @xmath50 , being @xmath51 and @xmath37 the gravitational constant and the planck mass , respectively . the potential term in eq . ( [ a12 ] ) is given by , @xmath52 where , @xmath53 , stands for hyperbolic , flat and closed spatial sections of the space - time , respectively , @xmath34 is the cosmological constant and @xmath54 is the potential of the matter field @xmath45 . thus , for homogeneous and isotropic models it is clear that the scale factor can play the role of a time variable . this can also be seen from the signature of the minisupermetric ( [ a13 ] ) . therefore , the third quantization procedure can formally be applied in a similar way to the case of one - dimensional universes . then the canonical momenta and hamiltonian , respectively , are given by @xmath55 = \mathcal{n } \mathcal{h}.\end{aligned}\ ] ] the hamiltonian constraint , @xmath56 , turns out to be @xmath57 , where @xmath58 is the wave function of the universe . finally , the third quantized action leading to the wheeler - de witt equation reads , @xmath59 as a particular example , let us consider a fluid with equation of state given by , @xmath60 , being @xmath61 a constant parameter , and @xmath62 and @xmath63 the pressure and the energy density of the fluid , respectively . these are given by , @xmath64 on the other hand , for an homogeneous and isotropic space - time the equation of conservation of the cosmic energy reads , @xmath65 it can be easily solved with the help of the equation of state , @xmath60 . the energy density turns out to be then , @xmath66 . therefore , combining eqs . ( [ a17 ] ) and ( [ a16 ] ) the hamiltonian constraint turns out to be , @xmath67 which is the friedmann equation , being @xmath68 . under canonical quantization , @xmath69 , and the wheeler - de witt equation can be written as , @xmath70 where , @xmath71 , is the wave function of the universe . ( [ wdw ] ) can be seen as the classical equation of motion of a harmonic oscillator . it can be derived from the following third quantized action , @xmath72 which formally corresponds to the action of a harmonic oscillator , where the scale factor @xmath48 formally plays the role of time . the _ time - dependent _ frequency , @xmath73 , is given by @xmath74 where the velocity of light has been introduced for completeness . in the general case the computations are actually more complicated . however , the wheeler - de witt equation can be seen like a wave equation and therefore , some of the developments used in quantum optics might also be applied to the third quantized formalism of the multiverse . for instance , let us make the change , @xmath75 . then , the wheeler - de witt equation can be written as @xcite @xmath76 this can be seen as the wave equation for a radiation field _ propagating in the void _ ( see , fig . [ comparative ] ) , at the speed @xmath77 , where the field is now the wave function of the universe , @xmath78 . the _ source term _ @xcite in eq . ( [ a21 ] ) , @xmath79 , is given by @xmath80 now , let us expand the modes of the _ free _ wave function of the universe ( @xmath81 , in eq . ( [ a21 ] ) ) as , @xmath82 then , the third quantization procedure eventually consists of promoting the wave function of the universe into an operator , @xmath83 . therefore , let @xmath84 and @xmath85 be , respectively , the creation and annihilation operators of universes in the mode @xmath86 , i.e. @xmath87 thus , the system described by eq . ( [ a21 ] ) can formally be viewed as the radiation emitted by a classical current distribution given by @xmath88 . following the analogy , the hamiltonian that would describe the interaction between the field @xmath15 and the current @xmath88 is given by @xcite , @xmath89 in the interaction picture , therefore , the state of the multiverse evolves ( as the scale factor is growing up ) into a coherent state defined by , @xmath90 where @xcite , @xmath91 with @xmath92 let us take as the boundary condition for the _ initial _ state of the multiverse that for a zero value of the scale factor there is no space , time and matter ( i.e. there is nothing ) . analogously , this is what strominger calls the _ void _ @xcite , i.e. @xmath93 . then , the state of the multiverse described by eq . ( [ a21 ] ) would evolve as the scale factor is growing up into a coherent state given by , @xmath94 this is an example of how far can be formally taken the analogy between the electromagnetic field and the third quantization formalism of the quantum multiverse , at least in the case of cosmological models of high symmetry . let us consider a quantum multiverse made up of homogeneous and isotropic universes filled up with a perfect fluid , for which the friedmann equation is given by eq . ( [ friedmann ] ) . following the third quantization procedure , the third quantized action for such a many - universe system is given by eq . ( [ 3rd action ] ) . it formally corresponds to the classical action of a harmonic oscillator , where it is the scale factor which plays the role of time . then , the quantum state of the multiverse , @xmath95 , can be defined as the solution of the third quantized schrdinger equation , @xmath96 where @xmath97 @xmath98 being the momentum conjugate to the wave function of a single universe , @xmath15 . the frequency @xmath73 is given by eq . ( [ frequency ] ) different disconnected regions of the whole space - time may undergo different expansion ( or contraction ) rates because they can be filled up with different energy - matter contents and have different spatial geometries , both features being incorporated in the value of the frequency @xmath73 . therefore , given an orthonormal basis of solutions of eq . ( [ schrodinger third quantized ] ) , @xmath99 , the general quantum state of a multiverse made up of disconnected regions , each being dominated by a particular kind of energy - matter content , can be expressed in terms of a linear combination of product states of the form @xmath100 where @xmath101 is the number of universes of type @xmath43 , represented by a second quantized wave function @xmath102 , which are filled up with an energy - matter content described by the frequency @xmath103 . let us notice that the quantum state given by eq . ( [ statemultiverse ] ) is a well - defined state for closed geometries and open geometries . the frequency @xmath73 of the third quantized harmonic oscillator which would represent the quantum state of the multiverse , contains not just the information of the second quantized theory but also that of the classical one . using the friedmann equation we can show that the frequency determines the expansion rate of a single universe . for instance , let us first consider the case of a flat de sitter universe with an exponential expansion . the solutions of eq . ( [ friedmann ] ) are then given by , @xmath104 , with @xmath105 , where the positive branch corresponds to the exponential expansion of a flat de sitter space - time . taking into account the negative branch , the solution for a closed de sitter space - time can be obtained @xcite , i.e. @xmath106 . furthermore , in conformal time , @xmath107 , the scale factor turns out to be , @xmath108 , which can be seen as the two - dimensional metric of a cosmological warp drive @xcite . next we consider the anti - de sitter space - time for which @xmath109 . analogously to the former case , the solution @xmath110 can be obtained , which in conformal time turns out to be , @xmath111 , and as a final example let us note that the tolman - hawking closed baby universe used in @xcite corresponds to the values , @xmath112 and @xmath113 , in eq . ( [ friedmann ] ) . in a similar way we can depict other kinds of baby and parent universes in the third quantized theory through the value of the frequency . quantum mechanically , in the case of a multiverse made up of flat universes filled up with a radiation like fluid , for which @xmath114 , the frequency @xmath73 turns out to be a constant given by @xmath115 , and the state of the multiverse is represented then by the state of a harmonic oscillator with constant frequency which is , of course , very well - known . in general , for universes with fluids for which @xmath116 , the quantum state of the multiverse can be expressed in terms of harmonic oscillators with a _ time_-dependent frequency , which have been studied in the past @xcite , and to which a renewed attention has been recently paid @xcite . using the method of invariants developed by lewis and riesenfeld @xcite , the solutions of eq . ( [ schrodinger third quantized ] ) can be expressed in terms of the eigenstates of an auxiliary operator , @xmath117 , which is invariant under the evolution ( i.e. scale factor invariant ) given by the hamiltonian , @xmath26 , i.e. @xmath118 = 0 $ ] , which implies that , @xmath119 , with @xmath120 being a schrdinger state vector . therefore , the action of an invariant operator @xmath117 on a schrdinger state vector produces another solution of the schrdinger equation . furthermore , if the invariant operator does not involve differentiation under the scale factor , relative phases for the eigenstates of @xmath117 can be found in such a way that such states themselves satisfy the schrdinger equation ( [ schrodinger third quantized ] ) , giving us an orthonormal basis for the space of solutions . thus , let us assume following @xcite the existence of a hermitian invariant operator , @xmath117 , with the homogeneous , quadratic form , @xmath121 $ ] . for such an operator to be invariant a set of equations has to be satisfied by the coefficients @xmath122 , @xmath123 and @xmath124 . solving those equations , the invariant operator can be written as , @xmath125 $ ] , where the function @xmath126 obeys the auxiliary equation @xmath127 eq . ( [ requation ] ) can also be solved in terms of the solutions of the wheeler - de witt equation , @xmath128 . let @xmath129 and @xmath130 be two independent solutions of the latter equation , then @xmath131 is a solution of eq . ( [ requation ] ) . moreover , the invariant operator @xmath117 can be expressed in terms of creation and annihilation operators , such as @xmath132 which gives , @xmath133 . for a flat space - time we can obtain analytical solutions of eq . ( [ requation ] ) and @xmath134 is given by eq . ( [ r ] ) with @xmath135 where @xmath136 and @xmath137 are bessel functions of first and second kind , respectively , of order @xmath33 . for other spatial geometries and energy - matter contents , the wheeler - de witt equation ( [ wdw ] ) may not be analytically solvable . however , it does not introduce any conceptual drawback as far as the existence of the solutions of eq . ( [ wdw ] ) is guaranteed , and they may be obtained analytically or numerically . nevertheless , analyticity makes clearer the computations and for that reason , we will mainly consider the case of flat universes for which such solutions are given by eq . ( [ fi1 ] ) . in any case the scale factor dependent creation and annihilation operators given by eqs . ( [ creationannihilation1 ] ) and ( [ creationannihilation2 ] ) give the expected results when they are applied on the eigenstates of the operator @xmath117 , @xmath138 , i.e. @xmath139 and , @xmath140 . therefore , the eigenvalues are invariant under scale factor transformations . furthermore , the scale factor dependent eigenstates , @xmath141 , can be linked to the eigenstates of a harmonic oscillator with constant frequency ( let us say , @xmath142 ) . in the schrdinger representation , both wave functions are related through the following unitary relation , @xmath143 where , @xmath144 , and @xmath145 are the eigenfunctions of the hamiltonian of a harmonic oscillator with constant frequency , i.e. @xmath146 , with @xmath147 . the unitary operator , @xmath148 , is given by @xmath149 with , @xmath126 , defined by eq . ( [ r ] ) . then , the most general solution @xmath150 of the schrdinger equation ( [ schrodinger third quantized ] ) , can be written as @xcite @xmath151 , are given by @xmath152 which has a well defined value because the zeros of the bessel functions @xmath136 and @xmath137 do not coincide , and therefore @xmath134 is not degenerate for any value of the scale factor . the number eigenstates @xmath138 of the invariant operator @xmath117 form an orthonormal basis for the space of solutions of the schrdinger equation , and the eigenfunctions @xmath153 in eq . ( [ state_multiverse ] ) can thus be interpreted as the probability amplitude for @xmath154 universes with a value of the scale factor @xmath155 , for which their second quantized wave functions are given by @xmath15 . if we consider the quantum state of a set of universes with different values of the scale factor , the probability amplitudes would be given by the product of harmonic wave functions ( [ state_multiverse ] ) , for each value of the scale factor . however , it seems to be more appropriate to consider probability amplitudes regardless of the value of the scale factor because , as it has been already noticed , the number of universes within the multiverse is a scale factor invariant quantity . in that case , we integrate over the scale factor to trace it out . on the other hand , since @xmath15 has a well - defined semi - classical regime @xcite , the quantum state of the multiverse given by eq . ( [ state_multiverse ] ) can represent both a multiverse of homogeneous and isotropic parent universes and a space - time foam made up of baby universes . moreover , let us notice that unlike in the second quantized theory , in which a quantum state of the universe can not in general be defined @xcite , the third quantized state of the multiverse given by eq . ( [ state_multiverse ] ) is indeed well - defined because there are no space - time singularities in the quantum multiverse . however , the orthonormal eigenstates @xmath156 , are not eigenstates of the hamiltonian given by eq . ( [ hamiltonian 3-quantized ] ) for all values of the scale factor ( except in the case @xmath157 ) , and therefore they are not stationary solutions for the quantum state of the multiverse . in fact , in terms of the creation and annihilation operators given by eqs . ( [ creationannihilation1 ] ) and ( [ creationannihilation2 ] ) , the hamiltonian ( [ hamiltonian 3-quantized ] ) can be written as @xmath158 , \ ] ] where , @xmath159 the quadratic terms in @xmath160 and @xmath161 in the hamiltonian given by eq . ( [ hamiltonian2 ] ) make the quantum state of the multiverse to evolve into a squeezed state @xcite . the squeezing effect is greater for the quantum state of a multiverse made up of accelerated universes @xcite . it might actually be related to the quantum nature of accelerated universes @xcite , in the sense that squeezed states are usually considered to be sharp quantum states with no classical analog @xcite . however , the squeezing effect vanishes as the dominant fluid of the single universes that form up the multiverse becomes more and more similar to a radiation fluid when @xmath162 . then , the state of the multiverse evolves into a set of conventional coherent states , which are usually taken to be the most classically possible states in the quantum theory . that can also be considered in the case of the multiverse , because for @xmath163 the wave function of a single universe turns out to be a plain wave whose argument is the classical action , i.e. @xmath164 , with @xmath165 . then , the hamilton - jacobi equation and the classical equation of motion for the scale factor are satisfied @xcite , being the distribution of momenta a delta function centered at the classical momentum , i.e. @xmath166 , with @xmath167 the classical momentum . in that sense , @xmath168 represents a classical universe and coherent states can associated with a multiverse consisting of classical universes ( see also ref . @xcite ) . therefore , for a general value of the scale factor and @xmath169 , the state of the multiverse is represented by a squeezed state . however , taking into account the asymptotic expansions of bessel functions for large arguments in the case of parent universes with a large value of the scale factor , in eqs . ( [ r ] ) and ( [ fi1 ] ) , it can be shown that @xmath170 , being @xmath171 . thus , the coefficients @xmath172 and @xmath173 in the hamiltonian ( [ hamiltonian2 ] ) can be approximated to @xmath174 so that the hamiltonian ( [ hamiltonian2 ] ) turns out to be the usual hamiltonian for a harmonic oscillator in terms of the number operator , @xmath175 , with a scale factor dependent frequency given by @xmath73 . then , the quantum correlations between number states disappear and they turn out to be an appropriate representation for the quantum state of the multiverse . in particular , at large values of the scale factor ( i.e. ; @xmath176 ) , the single universe approximation for which @xmath177 , can be considered . on the other hand , taking into account the expansion of bessel functions for small arguments in the case of virtual baby universes with a small value of the scale factor ( typically of planck length ) in eqs . ( [ r ] ) and ( [ fi1 ] ) , then , @xmath178 and @xmath179 , and therefore , @xmath180 and @xmath181 for @xmath182 . thus , from eqs . ( [ beta+ ] ) and ( [ beta0 ] ) , it follows that the coefficients @xmath172 and @xmath173 can be approximated as @xmath183 where @xmath184 is a constant parameter given by , @xmath185 if @xmath186 . for @xmath187 , the value of @xmath184 actually coincides with @xmath188 , but in that case @xmath189 in the hamiltonian given by eq . ( [ hamiltonian2 ] ) and no squeezing effect is present , as has been already shown . however , if we assume that baby universes are instead vacuum dominated universes or closed baby universes , for which @xmath190 and @xmath191 , respectively , then the non - vanishing value @xmath192 and @xmath193 in eq . ( [ hamiltonian2 ] ) when the scale factor decreases implies that high order correlations between number states are going to play an extremely important role in the space - time foam of baby universes at the planck length . it also means therefore that the number state representation is not an appropriate representation of the quantum gravitational vacuum . instead , it would be more properly represented by a squeezed state , which might be even experimentally observable because such a gravitational state would induce a loss of quantum coherence in the vacuum state of the matter field sector @xcite . in fact , from the hamiltonian ( [ hamiltonian2 ] ) one can clearly infer that baby universes should be continuously created and annihilated in pairs , at least under the assumptions made in the present model . that appears to enhance the result that euclidean doubly connected wormholes copiously nucleate , together perhaps with simply connected wormholes . this implication is relevant for the debate on the loss of quantum coherence in the matter field sector due to the gravitational vacuum ( see refs . @xcite and @xcite ) because it suggests that the loss of quantum coherence of matter fields might be unavoidable . on the other hand , the third quantized vacuum energy , @xmath194 , retains a marginal non zero value even when the scale factor decreases to @xmath195 , i.e when the universes are annihilated . this is a well - known feature of the quantum vacuum state with no classical analog , applied in the present case to the third quantized vacuum state of gravity . finally , let us briefly discuss the question on how and when the universes are quantum cosmologically created . in the third quantization formalism , an interaction scheme can be used to describe the space - time foam of a parent universe . as it was shown in sec . ii , an insertion operator is then introduced to define the space - time points of the parent universe in which the creation and annihilation operators of baby universes are located . thus , the interaction hamiltonian can in general be written as , @xmath196 , where the index @xmath43 accounts for the different species of baby universes being considered in the space - time foam . however , in the case of a multiverse made up of parent universes , which can not be placed in a common space - time , the above question becomes misleading . let us notice that space and time are properties with a well - defined meaning only within a single parent universe , and they are not therefore well - defined _ outside _ the universe . so the hilbert space corresponding to a multiverse made up of parent universes can not be described in terms of space - time coordinates but , instead , of statistical ones . however , correlated pairs of universes can still be considered because entanglement is a quantum property directly stemming from the superposition principle , and therefore , a pair of parent universes can be created at points of such a general statistical hilbert space . the most general quantum state of the multiverse is not represented by a pure state defined by a third quantized wave function , but rather by a mixed state described in terms of a density matrix operator , @xmath63 , for which the explicit scale factor dependence is given in the third quantization formalism by the schrdinger like equation , @xmath197 \rangle$ ] . therefore , let us consider the quantum state of a multiverse made up of flat space - times in the representation defined by the value of the second quantized wave function @xmath15 , i.e. the _ coordinate _ representation . then , taking into account the expression of the hamiltonian given by eq . ( [ hamiltonian 3-quantized ] ) , the schrdinger equation for the density matrix in such a representation can be written as , @xmath198 this can be solved by using the following gaussian ansatz , @xmath199 inserting eq . ( [ ansatz ] ) into the schrdinger equation ( [ schrodinger3 ] ) , we obtain the following equations for the coefficients , @xmath200 the last of these equations can actually be transformed into the wheeler - de witt equation by making the following change : @xmath201 ( i.e. @xmath202 ) . then , eq . ( [ 3 ] ) turns out to be @xmath203 , the solutions of which are given by eq . ( [ fi1 ] ) . the quantum state of the multiverse can therefore be expressed in terms of the two values of the second quantized wave functions of the universe , @xmath15 and @xmath204 respectively , as @xmath205 where the constants of integration have been chosen to satisfy the normalization criteria , @xmath206 , and , @xmath207 , is the general solution of the wheeler - de witt equation . @xmath208 and @xmath209 are two constants to be determined by the boundary conditions . the trace of the density matrix given by eq . ( [ generaldensity ] ) shows no divergences even when @xmath210 , it vanishes for an infinite value of the scale factor , avoiding in this way cosmic singularities , and it also satisfies the usual boundary conditions imposed to the second quantized wave function of the universe when the scale factor degenerates as @xmath211 . this can be checked by taking into account the limits of the bessel functions in eq . ( [ fi1 ] ) for a vanishing value of the scale factor , i.e. @xmath212 and @xmath213 . thus , if @xmath214 , the solution of the wheeler - de witt equation is given by @xmath215 , and the density matrix ( [ generaldensity ] ) tends to zero as the scale factor vanishes , satisfying in this way a boundary condition which corresponds to the hartle - hawking s no - boundary condition @xcite . on the other hand , if @xmath216 , then @xmath217 , and the density matrix ( [ generaldensity ] ) still represents a small oscillation even when the scale factor decreases to @xmath195 , satisfying then a boundary condition which corresponds to the vilenkin s proposal of a tunneling boundary condition @xcite . the state of the multiverse is therefore well - defined by eq . ( [ generaldensity ] ) , and the diagonal elements , @xmath218 , can be then interpreted as a probability distribution which , for a given value @xmath15 , is peaked around the solution of the wheeler - de witt equation , @xmath219 , for which the distribution @xmath218 is a maximum , i.e. @xmath220 . let us recall that the state of the multiverse represented by the density matrix ( [ generaldensity ] ) avoids cosmic singularities , at @xmath195 and @xmath221 , because it gives a zero value of the probability amplitude both when the scale factor degenerates and when it grows up to infinity . for instance , let us consider the case of a phantom energy dominated universe in which a bip rip singularity appears in the future of our proper time , splitting the universe into two causally disconnected parts before and after the singularity . in the context of the third quantized multiverse , such a model is rather described by two universes , provided that their quantum states are well defined by eq . ( [ generaldensity ] ) , which might be even entangled @xcite , and therefore no singularities appear in such a multiverse . that is an example of the way in which the third quantized formalism can also account for topological changes in the space - time , like the one that would happen , for instance , in the transition from a matter dominated universe into a phantom dominated one ( i.e. from one single universe to a pair of , may be entangled , universes ) . let us also notice that the quantum state represented by the density matrix ( [ generaldensity ] ) is a non - factorizable state for any value of the scale factor . however , it can not be inferred that the quantum state of the multiverse is then in a mixed state . it would only be so if the representation given by the second quantized wave function of a single universe is orthogonal . nevertheless , the solutions of the wheeler - de witt equation are not in general orthogonal @xcite , and therefore it is not at all clear in such a representation whether the multiverse stays in a pure or a mixed state . for instance , for large values of the scale factor the quantum state represented by the density matrix ( [ generaldensity ] ) is still a non - factorizable state . however , it has been shown that in the same limit the number state representation is orthogonal and diagonalizes the hamiltonian . thence , the quantum state of a single parent universe is described by a pure state defined by the value of the quantum number , @xmath222 . the number state representation could be seen then as an orthogonal or _ decoherenced _ coarse graining @xcite , given by a superposition of second quantized wave functions weighted along @xmath15 accordingly to the third quantized wave functions of a harmonic oscillator . finally , let us recall that the density matrix for the quantum state of the universe was already obtained in ref . our consideration of the density matrix in this section can be just viewed then for the aim of completeness concerning the properties of the quantum state of the multiverse in the third quantization formalism . let us consider now the number state representation defined by the relations given in eqs . ( [ numberstates1]-[numberstates3 ] ) . it has been shown that in the case of parent universes with a large value of the scale factor , or equivalently in the semi - classical limit where @xmath223 , such a number representation diagonalizes the hamiltonian @xmath224 , because @xmath225 in eq . ( [ hamiltonian2 ] ) . in that case , the scale factor dependence of the density matrix , @xmath226 , is given by the schrdinger equation which implies a differential equation for the density matrix elements , i.e. @xmath227 , where the dot stands for the derivative with respect to the scale factor . the solutions are given then by , @xmath228 where @xmath229 is again defined by eq . ( [ beta0 ] ) , with @xmath230 for large values of the scale factor . therefore , if the initial density matrix is diagonal in such a number representation , i.e. @xmath231 , then the diagonal elements remain constant with respect to the scale factor . for instance , for the pure state given by @xmath232 , which would represent a large single universe , there are no off - diagonal elements in the representation of the density matrix as the scale factor continues to increase and , therefore , once a large universe can be considered , transitions to other number of universes are asymptotically suppressed along the subsequent expansion of such a single universe . however , for small values of the scale factor in such a representation , non - diagonal elements appear in the schrdinger equation and the number states turn out to be squeezed states . let us note that in terms of the creation and annihilation operators for modes of a unity value of the frequency , @xmath233 and @xmath234 , the creation and annihilation operators given by eqs . ( [ creationannihilation1 ] ) and ( [ creationannihilation2 ] ) can be written as , @xmath235 where , @xmath236 with , @xmath237 , and @xmath126 defined by eq . ( [ r ] ) . it means therefore that the number eigenstates of the invariant operator @xmath117 do not diagonalize the hamiltonian of the multiverse , and the squeezing effect becomes then larger for small values of the scale factor . however , let us also notice that quantum effects like squeezing or entanglement crucially depend on the chosen representation @xcite . this can be easily shown in the realm of quantum optics with a simple example : one photon state in a given mode of constant frequency , namely @xmath238 , which is of course a number eigenstate in the above representation , turns out to be a squeezed state in the representation of the number states given for another different frequency , say @xmath239 , i.e. @xmath240 , with @xmath241 , being @xmath242 and @xmath243 the annihilation and creation operators of each mode , respectively . in quantum optics the appropriate representation is derived from the experimental setting , e.g. if the experimental devices of the setting measure the number of photons with energy @xmath244 , then the eigenstates of the number operator of that mode , @xmath245 , would become the appropriate representation . the choice of the representation is not that clear nevertheless in the case of the quantum multiverse . after all , identifying the relevant physical observables in the multiverse is not at all an easy task . in the case of the quantum multiverse , it can be assumed that the lewis number states , which are defined as the eigenstates of the invariant operator @xmath117 , represent the occupation number for the quantum state of the multiverse , i.e. the number of universes in the multiverse , which is scale factor invariant . however , other representations can be more useful in order to obtain the solutions of the schrdinger equation for the elements of the density matrix , i.e. @xmath197 \rangle$ ] . for instance , let us consider the diagonal representation of the hamiltonian ( [ hamiltonian 3-quantized ] ) . it can be defined by the scale factor dependent creation and annihilation operators which correspond to the modes of the hamiltonian with proper frequency , @xmath73 , i.e. @xmath246 where , @xmath247 . in such a representation , of course , the hamiltonian turns out to be diagonal and given by , @xmath248 . the annihilation and creation operators given by eqs . ( [ annihilationomega ] ) and ( [ creationomega ] ) can , moreover , be related to the annihilation and creation operators of modes of unity frequency , @xmath249 and @xmath250 , through the following squeezing transformation , @xmath251 where , @xmath252 and hence , @xmath253 . then , the new number states , @xmath254 , can be related to the number states of modes of unity frequency , @xmath255 , by @xmath256 , is defined by @xmath257 @xmath258 is the squeezing parameter which is related with eqs . ( [ muomega ] ) and ( [ nuomega ] ) by , @xmath259 , and , @xmath260 . as a matter of completeness , let us note that the number eigenstates of the hamiltonian can also be related to the lewis number states , @xmath261 , defined in sec . ( [ numberstates1])-([numberstates3 ] ) , as they are both linked to the number states of modes of unity frequency , @xmath255 . thus , if @xmath262 is the squeezing operator that relates the lewis number states , @xmath261 , with those of unity frequency , @xmath255 , defined by @xmath263 where , @xmath264 , with , @xmath265 , and , @xmath266 , being @xmath267 and @xmath184 given by eqs . ( [ mu0 ] ) and ( [ nu0 ] ) , then @xmath268 . therefore , the eigenstates of the hamiltonian , @xmath269 , are related to the number states , @xmath270 , through a more complicated squeezing transformation given by , @xmath271 , and therefore , they preserve the orthonormality relations among the number states of a given mode . thus , the hamiltonian eigenstates , @xmath269 , form also an orthonormal basis for the space of solutions . in order for these to satisfy the schrdinger equation , they have to be shifted by some phases which can be obtained following the same iterative , recurrent procedure as the one used in ref . @xcite . then , the orthogonal representation of hamiltonian eigenstates that satisfy the schrdinger equation is given by , @xmath272 where , @xmath273 let us now compute the scale factor dependence of the density matrix elements by solving the schrdinger equation in such a representation of hamiltonian eigenstates , @xmath274 , in which the density matrix operator can be written as , @xmath275 . the solutions to the schrdinger equation are given then by , @xmath276 with the diagonal elements taking on constant values . by employing another number representation it is possible to re - express the density matrix in terms of squeezing transformation operators . for instance , in terms of the number states of modes of unity frequency , @xmath255 , the elements of the density matrix operator are given by , @xmath277 where the involved matrix elements can no longer be trivially calculated . let us consider next the @xmath278 representation of the quantum state of the multiverse in terms of coherent states . these can be defined in terms of the number eigenstates of the hamiltonian as , @xmath279 , is a complex valued variable , and @xmath280 is given by eq . ( [ omega ] ) . the coherent states ( [ cs ] ) are the eigenstates of the annihilation operator for a mode of frequency @xmath281 , with scale factor eigenvalues given by , @xmath282 in the customary @xmath0 representation , the action of the creation operator upon the coherent state ( [ cs ] ) is given by , @xmath283 and the density matrix can be written then as , @xmath284 where , @xmath285 , are the so called bargmann coherent states @xcite . thus , using eq . ( [ rhop ] ) in the schrdinger equation , @xmath286 $ ] , a differential equation is obtained for the function @xmath287 , i.e. @xmath288 this equation can be easily solved by making the change , @xmath289 . the solutions can be written as , @xmath290 @xmath291 and @xmath292 are two constants to be determined by the boundary conditions and @xmath293 and @xmath294 are two independent solutions of the master equation ( [ schrodingerequation ] ) , given by @xmath295 where @xmath280 is once again given by eq . ( [ omega ] ) , and the integration constants have been chosen to ensure the real valued character of the function @xmath296 . ( [ p1 ] ) and ( [ p2 ] ) closely recall the form of the operator which represents the potential vector , @xmath297 , in the usual second quantization formalism of the electromagnetic field @xcite . thus , eq . ( [ p ] ) may be interpreted as an oscillating field of a particular scale factor dependent mode , @xmath280 , although the quantized field is now the wave function of a single universe . other solutions to eq . ( [ schrodingerequation ] ) can be found as well . for instance , by expanding in series the function @xmath296 , i.e. @xmath298 , the solution can be written as , @xmath299 where @xmath280 is given by eq . ( [ omega ] ) , and the integration constants have been again chosen to ensure a real value for the function @xmath296 . in terms of other number representations , it may become difficult to carry out the explicit computation of the function @xmath278 . for instance , let us consider a coherent state representation in terms of a general number representation , , i.e. @xmath300 where , @xmath301 are the number states of modes of a generic frequency , @xmath302 , defined in terms of the annihilation and creation operators , @xmath303 where for convenience we have retained the ( plausible ) complex character of the frequency of the mode . the phases in eq . ( [ csg ] ) , which are proportional to @xmath304 , account for the evolution of the number states like schrdinger vectors . therefore , in such an unspecified representation the schrdinger states evolve into squeezed states , and the hamiltonian is transformed into , @xmath305 where @xmath306 and @xmath307 are functions which in general depend on the scale factor , except in the case of the diagonal representation of the hamiltonian . they are given by , @xmath308 and therefore , @xmath309 . it can be checked that when the frequency of the representation , @xmath310 , equals the proper frequency of the hamiltonian , @xmath281 , then such a representation diagonalizes the hamiltonian . in general , the schrdinger equation for the density matrix , @xmath286 $ ] , can be written as , @xmath311 e^{-|\alpha|^2 } \partial_a p_g(\alpha , a).\end{aligned}\ ] ] taking into account that the relations ( [ annihilationp ] ) and ( [ creationp ] ) are satisfied now with @xmath312 , @xmath313 and @xmath304 instead of @xmath314 , @xmath315 and @xmath280 , the differential equation for the function @xmath316 can be written as , @xmath317 p_g(\alpha , a ) , \ ] ] where , @xmath318 , the drift vector , @xmath319 , is given by @xmath320 the diffusion matrix , @xmath321 , is given by @xmath322 with , @xmath323 , being , @xmath324 . in general , eq . ( [ general ] ) is not easy to be analytically solved . however , before going any further it should be first clarified the meaning of the different representations for a particular cosmological model in terms of what can be observed , which is not at this stage clear at all . the third quantization formalism has already been used in previous works on the multiverse and it has been criticised ( see ref . @xcite , and references therein ) . first of all , vilenkin argued that an infinite number of fields and interaction types had to be introduced in order to describe topology changes in a four - dimensional realm , i.e. for all the non - simply connected space - times that can generally be considered . let us notice , however , that such a non - simply connected manifold can actually be converted into a simply - connected one by cutting it with a finite number of hypersurfaces @xcite , and this argument can be applied in the four - dimensional case . hawking pointed out @xcite that the quantum state of the universe would then be given by a set of quantum oscillators , an idea similar to the multiverse scenario . in the third quantization formalism , on the other hand , the wave function which represents each of the disconnected regions might be given by that of a harmonic oscillator . in that case , the third quantization formalism adds a manageable procedure to study a multiverse made of disconnected regions , at least in the case of high symmetry . the second main objection of ref . @xcite was on the interpretation of the creation of universes in the third quantization formalism . vilenkin argued that the usual interpretation in terms of @xmath325 and @xmath326 states , corresponding to @xmath327 and @xmath328 , respectively , with @xmath154 the number of universes , does not match the meaning of the creation of a universe from nothing proposed by himsef some years before @xcite . the @xmath329 state would correspond to the `` nothing state '' when the scale factor decreases , whereas the @xmath330 state would describe a single universe with a large value of the scale factor , i.e. when @xmath331 . however , as vilenkin himself objected @xcite , the scale factor can not be used as a time variable because it is not a monotonic variable . moreover , taking a particular value of the scale factor is nothing but considering a particular time reparametrization , @xmath332 , and it is not expected that the number of universes in the whole multiverse would depend on a particular choice of a reference system within a single universe . on the other hand , we have shown that the number state which properly defines the number of universes is represented by the lewis number state defined by eq . ( [ numberstates3 ] ) and thus , the number of universes in the whole multiverse is scale factor invariant , i.e. @xmath333 . however , the concept of universes created from nothing can still be meaningfully kept because the vacuum state of the multiverse , i.e. the _ void _ , still represents the absence of space , time and matter , and is what vilenkin means when he uses the term `` nothing '' . finally , it was considered in ref . @xcite that in the context of eternal inflation a mosaic structure made up of causally disconnected thermalized regions should appear with different values of the effective coupling constants , with a probability distribution for their values . however , it is not at all clear that the general state of the multiverse has to be a thermal state because , unlike the inflationary model , a common space - time is not guaranteed at all in the case of a general multiverse . in our simplified model , on the other hand , a vanishing cosmological constant seems probable , in agreement with the result of ref . however , there may be a solution to this problem , because we have shown that in terms of the creation and annihilation operators for single universes , @xmath160 and @xmath161 , a non - zero value of the vacuum energy should be ascribed to each universe , even at @xmath334 . therefore , although the most probable value for the cosmological constant would be zero in the realm of the whole multiverse , if it is taken to be formed up by pairs of universes , a non - zero value for the vacuum energy could be present in each single universe due to entanglement with its partner , which could be responsible both for the currently observed accelerating expansion and for the value of the vacuum energy in the inflationary era . in this paper , we have applied a third quantization formalism to the case of a multiverse made up of homogeneous and isotropic space - times filled with a perfect fluid . a well - defined quantum state of such a multiverse has been obtained in terms of the states of a harmonic oscillator , and it satisfies the most familiar boundary conditions which are used in quantum cosmology . it can also be given in terms of a wavefunction or a density matrix , accounting for pure and mixed states . thus , the third quantization scheme can be completely well - defined in these models of high symmetry . the state of the multiverse turns out to be given by a squeezed state with no classical analog . on the one hand , the quantum state for a gravitational vacuum which is filled with a large number of popping baby universes and wormholes connecting them , is given by a squeezed state in which quantum correlations among the number states are dominant . this result agrees with that of ref . @xcite , in which the vacuum state of gravity is populated with relic gravitons which unavoidably evolve also into a highly squeezed state having no classical analog . on the other hand , in the case of large parent universes the quantum correlations among the number states disappear asymptotically , the number state representation turns out to be an appropriate one and the single universe approximation can be then considered . thus , once a single parent universe can be assumed , quantum transitions to larger number states are asymptotically suppressed . notwithstanding , if universes are created in pairs , then the correlations in the quantum state of a pair of universes would imply that each single universe is not independent . then , one of the main non - independence consequences might be the presence of an entanglement energy , which could account for both the large value of the cosmological constant in the inflationary period and for the small value of the vacuum energy now , because the entanglement between the pair of universes decreases as each single universe expands . finally , let us notice that the model presented in this paper can be naturally extended to include local matter fields as well as cosmological fields . in the first case , it is expected that their effect would represent a minor correction to the value of the frequency of the harmonic oscillator which quantum mechanically described the multiverse . however , in the latter case , derivatives with respect to the matter fields appear in the wheeler - de witt equation , implying that the state of the multiverse is given in terms of the modes of a wave equation in which the mass and the frequency might be scale factor dependent . nevertheless , it seems that it will be easier to work with the solutions of such a wave equation rather than with the general solutions of the second quantized theory . s. w. hawking , _ wormholes and non - simply connected manifolds _ , in _ quantum cosmology and baby universes _ , vol . 7 , ed . by s. coleman , j. b. hartle , t. piran and s. weinberg , world scientific , london ( 1990 )

we use the dynamical mean field method to study a model of electrons jahn - teller coupled to localized classical oscillators and ferromagnetically coupled to `` core spins '' , which , we argue , contains the essential physics of the `` colossal magnetoresistance '' manganites . we determine the different regimes of the model and present results for the temperature and frequency dependence of the conductivity , the electron spectral function and the root mean square lattice parameter fluctuations . we compare our results to data , and give a qualitative discussion of important physics not included in the calculation . extensive use is made of results from a companion paper titled : `` on the fermi liquid to polaron crossover i : general results '' . amssym.def amssym.tex

cond-mat9602155

the doped rare earth manganites have been studied for many years @xcite and interest in the materials has revived following the recent discovery of extremely large magnetoresistance in some members of the family @xcite . the chemical formula is @xmath1 , with re a rare earth element such as la or nd and a a divalent metal ion such as sr or ca . the electronically active orbitals are believed to be the mn d - orbitals @xcite and the mean d - occupancy is @xmath2 . each mn ion feels an approximately cubic crystal field which splits the mn d - levels into a @xmath3 triplet and an @xmath4 doublet @xcite . the @xmath3 levels are believed @xcite to lie substantially ( @xmath5 5ev ) below the @xmath4 levels . on - site coulomb interactions are apparently strong enough that no d - orbital may be occupied by more than one electron . further , all electron spins in mn-@xmath6 orbitals are aligned by a large ferromagnetic hunds - rule coupling . the coulomb and hunds - rule interaction energies have not been measured directly , but there is substantial indirect evidence that they are large . for example , at @xmath7 ( precise values depend on re and a ) the ground state is ferromagnetic , and the observed magnetization is consistent with all @xmath2 electrons on each mn being lined up in the maximal spin state @xcite , suggesting a large hunds coupling . also , @xmath8 undergoes a structural phase transition at @xmath9k , which has been shown @xcite to be due to a staggered ( @xmath10 ) ordering of jahn - teller distortion of locally @xmath4 symmetry . this would not occur unless the @xmath4 orbital were singly occupied , which in turn implies that the @xmath3 orbitals are also singly occupied , suggesting a large on - site coulomb interaction . the resulting physical picture is that @xmath11 of the @xmath12 d electrons fill up the @xmath3 levels , forming an electrically inert core spin @xmath13 of magnitude @xmath14 . the remaining @xmath15 electron goes into a linear combination of the @xmath4 levels , and is coupled to @xmath13 by a hunds rule coupling @xmath16 which is presumably large , but has not been directly measured . okimoto et al . have recently presented an interpretation of optical data implying that @xmath17ev @xcite . we shall argue below that their interpretation is not correct and that @xmath18 is rather larger . certainly , the conventional @xcite wisdom is that the limit @xmath19 is appropriate , so one only need consider configurations with @xmath4 electrons parallel to core spins . the @xmath1 materials display a wide range of interesting physics . for @xmath20 ( all x - values are approximate , and depend on re and a ) the materials are insulating at all temperatures and are antiferromagnetic or ferrimagnetic at low t. for @xmath21 the low - t phase is a fully polarized ferromagnetic metal . as the temperature is increased for @xmath21 , there is a ferromagnet to paramagnet transition , which may be of first or second order , at a @xmath22k . in the paramagnetic state the material may be either `` metallic '' ( in the sense that @xmath23 and @xmath24 ) or `` insulating '' ( in the sense that @xmath25 and @xmath26 ) . ( here @xmath27 , the mott `` maximum metallic resistivity '' , is about @xmath28-cm and corresponds to a mean free path of order @xmath29 @xcite ) . insulating behavior occurs at lower x and metallic behavior at higher x. there is for all x a very pronounced drop in @xmath30 as t is lowered through @xmath31 , and in this regime the resistivity has a very strong magnetic field dependence . the `` colossal '' magnetoresistance of interest here occurs for @xmath32 such that the material is in the `` insulating '' regime at @xmath33 but is a metallic ferromagnet at @xmath34 . finally , at @xmath35 the low t state is charge - ordered , antiferromagnetic and insulating @xcite . we do not address the physics of this regime here . a qualitative phase diagram is shown in fig most models @xcite of electron transport in @xmath1 have emphasized the `` double exchange '' phenomena caused by the large hunds coupling @xmath16 . the essence of double exchange is that when an electron hops from site @xmath36 to site @xmath37 it must also go from having its spin parallel to @xmath38 to having its spin parallel to @xmath39 ; the hopping amplitude , @xmath40 , thus depends upon relative spin orientation @xcite . for two fixed sites @xmath36 and @xmath37 it is possible to choose phase factors so that @xmath41 $ ] . the double exchange phenomenon gives an obvious connection between electron hopping and magnetic order : disorder in the spins implies randomness in @xmath40 , which decreases below @xmath31 or in a field . this effect seems very likely to be related to the `` colossal '' magnetoresistance observed near @xmath31 . however , two of us and p. b. littlewood have recently argued that models involving only double exchange can not explain the observed resistivity @xcite . the essential point is that in materials exhibiting `` colossal '' magnetoresistance the resistivity at @xmath42 is much larger than the mott limit and moreover rapidly increases as t decreases . indeed , as shown in appendix a , the observed @xmath43 resistivities are so large that a classical description involving particles incoherently hopping from site to site with a hopping probability @xmath44 is appropriate . in models involving only double exchange the scattering produced by spin disorder is simply not large enough to cause such insulating behavior . a straightforward calculation @xcite shows that if the spins are completely decorrelated one finds @xmath45 , i.e. @xmath46 . more sophisticated arguments involving localization and phase factors are shown in appendix b not to change this conclusion significantly . therefore we believe some additional physics not included in the double exchange - only model must be important . this conclusion is not universally accepted @xcite one possible source of this extra physics is the `` hubbard u '' effect of the on - site coulomb interaction which produced the hunds coupling in the first place . while this is undoubtedly quantitatively important , we do not believe it is the primary cause of the observed insulating behavior , essentially because away from commensurate densities ( such as one electron per site ) canonical mott insulating materials such as the high @xmath31 superconductors or other doped transition - metal oxides have resistivities which are rather less than the mott limit and which decrease with temperature @xcite , in stark contrast to the behavior observed at @xmath47 in @xmath1 . we proposed @xcite that the crucial additional physics is a strong electron - phonon coupling , which localizes the conduction electrons as polarons at @xmath33 and smaller @xmath32 , but is weakened in the @xmath34 ferromagnetic state , restoring metallic behavior . we argued that this is possible because the behavior of electron - phonon model is controlled by a dimensionless coupling parameter which is the ratio of an interaction energy to the electron kinetic energy . the double exchange physics implies that ferromagnetic order increases the electron kinetic energy , thereby decreasing the effective coupling strength . also , a recent analysis @xcite of the structural distortion observed @xcite in @xmath48 showed that the electron - phonon coupling is indeed strong . in this paper we present a detailed study of a model of electrons coupled to core spins and to phonons which we believe confirms the importance of electron - phonon interactions . we use a `` dynamical mean field '' method which has previously been extensively applied to interacting problems without double exchange @xcite and has been used by furukawa to study models involving only double exchange @xcite . the model we study does not capture all of the physics of @xmath1 ; in particular , coulomb effects and quantum and intersite terms in the phonon hamiltonian are omitted and an oversimplified electron - phonon coupling is used . we therefore can not quantitatively compare our results to experiment . the qualitative agreement we obtain is however compelling . other workers have also studied electron - phonon effects in manganites and related materials . emin , hillery and liu studied a theoretical model of a single bound polaron coupled to spin waves and found a temperature dependence of the polaron size which they argued could be related to transport anomalies at @xmath31 observed in euo @xcite . their work , however , is based on models in which double exchange physics is not relevant , reflecting the different physics of euo . roder , zang and bishop used variational wavefunction techniques to examine the interplay between electron - phonon interaction and double exchange @xcite . their work is in a sense complementary to ours . they have incorporated quantum phonons and have presented some results on intersite phonon correlations , but their technique seems to work best at low temperatures . to calculate properties at and above @xmath31 they resort to a dilute limit approximation which amounts to the study of a single carrier in a deformable medium . also , they have not presented results for transport and optical quantities ; their main result is a calculation of the coupling dependence and doping dependence of @xmath31 . their results for the coupling dependence are very similar to ours ; their results for the doping dependence are based on an assumption which , we argue , is not justified by the arguments in their paper a comparison of their results for @xmath31 to ours is given in section iv , and a further discussion is given in the conclusion . the rest of this paper is organized as follows . in section ii we define the model and the approximations used . in section iii we give the qualitative physics of the model , distinguishing different regimes and presenting the behavior of physical quantities in each . in section iv we present the results of a detailed numerical study of the model at half - filling . in section v we present and discuss results at other dopings . section vi is a conclusion , in which the relation of results to data is analysed and the effects of omitted interactions are outlined . appendix a discusses the theoretical interpretation of the observed resistivities , in particular the relation to the mott minimum metallic conductivity and to transport by classical particles while appendix b discusses in more detail the resistivity of the double exchange - only model . an announcement of some of the results of this paper has been submitted elsewhere @xcite . this paper relies heavily on results of a companion paper which uses the dynamical mean field method to study electron - phonon interactions in models without double exchange @xcite , to which we refer henceforth as i. we study a model of mn @xmath4 d - electrons coupled to core spins @xmath38 and phonons , with hamiltonian @xmath49 @xmath50 describes the hopping of d - electrons between sites @xmath51 of a lattice . we take the electrons to have two - fold orbital degeneracy labelled by a roman index ( a , b ) and two fold spin degeneracy labelled by a greek index @xmath52 . explicitly , @xmath53 the hopping matrix element @xmath54 is a real symmetric matrix whose form depends on the choice of basis in @xmath55 space and the direction of the @xmath56 bond . the precise form will not be important in what follows . the interaction of itinerant electrons with the core spins is given by @xmath57 as discussed below , we shall take the limit @xmath58 . we assume a jahn - teller form for the electron - phonon coupling . previous analysis @xcite has shown that this coupling is strong in @xmath48 , so it may be expected to be strong also in doped compounds . thus we focus on lattice distortions which split the on - site orbital degeneracy of the @xmath4 levels . physically , these correspond to @xmath4 symmetry distortions of the oxygen octahedra around an mn site . mathematically , we may parametrize a local @xmath4 distortion by a magnitude @xmath59 and an angle @xmath60 , and define a two - component vector @xmath61 with @xmath62 and @xmath63 . the coupling of this to the @xmath4 levels is prescribed by group theory @xcite to be @xmath64 here @xmath65 is a vector of pauli matrices acting in orbital space . it is important to note that the coupling written in eq . [ eq : helph ] is not the only physically relevant one . in ref @xcite it was argued that a mn site with no @xmath4 electrons would induce a breathing distortion of the surrounding oxygen ions , and that this breathing distortion played an important role in determining the x dependence of the structural phase boundary . we have not included this coupling in the present calculations , but will qualitatively discuss its effects in the conclusion . in order to obtain a tractable model we assume @xmath66 describes classical harmonic oscillators of stiffness @xmath67 which are furthermore independent from site to site . thus , @xmath68 despite the simplifying approximations , the model defined by eq . [ eq : h ] is not solvable except in certain limits . to obtain results , we adopt the `` dynamical mean field '' approximation , which becomes formally exact in a limit in which the spatial dimensionality @xmath69 and has been shown to provide a reasonable description of models of interacting electrons in @xmath70 @xcite . recently , the technique has also been applied to the double exchange - only model defined by eq . [ eq : h ] with @xmath71 @xcite . the dynamical mean field method is based on an assumption about the electron green function @xmath72 . in general this is a tensor in spin and orbital space which may be written @xmath73^{-1 } = \omega-\epsilon_p^{ab}+mu \sigma_{\alpha\beta}^{ab } ( p,\omega ) \label{eq : g}\ ] ] here @xmath74 is the dispersion implied by eq . [ eq : hband ] and @xmath75 is the self energy due in the present problem to @xmath76 and @xmath77 . the fundamental approximation of the dynamical mean field method is the neglect of the momentum ( p ) dependence of @xmath75 . this is a reasonable approximation because models of the form of eq . [ eq : h ] ( such as the usual migdal - eliashberg electron - phonon hamiltonian ) generally lead to a self energy with a weak momentum dependence in @xmath70 @xcite . if the momentum dependence of @xmath75 may be neglected , then all physical quantities may be expressed as functionals of the momentum - integrated green function @xmath78 , given by @xmath79 we shall assume that there is no long range order in orbital space , so @xmath78 and also @xmath80 must be proportional to the unit matrix in @xmath55 space . we shall allow for the possibility of ferromagnetic order , and shall take the ordered moment parallel to @xmath81 . we may then write @xmath82 and a similar equation for @xmath75 . we simplify eq . [ eq : glocab ] by first writing @xmath83 , then introducing at each @xmath84 the rotation matrix @xmath85 which diagonalizes @xmath74 , i.e. @xmath86 r_p^{-1 } \label{eq : diag}\ ] ] and finally exploiting the cyclic invariance of the trace . we obtain @xmath87^{-1 } \\ g_1(\omega ) & = \frac{1}{4 } tr_{ab } tr_{\alpha \beta } \int d\epsilon_p { \cal d } ( \epsilon_p ) \sigma^z [ \omega - \epsilon_p + \mu- \sigma_{\alpha \beta } ( \omega ) ] ^{-1 } \label{eq : g } \end{array}\ ] ] where @xmath88 is either of the eigenvalues of @xmath74 ( they are related by symmetry operations ) and @xmath89 is the density of states , which we take to be semicircular with width @xmath90 : @xmath91 because @xmath78 is momentum independent and involves two independent functions , it must be the green function of some effective single site model involving two mean field functions @xmath92 and @xmath93 . this model is described by the partition function @xmath94 \label{eq : zloc}\ ] ] here @xmath59 and @xmath95 are the classical oscillator coordinates introduced above eq . [ eq : helph ] , @xmath96 , and the integrals are simple integrals rather than functional integrals because we have taken @xmath59,@xmath95 and @xmath97 to be classical . the effective action @xmath98 is @xmath99 - \vec{h}_{ext } \cdot s_c \vec{\omega}_c /t . \label{eq : sloc } \end{array}\ ] ] here we have added a term coupling the core spin to an external field @xmath100 . one could also couple the external field to the @xmath4 electrons , but the factor of @xmath101 in size of moment means that this coupling is unimportant . the mean field parameters @xmath92 , @xmath93 in eq . [ eq : sloc ] are determined as follows @xcite . one obtains the local green functions @xmath102 / \delta a_{0,1}$ ] , defines from these self energies @xmath103 , and demands that @xmath104 , reproduces @xmath105 when used in eqs . [ eq : g ] and [ eq : gloc ] . for the semicircular density of states the resulting equations may be written @xmath106}{\delta a_0(\omega ) } , \\ a_1(\omega ) & = - \frac{t^2}{4 } \frac{\delta ln[z_{loc}]}{\delta a_1(\omega)}. \label{eq : mf } \end{array}\ ] ] the factor of four is that appearing in eq . [ eq : g ] . these equations simplify in the `` double exchange '' limit @xmath19 . the argument of the @xmath107 @xmath108 in eq . [ eq : sloc ] is a matrix in the direct product of spin and orbital space . it has four eigenvalues , @xmath109 with @xmath110 . these are , of course , independent of the angle @xmath95 describing the phonon . for @xmath111 the eigenvalues at @xmath112 correspond to high energy states which do not affect low energy phenomena . further , from eq . [ eq : mf ] it is clear that @xmath93 is of order @xmath113 , so @xmath114 and we may approximate @xmath115 . we may then absorb the constant term @xmath18 into @xmath92 and @xmath116 , rescale the parameters and define the spin angle @xmath60 via @xmath117 obtaining @xmath118 + h_0 cos \theta /t . \label{eq : hs}\ ] ] here @xmath119 , @xmath120 , @xmath121 , @xmath122 , and @xmath123 , @xmath124 and @xmath116 are measured in units of @xmath113 . the mean field equations become @xmath125 with @xmath126 \label{eq : p}\ ] ] these equations differ from those discussed in i by the presence of the angular integral and by the quantity @xmath127 , which expresses the spin dependence of g. expressions for physical quantities are also slightly different from those used in i because we must keep track of the spin dependence . the momentum integrated green functions has components parallel @xmath128 and antiparallel @xmath129 to the magnetization . the off - diagonal @xmath130 components vanish . we have @xmath131 we shall be interested in the spectral function @xmath132 the conductivity is given by @xmath133 \label{eq : sigma}\ ] ] where the factor of two comes from the trace over orbitals . here @xmath134 is a diagonal matrix in spin space and we have set @xmath135 . another interesting quantity is the electron kinetic energy , @xmath136 , defined by @xmath137 by use of the relation between the expectation value and the electron green function , of eq . [ eq : g ] with momentum independent self energy , and of the arguments leading from eq . [ eq : glocab ] to eq . [ eq : g ] and the mean field equations , we obtain @xmath138 ^ 2 + \left [ g_{loc}^{\downarrow \downarrow } ( \omega_n ) \right]^2 . \label{eq : k}\ ] ] the magnetization @xmath139 is given by @xmath140 in these units the @xmath141 value of @xmath142 . we shall also be interested in the mean - square lattice distortion , @xmath143 , given by @xmath144 we conclude this section by mentioning numerical methods . we use the procedures described in i , and handle the additional angular integral by a twenty - point legendre formula . computations are of course more time consuming because of the extra integral involved . we found it convenient first to locate the magnetic transition temperature @xmath31 and then to perform calculations at @xmath43 using equations obtained by forcing @xmath145 . convergence difficulties arise for temperatures near @xmath31 ; these are presumably related to critical slowing down near the magnetic phase transition . we found that an accurate value for @xmath31 was most conveniently obtained by computing several values of @xmath139 in the range @xmath146 @xmath147 and finding @xmath31 by fitting to @xmath148 with @xmath149 and @xmath31 fit parameters . in previous work @xcite we had also used an alternative method ( which we termed the _ projection method _ ) based on the observation that by choice of an appropriate local spin reference frame one may map the model into one of spinless fermions moving in a lattice with a spatially varying hopping determined by the local spin orientations . we further argued that within mean field theory one could approximate this hopping by @xmath150 , thereby simplifying the problem to one of spinless fermions , with hopping @xmath151 , coupled to phonons . finally we argued that one could construct a mean - field magnetic free energy by combining the m - dependence of the free energy of the auxiliary problem with the entropic term from the conventional mean field theory for heisenberg spins . this procedure leads to a @xmath152 almost identical to that shown in fig . 2 ; however the _ projection method _ @xmath31 is lower than those shown in fig . 2 . for example , the direct - integration @xmath31 at @xmath153 is @xmath154 , much less than the @xmath155 shown in fig . 2 . a numerical error originally led us to believe the @xmath31 s of the two approaches coincided . the discrepancy may most easily be understood by expanding @xmath156 $ ] to order @xmath157 . the result is a quadratic form @xmath158 . for example , at @xmath159 , @xmath160 . @xmath31 is the temperature at which @xmath161 first has a zero eigenvalue . the prjection method result corresponds to setting @xmath162 and @xmath163 ; in other words it produces a lower @xmath31 because it does not permit an optimal choice of @xmath164 . we have therefore not used the direct integration method in this paper . we note , however , that the projection method provides a transparent and physically appealing motivation for the result , found also in the detailed calculations presented below , that the @xmath31 is controlled by the kinetic energy at @xmath31 , is borne out by our detailed solution of the model . in this section we discuss the qualitative behavior of the solutions of eqs . [ eq : mf2 ] . much of the behavior is similar to that found in i. the new feature is the physics of double exchange , which is expressed via @xmath127 , via the angular integral and via the factors of @xmath165 on the right hand side of eqs . [ eq : mf2 ] . at @xmath166 , the @xmath60 integral is dominated by the regime @xmath167 , so @xmath168 . from eq . [ eq : glocup ] one sees that at @xmath141 the antialigned component of g vanishes , while the aligned component is determined by @xmath169 which is given by an equation identical to that considered in i. therefore , all of the results obtained in i for the @xmath166 limit hold also here . at @xmath43 , there is no long range magnetic order . thus @xmath145 , there is no @xmath60-dependence and @xmath170 is given by an equation which differs by a factor of @xmath165 from that treated in i. further insight into the quantity @xmath127 may be gained from the @xmath171 limit . at @xmath141 and @xmath171 the quantity @xmath172 is found from eqs . [ eq : mf ] to be @xmath173 \label{eq : nonint1}\ ] ] this is precisely the usual non - interacting solution : @xmath174 in a semicircular band of full width @xmath175 . in the present conventions , the fermi level is at @xmath176 and for @xmath171 the maximum of the spectral function is at @xmath177 . the self - energy for this solution vanishes . at @xmath43 and @xmath171 , @xmath178 and @xmath179 \label{eq : nonint2}\ ] ] here @xmath174 in a semicircular band of full width @xmath180 : the fact that neighboring spins are uncorrelated has reduced the bandwidth , and thus the kinetic energy , by a factor of @xmath181 . this may also be seen by a direct evaluation of @xmath136 from eq . [ eq : k ] . further , the self energy is @xmath182 and has a non - zero imaginary part at the fermi surface @xmath183 , corresponding physically to scattering by spin disorder . however , this scattering is not too strong . from eqs . [ eq : sigmanonint ] , [ eq : d ] one finds that the product of the imaginary part of the self energy and the density of states at the fermi level is @xmath184 . this number is rather less 1 , and implies a mean free path longer than @xmath29 . this spin disorder scattering decreases as @xmath123 is decreased below @xmath31 . the model with @xmath171 was studied in the dynamical mean field method by furukawa @xcite , who obtained eq . [ eq : sigmanonint ] . furukawa also used a method he referred to as solving the equations at constant magnetization to produce an interpolation formula describing the temperature dependence of @xmath185 for @xmath186 . we believe these results are similar but not quite equivalent to those we obtain by solving the equations [ eq : mf2 ] directly . however , the minor differences between furukawa s results and ours are not important . the main point is that the scattering at @xmath42 predicted by this calculation is much too small to explain the data . one may calculate @xmath31 at zero coupling by linearizing the second of eqs . [ eq : mf2 ] in @xmath127 . one finds that @xmath187 is given by the solution of @xmath188 where f is the fermi function . we now return to the issue of the effects of the electron - phonon coupling . at @xmath141 the mean field equation is identical to that considered in i. from this work we learn that there are three regimes : _ weak coupling _ , in which @xmath189 , @xmath190 and @xmath191 ; _ intermediate coupling _ , where @xmath192 , @xmath193 and @xmath194 may have either sign , and _ strong coupling _ , where @xmath195 and @xmath196 . here @xmath197 is the value of frozen in lattice distortion above which a gap appears in the electron spectral function . in the strong coupling regime one may think of the electrons as being localized as polarons . another crucial result of i is that the transitions between the different regimes are controlled by the values of an effective coupling determined by the ratio of an electron - phonon energy to a kinetic energy . as we have seen , the kinetic energy is temperature dependent because of double exchange ; thus as temperature is varied the behavior of the model may change from `` metallic '' @xmath198 to insulating ( @xmath199 . as t is decreased below @xmath31 there are two effects causing a decrease in the resistivity : the spin scattering freezes out and the effective electron - phonon coupling weakens . in the section we present and discuss results of numerical calculations for the particle - hole symmetric ( @xmath200 ) case . we begin with ferromagnetic @xmath31 shown in fig . one sees that @xmath31 decreases with increasing @xmath201 ; the variation is particularly rapid in the region @xmath202 which is shown below to be the critical value at which the model goes from metal to insulator . for @xmath200 and all @xmath203 we verified that the transitions were second order by comparing the @xmath31 obtained in this manner to the @xmath31 obtained by determing the temperature at which the non - magnetic state becomes linearly unstable . we also checked for metastability at various @xmath204 and @xmath201 by starting our iterations with saturated magnetization @xmath205 and with very small @xmath127 , and verifying that both initial conditions converged to the same solution . the magnetic transition was always found to be second order . it is interesting to compare our results to those of ref . the method used by these authors to treat the magnetic fluctuations is very similar to the `` projection method '' discussed at the end of section ii . we found that the method did not give an accurate value for @xmath31 but did reproduce the coupling dependence well . ref @xcite used a model with one orbital per site ; we should therefore compare their results for @xmath206 to ours for @xmath200 . their quantity @xmath207 corresponds precisely to our @xmath201 ; the factor of two comes from the orbital degeneracy as explained in our previous paper i. as far as can be determined from fig 1 of ref @xcite , their calculated @xmath208 agrees very well with our @xmath209 . the correspondence is interesting because the calculation of ref @xcite was done with quantum phonons with the rather high frequency @xmath210 ( units not specified , but presumably set by the electron hopping , @xmath113 ) . this supports our claim that quantum effects are not important at temperatures of the order of @xmath31 . we believe the results presented in @xcite for the doping dependence of @xmath31 are not physically relevant because they are based on a model with one orbital per site , which therefore has no kinetic energy at @xmath200 . they argue that the one - orbital model is justified by the existence of the jahn - teller splitting . our results show that this is not the case . we discuss the physics of the doping dependence of @xmath31 further in the conclusion . we now turn to the temperature dependence of the resistivity , shown in fig . 3 . the curves display kinks at the ferromagnetic @xmath31 . the resistivity drops as t is decreased below @xmath31 both because the magnetic contribution to the scattering begins to decrease at @xmath31 and because the effective electron - phonon interaction becomes weaker . from these curves we may distinguish `` metallic '' ( @xmath211 and `` insulating '' @xmath212 regimes . at @xmath33 @xmath213 marks the boundary between metallic and insulating regimes ; at @xmath213 , @xmath214 . for @xmath215 the crossover occurs at the somewhat larger @xmath216 . the difference in the critical @xmath201 required to produce insulating behavior reflects the effect of spin alignment on the electron kinetic energy . we also note that although it is difficult to perceive on the logarithmic scale used in fig . 3 , for @xmath217 , @xmath218 is neither zero nor infinite . for @xmath201 sufficiently close to 1.15 , @xmath219 . similar behavior was discussed at length in i. figure 4 shows the temperature dependence of the mean square lattice distortion @xmath143 . from this one may distinguish the low - t _ weak , intermediate _ and _ strong coupling _ regimes , based on the @xmath166 limit of @xmath220 . the regimes were discussed at length in i. roughly , in _ weak coupling _ @xmath221 ; in _ intermediate coupling _ @xmath222 and in _ strong coupling _ @xmath223 . in intermediate coupling there is a frozen - in lattice distortion which affects the @xmath141 physics but is not large enough to open a gap ; for strong coupling the distortion is large enough to open a gap and cause insulating behavior . for @xmath224 , @xmath225 even though if @xmath226 , @xmath227 is finite . the effects of double exchange may be seen in fig . @xmath31 is visible as a kink on each curve . at @xmath43 , @xmath228 decreases , implying a stronger electron - phonon coupling . for @xmath229 the @xmath141 values obtained by extrapolating the @xmath43 curves to 0 are non - zero , and are higher than the actual @xmath141 values , because the reduction of kinetic energy due to spin disorder has effectively made the electron - phonon coupling stronger . note that the @xmath141 extrapolation of @xmath43 portion of the curve corresponding to @xmath230 is about @xmath231 . this is slightly less than the critical value @xmath232 found in i to mark the boundary between finite and infinite @xmath233 at @xmath200 . from this we would infer at @xmath43 , @xmath234 changes sign at @xmath235 , as indeed is seen in fig . 3 . the curves presented in fig . 4 show @xmath236 in arbitrary units . to estimate the magnitude of the effect in @xmath1 we note that in @xmath48 each o - ion is displaced @xmath237 from its ideal perovskite position @xcite . the estimates obtained in @xcite imply @xmath238 in that material ; thus @xmath239 in fig . 4 corresponds to an rms displacement of an o - ion of about @xmath240 . we next consider the temperature dependence of the kinetic energy shown in fig . 5 . at @xmath171 the kinetic energy changes by about @xmath241% between @xmath141 and @xmath242 , and has a weak t dependence at @xmath43 . for @xmath243 the kinetic energy changes between 0 and @xmath31 by a somewhat larger amount ; for @xmath244 , by still larger amount , for @xmath245 , yet larger . these changes come from the previously discussed interplay between double exchange and electron - phonon coupling . as t is increased from zero , the spins disorder . this reduces the electron kinetic energy and permits the electron - phonon coupling to further localize the electrons , reducing their coupling yet more , etc . we also note that we found the ratio between @xmath31 and the kinetic energy at @xmath31 to be the same within a few percent for all @xmath204 and @xmath201 studied . for @xmath200 this can be seen by comparing figs 2 and 5 . these arguments also explain the magnetic field dependence of the resistivity . increasing the field aligns the spins , increases the kinetic energy , and decreases the effective electron - phonon coupling , leading to a large change in resistance as shown in figs 6a and 6b . when the decrease in effective electron - phonon coupling tunes the model across the `` metal''-``insulator '' transition , as in fig . 6b , the magnetoresistance is particularly large . further insight into the interplay of double exchange and localization comes from the optical conductivity shown in fig . panel 7a shows @xmath246 at different t for the weak coupling @xmath247 . at low t @xmath248 has approximately the drude form @xmath249 with scattering rate @xmath250 as expected from classical phonons . as t is increased through @xmath31 the drude peak broadens and acquires a t - independent part , due to spin scattering . panel 7b shows @xmath248 for the moderate coupling @xmath213 . at low t , @xmath248 has the drude form ; as t is increased a broad peak develops ; this is due to transitions between the two jahn - teller split levels . it is broad because the phonon coordinate is strongly fluctuating , so the level position is not well defined . as t increases beyond @xmath31 the peak broadens almost to indistinguishability . note also that as t is increased , the optical spectral weight @xmath251 decreases , reflecting the increasing localization of electrons by phonons . in models such as the present one which do not have galilean invariance and involve only a limited number of orbitals , the f - sum rule spectral weight is not constant and is indeed proportional to the kinetic energy @xcite . panel 7c shows that at a stronger coupling @xmath248 does not have the drude form , and the peak is already evident at @xmath252 . note that the maximum in @xmath248 has moved to a slightly higher frequency . recently published data of okimoto et al . on @xmath253 @xcite are similar to the curves shown in fig . 7c , although our use of classical phonons means that we can not obtain the very narrow drude peak found at low t. panel 7d shows @xmath248 at the still stronger coupling @xmath254 where the model has a large frozen - in lattice distortion even at @xmath141 . the @xmath248 has an insulating appearance , above and below @xmath31 , but as t is decreased the peak in @xmath248 shifts to a lower freqency and grows in intensity , reflecting the effectively weaker coupling . the nonmonotonic behavior of the dc conductivity is not reflected in @xmath255 at @xmath256 . the curves in fig 7d resemble data recently obtained by kaplan et . al . on @xmath257 @xcite . we will discuss the physical interpretation of @xmath258 in more detail in the next section and in the conclusion . here we note that in the strong - coupling regime the two d - states on a site are split . the peak in the optical conductivity corresponds roughly to a transition in which an electron moves from an occupied orbital on one site to an unoccupied orbital on an adjacent site . in our classical approximation , `` franck - condon '' transitions involving also emission or absorption of a phonon can not occur at all . in a more realistic model such effects would e.g. increase the low - frequency tails by a small amount . the width of the peak in @xmath259 is determined by the broadening of the localized states due to electron hopping and by thermal broadening , which leads to a range of lattice distortions and thus to a range of splittings . in this section we present and discuss results of numerical calculations for the particle - hole asymmetric case @xmath260 . as discussed in i ( see especially figs . 10 , 11 ) , at @xmath260 in the strong coupling limit the spectral function has a three - peaked structure . the outer two peaks represent the jahn - teller - split @xmath4 levels on occupied sites , and occur also for @xmath200 . the middle peak comes from unoccupied sites , on which there is no jahn - teller splitting . these states tend to fill in the gap created by the jahn - teller splitting and mean that stronger coupling is required to obtain insulating behavior at @xmath260 than at @xmath200 . further , in the strong coupling limit the temperature dependence of physical quantities is determined by the energy difference between filled and mid - gap states ; thus at fixed jahn - teller splitting the activation gap for physical properties is much less at @xmath260 than at @xmath200 . note also that @xmath31 is controlled by the electron kinetic energy which is in turn controlled by the jahn - teller splitting . therefore , in the strong coupling limit at fixed @xmath31 the activation gap characterizing the @xmath42 resistivity is much larger at @xmath200 than at @xmath260 . this physics is immediately apparent in the resistivity curves for @xmath261 and @xmath262 shown in fig . comparison to fig . 3 shows that much stronger couplings are required to obtain a @xmath25 for @xmath261 than for @xmath200 and stronger couplings yet are required for @xmath262 . the smaller value of the activation gap relative to @xmath31 means that the resistivity rises less before the behavior changes at @xmath31 for @xmath260 than for @xmath263 . for @xmath200 , @xmath264 we found an order - of - magnitude rise in @xmath30 as @xmath123 is decreased to @xmath31 , and we found metallic behavior below @xmath31 . at @xmath260 it is difficult to produce much of an up - turn in @xmath30 at @xmath265 for parameters such that the model is metallic at @xmath141 . this physics is due to the particular ( jahn - teller ) form of the electron - phonon coupling we have chosen to study . the same arguments mean that it is not possible to get as large a magnetoresistance at @xmath260 as at @xmath200 . figure 9 shows the temperature dependence of the resistance for several different coupling strengths and magnetic fields . to get even a moderately large effect one must choose a very strong coupling , such that the model is insulating for both @xmath42 and @xmath34 . finally , fig . 10 displays the temperature dependence of the optical conductivity at @xmath261 and moderate ( @xmath266 ) and strong ( @xmath267 ) coupling , and compares this to the momentum - integrated spectral functions . one sees by comparing energies that at strong coupling the two maxima in the conductivity may be associated with transitions from the lowest peak in the spectral function ( representing occupied orbitals on occupied sites ) to the middle feature ( representing unoccupied orbitals on unoccupied sites ) and to the higher feature ( representing unoccupied orbitals on occupied sites ) . at @xmath200 the middle peak in @xmath268 is absent and @xmath248 has only one peak , as seen in fig . 7d . of course , the on - site d - d transition is not optically active : the calculated conductivity involves electron motion from one site to another , and for this reason @xmath248 is not simply given by a convolution of two local spectral functions . one may see this in fig . the central peak in @xmath269 has less area than the upper one , yet the lower peak in the corresponding @xmath259 is the larger . this may be understood from the above arguments : a transition from the lower to the middle peak of @xmath270 necessarily involves moving an electron from one site to the other , but some of the transitions from lower to higher are on - site transitions which do not contribute to @xmath248 . note also that the @xmath271 spectral function shown in figs 10b , d has a sharp minimum at @xmath272 . this is a consequence of the fact , discussed in i , that the probability of a small - amplitude lattice distortion is small because of the @xmath273 measure , and decreases as @xmath123 increases , due to the shift to higher @xmath274 of @xmath275 . as can be seen in the corresponding optical conductivity curves in figs 10a , c , this minimum is of little significance for other physical quantities . we have used the `` dynamical mean field '' approximation to solve a model of electrons ferromagnetically coupled to classical spins and jahn - teller coupled to localized classical oscillators . in a companion paper ( i ) we considered electron - phonon coupling in a variety of models without double exchange . the results presented in section iv for the half - filled case bear a striking resemblance to data for the `` colossal magnetoresistance '' materials @xmath1 in the @xmath276 regime where the ground state is metallic . we believe the agreement supports the idea that the important physics of @xmath1 involves the interplay between a strong electron - phonon coupling and the `` double exchange '' effect of magnetic order on the electronic kinetic energy . specifically , the @xmath277 curves shown in fig . 3 are very similar to those shown , e.g in @xcite . varying the electron - phonon coupling produces changes very similar to those found experimentally by varying @xmath32 and the constituents @xmath278 and @xmath268 . the magnetic field dependences shown in fig 6 also bear a striking resemblance to data . fig 6b looks very like fig . 2 of @xcite , while fig 6a resembles magnetoresistance data which would be observed for @xmath279 . ( it should be noted , however , that the fields used to produce our curve , although very small compared to microscopic energies , are larger than experimental fields by a factor of about 5 ) . the variation of the rms lattice distortion shown in fig 4 has been observed via measurements of the @xmath4 component of the oxygen debye - waller factor @xcite . further , optical conductivity data of okimoto et . @xcite on @xmath253 bear a strong qualitative resemblance to fig 7b , while data obtained by kaplan et . @xcite strongly resemble fig 7c . as noted in sections iv and v , in our interpretation the higher frequency peak in @xmath280 is due to transitions between levels split by an electron - phonon coupling . okimoto et . al . interpreted the higher peak differently , attributing it to transitions from an inital @xmath4 state aligned to the core spin to a final @xmath4 state antialigned to the core spin . they argued that their identification was supported by the fact that in their data the higher peak was only visible at @xmath33 , and vanished at low t when all spins were aligned . however , the data of kaplan et . al . demonstrate that in some samples the high frequency peak does not vanish below @xmath31 and indeed grows in oscillator strength as t decreases . this rules out the interpretation of okimoto et . al . , at least for that sample . the detailed qualitative agreement between data and our model leaves little doubt that we have identified the important physics governing the @xmath1 materials . however , several very important issues remain unresolved . one concerns the origin of the experimentally observed material and doping dependence of the results , which are _ modelled _ in the @xmath200 calculations by varying the electron - phonon coupling . another is the degree of `` fine - tuning '' of parameters required . a third concerns the effects of omitted interactions and a fourth is that , as shown in section v , computations at different electron concentrations @xmath260 agree much less well with data . in the remainder of this paper we present a qualitative discussion of all of these issues , which , we argue , are closely related . we begin with the @xmath260 calculations . we showed that the differences between the @xmath200 and @xmath281 results are due to the presence , for @xmath260 , of mid - gap states in the spectral function ( shown , e.g. , in fig . 10 ) . these mid - gap states occur because we used a particular form of electron - phonon coupling , namely a jahn - teller coupling which splits the d - state degeneracy on a site if there is one electron on the site , and does nothing otherwise . in ref @xcite it is argued that the existence of the jahn - teller coupling justifies a model involving only one orbital per site . the results presented here suggest that this is oversimplified , because it does not take in to account the mid - gap states . a model with only jahn - teller coupling does not suffice . however , results presented in i strongly suggest that if the model were extended in a way which moved both the upper peak and the mid - gap states up in energy , then the model would become effectively a single - orbital model and results for @xmath260 would much more closely resemble those obtained for @xmath200 . one omitted piece of physics which will have precisely this effect is the breathing - mode distortion of the oxygen octahedron around an mn site . the breathing mode couples to charge fluctuations on the mn site . this coupling is likely to be at least as strong as the jahn - teller coupling , as may be seen from the following argument : the jahn - teller coupling is due to the dependence of the force exerted on an o ion on the orbital occupied by the outer - shell d - electron . whatever its magnitude , this force is unlikely to be larger than the force created by simply removing that d - electron , and making an unbalanced charge . the breathing - mode coupling was recently argued to be important for the small - x structural phase boundary @xcite . to understand the effects of the breathing mode , consider again fig . the central peak in the spectral function depicted in the low - t curve in fig 10d gives the states available for adding an electron to an unoccupied site . if such a site has a breathing distortion already present , the energy cost of adding an electron will be increased , thus the middle feature will also move up in energy , increasing the gap as required . another important piece of physics is the on - site coulomb interaction . this must be strong because if it were not , the hund s coupling @xmath16 would not be large @xcite . the coulomb interaction leads to two related effects . one is most easily discussed by reference to the spectral functions and optical conductivities shown in fig . 10 . now the @xmath282 part of the spectral function corresponds to states into which an electron maybe added ; the upper peak thus gives the states available for adding an electron onto a site which already has an electron . the coulomb interaction must move such states up in energy , and must similarly move up the second peak in @xmath246 . if the coulomb energy is of the order of the hunds coupling , then it is very likely that this effect will move the higher peak out beyond the physically interesting energy range @xmath283 ev . the combined effect of the breathing distortion and the coulomb interaction is therefore to lead to a spectral function with at most two peaks in the energy range of interest . the only difference between this realistic situation and the situation encountered in the @xmath200 calculations is that the realistic spectral function is not symmetric under the interchange of the two peaks . this asymmetry was shown in i not to be important . a second effect of a strong coulomb interaction is to localize the electrons . it is likely that the observed very strongly insulating behavior of @xmath8 is not due solely to the jahn - teller order , and that @xmath8 is to some degree a mott insulator . now the kinetic energy @xmath136 of a mott insulator has a pronounced doping dependence @xcite . for @xmath1 one would expect @xmath284 to increase with @xmath32 for @xmath285 . because , as we have argued at length , the properties of electron - phonon models are controlled by the ratio of a coupling energy and a kinetic energy , this will lead to an x - dependence of the effective coupling strength , with larger @xmath32 having a weaker effective coupling . we believe that this strong x - dependence of the effective coupling accounts for the ubiquity of the `` colossal '' magnetoresistance phenomenon . different materials have different bare electron hoppings and probably different electron - phonon couplings , but in all materials the variation of the electron kinetic energy with @xmath32 is large enough to sweep the effective coupling through the critical value at some x between @xmath286 and @xmath287 . the breathing distortion may be studied via the dynamical mean field formalism used here ; one must simply integrate over another variable in eq [ eq : zloc ] . the on - site coulomb interaction may also be included , but one must perform functional integrals rather than simple integrals . monte carlo techniques are required , the computational expense is greater and the accuracy is less . such an investigation would however be desirable . two other effects not included in the calculation should be mentioned . quantum fluctuations of the phonons have been omitted . as discussed in i , these will in the absence of long range order or commensurate density lead to metallic behavior at sufficiently low t , even in the strong coupling limit . the neglect of the phonon momentum and quantum fluctuations of core spins is not an important approximation because we are primarily interested in phenomena at temperatures of order room temperature ; however if needed they could be incorporated into the formalism . the neglect of intersite phonon correlations is potentially more serious . it is tempting to argue that they are unimportant because we are interested in optical phonons , which are usually weakly dispersing . however , in the @xmath8 structure each o is shared by two mn ; there must thus be a strong correlation between jahn - teller distortions on adjacent sites . in @xmath48 the jahn - teller distortions have long range order ; estimates presented in @xcite suggest that in @xmath1 the correlation length of the jahn - teller distortions is @xmath288 as long as the resistivity is well above the mott limit . extending the present calculations to include the effects of intersite correlations is an important open problem . it is worth addressing because in the present calculations the correlation length is zero and the strong - coupling physics is of polarons . in the infinite correlation length limit , the physics has to do with interband transitions in a bandstructure defined by jahn - teller order . the situation in the actual materials is presumably intermediate between these two limits . _ acknowledgements _ we thank a. sengupta and c. m. varma for helpful discussions and g. a. thomas and h. d. drew for sharing data in advance of publication and for helpful discussions . we are especially grateful to p. b. littlewood for stimulating our interest in the problem , collaborating in the early stages of our work , and providing continuing advice and encouragement . rm was supported in part by the studienstiftung des deutschen volkes . a. j. m. acknowledges the hospitality of the institute giamarchi- garnier during the early stages of this work , and of the institute for theoretical physics during the final stages of preparation of the manuscript . in this appendix we discuss the observed @xmath33 resistivities of @xmath1 . we note that the observed strong @xmath32-dependence suggests that the number of active carriers is @xmath32 . for @xmath32 classical particles hopping with _ probability _ @xmath289 on a cubic lattice of lattice constant @xmath290 , @xmath291 using @xmath292 as appropriate to @xmath1 we have @xmath293 > from this equation one may easily see that observed resistivities , which are typically greater than 0.01 @xmath294 at @xmath43 and @xmath295 , and increase rapidly with decreasing @xmath32 imply values of @xmath296 much less then unity . if @xmath297 , a particle has time to thermalize before it moves , and a classical model is appropriate . in this appendix we consider in more detail the resistivity of the double exchange only model . in this model , resistivity comes from spin disorder . it is maximal at @xmath298 and vanishes at @xmath141 . the resistivity has been calculated , using methods which are essentially perturbative in the amplitude of the spin disorder , by kubo and ohata @xcite and more recently by two of us and littlewood @xcite . the spin scattering was found not to be too strong . as discussed in the text , similar results have been obtained using the dynamical mean field method . these calculations omit several physical effects and have been questioned recently by varma @xcite . in this appendix we show that the omitted effects are not important . we begin by describing the omitted effects . in the double exchange only model , the scattering is due to spin disorder which , if the core spins are assumed to be classical , may be treated as static scattering with the important proviso that the disorder is annealed , not quenched . when applied to a model with static scattering , the dynamical mean field approximation with semicircular density of states is equivalent to the coherent potential approximation ( cpa ) for the bethe lattice @xcite . the cpa neglects localization ( as do the perturbative calculations @xcite ) . the lack of closed loops on the bethe lattice also means that berry phase effects arising from particle motion in a spin background are omitted . we consider the berry phase effects first . in the double exchange model the hopping matrix element between two sites @xmath36 and @xmath37 is with core spins characterized by polar angles @xmath299 , @xmath300 , is @xmath301 if closed loops are not important one may choose the @xmath302 independently on each site and recover the familiar double exchange result @xmath303 . in general the @xmath95 factors around a closed loop produce something like a magnetic field , which may scatter electrons . in the limit of strong ferromagnetic correlations the phase dependent term may be seen to be very small because all nearby sites have very similar angles , which may be taken to be near 0 . in the limit of uncorrelated spins we may estimate the size of the effective field by comparing the phase sensitive part of the hopping to the phase insensitive @xmath304 part . by integrating @xmath40 , around square placquette one finds that the phase sensitive part is @xmath305 while the phase insensitive part is @xmath306 . thus the rms deviation of the amplitude for an electron to move around a placquette is @xmath307 of the phase insensitive part . this combined with the relative insensitivity of three dimensional physics to closed loops suggest phase effects , while interesting , are too weak to cause the observed strongly insulating behavior . we now turn to localization . the problem at hand concerns electrons with random hopping , which has not received much attention . economou and antoniou @xcite have studied a bethe - lattice model in which the hopping amplitude @xmath113 has the symmetrical distribution @xmath308 for this model @xmath309 , the mean value of t , equals @xmath310 and the variance @xmath311 . the double exchange model at @xmath312 ( so the spins are completely disordered ) corresponds to the distribution @xmath313 the localization effects of the double exchange distribution have not been determined . we expect that because the most probable value is also the largest hopping , the double exchange distribution will produce fewer localized states than a semicircular distribution with the same mean and variance . now from eq [ eq : p_d - ex ] one sees that the double exchange distribution has mean @xmath314 and variance of @xmath315 . thus it should produce fewer localized states than the model of economou and antoniou with @xmath316 and @xmath317 , i.e. with @xmath318 . inspection of ref @xcite reveals that at this ratio of @xmath319 , a negligible fraction of the states are localized . we therefore conclude that localization effects are not important . ref @xcite on the contrary asserts that the double exchange model with completely disordered spins is better modelled by the economou - antoniou distribution with @xmath319 somewhat larger than unity , so a non - negligible fraction of the states are localized . drawing precise conclusions is somewhat difficult because one result of ref @xcite is that the number of localized states increases rapidly for @xmath320 . nevertheless , we believe the estimate @xmath321 obtained above shows that localization effects are unlikley to be important . e. d. wollan and w. c. koehler , phys . rev . * 100 * , 545 ( 1955 ) j. goodenough , phys 100 * , 564 ( 1955 ) . s. jin , t. h. tiefel , m. mccormack , r. a. fastnacht , r. ramesh and l. h. chen , science * 264 * , 413 ( 1994 ) . g. f. koster , a. dimmock , r. g. wheeler and h. statz , _ representations of the 32 point groups_,(m.i.t . press : cambrdige , ma ) 1963 . l. f. mattheiss , unpublished . j. kanamori , j. appl . phys . suppl . to vol.*31 * , 14s ( 1961 ) . j. b. a. a. ellemans , b. vanlaar , k. j. r. vanderveer and b. o. loopstra , j. sol . chem * 3 * , 238 ( 1971 ) . y. okimoto , t. katsufuji , t. ishikawa , a. urushibara , t. arima and y. tokura , phys . lett * 75 * 109 ( 1995 ) . see , e.g. n. f. mott , p. 3 in _ disordered semiconductors _ , m. a. kastner , g. a. thomas and s. r. ovshinsky , eds . ( plenum , new york : 1987 ) . c. t. chen , et al , unpublished . p. w. anderson and h. hasegawa , phys . rev . * 100 * , 675 ( 1955 ) . g. degennes , phys . rev . * 118 * , 141 ( 1960 ) . r. searle and c. wang , j. phys . * 48 * , 2023 ( 1970 ) . k. kubo and a. ohata , j. phys . soc jpn * 33 * , 21 ( 1972 ) . n. furukawa , j. phys . jpn , * 63 * , 3214 ( 1994 ) . a. j. millis , p. b. littlewood and b. i. shraiman , phys . lett . * 74 * , 5144 ( 1995 ) . c. m. varma , unpublished . for an example of the resistivity of @xmath322 see e.g. h. y. hwang , et . al . , phys . * 72 * , 2636 1994 . for examples of resistivities of electronically three dimensional transition metal oxides see e.g. y. tokura , y. taguchi , y. okada , y. fujishima , t. arima , k. kumagai and y. iye , phys . lett . * 70 * 2126 ( 1993 ) . a. j. millis , phys . b , in press , 1996 . for a review , see a. georges et . al . , rev . mod . phys . in press ( 1996 ) . d. emin , m. hillery and n - l . h. liu , phys . _ b33 _ 2933 ( 1986 ) . h. roder , j. zang and a. r. bishop , phys . rev . lett _ 76 _ 1356 ( 1996 ) . a. j. millis , b. i. shraiman and r. mueller , unpublished ( cond - mat/9507084 ) . a. j. millis , r. mueller and b. i. shraiman , unpublished . a. abrikosov , l. p. gorkov and i. e. dzyaloshnski , _ methods of quantum field theory in statistical physics _ , ( dover , new york ) , 1993 ) . see , e.g. a. j. millis and s. n. coppersmith , phys . b42 10807 ( 1990 ) . s. g. kaplan , u. m. quijade , h. d. drew , g. c. xiong , r. ramesh , c. meuon , t. venkatesan and d. b. tanner , unpublished . y. tokura , a. urushibara , y. moritomo , t. arima , a. asamitsu , g. kido and n. furukawa , j. phys . _ 63 _ , 3931 ( 1994 ) . h. y. hwang , s .- w . cheong , p. g. radaelli , m. marezio and b. batlogg , phys . * 75 * 914 ( 1995 ) and unpublished . p. dai , j zhang , h. a. mook , s - h liou , p. a. dowben and e. w. plummer , unpublished . p.g .. radaelli , m. marezio , h. y. hwang , s - w cheong and b. batlogg , unpublished . p. schiffer , a. ramirez , w. bao and s - w cheong , phys . * 75 * 3336 ( 1995 ) . e. n. economou and p. d. antoniou , sol . * 21 * , 285 ( 1977 ) . * qualitative temperature ( t ) - doping ( x ) phase digram of @xmath1 , with magnetic phases ( @xmath323ferromagnet , af@xmath324antiferromagnet , p@xmath324paramagnet ) , structural phases ( @xmath325jahn - teller order , no label@xmath324no order ) , and transport regimes ( @xmath326 `` metal '' , @xmath327 , @xmath328``insulator '' , @xmath329 indicated . the solid lines are magnetic phase boundaries , the heavy dashed line is the jahn - teller boundary and the light dotted line is the metal insulator crossover . for @xmath330 different physics , involving charge ordering , is important at low t. different materials may have phase diagrams differing in some details , and the magnetic and structural boundaries may not coincide at low @xmath123 . * dependence of ferromagnetic @xmath31 on coupling constant for @xmath200 ( heavy solid line ) , @xmath261 ( light solid line ) , @xmath262 ( light dashed line ) . the analytic zero - coupling results are indicated by dots ; the analytic strong coupling @xmath331 results by the heavy dotted line for @xmath200 . only for @xmath200 do the numerical calculations extend into the strong coupling regime . * temperature dependence of resistivity at @xmath200 for couplings @xmath332 ( lowest curve ) , 0.71 , 1 , 1.08 , 1.12 , 1.15 , 1.20 ( highest curve ) . * temperature - dependence of mean - square lattice distortion for @xmath200 and couplings @xmath243 ( lowest ) , 0.9 , 1.05 , 1.12 , 1.2 ( highest ) . * temperature ( t ) dependence of electron kinetic energy ( k ) for @xmath200 and @xmath171 ( second lowest curve ) , 0.71 , 1.12 and 1.29 ( highest curve ) . the lowest curve corresponds to @xmath171 in the model without double exchange . * temperature dependence of resistivity at different values of magnetic field , @xmath333 , for @xmath334 ( figure 6a ) and @xmath335 ( figure 6b ) . the parameter @xmath333 is related to the physical field @xmath336 by @xmath337 . using @xmath338 , @xmath339 ev and @xmath14 means @xmath340 corresponds to @xmath341 . * optical conductivity , @xmath200 , @xmath342 ( light solid line ) , @xmath343 ( light dashed line ) , @xmath344 ( light dotted line ) , @xmath242 ( heavy solid line ) , @xmath345 ( heavy dashed line ) . panel a : @xmath243 ( @xmath346 , panel b : @xmath213 ( @xmath347 ) , panel c : @xmath348 ( @xmath349 ) , panel d : @xmath254 ( @xmath350 . note that in panel a the lowest t is .025 not .02 , and @xmath351 for this curve is @xmath352 . * resistivity ( @xmath30 ) vs temperture ( t ) for @xmath353 ( upper panel ) and @xmath262 ( lower panel ) and couplings @xmath243 ( lowest ) , 1.12 , 1.41 , 1.49 , 1.58 ( highest ) . * magnetic field dependence of resistivity for @xmath261 and @xmath335 ( lower panel ) , @xmath354 ( middle panel ) and @xmath267 ( upper panel ) . * optical conductivities and spectral functions for @xmath261 , @xmath245 ( panels a , b ) and @xmath267 ( panels c , d ) . panel a : @xmath246 , @xmath245 and @xmath355 ( light solid line ) , @xmath356 ( light dashed line ) , @xmath357 ( light dotted line ) and @xmath358 ( heavy solid line ) . panel b : spectral function , @xmath245 and @xmath359 ( solid line ) and @xmath360 ( dashed line ) . panel c : @xmath259 , @xmath267 and @xmath342 ( light solid line ) , @xmath361 ( light dashed line ) @xmath362 ( light dotted line ) , and @xmath363 ( heavy solid line ) and @xmath364 ( heavy dashed line ) . panel d : spectral function , @xmath267 , @xmath342 ( solid line ) and @xmath364 ( dashed line ) .

we have studiedhe films adsorbed in two porous glasses , aerogel and vycor , using high precision torsional oscillator and dc calorimetry techniques . our investigation focused on the onset of superfluidity at low temperatures as thehe coverage is increased . torsional oscillator measurements of thehe - aerogel system were used to determine the superfluid density of films with transition temperatures as low as 20 mk . heat capacity measurements of thehe - vycor system probed the excitation spectrum of both non - superfluid and superfluid films for temperatures down to 10 mk . both sets of measurements suggest that the critical coverage for the onset of superfluidity corresponds to a mobility edge in the chemical potential , so that the onset transition is the bosonic analog of a superconductor - insulator transition . the superfluid density measurements , however , are not in agreement with the scaling theory of an onset transition from a gapless , bose glass phase to a superfluid . the heat capacity measurements show that the non - superfluid phase is better characterized as an insulator with a gap .

cond-mat9612133

for a wide variety of substrates , adsorbed films of @xmath0he thicker than approximately two atomic layers are superfluid at zero temperature . as the temperature is increased , a transition occurs to a non - superfluid phase at a critical temperature @xmath1 . although the details of this transition depend on the topology of the substrate,@xcite the phase diagrams for films adsorbed on substrates such as mylar , porous glasses , or packed powders are similar . in each case , there is a line of transition temperatures in the density temperature plane separating non - superfluid from superfluid films . at @xmath2 , this line terminates at a _ critical coverage _ @xmath3 , below which superfluidity does not occur . only cesium substrates , which are not wetted by @xmath0he,@xcite and the atomically ordered substrates graphite and molecular h@xmath4,@xcite lead to exceptions to this general picture . although the existence of a non - zero critical coverage @xmath3 has been known since the earliest experiments on unsaturated @xmath0he films , the onset of superfluidity as a function of coverage at @xmath2 has received much less experimental attention than the superfluid transition at @xmath5 . this has followed in part from the difficulty in making thermodynamic measurements on extremely thin @xmath0he films , but also from a conviction that the non - superfluid film of coverage @xmath3 forms an essentially inert pseudo - substrate for the overlying superfluid film . most aspects of thin - film superfluidity can indeed be understood without any consideration of the critical coverage @xmath3 , and the @xmath0he coverage is often renormalized by taking the effective density of the film to be @xmath6 . for example , the superfluid density of films with transition temperatures above 200 mk is observed to be proportional to @xmath6 for many substrates.@xcite in contrast to the inert - layer model used for the interpretation of most experiments , there has been considerable theoretical work suggesting that the onset of superfluidity at @xmath2 in @xmath0he films is analagous to a metal - insulator transition in fermi systems . this approach follows from a suggestion of hertz , fleishman , and anderson@xcite and has been pursued in depth by fisher _ et al._,@xcite who exploited the existence of a natural order parameter in bose systems to develop a scaling theory for the onset of superfluidity in strongly disordered systems . the spirit of this approach is that the onset transition is driven by the competition between exchange , which favors superfluidity , and the _ combined _ effects of disorder and the repulsive he he interactions , which favor localization . the localized phase in the case of a @xmath0he film adsorbed on a strongly disordered substrate is presumed to be a `` bose glass , '' with a gapless excitation spectrum and a correspondingly nonzero compressibility . this paper presents superfluid density measurements of @xmath0he films adsorbed in aerogel glass and heat capacity measurements of @xmath0he films adsorbed in vycor for coverages close to the critical coverage @xmath3 and temperatures down to 10 mk . our measurements demonstrate that the picture of a superfluid - insulator transition developed by fisher _ et al . _ and others@xcite is qualitatively appropriate . the heat capacity measurements , for example , indicate that a non - superfluid film is far from inert . in fact , the density of states of the non - superfluid film has a characteristic energy which vanishes as the onset transition is approached from _ below _ in coverage . we do not , however , find good quantitative agreement with the scaling predictions of fisher _ et al . _ for the temperature dependence of the heat capacity and the coverage dependence of the superfluid density and the superfluid transition temperature . we find that the non - superfluid film is best characterized as an insulator with a gap as opposed to a bose glass . the scaling behavior of thin superfluid films is complicated by both the large energy scale characteristic of physical adsorption and the existence of the thermodynamic superfluid transition at non - zero temperatures . the thermodynamic transition appears to be the dominant critical point , at least over the temperature and coverage range of our study . in fig . [ fig : generic ] we show a generic phase diagram for an unsaturated @xmath0he - film in the temperature - density plane . the film is superfluid at @xmath2 if the coverage exceeds a critical coverage @xmath3 , which is typically 1.5 to 2 atomic layers . ( all of the substrates discussed in this paper are disordered and do not support layer - by - layer growth . an `` atomic layer '' thus corresponds to an average coverage of one layer . ) a phase boundary , corresponding to the superfluid transition temperature @xmath7 , extends upward from the point ( @xmath8 , @xmath9 ) . the film is non - superfluid above this line . the phase diagram of fig . 1 is representative of the aerogel and vycor systems discussed in this paper as well as a variety of other disordered substrates . this paper is devoted to three questions regarding this general phase diagram . first , how can we better understand the onset transition at @xmath10 : does it have scaling properties like conventional second - order phase transitions ? second , what is the most appropriate description of non - superfluid films ? finally , what is our understanding of the thinnest superfluid films , within a few percent of the critical coverage @xmath3 ? one view of the onset transition is that the @xmath0he film is solid at densities below @xmath3 and thus can not support superfluidity . this viewpoint is equivalent to drawing a vertical line in fig . [ fig : generic ] at @xmath10 and therefore makes a distinction between the `` solid '' part of the film and the `` liquid '' overlayer , which is superfluid at low temperature . although this approach , often called the `` inert - layer model , '' is sufficient for interpreting most superflow measurements,@xcite it eschews particle statistics by ignoring exchange between the solid layer and the overlying superfluid film . in its most orthodox form , the inert - layer model predicts that the various thermodynamic properties of the superfluid should be extensive with respect to @xmath6 , where @xmath11 is the total density of the film . in practice , this does not appear to be the case . for example , the superfluid density @xmath12 is generally not linear in @xmath6 for coverages very close to the onset of superfluidity . this observation has to led to some variations on the inert - layer model , including the suggestion that the compression of the film as it becomes thicker leads to an increase in the density of the inert layer.@xcite a more rigorous approach is difficult , and , although outlined almost twenty years ago,@xcite has been pursued in detail only in the last several years . it is quite natural to draw an analogy between the metal - insulator transition in fermi systems and the onset transition in a @xmath0he film , which is a bose system . in both cases , one can argue that the onset of metallic ( or superfluid ) behavior corresponds to the appearance of extended states spanning the system above a critical density @xmath3 . unfortunately , the theoretical machinery developed for fermi systems is not readily adaptable to the bose case because of the strong repulsive interactions . there is no meaningful non - interacting limit , such as pure anderson localization , for a disordered bose system . it is possible , however , to consider a pure but strongly interacting system and then gradually turn on the disorder . the simplest representation of the interactions is a hard - core repulsion @xmath13 that prevents two atoms from occupying the same site . the effects of quantum exchange are introduced via a hopping matrix element @xmath14 between different sites . this _ bose - hubbard model _ has been studied for one - dimensional ( 1d ) and two - dimensional ( 2d ) lattice systems.@xcite in both cases , the system is `` superfluid '' ( meaning a many - body extended state exists ) for any density as long as the exchange energy @xmath14 is greater than a critical value @xmath15 . for weak but non - zero exchange , the system will be superfluid as long as there are sites free to accomodate hopping . this condition fails only at commensurate densities , at which exactly @xmath11 atoms occupy each site . in this case , the system is localized by the on - site repulsion and is entirely analogous to a mott insulator in the fermionic case . a gap for particle - hole excitations exists , corresponding to the energy needed to overcome the on - site repulsion when the next atom is added to the system . disorder , which can be thought of in the simplest case as a random fluctuation in the on - site repulsion @xmath13,@xcite has two effects . first , it leads to localization at _ incommensurate _ densities since atoms will prefer to occupy sites with the lowest on - site repulsion . this can be expected to destroy superfluidity in the case of sufficiently weak exchange . in addition , however , the disorder destroys the mott insulator at commensurate densities , since the distribution in @xmath13 implies that the penalty for double occupancy will be reduced at some sites , thus softening the gap . as the strength of disorder is increased further , the gap closes even for small exchange , and the mott insulator disappears completely . although the system is now gapless , it is still localized ( by disorder ) for small exchange strengths . this gapless bose insulator is often called a _ bose glass_.@xcite in the case of porous glasses , the adsorption potential is strongly disordered on the energy scale ( @xmath16 k ) characteristic of the exchange in a superfluid @xmath0he film . in this case , the onset of superfluidity within the framework of the bose - hubbard model is most likely to occur from the bose glass phase . a characteristic phase diagram in the plane of exchange strength @xmath14 and chemical potential @xmath17 for the case of strong disorder is shown in fig . [ fig : glass ] . in a real experiment , the chemical potential is tuned by changing the density , and one follows a nearly vertical line in fig . [ fig : glass ] as the @xmath0he coverage is increased . the onset of superfluidity occurs at the point @xmath18 . the situation sketched in fig . [ fig : glass ] is reminiscent of a conventional second order phase transition , with the chemical potential @xmath17 assuming the role of temperature . et al._@xcite have exploited this analogy to construct a scaling theory of the bose glass to superfluid transition , based on the ansatz that the correlation length @xmath19 for fluctuations in the phase of the order parameter diverges at the onset of superfluidity as @xmath20 where the reduced chemical potential @xmath21 parametrizes the distance to the critical point . since the number density and the phase of the order paramater are conjugate variables , phase fluctuations are linked to density fluctuations , which propagate at the speed of sound , introducing a time scale as well as the length scale @xmath19 into the problem . fisher _ et al . _ define the characteristic frequency @xmath22 , which is assumed to vanish at the onset of superfluidity according to the power law @xmath23 where @xmath24 is the _ dynamic exponent_. the existence of the dynamic exponent reflects the quantum mechanical origin of the fluctuations in this problem , as opposed to the thermal fluctuations at an ordinary critical point . although rigorous arguments are presented by fisher _ , we note here that conventional hyperscaling will hold if we assume that the correlation volume has a time dimension @xmath25 as well as a spatial part @xmath26 , where @xmath27 is the ordinary spatial dimensionality . this allows us to write down an asymptotic form for the singular part @xmath28 of the free energy density at the onset transition : @xmath29 an asymptotic form for the _ superfluid density _ @xmath12 on the superfluid side of the transition follows from finite size scaling:@xcite @xmath30 the singular part of the compressibility , @xmath31 is analogous to the singular part of the specific heat at a conventional phase transition . in order to make experimental predictions from these scaling laws , one needs to know the dynamic exponent @xmath24 . fisher _ et al . _ prove that @xmath32 for the case of a bose glass to superfluid transition . the proof depends critically on the bose glass having a non - zero compressibility , or , equivalently , a gapless low - energy excitation spectrum . the theory up to this point makes predictions for the scaling properties of the superfluid density only at @xmath2 . fisher _ et al . _ have extended their theory to thermodynamic properties at non - zero temperature using finite size scaling in @xmath33 . in implementing this argument , they assume that the only energy scale in the problem is set by the characteristic frequency @xmath22 . temperature will thus appear in the asymptotic free energy in the form @xmath34 . taking @xmath35 , the scaling - form for the free energy becomes @xmath36 the superfluid transition ( at non - zero temperature ) occurs at a universal value @xmath37 of the argument @xmath38 , so that @xmath39 in order to reduce any of the scaling laws into forms testable by experiment , we need to relate @xmath21 to the density @xmath11 , which can be tuned in the laboratory . assuming @xmath32 and using the inequality @xmath40 for the correlation length exponent in disordered systems,@xcite it can be shown that @xmath41 which means that @xmath42 can be replaced by @xmath6 in all of the above scaling laws . after doing so , we obtain the scaling laws of fisher _ et al._:@xcite @xmath43^x , & x \equiv z/(d+z-2),\label{eq : tcvsrhos}\end{aligned}\ ] ] where @xmath44 is the superfluid density at @xmath2 . the torsional oscillator experiment discussed below addresses the validity of these scaling laws for @xmath0he films adsorbed in aerogel glass . we now consider the low - temperature heat capacity on both sides of the onset transition . technically , the only requirement for the bose glass is that the excitation spectrum be gapless . this admits almost any power - law dependence of @xmath45 , although for a true glass , _ i.e. _ constant density of states , @xmath46 . on the superfluid side of the onset transition , the dominant low - energy excitations are long - wavelength third sound modes , leading to a debye specific heat , @xmath47 for @xmath48 . at onset ( @xmath10 ) , the specific heat can be inferred from the scaling form for the free energy ( eq . [ eq : fst ] ) . in order for @xmath49 to remain non - zero and finite at the transition , the temperature dependent part @xmath50 must assume the form @xmath51 since the sum of all the powers of @xmath42 in eq . [ eq : fst ] must equal zero . at onset , @xmath52 note that if @xmath32 , as is believed to be the case for a bose glass to superfluid transition , @xmath53 at onset . the temperature dependence at onset is thus the same as in the bose glass phase . fisher _ et al._@xcite demonstrate that in the mean - field limit , their scaling theory reduces to the inert - layer model , in which the superfluid film is taken to coexist with a solid - like @xmath0he underlayer . in this case , exchange between the superfluid film and the localized layer is assumed to be negligible , and the thermodynamics of the system near the onset of superfluidity can be deduced from consideration of an ideal bose gas coexisting with a solid @xmath0he pseudo - substrate . the superfluid density in this limit should be proportional to @xmath6 , where the critical coverage @xmath3 is the density of the solid layer . thus , @xmath54 , where @xmath55 for both @xmath56 and @xmath57 . two more exponents can be inferred directly from the density dependence of the bose - einstein condensation temperature : @xmath58^\frac{2}{d},\ ] ] where @xmath13 is the volume ( or area in two dimensions ) of the system . strictly , @xmath59 for @xmath56 because of the logarithmic divergence of the zeta function @xmath60 . this divergence can be removed through the introduction of a short - length cutoff which accounts for the finite thickness of the film . we thus find @xmath61 and @xmath62^x$ ] , where @xmath63 for @xmath56 , and @xmath64 for @xmath57 . we now propose low - temperature forms of the heat capacity for the inert - layer model . for @xmath65 , the heat capacity of the _ ideal _ bose gas has the limiting behavior @xmath66 as @xmath67 . ( the presence of superfluidity requires a debye contribution @xmath68 , but this vanishes more rapidly than the bose gas contribution in the limit @xmath69 . ) since the inert - layer model makes no assumptions about the non - superfluid film , it admits essentially any form of the heat capacity for densities @xmath70 . in table [ tab : scaling ] , we summarize the results of the scaling theory for both the bose glass and inert - layer models . four possible cases are considered , corresponding to the bose glass or inert - layer models in either 2 or 3 dimensions . in calculating the exponents from eqs . [ eq : rhosvsn]- [ eq : tcvsrhos ] and [ eq : catonset ] , we have used the result @xmath32 as well as the correlation length inequality @xmath40.@xcite two experiments will be discussed in this paper . the first is a systematic study of the superfluid density of @xmath0he films adsorbed in 91% porosity aerogel glass for temperatures between 10 mk and 200 mk . the second experiment is a high - resolution heat - capacity study of @xmath0he films adsorbed in vycor glass for temperatures from 10 to 200 mk . the heat capacity measurements were conducted for coverages on both sides of the onset transition . brief accounts of these experiments have been published previously.@xcite vycor and aerogel are both porous glasses . vycor@xcite is prepared by leaching out one phase of a borosilicate glass after spinodal decomposition . the resulting matrix has a distribution of pore diameters with a maximum at approximately 70 and a half - width of approximately 20 . aerogel@xcite is prepared by kinetic aggregation of silica particles in a gel . after hypercritical drying to remove the solvent , the resulting substrate consists of strands of silica approximately 50 - 100 in diameter . unlike vycor , there is no characteristic pore size , but the structure is correlated on length scales up to 600 for the sample studied in this work.@xcite in spite of the structural differences , the superfluid densities of thin @xmath0he films adsorbed in vycor and 91% porosity aerogel near @xmath1 are remarkably similar , although the heat capacity singularity at the superfluid transition seen in vycor does not appear in aerogel.@xcite as discussed in ref . , the correlated disorder of the substrate is apparently irrelevant for very thin films of superfluid @xmath0he adsorbed on either substrate , since the superfluid correlation length @xmath19 exceeds all structural length scales . the superfluid density of @xmath0he films adsorbed in 91% porosity aerogel was measured using the torsional oscillator technique.@xcite the experimental cell was the same as that described in ref . , which discusses the experimental aspects of the superfluid density measurements in detail . several important parameters are listed in table [ tab : cell ] . all of the torsional oscillator data in this paper will be presented in terms of the _ superfluid period shift _ @xmath71 defined below . these data can be converted to a superfluid mass by dividing by the _ superfluid mass sensitivity _ of the cell , 0.022 nsec/@xmath17 g . the measurements under discussion here can be divided into two classes . the first comprises temperature sweeps conducted in a manner identical to that described in ref . . for each @xmath0he coverage , the resonant period @xmath72 and the amplitude of the torsional oscillator were measured as a function of temperature . to determine the superfluid period shift @xmath71 , we first subtracted the temperature - dependent background of the empty cell from the raw data . the data above the superfluid transition temperature @xmath1 were fitted to a constant . the period data were then subtracted from this baseline , giving the superfluid period shift @xmath73 . the period shift at @xmath2 , @xmath74 , was determined by fitting @xmath73 to a horizontal line at the lowest temperatures . uncertainties due to the background subtraction and the fact that our lowest reliable measurements could be made only above 10 mk limited the precision of our @xmath74 measurements to about @xmath75 nsec . the superfluid transition temperature @xmath1 was determined with a typical uncertainty of @xmath76 mk by locating the break in @xmath73 at the transition . the most precise way to do this was to drift slowly ( @xmath77 mk / hr ) through the transition while recording the resonant period and amplitude of the torsional oscillator . the result of such a drift measurement for @xmath78 @xmath17moles / m@xmath79 is shown in fig . [ fig : drift ] , for which @xmath80 mk . these data also show two dips in the amplitude . the large feature , near 38 mk , is certainly a sound resonance since the period , which is measured using a phase - locked loop , shows an s - shaped distortion . the smaller peak , which seems to coincide with the superfluid transition , may be due to a cascade of resonances . we do not know why the resonance associated with the large peak is so prominent , since a simple calculation shows that it does not correspond to a low - order third sound resonance . for the thinnest films , the resonance could be seen even when @xmath73 was almost obscured by noise and uncertainties in the background subtraction . for films with transition temperatures above 50 mk , we could also determine @xmath1 by fitting the period shift data near the transition to a power law @xmath81 , where @xmath82.@xcite the @xmath1 determined in this fashion was typically 0.2 - 0.3 mk below the temperature at which @xmath73 was non - zero within the resolution of our measurements . in this sense , the superfluid transition for films on aerogel is considerably sharper than that observed in _ rods _ of vycor,@xcite where a `` foot '' in @xmath12 extends 1 - 2 mk above the superfluid @xmath1 as determined from power - law fits.@xcite the calorimeter , shown in fig . [ fig : calorimeter ] , was constructed from thin - walled silver . fourteen disks of thickness 0.3 - 0.6 mm were cut from a slab of vycor by first coring the slab with a slurry cutter and then slicing off the disks using a diamond - impregnated wire saw . cutting from a slab with the leaching plane parallel to the surface produces a more homogenous sample than is obtained from rods of vycor , in which the boron - rich phase is leached from the perimeter.@xcite after cutting , the disks were cleaned in hydrogen peroxide and then baked in vacuum . a silver coating , @xmath830.5 @xmath17 m thick , was then evaporated on both sides of each disk . the edges were left uncoated to allow the @xmath0he to penetrate the sample . each disk was then glued to two 0.025 mm thick ag foils , which had previously been diffusion - welded to the cap of the calorimeter . the calorimeter was sealed with emerson cumming stycast 1266 epoxy . a pt - w wire heater was wound around the body of the calorimeter , and a 1 k@xmath22 ruo@xmath4 resistor was glued to the cap of the cell . the calorimeter was supported by four vespel posts and was attached to the low - temperature stage of a prni@xmath84 demagnetization cryostat . the thermal contact to the stage was controlled using a tin - wire superconducting heat switch . a low thermal mass dc magnetization thermometer , shown in fig . [ fig : thermometer ] , was essential for measurements below 20 mk . the actual sensor in this thermometer is a miniaturized version of the design of greywall and busch.@xcite 2.5 mgrams of cerous magnesium nitrate ( cmn ) powder , consisting of grains less than 75 microns in diameter , was mixed with an equal volume of 2 - 4 micron silver powder and then pressed between eleven 0.025 mm thick silver foils . before pressing with the powder , the foils were annealed , perforated to enhance the adhesion of the ag - cmn powder , and their ends were diffusion welded together . after pressing , the assembly was potted in stycast 1266 epoxy and machined into a cylinder 3 mm in diameter . a 2 @xmath17h nbti wire coil , which was part of a superconducting flux transformer , was then wound around the cylinder , and the foils were attached to a ag wire using a crimp joint . this assembly was then repotted in epoxy and glued to a vespel post , which was attached to the low temperature stage of the refrigerator . two nb shields were used to trap a small ( 25 - 50 gauss ) magnetic field and for magnetic shielding . the wires for the flux transformer were twisted and then threaded through a nb capillary to a junction box on the still of the dilution refrigerator . the capillary was filled with silicon oil in order to prevent any vibrations of the twisted pair . the remainder of the flux transformer passed through a feedthrough into the bath space and was attached to the input coil of a commercial dc squid.@xcite the operation of the thermometer is identical in principle to the high - resolution thermometer developed for @xmath85-point experiments by lipa and chui.@xcite some details have been published previously by our group.@xcite the magnetization is measured by monitoring the current in the flux transformer , which changes so as to keep the total flux enclosed by the superconducting loop constant . the sensitivity of the thermometer is determined by the applied field . the price paid for the sensitivity of the dc magnetization thermometer is the need to count flux , since the dynamic range of the flux - locked loop is limited , and so the loop must therefore be reset periodically . in practice , we regularly exceeded the slew rate of the squid electronics and the flux count therefore became meaningless . this inconvenience was ameliorated in our case by the availability of the heat switch on the calorimeter , which could be closed to check the magnetic thermometer against a @xmath86he melting curve thermometer on the stage at any point during the experiment . the heat capacity measurements were conducted using the adiabatic calorimetry technique . a current pulse was applied to the heater , and the resulting thermal response of the cell was recorded . two response curves , recorded at temperatures of 102.1 and 18.6 mk for a coverage of 27.1 @xmath17moles / m@xmath79 , are shown in fig . [ fig : decay ] . these data are representative of two classes of response curves that we found . above 80 mk , the temperature of the cell decayed exponentially after the pulse , and the temperature step @xmath87 was determined simply by extrapolating the exponential decay back to the midpoint of the heat pulse . at lower temperatures , the decay became non - exponential , as can be seen in the lower panel of fig . [ fig : decay ] . since we could fit the long - time part of the decay to an exponential , we first believed that the short - time behavior reflected a thermal overshoot due to poor thermal contact between the @xmath0he film and the calorimeter . if we extrapolated the long - time behavior back to the midpoint of the pulse , however , the inferred heat capacities were clearly too high . this suggested that either heat was leaking out of the calorimeter on short time scales , hence leading to an overestimate of the heat capacity , or that the anomalous decay was due to some internal relaxation mechanism , perhaps reflecting a significant heat capacity in poor thermal contact with the rest of the calorimeter . the former possibility was unlikely since the superconducting heat switch provides its best thermal isolation at the lowest temperatures . we suspected that the vycor itself might harbor a thermal reservoir and modeled the thermal relaxation inside the cell assuming the existence of a reservoir that was weakly linked to the @xmath0he film.@xcite . this provided a reasonable fit to the decays , but we found that the thermal conductivity to the weakly linked reservoir depended on both coverage and temperature . a similar coverage - dependent coupling between the @xmath0he film and the vycor itself has been observed in ultrasound measurements.@xcite we decided to proceed on the assumption that the @xmath0he film equilibrated rapidly with the thermometers , in which case the temperature step @xmath87 for determining the specific heat should be extracted from the response of the calorimeter at short time scales . we tested this hypothesis by dosing approximately 2 monolayers of @xmath86he into the calorimeter at the end of the run . this film had a heat capacity approximately 60 times larger than any of the @xmath0he films in our experiment . the @xmath86he film equilibrated within 120 sec after each heat pulse , justifying our assumption about the short time - scale behavior of the calorimeter at low temperatures . the heat capacity for a coverage of 27.6 @xmath17moles / m@xmath79 is shown in fig . [ fig : cpexample ] after dividing by the temperature @xmath88 . these data , taken with the magnetization thermometer , show a cusp at 20 mk , which is a signature of the superfluid transition.@xcite ( since we did not take a bet isotherm for this sample , the surface area was determined by comparing our measured superfluid transition temperatures as a function of @xmath0he dosage with those of crooker _ et al._@xcite ) the ordinary low - temperature limit of our measurements was 10 mk . this was determined by the internal heat leak of the calorimeter , which decayed from approximately 0.8 to 0.4 nw over the course of three months . this large heat leak , which we suspect was due to the vycor in the calorimeter , required us to run with the heat switch partially open at the lowest temperatures . obtaining a proper tuning of the heat switch was very tedious , and this turned out to be the primary impediment to using the calorimeter below 10 mk . in fig . [ fig : alldat ] , we show the period shift data for twelve coverages of @xmath0he adsorbed on aerogel . the superfluid period shift at @xmath2 , @xmath74 , was obtained by fitting the low - temperature part of each curve to a horizontal line . the transition temperature @xmath1 was determined for each coverage using the drift technique described above . in fig . [ fig : deltap0 ] , we show @xmath1 and @xmath74 as a function of coverage . the curves show fits to power laws @xmath89 and @xmath90 where we find @xmath91 and @xmath92 . the critical coverage @xmath93 @xmath17moles / m@xmath79 was determined from the fit of @xmath1 versus coverage and then held fixed for the fit of the period shift data . the final relation of interest is the dependence of @xmath1 on @xmath74 , which is shown in fig . [ fig : tcvsdeltap0 ] . in this case , we have fitted the data to the power law @xmath94^x , \label{eq : tcp0powerlaw}\ ] ] where we find @xmath95 . the measured exponents are listed in table [ tab : exponents2 ] along with corresponding values we obtain from fits of the data of crooker@xcite for the @xmath0he - vycor system . we will discuss the exponents of table [ tab : exponents2 ] in more detail below . another point of interest is the size of the `` critical region '' for the onset transition . since the fits of the various power laws are poor ( by the standards generally applied to critical phenomena ) , the usual practice of looking at deviations from the power - law behavior is not particularly instructive . another approach is simply to see how the data near onset compare with those further away from the transition . we do this for @xmath74 as a function of coverage in fig . [ fig : dp0all ] . the highest coverage in this figure has a @xmath1 of approximately 1 k. a significant departure from linear behavior is seen only for films with @xmath96 nsec , or @xmath97 mk . the linear dependence of @xmath74 on @xmath11 at higher coverages is consistent with the inert - layer model of the onset transition , so the data of fig . [ fig : dp0all ] indicate only a small coverage regime in which fluctuations associated with the onset transition might be relevant . the evolution of @xmath1 with @xmath74 is even more striking in this regard , since , as shown in fig . [ fig : tcall ] , the nearly linear behavior observed for thin films continues for films with @xmath1 s up to 0.7 k. [ this linear dependence is suggestive of the kosterlitz - thouless - nelson ( ktn ) relation for a two - dimensional superfluid,@xcite for which @xmath98 @xmath17moles m@xmath99k@xmath100 , where @xmath101 is the limiting superfluid density as the transition temperature is approached from below . for reasons of comparison in fig . [ fig : tcall ] , we have converted the period shift to _ areal superfluid density _ using the calibration @xmath102 @xmath17moles m@xmath99nsec@xmath103 . the slope of the ktn line , which is drawn dashed in the figure , is about twice that of a linear fit to the data of fig . [ fig : tcall ] . although we expect @xmath101 to be reduced from @xmath44 by elementary excitations and vortex pair screening , the discrepancy between the two slopes is not resolved in a full fit of the superfluid density data to the finite size theory @xcite for a superfluid film adsorbed on a cylindrical strand . the best - fit coverage for each data set that is about half the actual coverage . @xcite we do not believe , however , that the failure of this naive application of the ktn relation says anything about the effective dimensionality of the film , which will be discussed further below . ] the @xmath0he - vycor heat capacity data , divided by temperature , are shown for ten coverages in fig . [ fig : alldat2 ] . each of the data sets in the lower panel of the figure shows a sharp cusp , corresponding to a superfluid transition.@xcite the superfluid transition temperatures determined from the locations of the cusps are shown as closed circles in fig . [ fig : tbtc ] . the critical coverage @xmath3 for the onset of superfluidity is determined by extrapolating the @xmath1 data to zero temperature . for temperatures @xmath104 , the heat capacity of the superfluid films is roughly quadratic in temperature , while it depends nearly linearly on temperature for @xmath105 , with a small quadratic correction that increases with coverage . in the vicinity of the superfluid transition , there is a small peak in @xmath106 due to critical fluctuations.@xcite the size of the peak decreases with decreasing thickness . as can be seen in fig . [ fig : cpexample ] , this peak can barely be resolved for the thinnest superfluid film ( @xmath1 = 20.5 mk ) . at high temperatures , the heat capacity of non - superfluid films , shown divided by temperature in the upper panel of fig . [ fig : alldat2 ] , depends linearly on temperature , but c / t drops rapidly at low temperatures . we define a crossover temperature @xmath107 separating these two regimes to be the point at which @xmath106 is half of its high - temperature value . @xmath107 is shown as a function of coverage in fig . [ fig : tbtc ] , along with the @xmath1 s measured for the superfluid films . this plot shows clearly that @xmath107 vanishes as the onset transition is approached from below . ( although our definition of @xmath107 is somewhat arbitrary , any reasonable choice of a characteristic temperature for the crossover would show the same coverage dependence . ) for @xmath108 , there is not much qualitative difference between the heat capacity of a non - superfluid film and that of a superfluid film for @xmath105 . this is emphasized in the three - dimensional representation of our data shown in fig . [ fig:3d ] , in which there is a wedge - shaped plateau in the vicinity of the onset coverage @xmath109 @xmath17moles / m@xmath79 . above 20 mk , the heat capacity isotherms , which are cross - sections of the 3d surface , show no feature at @xmath3 . the peaks in the isotherms become progressively narrower in coverage as the temperature is reduced , but we can not reliably extrapolate our data to @xmath2 . the data introduced above indicate that the concept of a superfluid - insulator transition in @xmath0he films is qualitatively appropriate . both sets of measurements indicate the existence of a critical coverage @xmath3 at which the superfluid transition temperature @xmath1 vanishes . the heat capacity data are particularly instructive in this regard , since there is a characteristic energy @xmath110 for nonsuperfluid films which also vanishes as @xmath3 is approached from _ below _ in coverage . as can be seen by comparing the exponents of table [ tab : exponents2 ] with the predictions of table [ tab : scaling ] , the experimentally determined exponents for the onset transition are not consistent with the scaling theory for either @xmath56 or @xmath57 . we have also seen in figs . [ fig : dp0all ] and [ fig : tcall ] that the departure from the behavior seen at higher coverages ( @xmath111 mk ) is actually quite small . some curvature is apparent in @xmath74 at the lowest coverages , but the relation between @xmath1 and @xmath74 remains nearly linear for @xmath1 s from 700 mk down to the lowest @xmath1 in our study ( 20 mk ) . these observations lead us to question whether our measurements are in the asymptotic regime of reduced coverage @xmath6 in which the scaling theory is valid . to address this question , we return to the asymptotic form of the free energy , eq . [ eq : fst ] . this scaling form comprises the product of the free energy at @xmath2 and a function of @xmath112 . this implies @xmath113 in the asymptotic regime , where @xmath114 is a universal function . thus , upon scaling the period shift @xmath71 by @xmath74 and the temperature by @xmath1 , all of the @xmath73 data in the asymptotic regime should collapse on to a universal curve . as can be seen in fig . [ fig : bgscale ] , our data do not scale in this fashion . in fact , the scaled data in this figure drift continuously toward the origin as the coverage is reduced . there is no indication that the data are even approaching a universal curve . the scaling plot demonstrates explicitly that the properties of superfluid films at non - zero temperature are _ not _ determined by the single energy scale @xmath22 . therefore , we should not expect the scaling laws ( eq . [ eq : tcvsn ] and [ eq : tcvsrhos ] ) for @xmath1 as a function of @xmath11 or @xmath44 to hold in the coverage range of our experiment . although we have shown explicitly that the scaling theory does not hold at non - zero temperature ( _ i.e. _ @xmath115 mk ) , we must still consider the possibility that it holds at @xmath2 . only one of the scaling laws , @xmath54 , applies in this case , but the experimental result @xmath116 is well below the theoretical bounds @xmath117 for @xmath57 or @xmath118 for @xmath56 . since , as shown in fig . [ fig : dp0all ] , the data approach the inert - layer prediction at higher coverages , it is reasonable to ask if our low - coverage data fall in a crossover regime in which quantum fluctuations are relevant but not yet dominant . this is certainly possible , but it leads one to ask why the critical region in coverage should be so small . in order to address this question , we convert the coverage to the chemical potential @xmath17 using the van der waals relation @xmath119 where @xmath120 k ( layer)@xmath86 for @xmath0he on glass.@xcite although the chemical potential is difficult to measure in this coverage range , third sound measurements , which determine @xmath121 indirectly , are consistent with the form of eq . [ eq : vdw].@xcite assuming a monolayer coverage of 13 @xmath17moles / m@xmath79 , @xmath122 2.6 layers for @xmath0he - aerogel and @xmath123 3.0 layers for a film with a @xmath1 of 200 mk . the chemical potential thus shifts from about -1.5 k at onset to approximately -1.0 k for films with @xmath1 s near 200 mk . given that the characteristic exchange energy for the @xmath0he films is on the order of 12 k,@xcite a chemical potential shift of 0.5 k can not be considered small . the rapid change of @xmath17 with coverage also leads to a technical problem in applying the scaling theory : the assumption @xmath124 is not correct , even in the narrow coverage range of our experiment . assuming that @xmath17 follows the form of eq . [ eq : vdw ] , we have converted coverage to chemical potential and have fitted our data to the power law @xmath125 in effect , this allows us to eliminate the assumptions implicit in eq . [ eq : nmu ] , with the caveat that we have not actually measured the chemical potential . we find an exponent @xmath126 by this approach . the case for the bose glass is thus improved in @xmath56 , although the inequality of eq . [ eq : rhosvsn ] is still not satisfied for @xmath57 . we note , however , that the theory can not be applied self - consistently in @xmath56 , since it assumes a diverging correlation length which will eventually exceed the strand size of the aerogel . in this case , topological arguments lead us to expect that the onset transition must be three - dimensional in character , although we can not exclude the possibility of an observable crossover regime between 2d and 3d behavior . we thus find that two conditions of the scaling theory do not apply in the case of @xmath0he adsorbed in aerogel in the coverage range of our study . first , there is no single energy scale that determines the thermodynamic properties at finite temperature . second , the chemical potential changes rapidly with coverage , which makes the asymptotic regime difficult to achieve and also implies that the reduced density @xmath6 is not simply proportional to @xmath127 . we recall at this point the third ( and most fundamental ) assumption of the theory , which is that the non - superfluid phase is compressible . ( this assumption is critical to establishing the relation @xmath32 for the dynamic exponent . ) since the compressibility is given by @xmath128 , eq . [ eq : vdw ] suggests that although @xmath129 will be non - zero , it will be considerably lower than for an idealized @xmath0he film such as that considered in the bose - hubbard model.@xcite in effect , the adsorption potential adds a third energy scale to the problem in addition to the exchange energy and the interparticle repulsion . the strong binding to the substrate makes the film much less compressible than it would be if it were free - standing."@xcite even if @xmath129 is technically non - zero , the film may be sufficiently incompressible that the softening due to quantum fluctuations is negligible . the heat capacity measurements allow us to test the predictions of the scaling theory below the onset of superfluidity . as shown in fig . [ fig : alldat2 ] , the salient feature of @xmath45 for coverages below onset is a characteristic temperature @xmath107 below which @xmath106 drops rapidly to zero and above which it approaches a constant . as shown in fig . [ fig : tbtc ] , @xmath107 vanishes as the onset of superfluidity is approached from below . at the onset transition , @xmath45 is approximately linear over the entire temperature range , @xmath130 mk , covered by our experiment . the fact that @xmath45 is linear in @xmath88 for @xmath10 is in apparent agreement with the scaling theory of fisher _ et al._@xcite ( see table [ tab : scaling ] ) . below onset , however , the data show only a single region of linear behavior , although the bose glass model predicts two regions : one at low temperature due to the bose glass and a second at higher temperatures due to quantum fluctuations . one possible explanation for the single linear region is that the bose glass contribution is immeasurably small . this would be consistent with the observation made above that the strong van der waals potential makes the @xmath0he film nearly incompressible . assuming for the moment that the bose glass contribution to the heat capacity is too small to be measured , we turn to the linear behavior that we do observe . at onset this is predicted by the scaling theory of fisher _ et al . _ , in which the contribution to @xmath131 from quantum fluctuations varies linearly with temperature . we note , however , that the linear behavior is observed over a very wide temperature range , extending up to at least 600 mk , while the superfluid density data for both aerogel and vycor indicate that the critical regime in which quantum fluctuations are relevant is confined to coverages with @xmath1 s less than 200 mk . although there is no reason why the asymptotic critical regions observed for two different thermodynamic quantities must be the same , it would be rather unusual to see critical behavior so readily in @xmath45 when all other measurements suggest that the critical regime is nearly inaccessible . as can be seen in fig . [ fig : alldat2 ] , the heat capacity of non - superfluid films drops off very rapidly , faster than @xmath132 , at low temperatures . this behavior is suggestive of a gap in the excitation spectrum of the film . the simplest density of states that includes a gap is shown in fig . [ fig : tait ] . the model , considered originally by tait and reppy,@xcite includes two bands of states . the lower band comprises localized states due to the heterogeneous adsorption potential . because of the strong repulsive he he interactions , each localized state can accommodate only a single he atom , leading to an effective fermi statistics in the lower band of fig . [ fig : tait ] . the upper band , which is unoccupied at @xmath2 , consists of extended states , which are assumed to obey bose statistics at low densities . this model leads to a heat capacity of the form @xmath133 for @xmath134 . the prefactor @xmath135 , where @xmath136 and @xmath137 are the densities of states in the lower and upper bands . the results of fits to eq . [ eq : activation ] for four non - superfluid films in the regime @xmath138 are shown in fig . [ fig : activationfits ] . the gap @xmath139 is shown as a function of coverage in the inset . for each coverage , we find a gap of order @xmath140 . our data are not precise enough to distinguish between a true gap in the density of states and a `` soft gap '' , in which a low - energy tail exists in the band of extended states . it is possible that a small number of extended states exist at energies well below @xmath139 . in principle , it would be possible to place a bound on the number of such states by determining how much of the classical entropy is _ not _ frozen out at low temperatures . unfortunately , this calculation is impossible because our data do not reach a limit in which @xmath45 is independent of temperature . ( this limit could not be reached without macroscopically populating the vapor phase , which is ignored in our model . ) there is no _ a priori _ reason for our having chosen a _ constant _ density of states in each of the bands shown in fig . [ fig : tait ] . as far as fitting the data of fig . [ fig : activationfits ] is concerned , the critical element of the excitation spectrum is the gap . additional information , however , comes from the high - temperature data on both sides of the onset transition . in both cases , the heat capacity depends linearly on temperature in this regime . for @xmath70 , only the extended states contribute to the heat capacity at high temperature , and a linear temperature dependence would follow from a constant density of states . the only difference for superfluid films @xmath141 is that some extended states are occupied at @xmath2 . for @xmath105 we would expect to see the same temperature dependence as observed for films below the critical coverage . based on fig . [ fig:3d ] , this appears to be the case , since the wedge - shaped plateau at high - temperatures extends to both sides of the critical coverage . we have also fitted the high - temperature ( @xmath108 or @xmath105 ) data to the form @xmath142 on both sides of the onset transition . as can be seen in fig . [ fig : abfits ] , the linear coefficient @xmath143 increases slowly with coverage , with no apparent feature at the onset transition , which is marked in the figure by the dashed line . the quadratic contribution , however , increases rapidly for coverages above onset , although it remains at least a factor of two smaller than the linear heat capacity for @xmath144 mk . the onset of superfluidity in @xmath0he films adsorbed in vycor thus corresponds to the closing of a gap in the excitation spectrum of the film as the coverage is increased . although this is a bose insulator to superfluid transition , the insulating phase is not gapless , in disagreement with the assumptions of the bose glass model . although this conclusion turns to some degree on whether or not we can resolve a small but nonzero density of states at low temperatures , there is also a significant philosophical distinction between the activation model we have employed here and the bose glass approach . in the bose glass model , superfluidity is destroyed at @xmath10 by order parameter phase fluctuations . the bose condensate ( essentially the microscopic order parameter ) is assumed to exist below onset , even though long - range order does not . in the simple activation model employed here , there is no order parameter below onset , since none of the extended states in the system are occupied at @xmath2 . a similar distinction appears for the case of the superconductor - insulator transition , for which two classes of transitions have been observed . for granular films , the order parameter is non - zero at the superconductor - insulator transition,@xcite and the superconductor - insulator transition is driven by fluctuations in the phase of the order parameter . there is also evidence for a bose glass phase in some homogeneously disordered films,@xcite but other experiments suggest that the amplitude of the order parameter vanishes at the superconductor - insulator transition in this case . @xcite at the moment , we can not probe the microscopic order parameter in our system , and so we can not determine directly whether or not a condensate exists for coverages below the onset of superfluidity . as discussed above , the rapid change of the chemical potential @xmath17 with coverage near @xmath3 tends to make quantum fluctuations less important in @xmath0he films than they would be if the adsorption energy were less strongly dependent on coverage . the large spread in energy scales due to the adsorption potential minimizes exchange between atoms in the localized layer and those in the overlying superfluid film . we believe this lies behind the qualitative success of the inert layer model for thicker films , and it is reasonable to ask how close to the onset transition one can neglect exchange with the non - superfluid film . if we ignore the @xmath0he atoms in the localized layer in our aerogel sample , we estimate that the inter - particle spacing @xmath27 for the atoms in the superfluid film is at least 10 for @xmath145 mk . this estimate follows from dividing the number of superfluid atoms by the _ surface area _ ( 9.2 m@xmath79 ) of the substrate . estimates based on the total _ volume _ occupied by the substrate ( 0.13 cm@xmath79 ) give an interparticle spacing on the order of 30 for a film with @xmath1 = 80 mk.@xcite both of these estimates indicate that it is not unreasonable to look for dilute bose - gas like behavior , since the @xmath147-wave scattering length characterizing the he - he interaction is of order 3 . a full formalism for the crossover from a strongly interacting superfluid to a dilute bose gas as the @xmath0he coverage is reduced has been developed by weichman and co - workers.@xcite the starting point for this theory is the observation that thin films of @xmath0he adsorbed on vycor show 3d xy critical behavior for coverages with transition temperatures above 120 mk.@xcite identical critical behavior ( for films with @xmath1 s down 50 mk ) has been observed in @xmath0he films adsorbed in 91% porosity aerogel glass.@xcite in both systems , power - law behavior of the superfluid density is observed over at least one order of magnitude in reduced temperature . _ derive a scaling function which in principle can be used to test for crossover from conventional 3d xy critical behavior to dilute bose gas behavior as the @xmath0he density decreases . the scaling function reduces to the dilute bose gas form @xmath148 $ ] as @xmath67 and to the 3d xy model form @xmath149 ( @xmath150 ) as @xmath151 . physically , the scaling function accounts for the reduction in the size of the 3d xy critical regime as @xmath1 decreases . unfortunately , the bose gas and 3d xy limits are sufficiently distinct that a scaling function capable of accounting for the crossover can be used to collapse data from essentially any superfluid film.@xcite the two limiting forms , however , suggest another way of testing for crossover to bose gas - like behavior.@xcite the quantity @xmath152 _ diverges _ in the strongly interacting limit but remains finite for the dilute bose gas . in fig . [ fig : dbg ] , we plot @xmath153 as a function of @xmath154 for four coverages of @xmath0he adsorbed on aerogel . the dilute bose gas prediction ( assuming the same ratio @xmath155 psec / mk found for the four films ) is shown as a dash - dotted line . the divergence near @xmath1 is pronounced for all of the films . there is thus no convincing evidence that @xmath0he films adsorbed in aerogel approach bose - gas like behavior for films with @xmath1 s down to 20 mk . the absence of any observable crossover may be due to exchange with the localized layer or simply the possibility that our films were not sufficiently dilute . studied films with @xmath1 s down to 7 mk in their study of the @xmath0he - vycor system and found a progressive decrease in the 3d xy critical regime as determined from the same analysis used here to produce fig . [ fig : dbg ] . unfortunately , the superfluid signal for a given substrate volume is smaller for aerogel than for vycor , and the thermal time constants are much longer in aerogel , so extending our measurements to lower coverages and temperatures would not be a trivial task . this is unfortunate , since the superfluid transitions in @xmath0he - aerogel are `` sharper '' than those in @xmath0he - vycor , in the sense that @xmath12 goes to zero only 0.2 - 0.3 mk above @xmath1 ( as determined from power - law fits@xcite ) as opposed to 1 - 2 mk in vycor . kotsubo and williams@xcite have noted that the smearing of the transition in the vycor case tends to mimic crossover to bose gas - like behavior . there is now convincing evidence that the 1 - 2 mk `` foot '' in the vycor data was due to macroscopic inhomogeneities introduced by the leaching process used to produce the porous glass.@xcite a new vycor study , with more homogeneous samples , would probably be much more conclusive than the experiment of crooker _ _ in testing for crossover to dilute bose gas behavior . porous gold substrates,@xcite which recent full - pore measurements suggest are as homogeneous as vycor , could be an even better substrate for ultra - low temperature measurements . the difficulties in interpreting the @xmath12 measurements for very thin superfluid films have been compounded by the absence of a complementary study of other thermodynamic quantities in this coverage regime . our heat capacity measurements for @xmath0he - vycor begin to address this problem , although the lowest transition temperature we were able to achieve was only 21 mk , significantly higher than the lowest @xmath1 ( @xmath156 mk ) reached in the @xmath12 study of crooker _ et al._@xcite in fig [ fig:21and85mk ] , we show heat capacity data for two superfluid coverages , @xmath157 @xmath17moles / m@xmath79 ( @xmath158 85 mk ) and @xmath159 @xmath17moles / m@xmath79 ( @xmath160 20.7 mk ) . the heat capacity is divided by temperature and is plotted as a function of the reduced temperature @xmath161 . the heat capacity for both coverages crosses over from a predominantly linear dependence on temperature above @xmath1 to a quadratic dependence for @xmath104 . in addition to these backgrounds " , there is a peak in the upper data set ( @xmath158 85 mk ) at the transition . a peak can only barely be resolved for the film with @xmath162 mk . this trend is consistent with that observed in earlier studies,@xcite in which the magnitude of the heat capacity peak for @xmath0he films adsorbed in vycor was shown to scale according to the hyperuniversality relation,@xcite @xmath163 for two systems , denoted primed and unprimed , within the same universality class . in eq . [ eq : tsfu1 ] , @xmath164 is the reduced temperature @xmath165 and @xmath19 is the correlation length . correlation lengths for the @xmath0he - vycor system can be estimated from the josephson relation,@xcite @xmath166 and @xmath12 data from the studies of bishop _ et al._,@xcite and crooker _ et al._@xcite based on these studies , we estimate that the correlation length for a given reduced temperature will be approximately twice as large for the film with @xmath162 mk than for the film with @xmath167 mk.@xcite according to the hyperuniversality relation eq . [ eq : tsfu1 ] , the heat capacity peak for the thinner film should therefore be eight times smaller than for the thicker film , which is consistent with fig . [ fig:21and85mk ] . a rigorous test of eq . [ eq : tsfu1 ] for @xmath0he films in vycor would require fitting each of the heat capacity data sets to power laws near @xmath1 . this is impractical at the moment due to both the noise in the data and the ambiguities in subtracting a non - singular background . although the noise could be reduced in a better experiment , the interpretation of the background heat capacity poses a more fundamental problem . in fig . [ fig:21and85mk ] , the data for @xmath162 mk show a cusp at @xmath1 , where @xmath45 crosses over abruptly from quadratic to linear dependence on temperature . this cusp may be the remnant of the fluctuation peak seen for thicker films , in which case it should round out as the superfluid coverage is further reduced . another possibility is that the heat capacity is singular even in the absence of order parameter fluctuations . although this type of behavior would emerge in any mean - field treatment , for an interacting @xmath0he film we would expect a jump discontinuity in @xmath45 in addition to a change in slope . ( the ideal bose gas should show only a simple cusp at @xmath1 , but of a different form than that seen in fig . [ fig:21and85mk ] . ) the superfluid density and heat capacity measurements presented in this paper indicate that a @xmath0he film adsorbed on a disordered substrate undergoes a transition from a bose insulator to a superfluid at the critical coverage @xmath3 . the gap in the insulating phase decreases to zero as the onset coverage is approached from below . our data show no evidence of a bose glass phase . in considering thin superfluid films , the rapid change of the chemical potential with coverage near @xmath3 appears to minimize the role of quantum fluctuations , so that the ordinary superfluid transition at @xmath1 remains the dominant critical point . evidence of critical fluctuations near @xmath1 is seen even for the thinnest superfluid films studied in these experiments . although some of the questions raised in this work could be addressed in experiments at even lower temperatures and smaller reduced coverages , the use of different substrates may be more rewarding . we have argued here that the strong substrate potential in porous glasses reduces the importance of quantum fluctuations near the critical coverage . weakly binding alkali metal substrates@xcite may ameliorate this effect . in principle , the use of ordered substrates would allow a test of the clean limit of the various dirty boson models , provided that the adsorption potential is sufficiently weak . such a system might be achieved by pre - plating a clean high - surface area substrate , such as exfoliated basal - plane graphite , with molecular hydrogen . finally , even on a strong - binding ordered substrate such as graphite , the adsorption potential changes only weakly within each monolayer , so that quantum effects should be more readily observable . the phase - separated superfluids which exist within the second and third monolayers of @xmath0he on graphite may therefore be reasonable analogs of granular superconductors.@xcite we thank mary lanzerotti for assistance with the superfluid density measurements . pac acknowledges financial support from at&t bell laboratories through its phd scholarship program . this work was supported by the national science foundation under grant nos . dmr-89 - 21733 and dmr-93 - 03855 and by the mrl program of the national science foundation under award no . dmr-91 - 21654 . m. ma , b.i . halperin , and p. lee , phys . b * 34*. 3136 ( 1986 ) ; t. giamarchi and h.j . schulz , phys . b * 37 * , 325 ( 1988 ) ; d.k.k . lee and j.m.f . gunn , j. phys . matter * 2 * , 7753 ( 1990 ) ; a. gold , z. phys . b * 83 * , 429 ( 1991 ) ; t. r. kirkpatrick and d. belitz , phys . rev lett * 68 * , 3232 ( 1992 ) ; k.g . singh and d.s . rokshar , phys . b * 46 * , 3002 ( 1992 ) ; a.p . kampf and g.t . zimanyi , phys . b * 47 * , 279 ( 1993 ) . vycor as discussed in this paper is corning glass 7930 , a precursor to the product sold commercially as vycor . for a review of the structural properties of vycor , see p. levitz , g ehret , s.k . sinha , and j.m . drake , j chem phys . * 95 * , 6151 ( 1991 ) . the films in the aerogel measurements were generally unannealed . once the critical coverage was exceeded , subsequent doses of @xmath0he were introduced into the cell at low temperature . h. cho and g.a . williams ( private communication ) have noted that annealing of films adsorbed on porous ceramic substrates removed a residual signal above @xmath1 that appeared in unannealed films . all of the coverages in the heat capacity measurements discussed here were annealed . d. finotello , k.a . gillis , a. wong and m.h.w . chan , phys . lett . * 61 * , 1954 ( 1988 ) ; m.h.w . _ , in _ excitations in two - dimensional and three - dimensional quantum fluids _ , ed . by a.f.g wyatt and h.j . lauter ( plenum , new york , 1991 ) . ceperley , rev . 67 * , 279 ( 1995 ) disucsses the exchange energy for bulk superfluid @xmath0he . we are not aware of a comparable calculation for films but expect that the exchange energy can not exceed the maximum value ( @xmath83 1.5k ) found for bulk superfluid . lccccc asymptotic form & exponent & & + & & d=2&d=3&d=2&d=3 + @xmath54 & @xmath168 & @xmath169 & @xmath170 & 1 & 1 + @xmath61 & @xmath171 & @xmath169 & @xmath169 & 1 & 1 + @xmath62^x$ ] & @xmath172 & 1 & 3/4 & 1 & 2/3 + @xmath173 & @xmath174 & 1 & 1 & & + @xmath175 & @xmath176 & 1 & 1 & & + @xmath177 & @xmath178 & 2 & 3 & 1 & 3/2 +

we show that the model checking problem for intuitionistic propositional logic with one variable is complete for logspace - uniform . as basic tool we use the connection between intuitionistic logic and heyting algebra , and investigate its complexity theoretical aspects . for superintuitionistic logics with one variable , we obtain-completeness for the model checking problem . universitt jena , institut fr informatik , jena , germany + \{martin.mundhenk , felix.weiss}@uni - jena.de

1012.3828

intuitionistic logic ( see e.g. @xcite ) is a part of classical logic that can be proven using constructive proofs e.g . by proofs that do not use _ reductio ad absurdum_. for example , the law of the excluded middle @xmath2 and the weak law of the excluded middle @xmath3 do not have constructive proofs and are not valid in intuitionistic logic . not surprisingly , constructivism has its costs . whereas the validity problem is @xmath4-complete for classical propositional logic @xcite , for intuitionistic propositional logic it is @xmath5-complete @xcite . the computational hardness of intuitionistic logic is already reached with the fragment that has only formulas with two variables : the validity problem for this fragment is already @xmath5-complete @xcite . recall that every fragment of classical propositional logic with a fixed number of variables has an @xmath1-complete validity problem ( follows from @xcite ) . in this paper , we consider the complexity of intuitionistic propositional logic @xmath6with one variable . the model checking problem i.e . the problem to determine whether a given formula is satisfied by a given intuitionistic kripke model for @xmath6is @xmath7-complete @xcite , even for the fragment with two variables only @xcite . more surprisingly , for the fragment with one variable @xmath8we show the model checking problem to be @xmath0-complete . to our knowledge , this is the first `` natural '' @xmath0-complete problem , whereas formerly known @xmath0-complete problems ( see e.g. @xcite ) have some explicit logarithmic bound in the problem definition . a basic ingredient for the @xmath0-completeness lies in normal forms for models and formulas as found by nishimura @xcite , that we reinvestigate under an algorithmic and complexity theoretical point of view . in contrast , the formula value problem for classical propositional logic is @xmath1-complete @xcite independent of the number of variables . classical propositional logic is the extension of @xmath6with the axiom @xmath9 . those proper extensions of intuitionistic logic are called superintuitionistic logics . the superintuitionistic logic @xmath10(see @xcite ) results from adding @xmath3 to @xmath6 . we show that the model checking problem for every superintuitionistic logic with one variable is @xmath1-complete ( and easier than that for @xmath8 ) . in contrast , for the superintuitionistic logic @xmath10with two variables it is known to be @xmath7-complete ( and as hard as for @xmath6with two variables ) @xcite . as a byproduct , we also obtain results for the validity problem for intuitionistic and superintuitionistic logics with one variable . this paper is organized as follows . in section [ sec : prelims ] we introduce the notations we use for intuitionistic logic and model checking . section [ sec : rnprelims ] is devoted to introduce the old results by nishimura @xcite and to upgrade them with a complexity analysis . the following section [ sec : ac1haerte ] presents our lower and upper bound for model checking for @xmath8 . section [ sec : superint1 ] deals with the complexity of the model checking problem and the validity problem for superintuitionistic logics with one variable . the implied completeness for the model checking for intuitionistic logic and conclusions are drawn in section [ sec : conclusion ] . the notion of reducibility we use is the logspace many - one reducibility @xmath11 , except for @xmath1-hardness , where we use first - order reducibility . @xmath1and @xmath0are the classes of sets that are decided by families of logspace - uniform circuits of polynomial size and logarithmic depth . the circuits consist of , or- , and not - gates . the not - gates have fan - in @xmath12 . for @xmath1 , the and- and or - gates have fan - in @xmath13 ( bounded fan - in ) , whereas for @xmath0there is no bound on the fan - in of the gates ( unbounded fan - in ) . @xmath14denotes the class of sets decided by alternating turing machines in logarithmic time , and we will use that @xmath15 ( see @xcite ) . @xmath16denotes the class of sets decidable in logarithmic space . we use @xmath17}$ ] to denote the class of sets decided by an alternating logspace turing machine that makes @xmath18 alternations , where @xmath19 is the length of the input . we will use that @xmath20}}\xspace}}$ ] ( see @xcite ) . @xmath21is the class of sets that are @xmath11-reducible to deterministic context - free languages . it is also characterized as the class of sets decidable by deterministic turing machines in polynomial - time and logarithmic space with additional use of a stack @xcite . the inclusion structure of the classes we use is as follows . @xmath22 let @xmath23 denote a countable set of _ variables_. the language @xmath24of intuitionistic propositional logic is the same as that of propositional logic @xmath25 , i.e. it is the set of all formulas of the form & : : = & p ( ) ( ) ( ) , where @xmath26 . for @xmath27 the languages @xmath28 are the subsets / fragments of @xmath24for which @xmath23 consists of @xmath29 variables . in this paper we mainly consider @xmath30 ( i.e. formulas with one variable ) . as usual , we use the abbreviations @xmath31 and @xmath32 . because of the semantics of intuitionistic logic , one can not express @xmath33 or @xmath34 using @xmath35 and @xmath36 . an _ intuitionistic kripke model _ for intuitionistic logic is a triple @xmath37 , where @xmath38 is a nonempty and finite set of _ states _ , @xmath39 is a preorder on @xmath38 ( i.e. a reflexive and transitive binary relation ) , and @xmath40 is a function denotes the powerset of @xmath38.]the _ valuation function_. informally speaking , for any variable it assigns the set of states in which this variable is satisfied . the valuation function @xmath41 is monotone in the sense that for every @xmath42 , @xmath43 : if @xmath44 and @xmath45 , then @xmath46 . @xmath47 can also be seen as a directed graph . given an intuitionistic kripke model @xmath48 and a state @xmath49 , the _ satisfaction relation for intuitionistic logics _ @xmath50 is defined as follows . @xmath51,s & & + @xmath51,s p & & s ( p ) , p , + @xmath51,s & & @xmath51,s @xmath51,s , + @xmath51,s & & @xmath51,s @xmath51,s , + @xmath51,s & & ns : , n , n a formula @xmath52 is _ satisfied _ by an intuitionistic kripke model @xmath53 in state @xmath54 if @xmath55 . a _ tautology _ is a formula that is satisfied by every intuitionistic kripke model . such formulas are also called _ valid_. from the monotonicity of @xmath41 and the definition of @xmath56 follows the monotonicity for every formula , i.e. for @xmath57 , @xmath58 and @xmath59 if @xmath60 , then @xmath61 . this paper examines the complexity of model checking problems for intuitionistic logics . .7em 1.5em by3em 0ex .2ex plus.1ex @xmath62 @xmath63 , where @xmath64 , @xmath53 is an intuitionistic kripke model , and @xmath54 is a state of @xmath53 @xmath65 ? we assume that formulas and intuitionistic kripke models are encoded in a straightforward way . this means , a formula is given as a text , and the graph @xmath47 of an intuitionistic kripke model is given by its adjacency matrix that takes @xmath66 bits . the set @xmath30 of formulas with at most one variable is partitioned into infinitely many equivalence is _ equivalent _ to @xmath67 if every state in every intuitionistic kripke model satisfies both or none formula . we write @xmath68 . ] classes @xcite . this was shown using the formulas that are inductively defined as follows ( see e.g.@xcite ) . we use @xmath69 for the only variable . @xmath70 the formulas @xmath71 are called _ rieger - nishimura formulas_. ( @xcite , cf . ( * ? ? ? * chap.6,thm.7))[thm : nish1 ] every formula in @xmath30 is equivalent to exactly one of the rieger - nishimura formulas . the function @xmath72 maps every formula to the index of its equivalent rieger - nishimura formula . we call this index _ rieger - nishimura index_. ( ) & = & \ { ccl ( i , ) , & & _ i + ( i , ) , & & _ i + ( 0 , ) , & & + ( 0 , ) , & & . in the following we analyse the complexity of @xmath72 . for @xmath64 let @xmath73 $ ] denote the equivalence class that contains @xmath52 . the equivalence classes of @xmath30 form a free heyting algebra over one generator ( for algebraic details see @xcite ) . this algebra is also called the _ rieger - nishimura lattice _ ( see fig . [ fig : rn - lattice ] ) . it is shown in @xcite that the lattice operations can be calculated using a big table look - up ( see appendix [ subsec : rn - lattice - operations ] ) . for @xmath74 , the binary lattice operators @xmath75 , @xmath76 and @xmath77 are defined as follows . @xmath78 \sqcap [ \beta ] = [ \alpha \wedge \beta]$ ] , @xmath78 \sqcup [ \beta ] = [ \alpha \vee \beta]$ ] , and @xmath78 \rightarrowtriangle [ \beta ] = [ \delta]$ ] , where @xmath79 $ ] is the largest element w.r.t . @xmath80 ( @xmath81 ) . ] with @xmath82,[\delta]\ } \sqsubseteq [ \beta]$ ] . is called the _ relative pseudo - complement operation_. ] we use the algebraic properties of @xmath8to give a lower bound on the length of formulas denotes the length of the formula @xmath83 , and it is the number of appearances of variables , connectives , and constants in @xmath83 . ] in the equivalence classes of @xmath30 ( lemma [ lem : formulalength ] ) , and to give an upper bound on the complexity of the problem to decide the rieger - nishimura index of a formula ( lemma [ lem : rnf_eqtest ] ) . let @xmath84 be the first element the integer of the @xmath85 pair . [ lem : formulalength ] for every @xmath64 it holds that @xmath86 , for a constant @xmath87 independent of @xmath52 . the proof relies on the following technical claim . let @xmath88 denote the @xmath19-th fibonacci number , @xmath89 , and @xmath90 for @xmath91 . ] . [ clm : fib ] let @xmath92 . then @xmath93 . _ proof of claim . _ for formulas @xmath94\cup[\top]$ ] it holds that @xmath95 . for formulas not in @xmath96\cup[\top]$ ] , we prove the claim by induction on the length of @xmath83 . the only relevant formula of length @xmath12 is @xmath97 . since @xmath98 and @xmath89 , the statement holds . for the induction step let @xmath99\cup[\top])$ ] with @xmath100 and @xmath101 with @xmath102 . then @xmath103 , and using the induction hypothesis we obtain @xmath104 . we have to distinguish the following cases . ( for the lattice operations see appendix [ subsec : rn - lattice - operations ] . ) * @xmath105 $ ] . * due to the fact that @xmath106\cup [ \top]$ ] it follows that @xmath107 . if @xmath108 , clearly @xmath109 $ ] and @xmath110 . with the induction hypothesis it follows that @xmath111 . otherwise if @xmath112 , it follows that @xmath113\cup [ { \upvarphi}_2 ] \cup [ { \uppsi}_1]$ ] and @xmath114 \cup [ { \upvarphi}_1]$ ] . hence @xmath115 . * @xmath116 $ ] . * this leads to @xmath108 and can be treated analogously to the case @xmath105 $ ] . * @xmath117 $ ] ( resp . * remember that @xmath119 $ ] , hence independent of the choice of @xmath120 it holds that @xmath78=[\gamma]$ ] ( resp . @xmath78=[\beta]$ ] and it follows that @xmath121 . * the remaining cases . * with the induction hypothesis it follows that @xmath122 . with respect to the rieger - nishimura lattice , we have to handle two cases . * * @xmath123 or @xmath124 . * * in this case it is not hard to see that @xmath125 . * * @xmath126 and @xmath127 . * * in this case it holds that one of the ranks of @xmath67 and @xmath128 needs to be @xmath129 and the other @xmath130 . ( see appendix [ subsec : rn - lattice - operations ] , for example @xmath131 respectively @xmath132 \rightarrowtriangle [ { \uppsi}_{k-2 } ] = [ { \upvarphi}_k]$ ] for @xmath133 . ) therefore it holds that @xmath134 . @xmath135 claim [ clm : fib ] shows @xmath136 . because of the exponential growth of the fibonacci numbers ( @xmath137 where @xmath138 denotes the golden ratio ) it follows that @xmath139 where @xmath87 is independent of @xmath52 . @xmath140 in order to analyse the complexity of the rieger - nishimura index computation , we define the following decision problem . .7em 1.5em by3em 0ex .2ex plus.1ex @xmath141 , where @xmath92 and @xmath142 is a rieger - nishimura index @xmath143 ? [ lem : rnf_eqtest ] @xmath144 is in @xmath145 . we form algorithm [ alg : eq_rn_formula ] based on the rieger - nishimura lattice of the equivalence classes of @xmath30 . the lattice and the lattice operations @xmath75 , @xmath76 and @xmath77 are described in appendix [ subsec : rn - lattice - operations ] . we can analogously define the lattice operations @xmath75 , @xmath76 and @xmath77 for the rieger - nishimura indices instead of the equivalence classes and @xmath146 . we set @xmath147 if @xmath78 \star [ \beta ] = [ \gamma]$ ] and @xmath148 . ] . the correctness of algorithm [ alg : eq_rn_formula ] is straightforward because the lattice operations for equivalence classes and indices are the same . with lemma [ lem : formulalength ] it follows that every variable value used in algorithm [ alg : eq_rn_formula ] can be stored in logarithmic space . the algorithm walks recursively through the formula and computes the index of every subformula once , hence running time is polynomial . all information that are necessary for recursion can be stored on the stack . therefore algorithm [ alg : eq_rn_formula ] can be implemented on a polynomial time logspace machine that uses an additional stack , i.e. a @xmath21-machine . @xmath140 a formula @xmath149 and a rieger - nishimura index @xmath142 * if * rnindex - calc@xmath150 * then * accept * else * reject * function * rnindex - calc(@xmath151 ) // returns a rieger - nishimura index similar as any formula can be represented by its index , intuitionistic kripke models can be represented , too . we give a construction of models the _ canonical models_that are also used to distinguish the formula equivalence classes ( theorem [ thm : rieger - nishimura ] ) . our definition differs a little bit from that in ( * ? ? ? * chap.6 , defi.5 ) . from theorems [ thm : rieger - nishimura ] and [ lem : hom_heytmodel ] it follows that every state @xmath54 in every intuitionistic kripke model @xmath53 over one variable has a unique corresponding canonical model @xmath152 in the sense that the state @xmath54 and the base state @xmath19 of @xmath152 satisfy exactly the same formulas . this was already shown in ( * ? ? ? * chap.6 , lemma 11 ) . further define a function @xmath153 that maps @xmath154 to @xmath19 . for @xmath155 , we define the canonical models @xmath156 as follows . w_n & : = & \{1,2, ,n-2}\{n } + & : = & \ { ( a , b ) a , b w_n , a = b a b+2 } + [ + 5px ] _ n(a ) & : = & , & n=2 + \{1 } , & see figure [ fig : heyting models ] for some examples . and @xmath157 ( reflexive and transitive edges are not depicted , @xmath158 is indicated by the double circle for state @xmath12 ) . ] @xmath159 and @xmath157 ( reflexive and transitive edges are not depicted , @xmath158 is indicated by the double circle for state @xmath12 ) . ] @xmath157 the formulas in @xmath30 can be distinguished using the canonical models as follows . ( @xcite , cf . * chap.6 , thm.8))[thm : rieger - nishimura ] for every @xmath155 and every @xmath160 it holds that : 1 . @xmath161 iff @xmath162 ( i.e. @xmath163 ) , and 2 . @xmath164 iff @xmath165 or @xmath166 ( i.e. @xmath167 ) . for analysing the complexity of the decision problem whether a canonical model is the corresponding model of a state of an arbitrary given intuitionistic kripke model we define a function @xmath153 . the function @xmath153 maps a given intuitionistic kripke model @xmath53 and state @xmath168 of @xmath53 to the index @xmath29 of the corresponding model @xmath169 . let @xmath170 be an intuitionistic kripke model and @xmath168 a state of @xmath53 . we define two abbreviations for @xmath171 . w_w & : = & \{vwwv } + w_w & : = & w_w \{w } the function @xmath153 is defined as follows . @xmath172 2 , & \mbox{~if } & \mbox{$w\not\in\zeta(a)$ and $ \forall v \in w_{w{\!\uparrow } } : v\not\in\zeta(a)$ } \\[3px ] 3 , & \mbox{~if } & \mbox{$w\not\in\zeta(a)$ and $ \forall v \in w_{w{\!\uparrow } } : { \mathpzc{h}}({\ensuremath{\mathcal{m}}},v ) \neq 2 $ and } \\ & & \mbox{$\exists u\in w_{w{\!\uparrow } } : { \mathpzc{h}}({\ensuremath{\mathcal{m}}},u)=1 $ } \\[3px ] n+2 , & \mbox{~if } & \mbox{$\forall v\in w_{w{\!\uparrow } } : { \mathpzc{h}}({\ensuremath{\mathcal{m}}},v ) \neq n+1 $ and}\\ & & \mbox{$\exists u_1,u_2 \in w_{w{\!\uparrow } } : { \mathpzc{h}}({\ensuremath{\mathcal{m}}},u_1)=n$ and $ { \mathpzc{h}}({\ensuremath{\mathcal{m}}},u_2)=n-1 $ } \\ \end{array } \right.\ ] ] we call @xmath173 the _ model index _ of @xmath168 in @xmath53 . the function @xmath153 is well defined because for every state @xmath168 it holds that @xmath174 . [ lem : hom_heytmodel][thm : hom_heytmodel ] let @xmath53 be an intuitionistic kripke model , @xmath168 a state of @xmath53 , and @xmath160 . then it holds that @xmath51,w _ k & & k ( @xmath51,w ) + @xmath51,w _ k & & k > ( @xmath51,w ) k = ( @xmath51,w)-1 . from theorem [ thm : rieger - nishimura ] follows that ( 1 ) is equivalent to the following claim . let @xmath53 be an intuitionistic kripke model and @xmath168 a state of @xmath53 . for every rieger - nishimura formula @xmath83 it holds that @xmath175 if and only if @xmath176 . _ proof of claim . _ we prove this by induction on the rank @xmath84 of @xmath83 . let @xmath170 be an intuitionistic kripke model , @xmath171 a state , and @xmath83 a rieger - nishimura formula . the case @xmath177 is clear . for the induction step we consider a formula @xmath83 with @xmath178 . we distinguish two cases . the case @xmath179 is clear because @xmath180 and the claim follows directly from the induction hypothesis . in the second case we have @xmath181 . & @xmath51,w _ k (= _ k-1 _ k-1 ) & ( 1 ) + & vw , w v : @xmath51,v _ k-1 @xmath51,v _ k-1 & ( 2 ) + & vw , w v : _ ( @xmath51,v),(@xmath51,v ) _ k-1 & + & _ ( @xmath51,v),(@xmath51,v ) _ k-1 & ( 3 ) + & xw_(@xmath51,w ) : _ x , x _ x , x _ k-1 & ( 4 ) + & xw_(@xmath51,w ) : _ ( @xmath51,w),x _ k-1 _ ( @xmath51,w),x _ k-1 & ( 5 ) + & _ ( @xmath51,w),(@xmath51,w ) _ k-1 _ k-1 (= _ k ) & ( 6 ) the equivalence between ( 1 ) and ( 2 ) is clear due to the definition of @xmath35 . from the induction hypothesis follows the equivalence between ( 2 ) and ( 3 ) . ( 3 ) and ( 4 ) are equivalent because @xmath182 . the definition of the canonical models , i.e. @xmath183 is a submodel of @xmath184 , causes the equivalence between ( 4 ) and ( 5 ) . the last equivalence between ( 5 ) and ( 6 ) comes from the definition of @xmath35 and the properties of @xmath184 . @xmath135@xmath140 we first define an @xmath0-hard graph problem , that is similar to the @xmath7-complete alternating graph accessibility problem @xcite , but has some additional simplicity properties . then we give a construction that transforms such a graph into an intuitionistic kripke model . this transformation is the basis for the reduction from the alternating graph accessibility problem to the model checking problem for @xmath8 . the alternating graph accessibility problem is shown to be @xmath7-complete in @xcite . we use the following restricted version of this problem that is very similar to boolean circuits with and- and or - gates ( and input - gates ) . an _ alternating slice graph _ @xcite @xmath185 is a directed bipartite acyclic graph with a bipartitioning @xmath186 , and a further partitioning @xmath187 ( @xmath188 _ slices _ , @xmath189 if @xmath190 ) where @xmath191 and @xmath192 , such that @xmath193 i.e. all edges go from slice @xmath194 to slice @xmath195 ( for @xmath196 ) . all nodes excepted those in the last slice @xmath197 have a positive outdegree . nodes in @xmath198 are called _ existential _ nodes , and nodes in @xmath199 are called _ universal _ nodes . alternating paths from node @xmath200 to node @xmath201 are defined as follows by the property @xmath202 . 1 ) : : @xmath203 holds for all @xmath204 2a ) : : for @xmath205 : @xmath202 iff @xmath206 2b ) : : for @xmath207 : @xmath202 iff @xmath208 the problem is similar to the alternating graph accessibility problem , but for the restricted class of alternating slice graphs . .7em 1.5em by3em 0ex .2ex plus.1ex @xmath209 , where @xmath210 is an alternating slice graph with slices @xmath211 , and @xmath212 , @xmath213 does @xmath214 hold ? similarly as the alternating graph accessibility problem , is @xmath7-complete ( * ? ? ? * lemma 2 ) . the following technical lemma is not hard to prove . [ lemma : asagap - ac1-hard ] for every set @xmath215 in ( logspace - uniform ) @xmath216 exists a function @xmath217 that maps instances @xmath200 of @xmath215 to instances @xmath218 of @xmath219 and satisfies the following properties . 1 . @xmath217 is computable in logspace . @xmath220 is an alternating slice graph of logarithmic depth ; i.e. if @xmath220 has @xmath19 nodes , then it has @xmath221 slices . 3 . for all instances @xmath200 of @xmath215 holds : @xmath222 if and only if @xmath223 . essentially , the function @xmath217 constructs the @xmath0circuit @xmath224 with input @xmath200 , and transforms it to an alternating slice graph @xmath220 . the goal node @xmath225 represents exactly the bits of @xmath200 that are @xmath12 . the start node @xmath226 corresponds to the output gate of @xmath224 , and @xmath227 expresses that @xmath224 accepts input @xmath200 . if we consider @xmath228 as the subset of where the slice graphs have logarithmic depth , this lemma would express that @xmath228 is @xmath216-hard under logspace reductions . our hardness results rely on a transformation of instances @xmath229 of to intuitionistic kripke models @xmath230 . let @xmath229 be an instance of for the slice graph @xmath231 with the @xmath188 slices @xmath232 and @xmath233 . for every @xmath234 , we construct two sets of new states w_i^ & : = & \{v^ vv_i } + w_i^ & : = & \{v^ vv_i } and let w & : = & _ i=0^m-1 ( w_i^ w_i^ ) . every edge @xmath235 from @xmath236 is transformed to an edge @xmath237 from an @xmath238-node to an @xmath239-node , and every @xmath239-node has an edge to its corresponding @xmath238-copy . this yields the set of edges e & : = & \{(u^,v^)(u , v)e_g } \{(v^,v^ ) vv_v _ } . let @xmath240 be the graph obtained in this way from @xmath241 . if we consider those nodes @xmath242 as @xmath243-nodes ( resp . @xmath244-nodes ) that come from nodes @xmath245 ( resp . @xmath246 ) , then @xmath247 if and only if @xmath248 . next , we add the nodes of the canonical model @xmath249 to @xmath250 as follows . add the nodes @xmath12 and @xmath13 to @xmath251 , the nodes @xmath252 and @xmath253 to @xmath254 , the nodes @xmath255 and @xmath256 to @xmath257 etc . formally , for @xmath258 , let s_i^ & : = & w_i^ \{4 i+1 , 4 i+2 } , + s_i^ & : = & w_i^ \{4 i+3 , 4 i+4 } + s_m-1^ & : = & w_m-1^ \{4m}. the set of states for our model is now u & : = & _ i=0^m-1 ( s_i^ s_i^ ) . note that @xmath259 is still a slice graph with slices @xmath260 . we yet have no edges that connect to nodes from the canonical model . first we add only those edges between these nodes that do not disturb the `` slice graph '' property , namely h & : = & \{(i , i-2 ) i\{3,4, ,4m-2}\{4 m } } + & & \{(i , i-3 ) i\{4,6, ,4m-2,4m}}. note that @xmath261 consists of the edges from @xmath262 that give the canonical model its typical structure , i.e. @xmath263 is the transitive closure of @xmath261 . second we add edges from every node in @xmath264 to a node in the neighboured slice @xmath265 from @xmath262 depending on whether @xmath266 or @xmath267 if @xmath267 and vice versa . ] . t _ & : = & \ { ( u,4i+2 ) uw^_i , i=0,1,2, ,m-1 } + t _ & : = & \ { ( u,4i-1 ) u w^_i , i=1,2, ,m-1 } notice that @xmath268 is still a slice graph with the slices mentioned above . it is depicted in figure [ fig : agap - example13 ] ) . an intuitionistic kripke model must be transitive and reflexive . the reduction function that transforms alternating slice graphs to intuitionistic kripke models must be computable in logarithmic space . within this space bound we can not compute the transitive closure of a graph . therefore , we make the graph transitive with brute force . we add all edges that jump over at least one slice we call these edges _ pseudotransitive_. p & : = & _ i = m-1 ^ 1 finally , we need to add all reflexive edges . t & : = & \{(u , u ) uu } notice that the subgraph induced by the states of the canonical model @xmath262 that consists of the edges in @xmath261 plus the pseudotransitive and the reflexive edges , is exactly @xmath262 . eventually , the relation @xmath39 for our model is r & : = & e h t _ pt , and the valuation function for our model is \(a ) & : = & \{t^,1 } , where @xmath269 is the copy of the goal node @xmath270 in @xmath251 , and @xmath271 is the node from @xmath262 . this yields the intuitionistic kripke model @xmath272 . an example of an instance @xmath209 and the corresponding intuitionistic kripke model @xmath273 constructed from it can be seen in figure [ fig : agap - example13 ] . ( left ) and the resulting intuitionistic kripke model @xmath273 ( right ) ; both the states in @xmath274 are drawn doubly ; pseudotransitive and reflexive edges in @xmath273 are not depicted . the value at state @xmath200 denotes its model index @xmath275 . for states in @xmath276 , their names and their model indices coincide . states @xmath277 and @xmath278 for which @xmath279 holds in @xmath241 are coloured grey . , title="fig : " ] ( left ) and the resulting intuitionistic kripke model @xmath273 ( right ) ; both the states in @xmath274 are drawn doubly ; pseudotransitive and reflexive edges in @xmath273 are not depicted . the value at state @xmath200 denotes its model index @xmath275 . for states in @xmath276 , their names and their model indices coincide . states @xmath277 and @xmath278 for which @xmath279 holds in @xmath241 are coloured grey . , title="fig : " ] the states from the canonical model were added to the slice graph in order to obtain control over the model indices of the other states ( w.r.t . the model @xmath273 ) . our controlling tool is the function @xmath153 which is defined in the previous section . it maps every state of an intuitionistic kripke model to its model index . this is described by proposition [ claim : h - werte - der - slices ] and proposition [ claim : knoteneigenschaften ] . [ claim : h - werte - der - slices ] for every @xmath234 and every @xmath280 holds @xmath281 we prove this by induction on the slices . for the base case we consider @xmath282 , where we have @xmath283 if @xmath284 , and @xmath285 if @xmath286 , and therefore @xmath287 . for the induction step , we consider the remaining slices . for @xmath288 , we have @xmath289 and @xmath290 for some @xmath291 . by the induction hypothesis it follows that @xmath292 for all @xmath293 . by the definition of @xmath153 it follows that @xmath294 . for @xmath295 , we have @xmath296 . by the induction hypothesis we know that for all @xmath290 with @xmath297 holds @xmath298 , and @xmath299 for all @xmath300 . now , if for some @xmath301 holds @xmath302 , then @xmath278 has successors with model indices @xmath303 and @xmath304 and by the definition of @xmath153 it follows that @xmath305 . otherwise , for all @xmath301 holds @xmath306 , and @xmath278 has no successor with model index @xmath303 but successors with model indices @xmath307 and @xmath304 . by the definition of @xmath153 it now follows that @xmath308 . @xmath140 [ claim : knoteneigenschaften ] for every @xmath234 and every @xmath280 holds : 1 . if @xmath29 is even ( @xmath244 slice ) : + @xmath279 if and only if @xmath308 , and + @xmath279 if and only if @xmath309 , + 2 . if @xmath29 is odd ( @xmath243 slice ) : + @xmath279 if and only if @xmath305 , and + @xmath279 if and only if @xmath310 . we prove this proposition by induction on @xmath29 . the initial step for @xmath311 follows directly from the definition of @xmath273 . now for the induction step . consider @xmath280 for even @xmath29 ( @xmath244 slice ) . & _ g(v , t ) & ( 1 ) + & wv_i-1 , ( v , w ) e_g : _ g(w , t ) & ( 2 ) + & w^w^_i-1 , ( v^ , w^)r : ( @xmath51_g , w^)=4i-2 & ( 3 ) + & \{(@xmath51_g , u ) uu_v^out } = \{1,2, ,4i-1 } & ( 4 ) + & ( @xmath51_g , v^)=4i+1 & ( 5 ) \(1 ) and ( 2 ) are equivalent by the definition of @xmath312 . the equivalence of ( 2 ) and ( 3 ) comes from the construction of @xmath273 and the induction hypothesis . to show the equivalence of ( 3 ) and ( 4 ) we prove both the directions separately . first we show ( 3 ) @xmath313 ( 4 ) . because of ( 3 ) there is no @xmath314 with @xmath315 . if @xmath316 , then ( 4 ) follows directly . for @xmath307 it follows directly from ( 3 ) . if @xmath317 , then @xmath318 . because of ( 3 ) , it is not possible that @xmath319 . and @xmath320 is inconsistent with proposition [ claim : h - werte - der - slices ] . hence @xmath321 . for the second direction , ( 4 ) @xmath313 ( 3 ) , assume that there is some @xmath314 with @xmath322 . then from proposition [ claim : h - werte - der - slices ] it follows that @xmath323 but this is inconsistent with ( 4 ) . hence ( 3 ) and ( 4 ) are equivalent . ( 4 ) equivalent ( 5 ) by the construction of @xmath273 and the definition of @xmath153 . by proposition [ claim : h - werte - der - slices ] we know that @xmath324 . remind that @xmath277 has @xmath278 and @xmath325 as its direct successors and @xmath308 . therefore , @xmath279 if and only if @xmath326 , where the latter is equivalent @xmath309 . finally , we consider @xmath280 for odd @xmath29 ( @xmath243 slice ) . & _ g(v , t ) & ( 1 ) + & wv_i-1 , ( v , w)e_g : _ g(w , t ) & ( 2 ) + & w^w^_i-1 , ( v^ , w^)r : ( @xmath51_g , w^)=4i & ( 3 ) + & \{4i-1,4i}\{(@xmath51_g , u ) uu_v^}\{1,2, ,4i } & ( 4 ) + & ( @xmath51_g , v^)=4i+2 & ( 5 ) \(1 ) and ( 2 ) are equivalent by the definition of @xmath312 . the equivalence of ( 2 ) and ( 3 ) comes from the construction of @xmath273 and the induction hypothesis . as in the case above ( @xmath29 is even ) the equivalence of ( 3 ) and ( 4 ) follows from the construction of @xmath273 and proposition [ claim : h - werte - der - slices ] . ( 4 ) equivalent ( 5 ) by the construction of @xmath273 and the definition of @xmath153 . by proposition [ claim : h - werte - der - slices ] we know that @xmath324 . remind that @xmath277 has @xmath278 and @xmath325 as its direct successors . therefore , @xmath279 if and only if @xmath327 , where the latter is equivalent @xmath310 . @xmath140 let @xmath328 denote the function that maps instances @xmath329 of @xmath219 to intuitionistic kripke models @xmath330 as described above . the following properties of @xmath328 are easy to verify . [ transformation - lemma ] 1 . @xmath328 is logspace computable . if @xmath329 for an alternating slice graph @xmath241 with @xmath19 nodes and @xmath331 slices , then @xmath332 is an intuitionistic kripke model with @xmath333 states and depth @xmath334 . we will use @xmath328 as part of the reduction functions for our hardness results . [ sec : upper bounds ] our first result states that the calculation of the model index of an intuitionistic kripke model is @xmath7-complete . it is already @xmath7-complete to decide the last bit of this model index . [ thm : hm - phard ] the following problems are @xmath7-complete . 1 . given an intuitionistic kripke model @xmath53 and a state @xmath168 , decide whether @xmath173 is even . 2 . given an intuitionistic kripke model @xmath53 , a state @xmath168 , and an integer @xmath29 , decide whether @xmath335 . in order to show the @xmath7-hardness of the problems , we give a reduction from the @xmath7-hard problem . from an instance @xmath209 of where @xmath241 is an alternating slice graph with @xmath188 slices , construct @xmath336 . then @xmath337 ( proposition [ claim : h - werte - der - slices ] ) , and @xmath214 if and only if @xmath338 ( proposition [ claim : knoteneigenschaften ] ) . therefore , @xmath339 if and only if @xmath340 is even respectively @xmath338 . for every intuitionistic kripke model @xmath341 it holds that @xmath342 . to decide for a given intuitionistic kripke model @xmath53 , a state w of @xmath53 , and an integer @xmath19 the problem `` does @xmath343 hold ? '' is in @xmath344}$ ] . the function @xmath153 can be implemented according to its definition straightforwardly as a logarithmically space bounded alternating algorithm . it requires an alternation depth of at most @xmath19 due to the construction of @xmath153 . using that @xmath345}}\xspace}}$ ] @xcite then it follows that both problems are in @xmath7 . @xmath140 in the construction of the above proof , the decision whether @xmath338 is the same as to decide whether @xmath346 , for the rieger - nishimura formula @xmath347 ( theorems [ thm : rieger - nishimura ] and [ lem : hom_heytmodel ] ) . unfortunately , the length of @xmath347 is exponential in @xmath188 ( lemma [ lem : formulalength ] ) , and therefore the mapping from @xmath209 ( with @xmath188 slices ) to the model checking instance @xmath348 can not in general be performed in logarithmic space . but if the depth @xmath188 of the slice graph is logarithmic , the respective formula @xmath347 has polynomial size only and the reduction works in logarithmic space . [ thm : ipc1-is - aci - hard ] the model checking problem for @xmath349 is @xmath0-hard . let @xmath350 be in @xmath216 . by lemma [ lemma : asagap - ac1-hard ] there exists a logspace computable function @xmath351 such that for all instances @xmath200 of @xmath350 , @xmath352 if and only if @xmath353 , where @xmath354 for an alternating slice graph @xmath220 with @xmath355 nodes and @xmath356 slices . the following function @xmath357 reduces @xmath350 to the model checking problem for @xmath8 . r(x ) & = & _ 4m_x+2 , g(f_b(x ) ) , s_x^ _ @xmath357 can be computed in logspace . _ since @xmath351 is logspace computable , it follows that @xmath358 and @xmath359 can be computed in logspace . the rieger - nishimura formula @xmath360 can also be computed in logspace , because @xmath361 is logarithmic in @xmath362 and therefore @xmath360 has length polynomial in @xmath362 . _ @xmath350 logspace reduces to the model checking problem for @xmath8via the reduction function @xmath357 . _ by propostion [ claim : knoteneigenschaften ] we have that @xmath363 if and only if @xmath364 . by the properties of the rieger - nishimura formulas ( theorem [ thm : rieger - nishimura ] ) this is equivalent to @xmath365 . this shows the correctness of the reduction . @xmath140 in the following theorem we show an upper bound for the @xmath8model checking problem . [ thm : ac1-algorithm ] the model checking problem for @xmath8is in @xmath0 . first we show that algorithm [ alg : ac1_mc ] decides the model checking problem and then we analyse its complexity . we show that algorithm [ alg : ac1_mc ] accepts the input @xmath366 if and only if @xmath367 . informally speaking algorithm [ alg : ac1_mc ] accepts the input if and only if @xmath368 and the model index @xmath369 of @xmath54 in @xmath53 match according to theorem [ lem : hom_heytmodel ] . instead of computing the equivalent rieger - nishimura formula , algorithm [ alg : ac1_mc ] only calculates its rieger - nishimura index . this is done in lines 1 and 2 . the trivial cases are handled in lines 3 and 4 . from theorem [ lem : hom_heytmodel ] we know for an arbitrary rieger - nishimura formula @xmath370 with @xmath371 the following . either @xmath372 and it holds that @xmath373 if and only if @xmath374 . this is checked in line 6 . or @xmath375 and it holds that @xmath376 or @xmath377 if and only if @xmath374 . this is checked in line 9 . if @xmath378 , then it holds that @xmath379 ( theorems [ thm : rieger - nishimura ] and [ lem : hom_heytmodel ] ) . in the following , we estimate the complexity of algorithm [ alg : ac1_mc ] it gets @xmath380 as input . in line 1 algorithm [ alg : ac1_mc ] guesses a rieger - nishimura index @xmath381 . the decision in line 2 whether @xmath382 can be done with the resources of @xmath21(lemma [ lem : rnf_eqtest ] ) . to decide for a given intuitionistic kripke model @xmath53 , a state w of @xmath53 , and an integer @xmath19 the problem `` does @xmath343 hold ? '' is in @xmath344}$ ] . the function @xmath153 can be implemented according to its definition straightforwardly as a logarithmically space bounded alternating algorithm . it requires an alternation depth of at most @xmath19 due to the construction of @xmath153 . hence the decision in line 6 ( resp . line 9 ) whether @xmath383 ( resp . @xmath384 ) can be done with @xmath357 ( resp . @xmath385 ) alternations . since @xmath357 is at most about @xmath386 ( lemma [ lem : formulalength ] ) , these decisions can be done with at most @xmath387 alternations . during the complete computation , the algorithm only needs to store a constant number of rieger - nishimura indices and model indices . according to lemma [ lem : formulalength ] and the fact that @xmath388 , algorithm [ alg : ac1_mc ] requires during the alternations logarithmic space . since @xmath389}}\xspace}}$ ] , we obtain the desired upper bound . @xmath140 a formula @xmath390 , an intuitionistic kripke model @xmath53 and a state @xmath54 superintuitionistic propositional logics are logics that have more valid formulas than @xmath6 . in this sense , classical propositional logic is a superintuitionistic logic , since it can be obtained as the closure under substitution and modus ponens of the tautologies from @xmath6plus @xmath9 as additional axiom . a well - studied superintuitionistic logic is @xmath10@xcite that results from adding the weak law of the excluded middle @xmath391 to @xmath6 . semantically , the intuitionistic kripke models for @xmath10are restricted to those intuitionistic kripke models @xmath392 where @xmath393 is a directed preorder . whereas @xmath30 over preorders has infinitely many equivalence classes of formulas , @xmath30 over directed preorders has only 7 equivalence classes represented by the rieger - nishimura formulas @xmath394that can be distinguished using the first 3 canonical models @xcite . this follows from @xmath395 . the function @xmath153 can be implemented for such models as an alternating turing machine that runs in logarithmic time , if the function value is fixed to a finite range that in this case is @xmath396independent of the input . for @xmath397 , the rieger - nishimura index of the formulas also has a finite range ( as mentioned above ) . therefore , it can be calculated by an alternating turing machine that runs in logarithmic time similar to the machine presented by buss @xcite that calculates the value of a boolean formula . instead of the boolean values @xmath398 and @xmath12 , here we have @xmath399 different rieger - nishimura indices . the rules how the index of a formula can be calculated from the indices of its subformulas and the connective , follow directly from the rieger - nishimura lattice operations see appendix [ subsec : rn - lattice - operations ] . if the indices are bound to a finite range , this big table yields an even bigger but finite table without index - variables . for example , the equivalence @xmath400 for all @xmath155 induces the three equivalences @xmath401 , @xmath402 , and @xmath403 for @xmath397 . this yields alternating logarithmic - time ( @xmath404 ) as upper bound for the validity problem for @xmath397 . there are infinitely many superintuitionistic logics ( with one variable ) that can be obtained by adding any not valid formula as axiom to @xmath8 . for example , if we add a formula equivalent to @xmath405 , then the superintuitionistic logic obtained has finitely many equivalence classes represented by @xmath406 . with similar arguments as for @xmath397 we can conclude that the model checking problems of these logics all are in @xmath1 . moreover , the formula value problem for boolean formulas without variables is @xmath1-hard @xcite . intuitionistic formulas without variables have the same values , if they are interpreted as classical boolean formulas . this means , the semantics of @xmath35 is the same for boolean formulas and for intuitionistic formulas without variables . therefore , the model checking problem for any superintuitionistic logic without variables is @xmath1-hard , too . the validity problem for superintuitionistic logic has the same complexity , since in order to decide whether a formula with one variable is valid it suffices to know its rieger - nishimura index . we consider computational problems that appear with intuitionistic propositional logic without variables and with one variable . we characterize the complexity of model checking for intuitionistic logic . [ theo : conclusion ] 1 . [ part1 ] the model checking problem for @xmath407 is @xmath1-complete . [ part2 ] the model checking problem for @xmath8 is @xmath0-complete . part([part1 ] ) follows from the fact that an intuitionistic formula that contains constants @xmath36 and @xmath408 but no variables can be evaluated like a boolean formula , whose evaluation problem is @xmath1-complete @xcite independently of the number of variables . part ( [ part2 ] ) follows from theorems [ thm : ipc1-is - aci - hard ] and [ thm : ac1-algorithm ] . it shows a difference between @xmath8and its modal companion @xmath409with one variable , for which the model checking problem is @xmath7-complete @xcite . intuitionistic logic with one variable turns out to be very interesting . there are infinitely many equivalence classes of formulas , and according to lemma [ lem : formulalength ] even the sequence of smallest formulas of these equivalence classes has an exponential growth with respect to the length of the formulas . such a fast growing sequence seems to appear rarely in `` natural '' problems , and it is a key ingredient for the @xmath0-completeness of the model checking problem . intuitionistic logic with one variable is strongly related to free heyting algebras with one generator . since heyting algebras are generalizations of boolean algebras , it would be interesting to investigate whether the difference between @xmath1and @xmath0is related to that between boolean algebras and heyting algebras . [ theo : super - int - mc ] the model checking problem for every superintuitionistic logic with one variable is @xmath1-complete . this follows from the discussion in section [ sec : superint1 ] . it is interesting to notice that the complexity results for @xmath6and for @xmath410 with at least two variables are the same for the model checking problem @xcite . but for the fragments with one variable , the complexity of @xmath8is higher than that of @xmath397 . the fragments of @xmath6with a restricted number of variables and @xmath35 as only connective have finitely many equivalence classes of formulas and models @xcite . the equivalence class of a given formula can be obtained with the resources of @xmath411 , using a technique from buss @xcite . this might indicate an upper bound lower than @xmath7for the model checking problem . for the implicational fragment with at most one variable , @xmath411-completeness follows from theorem [ sec : superint1 ] . but a general result for an arbitrary number of variables is open . for the validity problem we obtain the following results . [ thm : validity_conclusion ] 1 . [ cpart1 ] the validity problem for every superintuitionistic logic with one variable is @xmath1-complete . [ cpart2 ] the validity problem for @xmath349 is in @xmath412 . part ( [ cpart1 ] ) follows from the discussion in section [ sec : superint1 ] . part ( [ cpart2 ] ) is from svejdar @xcite and lemma [ lem : rnf_eqtest ] . the exact complexity of the validity problem for @xmath8is open . it is interesting to notice that superintuitionistic logics with one variable all have lower complexity than @xmath8 , whereas for superintuitionistic logics with two variables already @xmath10reaches the same complexity as @xmath6(follows from rybakov @xcite ) . if we consider other problems related to kripke models for @xmath8that are not `` out braked '' by a very fast growing part of the input , the complexity jumps up to @xmath7-completeness , as shown in theorem [ thm : hm - phard ] . model checking for @xmath8also gets @xmath7-hard if the instances @xmath366 allow the formula @xmath413 to be represented as a graph . this holds even for formulas without variables , and therefore it also holds for all superintuitionistic logics . if formulas are represented as graphs , the sequence of smallest representatives of the equivalence classes of @xmath8does not have exponential growth anymore . moreover , the calculation of the rieger - nishimura index gets @xmath7-hard . [ theo : p - complete - formulas ] if the formulas are represented as graphs , the following problems are @xmath7-complete : 1 . [ ppart1 ] the model checking problem for @xmath8 , 2 . [ ppart1a ] the model checking problem for every superintuitionistic logic with one variable , 3 . [ ppart2 ] the validity problem for @xmath8 , and 4 . [ ppart2a ] the validity problem for every superintuitionistic logic with one variable . parts ( [ ppart1 ] ) and ( [ ppart1a ] ) contrast the different upper bounds @xmath1and @xmath0 for the standard encodings of formulas ( theorem [ theo : super - int - mc ] resp . theorem [ theo : conclusion ] ) . parts ( [ ppart2 ] ) and ( [ ppart2a ] ) contrast the complexity of the validity problems for the logics under consideration ( theorem [ thm : validity_conclusion ] ) . * acknowledgements . * the authors thank vitek svejdar , heribert vollmer , and thomas schneider for helpful discussions . * remark . * this work is an extended version of @xcite . theorem 15 in @xcite and its proof can be found in ( * ? ? ? * theorem 3.6 ) . 10 m. beaudry and p. mckenzie . circuits , matrices , and nonassociative computation . , 50(3):441455 , 1995 . s. r. buss . the boolean formula value problem is in alogtime . in _ proc . 19th stoc _ , pages 123131 . acm press , 1987 . s. r. buss . algorithms for boolean formula evaluation and for tree contraction . in _ arithmetic , proof theory , and computational complexity _ , pages 96115 . oxford university press , 1993 . a. k. chandra , d. kozen , and l. j. stockmeyer . , 28:114133 , 1981 . s. a. cook . characterizations of pushdown machines in terms of time - bounded computers . 18:418 , 1971 . s. a. cook . the complexity of theorem proving procedures . in _ proc . 3rd stoc _ , pages 151158 . acm press , 1971 . s. a. cook . a taxonomy of problems with fast parallel algorithms . , 64:222 , 1985 . g. r. r. de lavalette , a. hendriks , and d. h. de jongh . intuitionistic implication without disjunction . . to appear , available at http://dx.doi.org/10.1093/logcom/exq058 m. dummett and e. lemmon . modal logics between s4 and s5 . , 14(24):250264 , 1959 . d. m. gabbay . . d.reidel , dordrecht , boston , london , 1981 . p. t. johnstone . . cambridge university press , cambridge , 1982 . d. makinson . there are infinitely many diodorean modal functions . , 31(3):406408 , 1966 . m. mundhenk and f. wei . the complexity of model checking for intuitionistic logics and their modal companions . in _ proc . of rp 2010 _ , volume 6227 of _ lncs _ , pages 146160 . springer , 2010 . m. mundhenk and f. wei . intuitionistic implication makes model checking @xmath414-hard . , abs/1107.1963v1 , 2011 . m. mundhenk and f. wei . the model checking problem for intuitionistic propositional logic with one variable is @xmath216-complete . in _ proc . 28th stacs _ , volume 9 of _ lipics _ , pages 368379 . schloss dagstuhl - leibniz - zentrum fr informatik , 2011 . i. nishimura . on formulas of one variable in intuitionistic propositional calculus . , 25:327331 , 1960 . w. l. ruzzo . on uniform circuit complexity . , 21:365383 , 1981 . m. n. rybakov . complexity of intuitionistic and visser s basic and formal logics in finitely many variables . in _ papers from the 6th conference on `` advances in modal logic '' _ , pages 393411 . college publications , 2006 . r. statman . intuitionistic propositional logic is polynomial - space complete . , 9:6772 , 1979 . v. svejdar . on the polynomial - space completeness of intuitionistic propositional logic , 42(7):711716 , 2003 . v. svejdar . the tautology problem for ipc@xmath415 is in space @xmath416 , 2009 . personal communication . a. urquhart . implicational formulas in intuitionistic logic . , 39(4):661664 , 1974 . d. van dalen . . springer , berlin , heidelberg , 4th edition , 2004 . h. vollmer . . texts in theoretical computer science . springer verlag , berlin heidelberg , 1999 . let @xmath417 $ ] denote the equivalence class that contains @xmath413 , for being @xmath418 . the equivalence classes of @xmath30 form a free heyting algebra over one generator ( see @xcite ) . this algebra is also called the _ rieger - nishimura lattice _ ( see fig . [ fig : rn - lattice ] ) , and is given by @xmath419 whereas @xmath69 denotes the only one variable that occurs in the formulas of @xmath30 . the induced partial order is denoted by @xmath80 ( @xmath81 ) . for @xmath74 , the binary lattice operators @xmath75 , @xmath76 and @xmath77 are defined as follows . @xmath78 \sqcap [ \beta ] = [ \alpha \wedge \beta]$ ] , @xmath78 \sqcup [ \beta ] = [ \alpha \vee \beta]$ ] , and @xmath78 \rightarrowtriangle [ \beta ] = [ \delta]$ ] , where @xmath79 $ ] is the largest element w.r.t . @xmath80 with @xmath82,[\delta]\ } \sqsubseteq [ \beta]$ ] . is called the _ relative pseudo - complement operation_. ] in @xcite the following properties of the operations of the rieger - nishimura lattice ( see figure [ fig : rn - lattice ] ) are shown . we describe these properties as equivalences of rieger - nishimura formulas . this is very similar to ( * ? ? ? * chap.6,thm.7 ) . for example because of @xmath420 it holds that @xmath421 \rightarrowtriangle [ { \uppsi}_n ] = [ { \upvarphi}_{n+2}]$ ] .

there are several plasma models intermediate in complexity between ideal magnetohydrodynamics ( mhd ) and two - fluid theory , with hall and extended mhd being two important examples . in this paper we investigate several aspects of these theories , with the ultimate goal of deriving the noncanonical poisson brackets used in their hamiltonian formulations . we present fully lagrangian actions for each , as opposed to the fully eulerian , or mixed eulerian - lagrangian , actions that have appeared previously . as an important step in this process we exhibit each theory s two advected fluxes ( in analogy to ideal mhd s advected magnetic flux ) , discovering also that with the correct choice of gauge they have corresponding lie - dragged potentials resembling the electromagnetic vector potential , and associated conserved helicities . finally , using the euler - lagrange map , we show how to derive the noncanonical eulerian brackets from canonical lagrangian ones .

1512.00942

ideal magnetohydrodynamics ( mhd ) , that reliable workhorse of plasma physics , has long been cast into noncanonical hamiltonian form @xcite . so has the theory from which it is usually derived , the two - fluid model @xcite . there are many advantages to a hamiltonian form : the discovery and classification of invariants ; the development of numerical algorithms that automatically preserve such invariants ; easily finding the equations of motion in curved coordinates ; conducting equilibrium and stability analysis . however , there are many theories intermediate in complexity between two - fluid theory and ideal mhd ; kimura and morrison @xcite describe eleven of them . two are particularly important : hall mhd , which accounts for the difference between the motion of the two species in a typical plasma , and lst s extended mhd @xcite , which includes all terms of first order in the ratio @xmath0 of species masses in the derivation from two - fluid electron - ion theory . recently , yoshida and hameiri @xcite formulated a noncanonical poisson bracket for hall mhd , and shortly later abdelhamid , kawazura and yoshida did so for extended mhd @xcite ; however , as often happens when working with hamiltonian systems , they had to simply posit a bracket and prove it satisfied all the desired attributes , such as antisymmetry and the jacobi identity . we will show how to derive these brackets , starting from action principles for each theory . these action principles have a long and distinguished history in fluids , originating with the work of lagrange in the 18th century @xcite . the action principle formulation has also been employed in plasma physics since the second half of the 20th century , as evident from the works of @xcite . for ideal magnetohydrodynamics ( mhd ) , the first action principle formulation was provided by newcomb in @xcite , shortly followed by other works in the same area @xcite . for extended mhd , an eulerian action principle was proposed by @xcite which was subsequently generalized to a eulerian - lagrangian action by @xcite . for recent overviews of action principle formulations of plasma models , we refer the reader to @xcite . the noncanonical hamiltonian formulations for these models can be found in the works of @xcite . in principle there is an easy process to construct a noncanonical poisson bracket , which goes as follows . first , construct an action whose variations give the correct equations of motion in some coordinate system . from this tangent - space action principle , derive a hamiltonian function via a legendre transform , and produce the corresponding phase - space action principle using the canonical momenta of the original action . the poisson bracket accompanying the phase - space action will be canonical . then simply change coordinates in order to produce the desired noncanonical bracket . this procedure is , indeed , what we use , but there are many complications along the way . to begin with , the canonical bracket for fluid theories requires lagrangian coordinates : those in which every fluid element is given a distinct label , and the equations of motion are expressed for a given labelled element , despite the fact that the element is changing position . however , fully lagrangian actions for hall and extended mhd have not yet been given . the closest are the mixed lagrangian - eulerian actions of ref . @xcite , `` eulerian '' coordinates being ones that observe fluid quantities at a fixed point rather than following a given element . in this paper we present fully lagrangian actions . another complication arises because the legendre transform fails to be invertible for either theory , and an expression for one of the velocities in terms of the phase - space variables must be inserted by hand . finally , the euler - lagrange map producing the noncanonical brackets requires prior knowledge of the generalized vorticities advected by the theories , so we must devote some time to their discovery and elucidation . these generalized vorticities turn out to be crucial to the structure of every hamiltonian mhd model . there are @xmath1 advected vorticities for a theory with @xmath1 distinct charged species , @xmath1 being two in our case . `` advected '' in this case means that the flux elements defined by the vorticities are carried along with the fluid , their corresponding two - forms obeying a lie - dragging equation . for ideal mhd , both generalized vorticities collapse down to the same quantity , the magnetic field , which is advected by the fluid velocity . for hall mhd , one generalized vorticity is the magnetic field , whose fluxes are carried along with the electron velocity , and the other is the magnetic field plus kinetic vorticity @xmath2 , advected by the ion velocity @xcite . for extended mhd they turn out to be almost the same , but differing from the hall mhd ones by terms of order @xmath0 in the curl of the current . both our actions and our derivations of noncanonical brackets would be impossible , but for the fact that we can eliminate the eulerian magnetic field terms in both the action and the euler - lagrange map in favor of fully lagrangian terms , an elimination wholly dependent on the existence of these advected fluxes . before moving on , we note a few ways in which our present work can be readily extended along the lines of past works that utilized these methods . one can incorporate finite larmor radius ( flr ) effects , such as the braginskii gyroviscosity @xcite . flr effects for reduced mhd @xcite and generalized fluid models @xcite were implemented via an action principle formulation , and evidently a similar treatment can be undertaken via our extended mhd action principles . further extensions include stability analyses @xcite , the systematic derivation of reduced extended mhd models ( with potential applications in collisionless reconnection ) , linear and nonlinear waves via a lagrangian approach @xcite , and mhd - like models ( ideal , hall , or extended ) for quasineutral plasmas with more than two charged species . the paper is organized as follows . section [ subsec : hamsys ] reviews the basic framework of hamiltonian systems , while section [ subsec : hammhd ] presents a specific example of that framework for ideal mhd , allowing comparisons with the related , but more complex constructions for hall and extended mhd . we begin our new material by focusing on the simpler theory , hall mhd , in section [ sec : hallmhd ] . section [ subsec : hallflux ] lays out the needed terminology and facts about hall mhd , which are then used in section [ subsec : hallactions ] to construct both tangent - space and phase - space actions . section [ subsec : hallliegauge ] is an interesting digression in which we lay out a useful gauge , producing not only advected fluxes but corresponding advected one - forms . finally , we reach the goal which motivates our entire paper , the derivation in section [ subsec : hallelmap ] of the noncanonical bracket . this derivation is carried out in more algebraic detail than might be necessary , in light of its unfamiliarity to many readers . we then pivot to extended mhd in section [ sec : exmhd ] , starting with a derivation of its fluxes in section [ subsec : exmhdflux ] . we give its actions in section [ subsec : exmhdaction ] , and derive its noncanonical bracket in section [ subsec : exmhdelmap ] . this derivation takes more work than that in [ subsec : hallelmap ] , but the procedure is identical , so this time we omit the details . we conclude in section [ sec : conclusion ] . as mentioned , all four models have been put into hamiltonian form . by that we mean that there is a functional @xmath3 and a bracket @xmath4 so that the time derivative of an arbitrary functional is given by @xmath5 the functionals are integral expressions of the field variables ; for instance , the hamiltonian functional is @xmath6 , where @xmath7 is an energy density . to separate out the time evolution of the field variables ( like a momentum @xmath8 ) , one can use a test functional such as @xmath9 the bracket can be expressed as @xmath10 where the @xmath11 denotes functional differentiation and the components of @xmath12 represent the field variables . the differential operators @xmath13 must be chosen so that the bracket satisfies its usual properties ( here @xmath14 , @xmath15 , and @xmath16 are functionals , and @xmath17 and @xmath18 are real numbers ) : @xmath19 of these , only the last , the jacobi identity , proves difficult to confirm . thankfully , the method of this paper provides a relatively easy way to confirm it . in lagrangian coordinates , where every fluid element is given a distinct label @xmath20 and the equations of motion are evaluated at fixed label ( i.e. for a given fluid element ) , the bracket will be canonical : @xmath21 where @xmath22 is the coordinate at fixed label @xmath20 , and @xmath23 is its conjugate momentum , which can be obtained via an action . this paper has such actions for each of its models . the jacobi identity is fairly easy to prove for the canonical bracket , relying only on the commutation of functional derivatives , @xmath24 . however , the map converting lagrangian to eulerian coordinates , in which equations are expressed at fixed spatial coordinates , produces a noncanonical bracket ; for example , even the straightforward definition @xmath25 gives an eulerian momentum dependent on both lagrangian position and momentum . the jacobi identity can be directly proven for the eulerian bracket , as was done in @xcite for relativistic mhd and @xcite for hall mhd . however , when such a bracket is produced from the canonical lagrangian one , the jacobi identity is assured , as it is invariant under coordinate changes and reductions . in this section we exhibit the already - discovered hamiltonian forms for ordinary , hall and extended mhd . it will be our goal in secs . [ sec : hallmhd ] and [ sec : exmhd ] to derive the latter two . these theories are all expressed here in terms of eulerian variables , and lagrangian equivalents will be postponed until our later discussion of their actions , where they occur more naturally . in mhd the eulerian field variables are density @xmath26 , specific entropy @xmath27 , fluid velocity @xmath28 and magnetic field @xmath29 . in the barotropic case , one can express @xmath27 as a function of @xmath26 and thereby eliminate it , but we consider the more general case . one can also use as a supplementary variable the current density @xmath30 , using ampre s law in the absence of displacement current . particle number and entropy are conserved and advected , respectively : @xmath31 the fluid velocity obeys the following momentum equation : @xmath32 while the magnetic field s evolution is determined by ohm s law for a perfect conductor : @xmath33 as can be seen by taking its curl and applying faraday s law : @xmath34 in hamiltonian mhd , while one can express the hamiltonian and bracket in terms of @xmath28 and @xmath27 ( along with @xmath29 and @xmath26 ) , their derivation turns out to be simpler when using the momentum density @xmath35 and the entropy density @xmath36 . in terms of these variables the hamiltonian is the total energy @xmath37 and the bracket is @xmath38 this bracket also constitutes the bulk of the hall and extended mhd brackets . it was first given in ref . @xcite , and the sign convention used is from that paper . hall mhd differs from ordinary mhd in that the difference between ion and electron velocities is no longer neglected in ohm s law . the derivation then modifies that law to @xmath39 where @xmath40 is the number density , @xmath41 is the particle mass ( here equal to the ion mass ) and @xmath42 is the electron pressure . the entropy , continuity , and momentum equations are unchanged , as is the hamiltonian . the bracket , in turn , now has the additional term : @xmath43 this bracket was first described by @xcite , and it requires the assumption of a barotropic electron pressure . later we will see that this assumption gives us an advected magnetic flux , while barotropic ion pressure gives a second advected quantity . these will be a necessary part of our construction of the bracket . in extended mhd , one additionally retains terms of first order in @xmath44 during the derivation , producing a new momentum equation : @xmath45 and a new version of ohm s law : @xmath46 here the term @xmath47 refers to the symmetric tensorial outer product . the bracket and hamiltonian , however , are more compactly expressed in terms of @xmath48 . the hamiltonian , which now includes a term for the electron kinetic energy , is @xmath49 in light of the mhd ampre s law @xmath50 . its bracket , in turn , is @xmath51 this bracket was first given in ref . it also requires barotropic electron pressure @xmath52 . the structural similarity between the hall and extended mhd poisson brackets was investigated in ref . @xcite , and will be elucidated further below . the essential difference between the various mhd models lies in their flux conservation laws , each one having a different version . the archetypal flux conservation law is that of ordinary mhd , @xmath53 @xcite . here the @xmath20 variables are coordinates in a label space @xmath54 , whose continuous values identify fluid elements at @xmath55 ( this condition can be relaxed , as in @xcite ) . meanwhile , the coordinates @xmath56 describe the point to which a specific element flows ; thus , @xmath57 . in addition , @xmath58 while @xmath59 . more explicitly , we write the flux conservation law as @xmath60 this expression can be manipulated into a transformation rule for the magnetic field : @xmath61 where @xmath62 is the jacobian determinant of the invertible transformation from @xmath20 to @xmath63 . , while the magnetic flux is dragged to @xmath64 . their difference is @xmath65 . ] there are two distinct ways one can modify the flux conservation law . first , one can advect a flux different from that of @xmath29 ; with an appropriate choice of this flux , one then gets 2d inertial mhd @xcite . second , the same flux can be advected , but along a path distinct from that of the fluid . this second approach gives hall mhd . specifically , while the fluid itself flows from @xmath20 to a point @xmath56 , the flux element moves from @xmath20 to a distinct point @xmath64 , as illustrated by figure [ labelspacefigure1 ] . flux conservation is now @xmath66 which gives rise to the transformation rule @xmath67 the flux jacobian @xmath68 is also invertible , and can be written @xmath69 from which one can derive the expression @xmath70 . taking a full time derivative of @xmath71 in equation gives @xmath72 this equation shows that @xmath29 is advected along @xmath73 as the vector dual to a 2-form , as desired . since @xmath29 is divergenceless , we can add a term proportional to @xmath74 and put the equation in the more familiar faraday form @xmath75 so far , so good . however , complications arise when you look for the other equations of motion . some fluid attributes ( density , specific entropy ) are transported along the flow lines @xmath63 , not @xmath73 : mass conservation is described by @xmath76 , and entropy conservation by @xmath77 , recalling that no dissipative terms have been added . as a result , the label corresponding to the magnetic field will differ from the label on the other quantities . this situation is shown in figure [ labelspacefigure2 ] . , where a flux has also flowed from a different initial point @xmath78 ; thus @xmath79 . ] in this figure , the fluid element labelled by @xmath20 flows to @xmath56 , while a different label @xmath78 shows the origin of the flux element that has been advected to @xmath80 . for future use we will need two additional quantities : the point @xmath81 , to which the @xmath78 element flows , and the difference @xmath82 between @xmath83 and @xmath81 . all these quantities are related via @xmath84 more relations are available , for example @xmath85 . in principle we could eliminate all but two of the quantities , but it is simpler to keep the extras around . we also note that , for hall mhd , @xmath63 corresponds to ion flow and @xmath73 to electron flow , so they might also have been written @xmath86 and @xmath87 ; however , in extended mhd , we will use similar quantities , but they will now differ slightly from the electron and ion paths . thus we use a convention that will be appropriate for both models . we will also adopt the convention of using superscripts to show vectorial indices , and subscripts to show other attributes , like species identity or initial condition . there are cases where the distinction between vectors and covectors ( and thus raised and lowered indices ) matters , such as when using curved coordinates , but it may easily be reinstated when needed . we have , however , left the distinction intact in the expressions @xmath88 and @xmath89 , where it improves readability . every point corresponds to two labels . in hall mhd , using the convention described above , unprimed labels will denote ion quantities : for example , the number density is advected along the fluid lines , which in our approximation are the ion flow lines . meanwhile , primed labels will denote electron quantities , such as the magnetic flux density advected along electron flow lines . in light of fig . 2 above , the variable @xmath63 will appear as both @xmath56 and @xmath81 , as will some quantities ( namely the potentials @xmath90 and @xmath91 ) dependent on them . to simplify expressions we will write @xmath92 , @xmath93 , @xmath94 and @xmath95 , with unprimed expressions such as @xmath63 denoting unprimed quantities like @xmath56 . if we treat primed and unprimed quantities separately , then the full euler - lagrange equations , using lagrangian density @xmath96 , will be @xmath97_{a ' = q_f^{-1}(q(a , t),t ) } & = 0 \ ; . \nonumber\end{aligned}\ ] ] with a similar expression for @xmath98 . many of the terms in the euler - lagrange equations are superfluous : only the first four terms will contribute in the @xmath63 variation , and only the second and fourth terms in the @xmath98 one . these euler - lagrange equations can be obtained via dirac delta function manipulations on a six - dimensional label space : @xmath99 but for the most part , we will omit this consideration and work with . however , we emphasize one peculiarity of the action : it only works if the delta function integral is performed after the variations . if one does so before varying , collapsing back down to a single label space , the variational principle no longer gives the correct expressions . this peculiarity is shared by the mixed lagrangian - eulerian approach of ref . @xcite . if it were written in terms of ion and electron velocities @xmath86 and @xmath87 , the lagrangian density would be standard : @xmath100 in hall mhd , we treat electron velocity as being different from ion velocity ( unlike in regular mhd ) , but nonetheless neglect terms of order @xmath101 . the variables used will be center - of - mass velocity @xmath102 , and the drift velocity of electrons relative to ions , @xmath103 . in terms of ion and electron velocities we have @xmath104 inverting these equations and neglecting terms of the order of the mass ratio , we have @xmath105 thus , rewriting , setting @xmath106 , and noting the distinction between primed and unprimed labels , the lagrangian density becomes @xmath107 \nonumber \\ & - e n_0 \left[\phi(q , t ) - \phi(q ' + q'_d , t ) \right ] - n_{0 } \left [ u_i \left ( \frac{n_{0}}{\mathcal{j } } , s_{(0)i } \right ) + u_e \left(\frac{n_0}{\mathcal{j}_f } , s_{(0)e } \right ) \right ] \ ; . \label{tangentspaceaction}\end{aligned}\ ] ] in the @xmath63 equation of motion , the terms arising from @xmath108 cancel , plus most of the terms coming from @xmath109 , due to the @xmath110 evaluation . the only surviving term comes from the advective parts of @xmath111 , which are different for the two terms . an additional term arises from the @xmath112 partial derivative on @xmath113 . setting @xmath114 , @xmath115 , and @xmath116 , we have , for the @xmath63 equation of motion , @xmath117_{a ' = q_f^{-1}(q(a , t),t ) } = 0\ ] ] which can , by multiplying with @xmath118 and using @xmath119 , be simplified to @xmath120 which is the lagrangian equivalent of . in the @xmath98 equation of motion , the three final terms come from the full derivative @xmath121 , and the pressure term comes from the @xmath122 dependence of @xmath68 : @xmath123 with the whole thing evaluated at @xmath124 as usual . reordering and simplifying , one finds @xmath125 which is ohm s law for hall mhd . finally , the canonical momenta are @xmath126 with the lagrangian function @xmath127 defined by @xmath128 . the expression for @xmath129 will allow us to convert the thus - far omitted field term @xmath130 into a term expressed by fluid quantities , once we switch to a phase - space action . it turns out to be easy to translate the mixed eulerian - lagrangian terms of ref . @xcite into the appropriate terms of , when one minds the difference between primed and unprimed labels . we translate that paper s @xmath131 and @xmath132 into our variables using @xmath133 and @xmath134 . then its mixed ion terms in the lagrangian are @xmath135 and its electron terms are @xmath136 however , the action in ref . @xcite also contains a fully eulerian term @xmath137 which is used to produce @xmath138 , a missing piece in our fully lagrangian tangent space action . we also can not perform the usual legendre transform , because we have no expression @xmath139 . fortunately , we can solve these problems by switching to a phase space action and invoking . the four variations of this action give all the needed equations . the needed action density is @xmath140 \ ; . \label{phasespacedensity}\end{aligned}\ ] ] the middle term , note , is simply @xmath130 . we have expanded it using to express the magnetic field in terms of its initial value , and then applying to express this initial value as the curl of that of a canonical momentum . there are four phase space variations ; as when using , one sets @xmath110 after taking variations . thus the @xmath141 variation gives @xmath142 the @xmath129 variation involves an integration by parts on the middle term of the density , giving @xmath143 i.e. @xmath30 , the missing piece of our earlier tangent space action . note that the @xmath144 term requires varying @xmath145 , which is the value of @xmath129 on the boundary at @xmath146 . this is permitted since the action principle only requires @xmath147 on the boundary in order to perform an integration by parts , while the momenta are free to vary at @xmath55 . once again , most of the terms vanish in the @xmath63 variation . the @xmath148 terms in the middle term of give two factors of @xmath149 , and the @xmath68 in the same term gives a factor of @xmath150 . the remaining terms proceed similarly as in our tangent space calculation . the overall result is @xmath151 + \mathcal{j } \partial^i p = 0 \ ; , \ ] ] where @xmath152 is the cofactor matrix to @xmath153 . given @xmath154 and the @xmath155-@xmath155 identity , is the same as . finally , the @xmath98 variation gives @xmath156 + n_0 e\partial^i \phi + \mathcal{j}_f \partial^i p_e = 0 \ ; .\ ] ] considering that @xmath157 , and @xmath158 will thus have two terms , this equation is identical to . however , the action is still slightly unsatisfactory , because we use the quantities @xmath91 and @xmath90 , which are not fully determined by : namely , their gauge freedom remains . we did use the relation ( [ hallcanonicalmomenta ] ) , viz . @xmath157 , from the tangent space action to construct the phase space action ; however , does not produce this relation , and neither action gives us the evolution of @xmath90 . [ subsec : hallliegauge ] develops a gauge condition which resolves this problem in an elegant manner . look at the hall mhd ohm s law in eulerian coordinates : @xmath159 using @xmath160 and reordering , it becomes @xmath161 and , for a barotropic plasma in which @xmath52 , taking the curl renders it into the form @xmath162 with @xmath163 . this equation is in the form of , showing that the components of @xmath29 are dual to those of a two - form which is lie - dragged by @xmath164 . it would be even more convenient for @xmath91 to be the components of a lie - dragged one - form , with @xmath29 dual to the components of its exterior derivative . because the last two terms of are curl - free , they can be expressed as a gradient : @xmath165 . we then use the gauge freedom in @xmath90 to set @xmath166 which we call the lie gauge , due to the fact that it will produce a lie - dragging equation . with this gauge equation becomes @xmath167 or @xmath168 so that the vector potential is now a lie - dragged one - form , as desired . in fact , there exists an entire family of gauges that result in a lie - dragged one - form . suppose @xmath91 is one such member ( like that already provided ) , and @xmath169 is a lie - dragged zero - form , so that @xmath170 and @xmath171 let @xmath172 . then , starting from @xmath173 we have @xmath174 collecting the @xmath169 terms inside an overall gradient operator and applying eliminates all of them , showing that @xmath175 is also an advected one - form . lie - dragging of @xmath91 as a one - form implies that @xmath176 , thus @xmath177 where @xmath178 is the cofactor matrix of the coordinate transformation @xmath179 . because of the relation @xmath180 , the canonical momentum also transforms as a one - form : @xmath181 using the lie gauge , one can eliminate @xmath90 from the phase space action , and using one can also eliminate @xmath182 in favor of its initial value @xmath183 at @xmath146 . however , the other appearance of the vector potential is @xmath184 in is written in terms of `` ion quantities '' ( i.e. unprimed variables ) , whereas expresses it using solely the `` electron quantity '' ( i.e. primed variable ) @xmath185 . thus , we ve only solved half the problem : we ve expressed @xmath186 in terms of @xmath187 and eliminated @xmath188 with a gauge condition , but we re still left with @xmath189 and @xmath190 . thankfully , there is a general result showing that , in a system of @xmath1 charged fluid species with barotropic equations of state , there are @xmath1 conserved helicities @xcite and @xmath1 lie - dragged two - forms @xcite . in mhd , this duality is of no concern , because the two collapse and give rise to a single magnetic helicity @xmath191 ; however , in more general models they remain distinct . we can use the other helicity to eliminate the last two extraneous variables . hall mhd has the following variable as its second lie - dragged two - form : @xmath192 its advection is straightforward to prove , using eulerian variables . taking the time derivative , and remembering our assumption of barotropic pressures , @xmath193 \\ = & \ ; \nabla \times \left [ \mathbf{v } \times \left(\mathbf{b } + \frac{cm}{e } \nabla \times \mathbf{v } \right ) \right ] = \nabla \times \left(\mathbf{v } \times \mathbf{\mathcal{b } } \right ) = -\pounds_{\mathbf{v } } \mathbf{\mathcal{b } } \ ; .\end{aligned}\ ] ] since @xmath194 is divergenceless , it can be expressed as the curl of a vector @xmath195 . a fully general expression for such a vector is @xmath196 just as @xmath90 was for @xmath91 , @xmath169 can be chosen to make @xmath195 a lie - dragged one - form , expressed as @xmath197 for we have @xmath198 \ ; . \nonumber\end{aligned}\ ] ] following the reasoning that motivated the lie gauge , we note that the expression in square brackets is the equivalent of @xmath199 from before . we can thus get a lie - dragged one - form @xmath195 by setting @xmath200 + \frac{1}{ne } \nabla p_e - \frac{1}{ne } \nabla p = - \nabla \left(\frac{\mathbf{v } \cdot \mathbf{\mathcal{a}}}{c}\right ) \ ; .\ ] ] using the lie gauge , this equation simplifies to @xmath201 solved by the action - like quantity @xmath202 just as before , you can add a lie - dragged zero - form to @xmath169 and still have @xmath195 lie - dragged . however , tantalizing though the action - like expression is , we needed an action expressed entirely in terms of ion quantities and electron quantities , while the above one is mixed due to @xmath203 being an electron quantity . thus we go back to line , which can be simplified slightly to @xmath204 where @xmath205 is the ion pressure . except for @xmath90 , which is still an electron quantity , and @xmath169 , which is mixed , these are all ion quantities . thus we can create the following quantity : @xmath206 which is an ion quantity because all the terms on the right hand side are . with the four quantities @xmath91 , @xmath207 , @xmath90 and @xmath208 , obeying transformation rules like and subject to the lie gauge , the potential terms in the action can be expressed entirely in terms of their initial conditions , solving the problem mentioned at the end of sec . [ subsec : hallactions ] . the new ion variables @xmath207 and @xmath208 introduced in this section deserve a bit more attention . by writing the expression @xmath209 we can say , loosely , that @xmath208 is to @xmath195 what @xmath90 is to @xmath91 . the parallel is further reinforced by the equivalent of faraday s law , @xmath210 and an easily derived expression @xmath211 reminiscent of ohm s law , but with an extra gradient term . were all the time - dependent terms to be removed from @xmath212 , would be a generalization of bernoulli s law . we conclude this section with two observations . first , we can combine , , and @xmath213 to show that @xmath214 which is a nicely compact proof of the conservation of magnetic helicity . the similar expressions derivable for the ion quantities can similarly be combined to read @xmath215 second , everything that has been said in this section applies to ideal mhd as well , which simply requires one to use @xmath28 instead of @xmath164 and to remember that there is only one distinct helicity . our phase - space action principle may equivalently be expressed as the set of hamilton s equations @xmath216 , for arbitrary functionals @xmath14 of the phase - space variables . the bracket in this case is the canonical one : @xmath217 in this section we will show how to convert this bracket into the noncanonical bracket . the eulerian quantities @xmath26 , @xmath218 , and @xmath219 are defined via standard euler - lagrange maps : @xmath220 the variable @xmath218 is superfluous for barotropic hall mhd , but it is included here for the sake of generality . when we induce variations later on , the quantities @xmath26 and @xmath218 will only have @xmath221 variations ( from the delta functions ) , while @xmath222 will have a @xmath221 and @xmath223 variation . the odd one is the magnetic euler - lagrange map : @xmath224 this will have @xmath63 and @xmath98 dependence via @xmath73 , and @xmath129 dependence via @xmath225 . we can now show how the eulerian variables change under variations in the lagrangian phase - space ones , using and : @xmath226 note that the addition of @xmath98 and @xmath129 , which do not appear in regular mhd , nonetheless do not require us to add any new eulerian variables . they do , however , add a new term in the variation @xmath227 that does not appear in ideal mhd , because now @xmath29 has a @xmath129 dependence via @xmath228 . the variation induced by an arbitrary function @xmath14 , in both lagrangian and eulerian variables , is @xmath229 substituting the various , except for the one term involving @xmath230 ( which will require more careful attention ) , into the left side of gives the expression @xmath231 d^3 x \ : d^3 a \ ; .\end{aligned}\ ] ] in this expression , the disappearance of @xmath78 is rather startling , but it is still there implicitly via the delta functions , for at a fixed @xmath232 they will pick out values of @xmath20 for the magnetic terms distinct from those of the other terms . meanwhile , the term that we omitted is , by using @xmath233 , given by @xmath234 here the @xmath235 term in the integration by parts vanishes because it is a symmetric object contracted with an antisymmetric one , and the second factor of @xmath236 appears because we want the delta - function derivative to give a derivative with respect to @xmath63 ( and thus @xmath232 ) . these factors may be eliminated in the following manner : @xmath237 thus , using , that portion of the @xmath238 variation becomes @xmath239 comparison of the expanded eulerian @xmath238 with the right side of then gives expressions for the lagrangian functional derivatives in terms of the eulerian ones : @xmath240 \delta(x - q ) \\ & + \mathcal{j}_f \left [ b^j \frac{\partial}{\partial x^i } \left(\frac{\delta f}{\delta b^j}\right ) - b^j \frac{\partial}{\partial x^j } \left(\frac{\delta f}{\delta b^i}\right ) \right ] \delta(x - q - q_d ) \ : d^3 x \\ f}{\delta q_d^i } = & - \int \frac{\delta f}{\delta b^j } b^k_0 \frac{\partial q^j_f}{\partial a^k } \delta'_i \left(x - q - q_d \right ) - \frac{\delta f}{\delta b^i } b^k_0 \frac{\partial q^j_f}{\partial a^k } \delta'_j \left(x - q - q_d \right ) d^3 x \\ = & \int \mathcal{j}_f \left [ b^j \frac{\partial}{\partial x^i } \left(\frac{\delta f}{\delta b^j}\right ) - b^j \frac{\partial}{\partial x^j } \left(\frac{\delta f}{\delta b^i}\right ) \right ] \delta(x - q - q_d ) \ : d^3 x \\ f}{\delta \pi^i_d } = & \int \frac{\delta f}{\delta b^j } \frac{c}{2 n_0 e } \mathcal{j}_f \ : \epsilon^{ijk } \delta'_{k } ( x - q - q_d ) \ : d^3 x \\ = & \ ; \frac{c}{2 n e } \int \left(\nabla \times \frac{\delta f}{\delta \mathbf{b } } \right ) _ i \ : \delta ( x - q - q_d ) \ : d^3 x = - \frac{c}{2 n e } \left . \left(\nabla \times \frac{\delta f}{\delta \mathbf{b } } \right ) _ i \ ; \right|_{x = q(a , t ) + q_d(a , t)}\end{aligned}\ ] ] finally , we can insert these functional derivatives into the canonical lagrangian bracket . evaluating the delta function introduces a factor of @xmath241 or @xmath242 , eliminates the @xmath243 integral and converts the remaining lagrangian quantities into eulerian ones : @xmath244 d^3 x \nonumber \\ \equiv & \quad \{f , g\}_{mhd } + \{f , g\}_{hall } \nonumber \ ; .\end{aligned}\ ] ] here the @xmath245 terms are those in the square bracket , and the remaining @xmath246 terms are familiar from ordinary mhd . the hall portion of the bracket can be greatly simplified . take the two terms involving the curl of @xmath247 . they become @xmath248 \\ = \ ; & \frac{c}{2 n e } \left[b^j \left(\nabla \times \frac{\delta f}{\delta \mathbf{b}}\right)^i \ ; \epsilon_{mij}\epsilon^{mkl } \frac{\partial}{\partial x^k}\left(\frac{\delta g}{\delta b^l}\right)\right ] \\ = \ ; & - \frac{c}{2 n e } \mathbf{b } \cdot \left[\left(\nabla \times \frac{\delta f}{\delta \mathbf{b}}\right ) \times \left(\nabla \times \frac{\delta g}{\delta \mathbf{b}}\right)\right ] \ ; .\end{aligned}\ ] ] the other two terms give an identical expressions ; together , they eliminate the factor of @xmath249 and reproduce the hall mhd bracket . before we move on to produce results for extended mhd , we should pause a moment to discuss our peculiar method of introducing a phase - space constraint . the simplest phase - space action is a finite - dimensional particle one extremizing @xmath250 with fixed endpoints @xmath251 and @xmath252 . when doing the @xmath63 variations , an integration by parts must be performed , so @xmath253 on the endpoints of the action integral ; however , when varying @xmath254 , no integration by parts is required , so the momenta can vary on the endpoints . in our lagrangian density , the @xmath63 , @xmath141 , and @xmath98 variations occur as normal . however , @xmath129 has been expressed entirely in terms of its initial value @xmath145 . thus , when doing the @xmath129 variation , one _ only _ varies at the endpoints ( here the initial surface @xmath146 ) , with the variation at @xmath255 determined , ultimately , by via . the same substitution for @xmath129 in terms of its initial value occurs in the magnetic euler - lagrange map , making it crucial for the derivation of the hall mhd bracket . we consider the successful derivation of the bracket to be a sign of this constraint s validity . however , viewing it as a specific instance of a more general method ( hopefully with applications elsewhere in hamiltonian physics ) , it is clear that we have not established the full conditions under which this method may be applied . we hope to do so in future work . in the course of writing an action for extended mhd , we will need to write the field portions in terms of an advected two - form , which will be a vorticity - like quantity . unfortunately , this time around the magnetic field @xmath29 is not such a quantity . we will thus begin by showing how to derive a pair of vorticity equations in extended mhd . written in a standard fashion , the two central equations for extended mhd are the momentum equation and the generalized ohm s law , rounded off with ampere s law and the continuity equation , plus the isentropy equation if needed . our goal will be to use these equations to derive a pair of vorticity equations @xmath256 previous experience with ideal and hall mhd suggests that there should be such equations , along with the result that a theory of @xmath1 charged fluids will have @xmath1 such vorticities . moreover , it was shown in ref . @xcite that they do exist , by exploiting a map between the poisson brackets for hall and extended mhd . however , our purpose in this paper is to derive those very brackets , so we should not rely on knowledge drawn from those brackets . thus we will show how to derive the equations directly . in ref . @xcite , the form of the generalized ohm s law is derived from the equivalent expression @xmath257 by combining the last two terms and adding a term proportional to @xmath258 ( which is zero ) . instead of doing that , we combine the first term , the last term , and the term proportional to @xmath259 into @xmath260 by using the continuity equation . we can then replace all occurrences of @xmath261 with @xmath262 , which has units of velocity , producing @xmath263 we next apply the @xmath264 identity and switch to the new field variable @xmath265 to produce @xmath266 \\ & \ ; + \frac{\mathbf{u}\times\mathbf{b}_\star}{c } - \frac{\nabla p_e}{n e } \ ; . \label{ohmslawu}\end{aligned}\ ] ] performing similar operations on the momentum equation produces @xmath267 equations and , being slightly more compact than and , suggest that @xmath268 and @xmath269 are indeed more natural variables than @xmath261 and @xmath29 . the various gradients in and are crying out for us to take a curl , so we will . using faraday s law and collecting the time derivative terms in one place , becomes @xmath270 - \frac{c}{ne^2 } \left(\nabla p_e \times \nabla n\right)\ ] ] while becomes @xmath271 + \frac1{mn^2 } \left(\nabla p \times \nabla n\right ) \ ; .\ ] ] going forward we assume barotropic equations of state for both the electrons and ions , because then @xmath272 and the pressure terms drop . in and , shorn of their pressure terms , we have all the ingredients we need to make two equations of the form . we assume that the quantities appearing in the vorticity equations are linear combinations of those that have appeared in deriving and : @xmath273 the coefficients are all unitless . expanding , we have @xmath274 \ ; . \label{vorticityexpansion}\end{aligned}\ ] ] on the other hand , we can also use linear combinations of and to express @xmath275 . equating the resulting coefficients with what we find in , we have the system of equations @xmath276 one equation is redundant , which should be no surprise since @xmath277 is only established up to an overall scale . we use this extra freedom to set @xmath278 . then the solutions are @xmath279 , and @xmath280 where @xmath281 is the electron - ion mass ratio . this confirms what was found in ref . we can also invert to get @xmath282 because the @xmath277 are divergenceless , they can each be written as the curl of a vector potential @xmath283 . accounting for the gauge freedom , these potentials are @xmath284 thus , taking appropriate linear combinations of and , we find @xmath285 taking a cue from the hall mhd results , we set @xmath286 and @xmath287 . we also note @xmath288 , @xmath289 , and @xmath290 . together , all these identities allow a considerable simplification : @xmath291 thus , instead of the differing and , or the greatly different and , we have two highly symmetric versions of ohm s law expressed using our beloved advected quantities . because we derived only by taking linear combinations and applying lots of vector identities , no information has been lost , and the pleasant equations are fully equivalent to the messy and . finally , in the @xmath292 limit , @xmath293 and @xmath294 , which explains why the @xmath295 term appears in the hall mhd momentum expression but not in its standard ohm s law . we now turn to the problem of writing a fully lagrangian action principle for extended mhd . as in ref . @xcite , this is done by retaining terms up to first order in @xmath281 in the action , and taking @xmath296 after variations . the coordinate change expressed in now becomes @xmath297 expanding the standard lagrangian and keeping terms up to first order in @xmath0 gives a new lagrangian density @xmath298 \label{extendedtangentaction } \\ & + \frac{e n_0}{c}\left [ \mathbf{\dot{q}}\cdot\mathbf{a } - \mu \mathbf{\dot{q}}_d \cdot \mathbf{a } - \mathbf{\dot{q } } ' \cdot \mathbf{a } ' - ( 1-\mu ) \mathbf{\dot{q}}'_d \cdot \mathbf{a } ' \right ] \nonumber \\ & - e n_0 \left[\phi(q , t ) - \phi(q ' + q'_d , t ) \right ] - n_{0 } \left [ u_i \left ( \frac{n_{0}}{\mathcal{j } } , s_{(0)i } \right ) + u_e \left(\frac{n_0}{\mathcal{j}_f } , s_{(0)e } \right ) \right ] \nonumber \ ; .\end{aligned}\ ] ] here @xmath299 and @xmath300 . while there are many new terms in this lagrangian , remarkably few make it to the actual equations of motion . this is partly due to the cancellations that occur ( as in hall mhd ) after setting @xmath301 , and partly due to the ordering @xmath296 imposed after variations . the @xmath63 variation gives @xmath302 and the @xmath98 variation gives @xmath303 these are the correct equations , given that @xmath304 and @xmath305 will be complicated expressions when eulerianized , particularly the latter . see sec iv.a of ref . @xcite for a detailed explanation on how to convert them into eulerian form . the lagrangian density produces canonical momenta @xmath306 thus @xmath307 the @xmath277 are advected by @xmath308 , whose lagrangian equivalents we will call @xmath309 , where the minus sign comes because @xmath268 and @xmath310 differ by a sign ( see ) . thus we have these fully expanded expressions for @xmath277 , in analogy with : @xmath311 where @xmath312 are the integral lines of @xmath313 , and @xmath314 are the determinants of the matrices @xmath315 . producing a phase - space action will involve bringing in the omitted term @xmath316 ( see the note about this term in the hall mhd section ) , and converting the terms @xmath317 and @xmath318 to field terms by anticipating the relation @xmath319 . by integrating by parts on the @xmath320 term , one can combine both of them into the single term @xmath321 . but first we need to show how to actually perform a variation on such a term . in eulerian variables we have @xmath322 so that @xmath269 is a function of @xmath29 and @xmath1 . thankfully , the situation is simpler in lagrangian variables . the variable appears inside a @xmath323 integral , contracted with another vector . one can integrate by parts to remove a curl , swap in @xmath324 ( any term involving @xmath269 will be an electron quantity ) , and integrate by parts again , so that @xmath325 whenever it appears inside a label space integral . thus we have @xmath326 in lagrangian variables . we assume the differential equation can be inverted to produce @xmath327 for some tensorial green s function @xmath328 . therefore , when varying @xmath329 , the result is @xmath330 where we have used the symmetry @xmath331 of green s functions . the variation did not affect @xmath332 despite the dependence on @xmath63 because the green s function is translation invariant , i.e. @xmath333 . we are now in a position to write the full phase - space lagrangian . it is @xmath334 \delta(a ' - q_f^{-1}(q(a , t),t ) ) \nonumber \\ & + \ ; \frac{j_f}{8\pi ( \delta \beta)^2 } \left [ \frac{\beta_-}{\mathcal{j}_+ } \left(\nabla \times \pi_d - \beta_+ \nabla \times \pi \right)^k \left(\frac{\partial q^i}{\partial a^k } - \alpha_+ \frac{\partial q^i_d}{\partial a^k}\right ) \right . \nonumber \\ & \ ; \left . \quad - \frac{\beta_+}{\mathcal{j}_- } \left(\nabla \times \pi_d - \beta_- \nabla \times \pi \right)^k \left(\frac{\partial q^i}{\partial a^k } - \alpha_- \frac{\partial q^i_d}{\partial a^k}\right ) \right ] \label{extendedphasespaceaction } \\ & \ ; \times \left [ \frac{\beta_-}{\mathcal{j}_+ } \left(\nabla \times \pi_d - \beta_+ \nabla \times \pi \right)^l \left(\frac{\partial q^j}{\partial a^l } - \alpha_+ \frac{\partial q^j_d}{\partial a^l}\right ) \right . \nonumber \\ & \ ; \left . \quad - \frac{\beta_+}{\mathcal{j}_- } \left(\nabla \times \pi_d - \beta_- \nabla \times \pi \right)^l \left(\frac{\partial q^j}{\partial a^l } - \alpha_- \frac{\partial q^j_d}{\partial a^l}\right ) \right ] g^{ij}(q;q ' ) \ ; \mathrm{d}^3 a \ ; \mathrm{d}^3 a ' \ ; , \nonumber\end{aligned}\ ] ] where we have set @xmath335 and expanded @xmath329 using , the first equation of , and . thankfully , variations are simplified considerably by the result . it is also interesting that the usual delta function is replaced by a more specialized green s function in the @xmath329 term . it is worth pointing out that , while we assumed barotropic equations of state in our development , the above action works for the more general equations of state @xmath336 . the @xmath141 variation gives @xmath337 the @xmath141 variations occurring in the big field term all cancel , which is not surprising since @xmath269 has no @xmath141 dependence . the @xmath129 variation gives @xmath338 as it should , with the factor of @xmath68 being absorbed back into @xmath339 when invoking . after some work , the @xmath63 variation gives @xmath340 + \mathcal{j } \nabla^i p = 0 \ ; .\ ] ] imposing the @xmath296 condition turns this into @xmath341 + \mathcal{j } \nabla^i p = 0 \ ; .\ ] ] finally , the @xmath98 variation gives @xmath342 \ ; .\end{aligned}\ ] ] to simplify this , note that @xmath343 now @xmath344 , so we end up dropping the @xmath345 term . next @xmath346 , so we get a plain @xmath347 term , which again gets reduced to @xmath348 . in the end , @xmath349 + n_0 e\partial^i \phi + \mathcal{j}_f \partial^i p_e = 0\ ] ] remarkably , equations , , , and are identical to their equivalents , , and from hall mhd . this is likely the source of the maps discovered in ref . the differences between hall and extended mhd arise when you switch to eulerian variables . the field variable @xmath350 will be written as a linear combination of the two - forms @xmath351 , each of which is advected along a linear combination of @xmath63 and @xmath98 . thus will be rewritten as @xmath352 the coefficients @xmath353 are given by , and those in front of the @xmath354 come from inverting , writing @xmath355 . the minus signs in front of the various @xmath353 arise because @xmath356 and @xmath268 ( via ) differ by a sign . equation shows that @xmath357 however , despite the appearance of @xmath358 here , the expression @xmath359 is unchanged , because the terms arising from the two parts of @xmath360 cancel each other . however , @xmath361 is changed , converting @xmath245 into its extended mhd counterpart . the following changes appear in the previous calculation : ( i ) all functional derivatives with respect to @xmath29 are now done with respect to @xmath269 ; ( ii ) in the magnetic portion of @xmath246 , @xmath29 is replaced by @xmath269 ; ( iii ) in @xmath245 , @xmath29 is replaced by @xmath362 ( \beta_- \alpha_+ \mathbf{b}_+ - \beta_+ \alpha_- \mathbf{b}_-)$ ] . this quantity works out to be @xmath363 . thus we derive the following bracket : @xmath364 with @xmath246 given by the first part of and @xmath365 \ : d^3x,\ ] ] matching . as expected , the limit @xmath366 reduces @xmath269 to @xmath29 , eliminates the @xmath2 term , and thus reduces @xmath367 to @xmath245 . we have accomplished many things in the course of deriving the noncanonical brackets for hall and extended mhd . the need for canonical momenta to serve as the backbone of the brackets led to actions for both theories : first the tangent - space ones and , then the phase - space ones and . essential to the actions were the advected generalized vorticities : the magnetic field and for hall mhd , and the two expressions for extended mhd . to go with these advected two - forms , we found advected one - form potentials and . for hall mhd , we found that the momentum equation could be restated in the form of an additional equation resembling ohm s law . similarly , in extended mhd , after recasting the complex original equations and into variables based around the various advected forms , we could produce the equivalent , but much simpler expressions . knowing these forms , we were also able to define the natural euler - lagrange maps and , and at last derive the noncanonical brackets and . a number of interesting concepts had to be used in order to produce these results . to begin with , the actions required a double label space in order to be fully lagrangian , and we may speculate that lagrangian theories incorporating @xmath1-fluid effects will require @xmath1 label spaces . moreover , padhye and morrison @xcite showed that , in ideal mhd , magnetic helicity is the noether invariant corresponding to the symmetry of relabelling . in each of hall and extended mhd , we have two label spaces , and we speculate that helicities corresponding to each theory s two generalized vorticities will arise from distinct relabelling symmetries on the doubled label space . this is a matter for future research . next , our expanded inventory of two - forms and one - forms , defined by their advection properties , allowed us to greatly simplify extended mhd . this shows that a firm understanding of the geometric nature of the objects appearing in a physical system will allow one to cut away much of its seeming complexity , as briefly noted in @xcite . finally , our implementation of the euler - lagrange map ( and the phase - space action principle ) required an unusual method of implementing a constraint , a method which may turn out to have broader applicability . prior to attempting work such as ours , one might have objected that hall and extended mhd are theories too specialized and inelegant to be a fruitful topic for mathematical physics . fortunately , we have found that it is precisely when investigating such specialized problems that one may discover ideas and methods useful in a broader context . from a practical perspective , we emphasize that our action principle for extended mhd ( and the concomitant noncanonical hamiltonian formulation ) can be applied to many problems of interest and relevance in diverse areas . we list a few examples in this category : topological invariants @xcite , particle relabelling symmetries @xcite , reconnection based on hamiltonian models @xcite , tearing modes @xcite , hamiltonian closures @xcite , nonlinear waves @xcite , weakly nonlinear dynamics @xcite , the derivation of gyrofluid and hybrid fluid - kinetic models @xcite , the properties of the equatorial electrojet @xcite , and the rapidly burgeoning field of variational integrators @xcite . thus , our work serves to advance and flesh out the mathematical foundations of extended mhd , whilst also paving the way for the applications of our methodology in fusion , space and astrophysical plasmas . pjm and ecd were supported by doe office of fusion energy sciences , under de - fg02 - 04er-54742 . ml was supported by the nsf grant no . ags-1338944 and the doe grant no . de - ac02 - 09ch-11466 . 59ifxundefined [ 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 link:\doibase 10.1103/physrevlett.45.790 [ * * , ( ) ] link:\doibase 10.1103/physreva.25.2437 [ * * , ( ) ] link:\doibase 10.1063/1.4890955 [ * * , ( ) ] link:\doibase 10.1002/prop.19590070902 [ * * , ( ) ] link:\doibase 10.1088/1751 - 8113/46/33/335502 [ * * , ( ) ] link:\doibase 10.1088/1751 - 8113/48/23/235502 [ * * , ( ) ] @noop _ _ ( , ) link:\doibase 10.1098/rspa.1958.0244 [ * * , ( ) ] link:\doibase 10.1016/0003 - 4916(58)90051 - 4 [ * * , ( ) ] link:\doibase 10.1063/1.1706227 [ * * , ( ) ] link:\doibase 10.1063/1.1761820 [ * * , ( ) ] link:\doibase 10.1017/s002237780002568x [ * * , ( ) ] link:\doibase 10.1063/1.1692854 [ * * , ( ) ] link:\doibase 10.1017/s0022377800006590 [ * * , ( ) ] link:\doibase 10.1017/s0022377800024685 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1063/1.1706844 [ * * , ( ) ] link:\doibase 10.1139/p63 - 216 [ * * , ( ) ] link:\doibase 10.1063/1.1692333 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1063/1.4896336 [ * * , ( ) ] link:\doibase 10.1063/1.1882353 [ * * , ( ) ] in link:\doibase 10.1063/1.3266810 [ _ _ ] , vol . , ( ) pp . link:\doibase 10.1063/1.4891321 [ * * , ( ) ] link:\doibase 10.1016/0167 - 2789(83)90017 - 9 [ * * , ( ) ] link:\doibase 10.1016/j.physleta.2014.12.008 [ * * , ( ) ] link:\doibase 10.1063/1.4926821 [ * * , ( ) ] link:\doibase 10.1016/j.physleta.2016.05.024 [ ( ) , 10.1016/j.physleta.2016.05.024 ] link:\doibase 10.1103/physrevlett.81.4863 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1016/j.physleta.2014.10.013 [ * * , ( ) ] in link:\doibase 10.1063/1.33633 [ _ _ ] , vol . , ( ) pp . link:\doibase 10.1016/0370 - 1573(85)90028 - 6 [ * * , ( ) ] @noop _ _ ( , ) link:\doibase 10.1103/physrevd.91.084050 [ * * , ( ) ] link:\doibase 10.1063/1.4819779 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.88.095001 [ * * , ( ) ] link:\doibase 10.1063/1.4931069 [ * * , ( ) ] link:\doibase 10.1016/0375 - 9601(96)00472 - 0 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physreve.92.063106 [ * * , ( ) ] @noop ( ) link:\doibase 10.1088/0029 - 5515/53/6/063024 [ * * , ( ) ] link:\doibase 10.1063/1.4821840 [ * * , ( ) ] link:\doibase 10.1063/1.4921329 [ * * , ( ) ] link:\doibase 10.1088/1751 - 8113/45/36/365502 [ * * , ( ) ] link:\doibase 10.1016/j.aop.2014.05.011 [ * * , ( ) ] link:\doibase 10.1088/1751 - 8113/48/27/275501 [ * * , ( ) ] link:\doibase 10.1016/j.cnsns.2011.04.010 [ * * , ( ) ] link:\doibase 10.1063/1.4941596 [ * * , ( ) ] link:\doibase 10.1017/s0022377810000760 [ * * , ( ) ] link:\doibase 10.1016/j.aop.2016.02.003 [ * * , ( ) ] link:\doibase 10.1088/1751 - 8113/43/37/375501 [ * * , ( ) ] link:\doibase 10.1088/0741 - 3335/56/9/095008 [ * * , ( ) ] link:\doibase 10.1063/1.4936102 [ * * , ( ) ] @noop ( ) link:\doibase 10.1063/1.4765691 [ * * , ( ) ] link:\doibase 10.1016/j.jcp.2013.01.017 [ * * , ( ) ] link:\doibase 10.1063/1.4897372 [ * * , ( ) ]